Discussion:
Electronic design
(too old to reply)
Phil Hobbs
2024-01-21 23:16:01 UTC
Permalink
JL wrote an interesting post in the depths of the "better
microelectronics from coal" thread that I thought was worth pulling out
on its own.

On 2024-01-21 10:12, John Larkin wrote:>

"...what IS electronic
design, and what's the best way to do it? <snip>
Short answer, cobbling. When presented with a problem or an
opportunity to design electronics, the most efficient way to do that
is to grab a piece of paper and immediately sketch a circuit or an
assembly. Sometimes one can do that instantly, without thinking, or
sometimes one can ignore the issue for a few days and then the design
pops up. Sometimes brainstorming and whiteboarding help. Sometimes
fiddling with Spice helps.
All that literature research and math analysis and simulation and
breadboarding and prototyping are just slow and expensive follow-up
chores for people who don't have 100% confidence in their instincts.
Analysis, sometimes prudent to do, but not design.
Design is subconsious and instinctive. And it's free! And to some
extent, it can be taught, but seldom is.
Most of us design things to sell, so do whatever works. We're selling
stuff, not publishing papers.
Hmm. I don't think that I agree in general, because you make it sound
as though the process were just intuitively plucking one idea out of
somewhere-or-other and cranking it out.

You've often argued in favor of brainstorming, where you get a few smart
people in front of a white board and try out ideas to find the best one
and flesh it out. We've done that together, very fruitfully.

It's possible to do more or less the same thing by oneself, but it
requires the ability to tolerate uncertainty for extended periods.
(That's a skill well worth developing, which most people are really,
really bad at, IME.)

I sometimes need to do a family of designs, rather than just one.
Recently I've been working on some very fast, very cheap SPAD preamps,
intended to go in the guts of positron-emission scanners.

Designs with lots of real-world constraints are often the most fun, and
this one's specs include: 300-ps edges with 100-ps timing repeatability
from unit to unit; no magnetics allowed; and a BOM cost of $1 or less.
(You need a whole lot of channels, and PET and MRI machines are often
combined.)

I do a fair amount of analysis of circuits of that sort, to figure out
what actually limits their performance. It isn't super detailed--in
this case, just enough to figure out whether it'll be the base-emitter
time constant, the Miller effect, or the SPAD's series resistance that
will be the limiting factor.

Miller, I can deal with using circuit hacks. The BE time constant is
Rbb' * Cbe, which gets slightly worse at high current, but is mainly a
device parameter--to get a big improvement you have to change
transistors. The SPAD can be negotiable depending on whose process
you're making them on--when each machine needs thousands of them,
vendors tend to listen.

Eventually, of course, you have to pick one and go with it, but picking
a topology usually takes me an iteration or two.

Cheers

Phil Hobbs
--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC / Hobbs ElectroOptics
Optics, Electro-optics, Photonics, Analog Electronics
Briarcliff Manor NY 10510

http://electrooptical.net
http://hobbs-eo.com
John Larkin
2024-01-22 00:05:07 UTC
Permalink
On Sun, 21 Jan 2024 18:16:01 -0500, Phil Hobbs
Post by Phil Hobbs
JL wrote an interesting post in the depths of the "better
microelectronics from coal" thread that I thought was worth pulling out
on its own.
On 2024-01-21 10:12, John Larkin wrote:>
"...what IS electronic
design, and what's the best way to do it? <snip>
Short answer, cobbling. When presented with a problem or an
opportunity to design electronics, the most efficient way to do that
is to grab a piece of paper and immediately sketch a circuit or an
assembly. Sometimes one can do that instantly, without thinking, or
sometimes one can ignore the issue for a few days and then the design
pops up. Sometimes brainstorming and whiteboarding help. Sometimes
fiddling with Spice helps.
All that literature research and math analysis and simulation and
breadboarding and prototyping are just slow and expensive follow-up
chores for people who don't have 100% confidence in their instincts.
Analysis, sometimes prudent to do, but not design.
Design is subconsious and instinctive. And it's free! And to some
extent, it can be taught, but seldom is.
Most of us design things to sell, so do whatever works. We're selling
stuff, not publishing papers.
Hmm. I don't think that I agree in general, because you make it sound
as though the process were just intuitively plucking one idea out of
somewhere-or-other and cranking it out.
If an idea is new, where else would come from?
Post by Phil Hobbs
You've often argued in favor of brainstorming, where you get a few smart
people in front of a white board and try out ideas to find the best one
and flesh it out. We've done that together, very fruitfully.
It's possible to do more or less the same thing by oneself, but it
requires the ability to tolerate uncertainty for extended periods.
(That's a skill well worth developing, which most people are really,
really bad at, IME.)
The uncertainty period is probably necessary, to let ones neurons
prowl the noisy solution space. The period is usually a day or two,
but can be years.

Some engineers are uncomfortable with uncertainty, and want to lock
down a design as soon as possible, preferably something sanctioned by
some authority. I like to stay confused for a while.
Post by Phil Hobbs
I sometimes need to do a family of designs, rather than just one.
Recently I've been working on some very fast, very cheap SPAD preamps,
intended to go in the guts of positron-emission scanners.
Designs with lots of real-world constraints are often the most fun, and
this one's specs include: 300-ps edges with 100-ps timing repeatability
from unit to unit; no magnetics allowed; and a BOM cost of $1 or less.
(You need a whole lot of channels, and PET and MRI machines are often
combined.)
I do a fair amount of analysis of circuits of that sort, to figure out
what actually limits their performance. It isn't super detailed--in
this case, just enough to figure out whether it'll be the base-emitter
time constant, the Miller effect, or the SPAD's series resistance that
will be the limiting factor.
Certainly quantitative reality should filter the solution space. But
even that can be mostly intuitive. I was talking about that with C on
Friday, about how some people have good quantitative intuition and
some don't. She can look at soup in a round pot and know if it will
fit into a square plastic container, to about 10%. I can do that.
Neither of our spouses can.
Post by Phil Hobbs
Miller, I can deal with using circuit hacks. The BE time constant is
Rbb' * Cbe, which gets slightly worse at high current, but is mainly a
device parameter--to get a big improvement you have to change
transistors. The SPAD can be negotiable depending on whose process
you're making them on--when each machine needs thousands of them,
vendors tend to listen.
Eventually, of course, you have to pick one and go with it, but picking
a topology usually takes me an iteration or two.
Sometimes a circuit takes me dozens, lots of sheets in the trash can.
I think it's important to give as many ideas as possible a chance.

See Barrie Gilbert's essay "Where do little circuits come from?"

"Prod and poke" and "doodling" are suggested.
Post by Phil Hobbs
Cheers
Phil Hobbs
Anthony William Sloman
2024-01-22 00:25:46 UTC
Permalink
<snip>
Post by John Larkin
Post by Phil Hobbs
Hmm. I don't think that I agree in general, because you make it sound
as though the process were just intuitively plucking one idea out of
somewhere-or-other and cranking it out.
If an idea is new, where else would come from?
Several people seem to patent the same idea at much the same time.

On at least one occasion I knew exactly why - and could put my finger on the paper that had inspired me and the guy who had got there first - who turned out to have edited the journal that published the paper, and had had to put in quite a lot of work to get the author to get the bugs out of the paper
Post by John Larkin
Post by Phil Hobbs
You've often argued in favor of brainstorming, where you get a few smart
people in front of a white board and try out ideas to find the best one
and flesh it out. We've done that together, very fruitfully.
It's possible to do more or less the same thing by oneself, but it
requires the ability to tolerate uncertainty for extended periods.
(That's a skill well worth developing, which most people are really,
really bad at, IME.)
The uncertainty period is probably necessary, to let ones neurons
prowl the noisy solution space.
Neurons don't prowl. They accept inputs and generate outputs which they pass on to other neurons. Look up neural nets sometime.
Post by John Larkin
The period is usually a day or two, but can be years.
Years usually means that the technology has had a chance to move on. A good idea I had in 1975 had to wait until 1993 until I could get my hands on a big-enough cheap programmable device to make it work cheaply.
Post by John Larkin
Some engineers are uncomfortable with uncertainty, and want to lock
down a design as soon as possible, preferably something sanctioned by
some authority. I like to stay confused for a while.
Making a virtue of necessity,

<snip>
Post by John Larkin
Sometimes a circuit takes me dozens, lots of sheets in the trash can.
I think it's important to give as many ideas as possible a chance.
There are an infinite number of ideas. You do have to be selective.
Post by John Larkin
See Barrie Gilbert's essay "Where do little circuits come from?"
"Prod and poke" and "doodling" are suggested.
Whatever works for you.
--
Bill Sloman, Sydney
legg
2024-01-22 00:38:37 UTC
Permalink
Post by John Larkin
On Sun, 21 Jan 2024 18:16:01 -0500, Phil Hobbs
Post by Phil Hobbs
JL wrote an interesting post in the depths of the "better
microelectronics from coal" thread that I thought was worth pulling out
on its own.
On 2024-01-21 10:12, John Larkin wrote:>
"...what IS electronic
design, and what's the best way to do it? <snip>
Short answer, cobbling. When presented with a problem or an
opportunity to design electronics, the most efficient way to do that
is to grab a piece of paper and immediately sketch a circuit or an
assembly. Sometimes one can do that instantly, without thinking, or
sometimes one can ignore the issue for a few days and then the design
pops up. Sometimes brainstorming and whiteboarding help. Sometimes
fiddling with Spice helps.
All that literature research and math analysis and simulation and
breadboarding and prototyping are just slow and expensive follow-up
chores for people who don't have 100% confidence in their instincts.
Analysis, sometimes prudent to do, but not design.
Design is subconsious and instinctive. And it's free! And to some
extent, it can be taught, but seldom is.
Most of us design things to sell, so do whatever works. We're selling
stuff, not publishing papers.
Hmm. I don't think that I agree in general, because you make it sound
as though the process were just intuitively plucking one idea out of
somewhere-or-other and cranking it out.
If an idea is new, where else would come from?
It's the same idea that your FNG suggested in the design review
two weeks ago; the one that you shot down in flames.

RL
John Larkin
2024-01-22 03:00:40 UTC
Permalink
Post by legg
Post by John Larkin
On Sun, 21 Jan 2024 18:16:01 -0500, Phil Hobbs
Post by Phil Hobbs
JL wrote an interesting post in the depths of the "better
microelectronics from coal" thread that I thought was worth pulling out
on its own.
On 2024-01-21 10:12, John Larkin wrote:>
"...what IS electronic
design, and what's the best way to do it? <snip>
Short answer, cobbling. When presented with a problem or an
opportunity to design electronics, the most efficient way to do that
is to grab a piece of paper and immediately sketch a circuit or an
assembly. Sometimes one can do that instantly, without thinking, or
sometimes one can ignore the issue for a few days and then the design
pops up. Sometimes brainstorming and whiteboarding help. Sometimes
fiddling with Spice helps.
All that literature research and math analysis and simulation and
breadboarding and prototyping are just slow and expensive follow-up
chores for people who don't have 100% confidence in their instincts.
Analysis, sometimes prudent to do, but not design.
Design is subconsious and instinctive. And it's free! And to some
extent, it can be taught, but seldom is.
Most of us design things to sell, so do whatever works. We're selling
stuff, not publishing papers.
Hmm. I don't think that I agree in general, because you make it sound
as though the process were just intuitively plucking one idea out of
somewhere-or-other and cranking it out.
If an idea is new, where else would come from?
It's the same idea that your FNG suggested in the design review
two weeks ago; the one that you shot down in flames.
RL
What's an FNG?

But yes, people are sometimes resistant to new ideas but if you don't
push too hard, they may come around in time, and maybe think it is
their idea.

It's best to not shoot down ideas unless they are really impossible.
But if played with, they might lead to something good.
Joe Gwinn
2024-01-22 03:08:08 UTC
Permalink
Post by John Larkin
On Sun, 21 Jan 2024 18:16:01 -0500, Phil Hobbs
Post by Phil Hobbs
JL wrote an interesting post in the depths of the "better
microelectronics from coal" thread that I thought was worth pulling out
on its own.
On 2024-01-21 10:12, John Larkin wrote:>
"...what IS electronic
design, and what's the best way to do it? <snip>
Short answer, cobbling. When presented with a problem or an
opportunity to design electronics, the most efficient way to do that
is to grab a piece of paper and immediately sketch a circuit or an
assembly. Sometimes one can do that instantly, without thinking, or
sometimes one can ignore the issue for a few days and then the design
pops up. Sometimes brainstorming and whiteboarding help. Sometimes
fiddling with Spice helps.
All that literature research and math analysis and simulation and
breadboarding and prototyping are just slow and expensive follow-up
chores for people who don't have 100% confidence in their instincts.
Analysis, sometimes prudent to do, but not design.
Design is subconsious and instinctive. And it's free! And to some
extent, it can be taught, but seldom is.
Most of us design things to sell, so do whatever works. We're selling
stuff, not publishing papers.
Hmm. I don't think that I agree in general, because you make it sound
as though the process were just intuitively plucking one idea out of
somewhere-or-other and cranking it out.
If an idea is new, where else would come from?
Post by Phil Hobbs
You've often argued in favor of brainstorming, where you get a few smart
people in front of a white board and try out ideas to find the best one
and flesh it out. We've done that together, very fruitfully.
It's possible to do more or less the same thing by oneself, but it
requires the ability to tolerate uncertainty for extended periods.
(That's a skill well worth developing, which most people are really,
really bad at, IME.)
The uncertainty period is probably necessary, to let ones neurons
prowl the noisy solution space. The period is usually a day or two,
but can be years.
Some engineers are uncomfortable with uncertainty, and want to lock
down a design as soon as possible, preferably something sanctioned by
some authority. I like to stay confused for a while.
Post by Phil Hobbs
I sometimes need to do a family of designs, rather than just one.
Recently I've been working on some very fast, very cheap SPAD preamps,
intended to go in the guts of positron-emission scanners.
Designs with lots of real-world constraints are often the most fun, and
this one's specs include: 300-ps edges with 100-ps timing repeatability
from unit to unit; no magnetics allowed; and a BOM cost of $1 or less.
(You need a whole lot of channels, and PET and MRI machines are often
combined.)
I do a fair amount of analysis of circuits of that sort, to figure out
what actually limits their performance. It isn't super detailed--in
this case, just enough to figure out whether it'll be the base-emitter
time constant, the Miller effect, or the SPAD's series resistance that
will be the limiting factor.
Certainly quantitative reality should filter the solution space. But
even that can be mostly intuitive. I was talking about that with C on
Friday, about how some people have good quantitative intuition and
some don't. She can look at soup in a round pot and know if it will
fit into a square plastic container, to about 10%. I can do that.
Neither of our spouses can.
Post by Phil Hobbs
Miller, I can deal with using circuit hacks. The BE time constant is
Rbb' * Cbe, which gets slightly worse at high current, but is mainly a
device parameter--to get a big improvement you have to change
transistors. The SPAD can be negotiable depending on whose process
you're making them on--when each machine needs thousands of them,
vendors tend to listen.
Eventually, of course, you have to pick one and go with it, but picking
a topology usually takes me an iteration or two.
Sometimes a circuit takes me dozens, lots of sheets in the trash can.
I think it's important to give as many ideas as possible a chance.
See Barrie Gilbert's essay "Where do little circuits come from?"
"Prod and poke" and "doodling" are suggested.
I agree with both of you. What Phil is doing is figuring out where to
focus the brainstorming and fiddling, and the resulting wild
alternatives can easily be assessed. It's at the very least an
orthogonal method.

My personal experience is that iterations and inspirations require
studying extensively followed by sleeping on it, so the metric isn't a
few days, it's a few nights.

The bit about the necessity of nights was pointed out by J. Hadamard
in his famous book on this issue. The book has become hard to find
and expensive, but has now been reissued:

.<https://www.amazon.com/Mathematicians-Mind-Jacques-Hadamard/dp/0691029318/ref=sr_1_1>

Joe Gwinn
John Larkin
2024-01-22 03:43:18 UTC
Permalink
Post by Joe Gwinn
Post by John Larkin
On Sun, 21 Jan 2024 18:16:01 -0500, Phil Hobbs
Post by Phil Hobbs
JL wrote an interesting post in the depths of the "better
microelectronics from coal" thread that I thought was worth pulling out
on its own.
On 2024-01-21 10:12, John Larkin wrote:>
"...what IS electronic
design, and what's the best way to do it? <snip>
Short answer, cobbling. When presented with a problem or an
opportunity to design electronics, the most efficient way to do that
is to grab a piece of paper and immediately sketch a circuit or an
assembly. Sometimes one can do that instantly, without thinking, or
sometimes one can ignore the issue for a few days and then the design
pops up. Sometimes brainstorming and whiteboarding help. Sometimes
fiddling with Spice helps.
All that literature research and math analysis and simulation and
breadboarding and prototyping are just slow and expensive follow-up
chores for people who don't have 100% confidence in their instincts.
Analysis, sometimes prudent to do, but not design.
Design is subconsious and instinctive. And it's free! And to some
extent, it can be taught, but seldom is.
Most of us design things to sell, so do whatever works. We're selling
stuff, not publishing papers.
Hmm. I don't think that I agree in general, because you make it sound
as though the process were just intuitively plucking one idea out of
somewhere-or-other and cranking it out.
If an idea is new, where else would come from?
Post by Phil Hobbs
You've often argued in favor of brainstorming, where you get a few smart
people in front of a white board and try out ideas to find the best one
and flesh it out. We've done that together, very fruitfully.
It's possible to do more or less the same thing by oneself, but it
requires the ability to tolerate uncertainty for extended periods.
(That's a skill well worth developing, which most people are really,
really bad at, IME.)
The uncertainty period is probably necessary, to let ones neurons
prowl the noisy solution space. The period is usually a day or two,
but can be years.
Some engineers are uncomfortable with uncertainty, and want to lock
down a design as soon as possible, preferably something sanctioned by
some authority. I like to stay confused for a while.
Post by Phil Hobbs
I sometimes need to do a family of designs, rather than just one.
Recently I've been working on some very fast, very cheap SPAD preamps,
intended to go in the guts of positron-emission scanners.
Designs with lots of real-world constraints are often the most fun, and
this one's specs include: 300-ps edges with 100-ps timing repeatability
from unit to unit; no magnetics allowed; and a BOM cost of $1 or less.
(You need a whole lot of channels, and PET and MRI machines are often
combined.)
I do a fair amount of analysis of circuits of that sort, to figure out
what actually limits their performance. It isn't super detailed--in
this case, just enough to figure out whether it'll be the base-emitter
time constant, the Miller effect, or the SPAD's series resistance that
will be the limiting factor.
Certainly quantitative reality should filter the solution space. But
even that can be mostly intuitive. I was talking about that with C on
Friday, about how some people have good quantitative intuition and
some don't. She can look at soup in a round pot and know if it will
fit into a square plastic container, to about 10%. I can do that.
Neither of our spouses can.
Post by Phil Hobbs
Miller, I can deal with using circuit hacks. The BE time constant is
Rbb' * Cbe, which gets slightly worse at high current, but is mainly a
device parameter--to get a big improvement you have to change
transistors. The SPAD can be negotiable depending on whose process
you're making them on--when each machine needs thousands of them,
vendors tend to listen.
Eventually, of course, you have to pick one and go with it, but picking
a topology usually takes me an iteration or two.
Sometimes a circuit takes me dozens, lots of sheets in the trash can.
I think it's important to give as many ideas as possible a chance.
See Barrie Gilbert's essay "Where do little circuits come from?"
"Prod and poke" and "doodling" are suggested.
I agree with both of you. What Phil is doing is figuring out where to
focus the brainstorming and fiddling, and the resulting wild
alternatives can easily be assessed. It's at the very least an
orthogonal method.
My personal experience is that iterations and inspirations require
studying extensively followed by sleeping on it, so the metric isn't a
few days, it's a few nights.
Actually, it is a few showers.
Post by Joe Gwinn
The bit about the necessity of nights was pointed out by J. Hadamard
in his famous book on this issue. The book has become hard to find
.<https://www.amazon.com/Mathematicians-Mind-Jacques-Hadamard/dp/0691029318/ref=sr_1_1>
Joe Gwinn
Joe Gwinn
2024-01-22 05:12:36 UTC
Permalink
Post by John Larkin
Post by Joe Gwinn
Post by John Larkin
On Sun, 21 Jan 2024 18:16:01 -0500, Phil Hobbs
Post by Phil Hobbs
JL wrote an interesting post in the depths of the "better
microelectronics from coal" thread that I thought was worth pulling out
on its own.
On 2024-01-21 10:12, John Larkin wrote:>
"...what IS electronic
design, and what's the best way to do it? <snip>
Short answer, cobbling. When presented with a problem or an
opportunity to design electronics, the most efficient way to do that
is to grab a piece of paper and immediately sketch a circuit or an
assembly. Sometimes one can do that instantly, without thinking, or
sometimes one can ignore the issue for a few days and then the design
pops up. Sometimes brainstorming and whiteboarding help. Sometimes
fiddling with Spice helps.
All that literature research and math analysis and simulation and
breadboarding and prototyping are just slow and expensive follow-up
chores for people who don't have 100% confidence in their instincts.
Analysis, sometimes prudent to do, but not design.
Design is subconsious and instinctive. And it's free! And to some
extent, it can be taught, but seldom is.
Most of us design things to sell, so do whatever works. We're selling
stuff, not publishing papers.
Hmm. I don't think that I agree in general, because you make it sound
as though the process were just intuitively plucking one idea out of
somewhere-or-other and cranking it out.
If an idea is new, where else would come from?
Post by Phil Hobbs
You've often argued in favor of brainstorming, where you get a few smart
people in front of a white board and try out ideas to find the best one
and flesh it out. We've done that together, very fruitfully.
It's possible to do more or less the same thing by oneself, but it
requires the ability to tolerate uncertainty for extended periods.
(That's a skill well worth developing, which most people are really,
really bad at, IME.)
The uncertainty period is probably necessary, to let ones neurons
prowl the noisy solution space. The period is usually a day or two,
but can be years.
Some engineers are uncomfortable with uncertainty, and want to lock
down a design as soon as possible, preferably something sanctioned by
some authority. I like to stay confused for a while.
Post by Phil Hobbs
I sometimes need to do a family of designs, rather than just one.
Recently I've been working on some very fast, very cheap SPAD preamps,
intended to go in the guts of positron-emission scanners.
Designs with lots of real-world constraints are often the most fun, and
this one's specs include: 300-ps edges with 100-ps timing repeatability
from unit to unit; no magnetics allowed; and a BOM cost of $1 or less.
(You need a whole lot of channels, and PET and MRI machines are often
combined.)
I do a fair amount of analysis of circuits of that sort, to figure out
what actually limits their performance. It isn't super detailed--in
this case, just enough to figure out whether it'll be the base-emitter
time constant, the Miller effect, or the SPAD's series resistance that
will be the limiting factor.
Certainly quantitative reality should filter the solution space. But
even that can be mostly intuitive. I was talking about that with C on
Friday, about how some people have good quantitative intuition and
some don't. She can look at soup in a round pot and know if it will
fit into a square plastic container, to about 10%. I can do that.
Neither of our spouses can.
Post by Phil Hobbs
Miller, I can deal with using circuit hacks. The BE time constant is
Rbb' * Cbe, which gets slightly worse at high current, but is mainly a
device parameter--to get a big improvement you have to change
transistors. The SPAD can be negotiable depending on whose process
you're making them on--when each machine needs thousands of them,
vendors tend to listen.
Eventually, of course, you have to pick one and go with it, but picking
a topology usually takes me an iteration or two.
Sometimes a circuit takes me dozens, lots of sheets in the trash can.
I think it's important to give as many ideas as possible a chance.
See Barrie Gilbert's essay "Where do little circuits come from?"
"Prod and poke" and "doodling" are suggested.
I agree with both of you. What Phil is doing is figuring out where to
focus the brainstorming and fiddling, and the resulting wild
alternatives can easily be assessed. It's at the very least an
orthogonal method.
My personal experience is that iterations and inspirations require
studying extensively followed by sleeping on it, so the metric isn't a
few days, it's a few nights.
Actually, it is a few showers.
So, you're all wet?

Actually, I also get ideas in the shower, probably because I stopped
focusing so hard.

I used to keep a waterproof dictation recorder handy, and on my
bedside table, so I wouldn't lose the ideas, but don't need the
recorder any more.

But the key is to stop trying for a while and think irrelevant things.

Joe Gwinn
Post by John Larkin
Post by Joe Gwinn
The bit about the necessity of nights was pointed out by J. Hadamard
in his famous book on this issue. The book has become hard to find
.<https://www.amazon.com/Mathematicians-Mind-Jacques-Hadamard/dp/0691029318/ref=sr_1_1>
Joe Gwinn
John Larkin
2024-01-22 06:14:34 UTC
Permalink
Post by Joe Gwinn
Post by John Larkin
Post by Joe Gwinn
Post by John Larkin
On Sun, 21 Jan 2024 18:16:01 -0500, Phil Hobbs
Post by Phil Hobbs
JL wrote an interesting post in the depths of the "better
microelectronics from coal" thread that I thought was worth pulling out
on its own.
On 2024-01-21 10:12, John Larkin wrote:>
"...what IS electronic
design, and what's the best way to do it? <snip>
Short answer, cobbling. When presented with a problem or an
opportunity to design electronics, the most efficient way to do that
is to grab a piece of paper and immediately sketch a circuit or an
assembly. Sometimes one can do that instantly, without thinking, or
sometimes one can ignore the issue for a few days and then the design
pops up. Sometimes brainstorming and whiteboarding help. Sometimes
fiddling with Spice helps.
All that literature research and math analysis and simulation and
breadboarding and prototyping are just slow and expensive follow-up
chores for people who don't have 100% confidence in their instincts.
Analysis, sometimes prudent to do, but not design.
Design is subconsious and instinctive. And it's free! And to some
extent, it can be taught, but seldom is.
Most of us design things to sell, so do whatever works. We're selling
stuff, not publishing papers.
Hmm. I don't think that I agree in general, because you make it sound
as though the process were just intuitively plucking one idea out of
somewhere-or-other and cranking it out.
If an idea is new, where else would come from?
Post by Phil Hobbs
You've often argued in favor of brainstorming, where you get a few smart
people in front of a white board and try out ideas to find the best one
and flesh it out. We've done that together, very fruitfully.
It's possible to do more or less the same thing by oneself, but it
requires the ability to tolerate uncertainty for extended periods.
(That's a skill well worth developing, which most people are really,
really bad at, IME.)
The uncertainty period is probably necessary, to let ones neurons
prowl the noisy solution space. The period is usually a day or two,
but can be years.
Some engineers are uncomfortable with uncertainty, and want to lock
down a design as soon as possible, preferably something sanctioned by
some authority. I like to stay confused for a while.
Post by Phil Hobbs
I sometimes need to do a family of designs, rather than just one.
Recently I've been working on some very fast, very cheap SPAD preamps,
intended to go in the guts of positron-emission scanners.
Designs with lots of real-world constraints are often the most fun, and
this one's specs include: 300-ps edges with 100-ps timing repeatability
from unit to unit; no magnetics allowed; and a BOM cost of $1 or less.
(You need a whole lot of channels, and PET and MRI machines are often
combined.)
I do a fair amount of analysis of circuits of that sort, to figure out
what actually limits their performance. It isn't super detailed--in
this case, just enough to figure out whether it'll be the base-emitter
time constant, the Miller effect, or the SPAD's series resistance that
will be the limiting factor.
Certainly quantitative reality should filter the solution space. But
even that can be mostly intuitive. I was talking about that with C on
Friday, about how some people have good quantitative intuition and
some don't. She can look at soup in a round pot and know if it will
fit into a square plastic container, to about 10%. I can do that.
Neither of our spouses can.
Post by Phil Hobbs
Miller, I can deal with using circuit hacks. The BE time constant is
Rbb' * Cbe, which gets slightly worse at high current, but is mainly a
device parameter--to get a big improvement you have to change
transistors. The SPAD can be negotiable depending on whose process
you're making them on--when each machine needs thousands of them,
vendors tend to listen.
Eventually, of course, you have to pick one and go with it, but picking
a topology usually takes me an iteration or two.
Sometimes a circuit takes me dozens, lots of sheets in the trash can.
I think it's important to give as many ideas as possible a chance.
See Barrie Gilbert's essay "Where do little circuits come from?"
"Prod and poke" and "doodling" are suggested.
I agree with both of you. What Phil is doing is figuring out where to
focus the brainstorming and fiddling, and the resulting wild
alternatives can easily be assessed. It's at the very least an
orthogonal method.
My personal experience is that iterations and inspirations require
studying extensively followed by sleeping on it, so the metric isn't a
few days, it's a few nights.
Actually, it is a few showers.
So, you're all wet?
That's the idea.
Post by Joe Gwinn
Actually, I also get ideas in the shower, probably because I stopped
focusing so hard.
I think sleepytime ideas get delivered in a morning shower. I don't
have ideas if I shower later in the day.
Post by Joe Gwinn
I used to keep a waterproof dictation recorder handy, and on my
bedside table, so I wouldn't lose the ideas, but don't need the
recorder any more.
Sometimes I have ideas at around 3AM. I scribble them on a pad so I
don't forget.
Post by Joe Gwinn
But the key is to stop trying for a while and think irrelevant things.
Joe Gwinn
Post by John Larkin
Post by Joe Gwinn
The bit about the necessity of nights was pointed out by J. Hadamard
in his famous book on this issue. The book has become hard to find
.<https://www.amazon.com/Mathematicians-Mind-Jacques-Hadamard/dp/0691029318/ref=sr_1_1>
Joe Gwinn
john larkin
2024-02-05 22:47:33 UTC
Permalink
Post by John Larkin
Post by Joe Gwinn
Post by John Larkin
Post by Joe Gwinn
Post by John Larkin
On Sun, 21 Jan 2024 18:16:01 -0500, Phil Hobbs
Post by Phil Hobbs
JL wrote an interesting post in the depths of the "better
microelectronics from coal" thread that I thought was worth pulling out
on its own.
On 2024-01-21 10:12, John Larkin wrote:>
"...what IS electronic
design, and what's the best way to do it? <snip>
Short answer, cobbling. When presented with a problem or an
opportunity to design electronics, the most efficient way to do that
is to grab a piece of paper and immediately sketch a circuit or an
assembly. Sometimes one can do that instantly, without thinking, or
sometimes one can ignore the issue for a few days and then the design
pops up. Sometimes brainstorming and whiteboarding help. Sometimes
fiddling with Spice helps.
All that literature research and math analysis and simulation and
breadboarding and prototyping are just slow and expensive follow-up
chores for people who don't have 100% confidence in their instincts.
Analysis, sometimes prudent to do, but not design.
Design is subconsious and instinctive. And it's free! And to some
extent, it can be taught, but seldom is.
Most of us design things to sell, so do whatever works. We're selling
stuff, not publishing papers.
Hmm. I don't think that I agree in general, because you make it sound
as though the process were just intuitively plucking one idea out of
somewhere-or-other and cranking it out.
If an idea is new, where else would come from?
Post by Phil Hobbs
You've often argued in favor of brainstorming, where you get a few smart
people in front of a white board and try out ideas to find the best one
and flesh it out. We've done that together, very fruitfully.
It's possible to do more or less the same thing by oneself, but it
requires the ability to tolerate uncertainty for extended periods.
(That's a skill well worth developing, which most people are really,
really bad at, IME.)
The uncertainty period is probably necessary, to let ones neurons
prowl the noisy solution space. The period is usually a day or two,
but can be years.
Some engineers are uncomfortable with uncertainty, and want to lock
down a design as soon as possible, preferably something sanctioned by
some authority. I like to stay confused for a while.
Post by Phil Hobbs
I sometimes need to do a family of designs, rather than just one.
Recently I've been working on some very fast, very cheap SPAD preamps,
intended to go in the guts of positron-emission scanners.
Designs with lots of real-world constraints are often the most fun, and
this one's specs include: 300-ps edges with 100-ps timing repeatability
from unit to unit; no magnetics allowed; and a BOM cost of $1 or less.
(You need a whole lot of channels, and PET and MRI machines are often
combined.)
I do a fair amount of analysis of circuits of that sort, to figure out
what actually limits their performance. It isn't super detailed--in
this case, just enough to figure out whether it'll be the base-emitter
time constant, the Miller effect, or the SPAD's series resistance that
will be the limiting factor.
Certainly quantitative reality should filter the solution space. But
even that can be mostly intuitive. I was talking about that with C on
Friday, about how some people have good quantitative intuition and
some don't. She can look at soup in a round pot and know if it will
fit into a square plastic container, to about 10%. I can do that.
Neither of our spouses can.
Post by Phil Hobbs
Miller, I can deal with using circuit hacks. The BE time constant is
Rbb' * Cbe, which gets slightly worse at high current, but is mainly a
device parameter--to get a big improvement you have to change
transistors. The SPAD can be negotiable depending on whose process
you're making them on--when each machine needs thousands of them,
vendors tend to listen.
Eventually, of course, you have to pick one and go with it, but picking
a topology usually takes me an iteration or two.
Sometimes a circuit takes me dozens, lots of sheets in the trash can.
I think it's important to give as many ideas as possible a chance.
See Barrie Gilbert's essay "Where do little circuits come from?"
"Prod and poke" and "doodling" are suggested.
I agree with both of you. What Phil is doing is figuring out where to
focus the brainstorming and fiddling, and the resulting wild
alternatives can easily be assessed. It's at the very least an
orthogonal method.
My personal experience is that iterations and inspirations require
studying extensively followed by sleeping on it, so the metric isn't a
few days, it's a few nights.
Actually, it is a few showers.
So, you're all wet?
That's the idea.
Post by Joe Gwinn
Actually, I also get ideas in the shower, probably because I stopped
focusing so hard.
I think sleepytime ideas get delivered in a morning shower. I don't
have ideas if I shower later in the day.
I opened an email in the morning and took a shower and had a bunch of
ideas. So ideas both get delivered in the suds, and happen there too.

Other people have noted the creative powers of hot water falling on
your head.

Good book, First Steps by Jeremy DeSilva. It's about the evolution of
upright walking, but he mentions that various great thinkers had ideas
sleeping, showering, or walking. Walking works best in the woods, not
on city streets.
Wanderer
2024-02-05 13:04:59 UTC
Permalink
"In London, where Southampton Row passes Russell Square,
across from the British Museum in Bloomsbury, Leo
Szilard waited irritably one gray Depression morning
for the stoplight to change. A trace of rain had fallen
during the night; Tuesday, September 12, 1933, dawned
cool, humid and dull. Drizzling rain would begin again
in early afternoon. When Szilard told the story later he
never mentioned his destination that morning. He may
have had none; he often walked to think. In any case
another destination intervened. The stoplight changed to
green. Szilard stepped off the curb. As he crossed the
street time cracked open before him and he saw a way to
the future, death into the world and all our woe,
the shape of things to come."


https://en.wikipedia.org/wiki/Leo_Szilard
John Larkin
2024-02-06 00:13:31 UTC
Permalink
Post by Wanderer
"In London, where Southampton Row passes Russell Square,
across from the British Museum in Bloomsbury, Leo
Szilard waited irritably one gray Depression morning
for the stoplight to change. A trace of rain had fallen
during the night; Tuesday, September 12, 1933, dawned
cool, humid and dull. Drizzling rain would begin again
in early afternoon. When Szilard told the story later he
never mentioned his destination that morning. He may
have had none; he often walked to think. In any case
another destination intervened. The stoplight changed to
green. Szilard stepped off the curb. As he crossed the
street time cracked open before him and he saw a way to
the future, death into the world and all our woe,
the shape of things to come."
https://en.wikipedia.org/wiki/Leo_Szilard
Nice story. Richards Rhodes is a great writer.
Jan Panteltje
2024-02-06 05:21:10 UTC
Permalink
On a sunny day (Mon, 05 Feb 2024 14:47:33 -0800) it happened john larkin
Post by john larkin
Post by John Larkin
Post by Joe Gwinn
Post by John Larkin
Post by Joe Gwinn
Post by John Larkin
On Sun, 21 Jan 2024 18:16:01 -0500, Phil Hobbs
Post by Phil Hobbs
JL wrote an interesting post in the depths of the "better
microelectronics from coal" thread that I thought was worth pulling out
on its own.
On 2024-01-21 10:12, John Larkin wrote:>
"...what IS electronic
design, and what's the best way to do it? <snip>
Short answer, cobbling. When presented with a problem or an
opportunity to design electronics, the most efficient way to do that
is to grab a piece of paper and immediately sketch a circuit or an
assembly. Sometimes one can do that instantly, without thinking, or
sometimes one can ignore the issue for a few days and then the design
pops up. Sometimes brainstorming and whiteboarding help. Sometimes
fiddling with Spice helps.
All that literature research and math analysis and simulation and
breadboarding and prototyping are just slow and expensive follow-up
chores for people who don't have 100% confidence in their instincts.
Analysis, sometimes prudent to do, but not design.
Design is subconscious and instinctive. And it's free! And to some
extent, it can be taught, but seldom is.
Most of us design things to sell, so do whatever works. We're selling
stuff, not publishing papers.
Hmm. I don't think that I agree in general, because you make it sound
as though the process were just intuitively plucking one idea out of
somewhere-or-other and cranking it out.
If an idea is new, where else would come from?
Post by Phil Hobbs
You've often argued in favor of brainstorming, where you get a few smart
people in front of a white board and try out ideas to find the best one
and flesh it out. We've done that together, very fruitfully.
It's possible to do more or less the same thing by oneself, but it
requires the ability to tolerate uncertainty for extended periods.
(That's a skill well worth developing, which most people are really,
really bad at, IME.)
The uncertainty period is probably necessary, to let ones neurons
prowl the noisy solution space. The period is usually a day or two,
but can be years.
Some engineers are uncomfortable with uncertainty, and want to lock
down a design as soon as possible, preferably something sanctioned by
some authority. I like to stay confused for a while.
Post by Phil Hobbs
I sometimes need to do a family of designs, rather than just one.
Recently I've been working on some very fast, very cheap SPAD preamps,
intended to go in the guts of positron-emission scanners.
Designs with lots of real-world constraints are often the most fun, and
this one's specs include: 300-ps edges with 100-ps timing repeatability
from unit to unit; no magnetics allowed; and a BOM cost of $1 or less.
(You need a whole lot of channels, and PET and MRI machines are often
combined.)
I do a fair amount of analysis of circuits of that sort, to figure out
what actually limits their performance. It isn't super detailed--in
this case, just enough to figure out whether it'll be the base-emitter
time constant, the Miller effect, or the SPAD's series resistance that
will be the limiting factor.
Certainly quantitative reality should filter the solution space. But
even that can be mostly intuitive. I was talking about that with C on
Friday, about how some people have good quantitative intuition and
some don't. She can look at soup in a round pot and know if it will
fit into a square plastic container, to about 10%. I can do that.
Neither of our spouses can.
Post by Phil Hobbs
Miller, I can deal with using circuit hacks. The BE time constant is
Rbb' * Cbe, which gets slightly worse at high current, but is mainly a
device parameter--to get a big improvement you have to change
transistors. The SPAD can be negotiable depending on whose process
you're making them on--when each machine needs thousands of them,
vendors tend to listen.
Eventually, of course, you have to pick one and go with it, but picking
a topology usually takes me an iteration or two.
Sometimes a circuit takes me dozens, lots of sheets in the trash can.
I think it's important to give as many ideas as possible a chance.
See Barrie Gilbert's essay "Where do little circuits come from?"
"Prod and poke" and "doodling" are suggested.
I agree with both of you. What Phil is doing is figuring out where to
focus the brainstorming and fiddling, and the resulting wild
alternatives can easily be assessed. It's at the very least an
orthogonal method.
My personal experience is that iterations and inspirations require
studying extensively followed by sleeping on it, so the metric isn't a
few days, it's a few nights.
Actually, it is a few showers.
So, you're all wet?
That's the idea.
Post by Joe Gwinn
Actually, I also get ideas in the shower, probably because I stopped
focusing so hard.
I think sleepytime ideas get delivered in a morning shower. I don't
have ideas if I shower later in the day.
I opened an email in the morning and took a shower and had a bunch of
ideas. So ideas both get delivered in the suds, and happen there too.
Other people have noted the creative powers of hot water falling on
your head.
Good book, First Steps by Jeremy DeSilva. It's about the evolution of
upright walking, but he mentions that various great thinkers had ideas
sleeping, showering, or walking. Walking works best in the woods, not
on city streets.
I have been doing 1 to 2 hours meditation every day since the mid seventies.
Had ideas in the shower too.
Walking is good, I had a box full of medals as a kid for completing long marches.
Sill running faster than everybody here it seems, yesterday big storm
was on the bike, flying with wind in back.
On the way back walking with bike in hand.. now way with this bike against the wind
no gears...
More storm coming...
Satellite dish still works...

Design subconscious?
I dunno, it is like language, you combine words to express what you want,
like sub-circuits you have build and tested or code you have written and tested.
The vocabulary gets ever bigger, I speak 4 languages, French Dutch German and English
The learning never stops.
Several computer languages, several asm for different controllers I came about over the years
Life is a learning adventure, like living in the wild or in the civilized?? world
Better stop here.
John Larkin
2024-02-06 10:40:39 UTC
Permalink
Post by Jan Panteltje
On a sunny day (Mon, 05 Feb 2024 14:47:33 -0800) it happened john larkin
Post by john larkin
Post by John Larkin
Post by Joe Gwinn
Post by John Larkin
Post by Joe Gwinn
Post by John Larkin
On Sun, 21 Jan 2024 18:16:01 -0500, Phil Hobbs
Post by Phil Hobbs
JL wrote an interesting post in the depths of the "better
microelectronics from coal" thread that I thought was worth pulling out
on its own.
On 2024-01-21 10:12, John Larkin wrote:>
"...what IS electronic
design, and what's the best way to do it? <snip>
Short answer, cobbling. When presented with a problem or an
opportunity to design electronics, the most efficient way to do that
is to grab a piece of paper and immediately sketch a circuit or an
assembly. Sometimes one can do that instantly, without thinking, or
sometimes one can ignore the issue for a few days and then the design
pops up. Sometimes brainstorming and whiteboarding help. Sometimes
fiddling with Spice helps.
All that literature research and math analysis and simulation and
breadboarding and prototyping are just slow and expensive follow-up
chores for people who don't have 100% confidence in their instincts.
Analysis, sometimes prudent to do, but not design.
Design is subconscious and instinctive. And it's free! And to some
extent, it can be taught, but seldom is.
Most of us design things to sell, so do whatever works. We're selling
stuff, not publishing papers.
Hmm. I don't think that I agree in general, because you make it sound
as though the process were just intuitively plucking one idea out of
somewhere-or-other and cranking it out.
If an idea is new, where else would come from?
Post by Phil Hobbs
You've often argued in favor of brainstorming, where you get a few smart
people in front of a white board and try out ideas to find the best one
and flesh it out. We've done that together, very fruitfully.
It's possible to do more or less the same thing by oneself, but it
requires the ability to tolerate uncertainty for extended periods.
(That's a skill well worth developing, which most people are really,
really bad at, IME.)
The uncertainty period is probably necessary, to let ones neurons
prowl the noisy solution space. The period is usually a day or two,
but can be years.
Some engineers are uncomfortable with uncertainty, and want to lock
down a design as soon as possible, preferably something sanctioned by
some authority. I like to stay confused for a while.
Post by Phil Hobbs
I sometimes need to do a family of designs, rather than just one.
Recently I've been working on some very fast, very cheap SPAD preamps,
intended to go in the guts of positron-emission scanners.
Designs with lots of real-world constraints are often the most fun, and
this one's specs include: 300-ps edges with 100-ps timing repeatability
from unit to unit; no magnetics allowed; and a BOM cost of $1 or less.
(You need a whole lot of channels, and PET and MRI machines are often
combined.)
I do a fair amount of analysis of circuits of that sort, to figure out
what actually limits their performance. It isn't super detailed--in
this case, just enough to figure out whether it'll be the base-emitter
time constant, the Miller effect, or the SPAD's series resistance that
will be the limiting factor.
Certainly quantitative reality should filter the solution space. But
even that can be mostly intuitive. I was talking about that with C on
Friday, about how some people have good quantitative intuition and
some don't. She can look at soup in a round pot and know if it will
fit into a square plastic container, to about 10%. I can do that.
Neither of our spouses can.
Post by Phil Hobbs
Miller, I can deal with using circuit hacks. The BE time constant is
Rbb' * Cbe, which gets slightly worse at high current, but is mainly a
device parameter--to get a big improvement you have to change
transistors. The SPAD can be negotiable depending on whose process
you're making them on--when each machine needs thousands of them,
vendors tend to listen.
Eventually, of course, you have to pick one and go with it, but picking
a topology usually takes me an iteration or two.
Sometimes a circuit takes me dozens, lots of sheets in the trash can.
I think it's important to give as many ideas as possible a chance.
See Barrie Gilbert's essay "Where do little circuits come from?"
"Prod and poke" and "doodling" are suggested.
I agree with both of you. What Phil is doing is figuring out where to
focus the brainstorming and fiddling, and the resulting wild
alternatives can easily be assessed. It's at the very least an
orthogonal method.
My personal experience is that iterations and inspirations require
studying extensively followed by sleeping on it, so the metric isn't a
few days, it's a few nights.
Actually, it is a few showers.
So, you're all wet?
That's the idea.
Post by Joe Gwinn
Actually, I also get ideas in the shower, probably because I stopped
focusing so hard.
I think sleepytime ideas get delivered in a morning shower. I don't
have ideas if I shower later in the day.
I opened an email in the morning and took a shower and had a bunch of
ideas. So ideas both get delivered in the suds, and happen there too.
Other people have noted the creative powers of hot water falling on
your head.
Good book, First Steps by Jeremy DeSilva. It's about the evolution of
upright walking, but he mentions that various great thinkers had ideas
sleeping, showering, or walking. Walking works best in the woods, not
on city streets.
I have been doing 1 to 2 hours meditation every day since the mid seventies.
Had ideas in the shower too.
Walking is good, I had a box full of medals as a kid for completing long marches.
Sill running faster than everybody here it seems, yesterday big storm
was on the bike, flying with wind in back.
On the way back walking with bike in hand.. now way with this bike against the wind
no gears...
More storm coming...
Satellite dish still works...
We are having big storms here, and it's worse in southern California.

On Saturday I was climbing the ladder up the side of the house, to
check something on the roof, and couldn't make it to the top. Turns
out the wind gusts were peaking at 80 MPH. That's about 130 KPH in
your quaint units.

Yesterday a giant tree limb broke off from the tree just outside my
office window, blocking a lane of Potrero Avenue. We have trees down
all over the state.
Post by Jan Panteltje
Design subconscious?
I dunno, it is like language, you combine words to express what you want,
like sub-circuits you have build and tested or code you have written and tested.
The vocabulary gets ever bigger, I speak 4 languages, French Dutch German and English
The learning never stops.
Several computer languages, several asm for different controllers I came about over the years
Life is a learning adventure, like living in the wild or in the civilized?? world
Better stop here.
Stay crazy.
Fred Bloggs
2024-02-07 14:54:16 UTC
Permalink
Post by John Larkin
Post by Jan Panteltje
On a sunny day (Mon, 05 Feb 2024 14:47:33 -0800) it happened john larkin
Post by john larkin
Post by John Larkin
Post by Joe Gwinn
Post by John Larkin
Post by John Larkin
On Sun, 21 Jan 2024 18:16:01 -0500, Phil Hobbs
Post by Phil Hobbs
JL wrote an interesting post in the depths of the "better
microelectronics from coal" thread that I thought was worth pulling out
on its own.
On 2024-01-21 10:12, John Larkin wrote:>
"...what IS electronic
design, and what's the best way to do it? <snip>
Short answer, cobbling. When presented with a problem or an
opportunity to design electronics, the most efficient way to do that
is to grab a piece of paper and immediately sketch a circuit or an
assembly. Sometimes one can do that instantly, without thinking, or
sometimes one can ignore the issue for a few days and then the design
pops up. Sometimes brainstorming and whiteboarding help. Sometimes
fiddling with Spice helps.
All that literature research and math analysis and simulation and
breadboarding and prototyping are just slow and expensive follow-up
chores for people who don't have 100% confidence in their instincts.
Analysis, sometimes prudent to do, but not design.
Design is subconscious and instinctive. And it's free! And to some
extent, it can be taught, but seldom is.
Most of us design things to sell, so do whatever works. We're selling
stuff, not publishing papers.
Hmm. I don't think that I agree in general, because you make it sound
as though the process were just intuitively plucking one idea out of
somewhere-or-other and cranking it out.
If an idea is new, where else would come from?
Post by Phil Hobbs
You've often argued in favor of brainstorming, where you get a few smart
people in front of a white board and try out ideas to find the best one
and flesh it out. We've done that together, very fruitfully.
It's possible to do more or less the same thing by oneself, but it
requires the ability to tolerate uncertainty for extended periods.
(That's a skill well worth developing, which most people are really,
really bad at, IME.)
The uncertainty period is probably necessary, to let ones neurons
prowl the noisy solution space. The period is usually a day or two,
but can be years.
Some engineers are uncomfortable with uncertainty, and want to lock
down a design as soon as possible, preferably something sanctioned by
some authority. I like to stay confused for a while.
Post by Phil Hobbs
I sometimes need to do a family of designs, rather than just one.
Recently I've been working on some very fast, very cheap SPAD preamps,
intended to go in the guts of positron-emission scanners.
Designs with lots of real-world constraints are often the most fun, and
this one's specs include: 300-ps edges with 100-ps timing repeatability
from unit to unit; no magnetics allowed; and a BOM cost of $1 or less.
(You need a whole lot of channels, and PET and MRI machines are often
combined.)
I do a fair amount of analysis of circuits of that sort, to figure out
what actually limits their performance. It isn't super detailed--in
this case, just enough to figure out whether it'll be the base-emitter
time constant, the Miller effect, or the SPAD's series resistance that
will be the limiting factor.
Certainly quantitative reality should filter the solution space. But
even that can be mostly intuitive. I was talking about that with C on
Friday, about how some people have good quantitative intuition and
some don't. She can look at soup in a round pot and know if it will
fit into a square plastic container, to about 10%. I can do that.
Neither of our spouses can.
Post by Phil Hobbs
Miller, I can deal with using circuit hacks. The BE time constant is
Rbb' * Cbe, which gets slightly worse at high current, but is mainly a
device parameter--to get a big improvement you have to change
transistors. The SPAD can be negotiable depending on whose process
you're making them on--when each machine needs thousands of them,
vendors tend to listen.
Eventually, of course, you have to pick one and go with it, but picking
a topology usually takes me an iteration or two.
Sometimes a circuit takes me dozens, lots of sheets in the trash can.
I think it's important to give as many ideas as possible a chance.
See Barrie Gilbert's essay "Where do little circuits come from?"
"Prod and poke" and "doodling" are suggested.
I agree with both of you. What Phil is doing is figuring out where to
focus the brainstorming and fiddling, and the resulting wild
alternatives can easily be assessed. It's at the very least an
orthogonal method.
My personal experience is that iterations and inspirations require
studying extensively followed by sleeping on it, so the metric isn't a
few days, it's a few nights.
Actually, it is a few showers.
So, you're all wet?
That's the idea.
Post by Joe Gwinn
Actually, I also get ideas in the shower, probably because I stopped
focusing so hard.
I think sleepytime ideas get delivered in a morning shower. I don't
have ideas if I shower later in the day.
I opened an email in the morning and took a shower and had a bunch of
ideas. So ideas both get delivered in the suds, and happen there too.
Other people have noted the creative powers of hot water falling on
your head.
Good book, First Steps by Jeremy DeSilva. It's about the evolution of
upright walking, but he mentions that various great thinkers had ideas
sleeping, showering, or walking. Walking works best in the woods, not
on city streets.
I have been doing 1 to 2 hours meditation every day since the mid seventies.
Had ideas in the shower too.
Walking is good, I had a box full of medals as a kid for completing long marches.
Sill running faster than everybody here it seems, yesterday big storm
was on the bike, flying with wind in back.
On the way back walking with bike in hand.. now way with this bike against the wind
no gears...
More storm coming...
Satellite dish still works...
We are having big storms here, and it's worse in southern California.
On Saturday I was climbing the ladder up the side of the house, to
check something on the roof, and couldn't make it to the top. Turns
out the wind gusts were peaking at 80 MPH. That's about 130 KPH in
your quaint units.
LOL- you must have death wish climbing that height under those conditions. The moisture lubricates the contact points of the ladder against the wall/ roof, and possibly softens the ground under the base enough to make things even more hazardous. Then even if you could make it to the roof, you would slip and slide off due to wind ( gets worse with height ) and moisture. You're not going to survive a 30 ft fall. If you have one of those permanent structural ladders bolted to the side of the building, that would better, as long as you don't get out on the roof.
Post by John Larkin
Yesterday a giant tree limb broke off from the tree just outside my
office window, blocking a lane of Potrero Avenue. We have trees down
all over the state.
Post by Jan Panteltje
Design subconscious?
I dunno, it is like language, you combine words to express what you want,
like sub-circuits you have build and tested or code you have written and tested.
The vocabulary gets ever bigger, I speak 4 languages, French Dutch German and English
The learning never stops.
Several computer languages, several asm for different controllers I came about over the years
Life is a learning adventure, like living in the wild or in the civilized?? world
Better stop here.
Stay crazy.
Fred Bloggs
2024-01-22 19:47:54 UTC
Permalink
Post by John Larkin
On Sun, 21 Jan 2024 18:16:01 -0500, Phil Hobbs
Post by Phil Hobbs
JL wrote an interesting post in the depths of the "better
microelectronics from coal" thread that I thought was worth pulling out
on its own.
On 2024-01-21 10:12, John Larkin wrote:>
"...what IS electronic
design, and what's the best way to do it? <snip>
Short answer, cobbling. When presented with a problem or an
opportunity to design electronics, the most efficient way to do that
is to grab a piece of paper and immediately sketch a circuit or an
assembly. Sometimes one can do that instantly, without thinking, or
sometimes one can ignore the issue for a few days and then the design
pops up. Sometimes brainstorming and whiteboarding help. Sometimes
fiddling with Spice helps.
All that literature research and math analysis and simulation and
breadboarding and prototyping are just slow and expensive follow-up
chores for people who don't have 100% confidence in their instincts.
Analysis, sometimes prudent to do, but not design.
Design is subconsious and instinctive. And it's free! And to some
extent, it can be taught, but seldom is.
Most of us design things to sell, so do whatever works. We're selling
stuff, not publishing papers.
Hmm. I don't think that I agree in general, because you make it sound
as though the process were just intuitively plucking one idea out of
somewhere-or-other and cranking it out.
If an idea is new, where else would come from?
Within 5 years, all this manual fiddling, and so-called brainstorming, will be reduced to an AI-app resident on a $ phone. It may not be optimum, but it will work.
Post by John Larkin
Post by Phil Hobbs
You've often argued in favor of brainstorming, where you get a few smart
people in front of a white board and try out ideas to find the best one
and flesh it out. We've done that together, very fruitfully.
It's possible to do more or less the same thing by oneself, but it
requires the ability to tolerate uncertainty for extended periods.
(That's a skill well worth developing, which most people are really,
really bad at, IME.)
The uncertainty period is probably necessary, to let ones neurons
prowl the noisy solution space. The period is usually a day or two,
but can be years.
Some engineers are uncomfortable with uncertainty, and want to lock
down a design as soon as possible, preferably something sanctioned by
some authority. I like to stay confused for a while.
Post by Phil Hobbs
I sometimes need to do a family of designs, rather than just one.
Recently I've been working on some very fast, very cheap SPAD preamps,
intended to go in the guts of positron-emission scanners.
Designs with lots of real-world constraints are often the most fun, and
this one's specs include: 300-ps edges with 100-ps timing repeatability
from unit to unit; no magnetics allowed; and a BOM cost of $1 or less.
(You need a whole lot of channels, and PET and MRI machines are often
combined.)
I do a fair amount of analysis of circuits of that sort, to figure out
what actually limits their performance. It isn't super detailed--in
this case, just enough to figure out whether it'll be the base-emitter
time constant, the Miller effect, or the SPAD's series resistance that
will be the limiting factor.
Certainly quantitative reality should filter the solution space. But
even that can be mostly intuitive. I was talking about that with C on
Friday, about how some people have good quantitative intuition and
some don't. She can look at soup in a round pot and know if it will
fit into a square plastic container, to about 10%. I can do that.
Neither of our spouses can.
Post by Phil Hobbs
Miller, I can deal with using circuit hacks. The BE time constant is
Rbb' * Cbe, which gets slightly worse at high current, but is mainly a
device parameter--to get a big improvement you have to change
transistors. The SPAD can be negotiable depending on whose process
you're making them on--when each machine needs thousands of them,
vendors tend to listen.
Eventually, of course, you have to pick one and go with it, but picking
a topology usually takes me an iteration or two.
Sometimes a circuit takes me dozens, lots of sheets in the trash can.
I think it's important to give as many ideas as possible a chance.
See Barrie Gilbert's essay "Where do little circuits come from?"
"Prod and poke" and "doodling" are suggested.
Post by Phil Hobbs
Cheers
Phil Hobbs
john larkin
2024-01-22 20:43:18 UTC
Permalink
On Mon, 22 Jan 2024 11:47:54 -0800 (PST), Fred Bloggs
Post by Fred Bloggs
Post by John Larkin
On Sun, 21 Jan 2024 18:16:01 -0500, Phil Hobbs
Post by Phil Hobbs
JL wrote an interesting post in the depths of the "better
microelectronics from coal" thread that I thought was worth pulling out
on its own.
On 2024-01-21 10:12, John Larkin wrote:>
"...what IS electronic
design, and what's the best way to do it? <snip>
Short answer, cobbling. When presented with a problem or an
opportunity to design electronics, the most efficient way to do that
is to grab a piece of paper and immediately sketch a circuit or an
assembly. Sometimes one can do that instantly, without thinking, or
sometimes one can ignore the issue for a few days and then the design
pops up. Sometimes brainstorming and whiteboarding help. Sometimes
fiddling with Spice helps.
All that literature research and math analysis and simulation and
breadboarding and prototyping are just slow and expensive follow-up
chores for people who don't have 100% confidence in their instincts.
Analysis, sometimes prudent to do, but not design.
Design is subconsious and instinctive. And it's free! And to some
extent, it can be taught, but seldom is.
Most of us design things to sell, so do whatever works. We're selling
stuff, not publishing papers.
Hmm. I don't think that I agree in general, because you make it sound
as though the process were just intuitively plucking one idea out of
somewhere-or-other and cranking it out.
If an idea is new, where else would come from?
Within 5 years, all this manual fiddling, and so-called brainstorming, will be reduced to an AI-app resident on a $ phone. It may not be optimum, but it will work.
I don't think so. Just a few words are not enough to specify and
generate a specific, reliable design.

There's not much I in AI. It's mostly a silly fad.

But brainstorming isn't so-called. Done right, it really works.
Anthony William Sloman
2024-01-22 23:29:27 UTC
Permalink
I don't think so. Just a few words are not enough to specify and generate a specific, reliable design.
They are enough to specify a patentable idea. Reducing it to practice takes a lot more work, and documenting a complete system takes a lot of words (and pictures). Oddly enough, software can do a lot of the documentation.
There's not much I in AI. It's mostly a silly fad.
For particular problems it can already find solutions that humans can't

https://en.wikipedia.org/wiki/Protein_folding

AlphaFold is well ahead of any expert, and some theorem-proving programs operate without making mistakes to a degree that allows them to outperform human mathematicians on specific complex problems. It really isn't any kid of silly fad.
But brainstorming isn't so-called. Done right, it really works.
Mainly by discouraging status-seeking creeps from insisting on concentrating on their own ideas.
--
Bill Sloman, Sydney
Fred Bloggs
2024-01-24 16:00:15 UTC
Permalink
Post by Anthony William Sloman
I don't think so. Just a few words are not enough to specify and generate a specific, reliable design.
They are enough to specify a patentable idea. Reducing it to practice takes a lot more work, and documenting a complete system takes a lot of words (and pictures). Oddly enough, software can do a lot of the documentation.
There's not much I in AI. It's mostly a silly fad.
For particular problems it can already find solutions that humans can't
https://en.wikipedia.org/wiki/Protein_folding
AlphaFold is well ahead of any expert, and some theorem-proving programs operate without making mistakes to a degree that allows them to outperform human mathematicians on specific complex problems. It really isn't any kid of silly fad.
But brainstorming isn't so-called. Done right, it really works.
Mainly by discouraging status-seeking creeps from insisting on concentrating on their own ideas.
To be replaced by group concentration on no ideas at all.

https://resources.pcb.cadence.com/blog/2022-the-role-of-machine-learning-in-analog-circuit-design

https://www.synopsys.com/blogs/chip-design/ai-analog-design-migration-samsung-safe-forum-2023.html

https://www.planetanalog.com/what-can-ai-do-for-analog-design/

https://www.electronicdesign.com/technologies/embedded/article/21272567/electronic-design-ai-lends-a-helping-hand-with-analog-and-custom-ic-design

https://semiengineering.com/ai-for-circuit-design-quality-productivity-and-advanced-node-mapping/

http://ieeexplore.ieee.org/document/31523/

The list is endless. Humans are not as unique and special as they make themselves out to be. They'll all be replaced by AI before long.
Post by Anthony William Sloman
--
Bill Sloman, Sydney
John Larkin
2024-01-24 16:27:12 UTC
Permalink
On Wed, 24 Jan 2024 08:00:15 -0800 (PST), Fred Bloggs
Post by Fred Bloggs
Post by Anthony William Sloman
I don't think so. Just a few words are not enough to specify and generate a specific, reliable design.
They are enough to specify a patentable idea. Reducing it to practice takes a lot more work, and documenting a complete system takes a lot of words (and pictures). Oddly enough, software can do a lot of the documentation.
There's not much I in AI. It's mostly a silly fad.
For particular problems it can already find solutions that humans can't
https://en.wikipedia.org/wiki/Protein_folding
AlphaFold is well ahead of any expert, and some theorem-proving programs operate without making mistakes to a degree that allows them to outperform human mathematicians on specific complex problems. It really isn't any kid of silly fad.
But brainstorming isn't so-called. Done right, it really works.
Mainly by discouraging status-seeking creeps from insisting on concentrating on their own ideas.
To be replaced by group concentration on no ideas at all.
The people have to be right for the process to be productive of ideas.
Some people will poison a brainstorming session, and too much general
sociability in the room will reinforce conventional thinking.
Post by Fred Bloggs
https://resources.pcb.cadence.com/blog/2022-the-role-of-machine-learning-in-analog-circuit-design
That's absurd. Sounds like they are trying to sell cad options to
beginners.
Post by Fred Bloggs
https://www.synopsys.com/blogs/chip-design/ai-analog-design-migration-samsung-safe-forum-2023.html
Certainly a lot of computing helps design digital ICs, but I wouldn't
call that intelligence. Smart people wrote very specialized software.
I sometimes write software to solve circuit problems, but the software
just does what I told it to do.
Post by Fred Bloggs
https://www.planetanalog.com/what-can-ai-do-for-analog-design/
I'd love to have a good component selection tool. The intelligence
would be in inferring things from bad data sheets that have no
standards. It would of course have to read and understand application
schematics and mechanical drawings and find gotchas buried in
footnotes and graphs.

Find me a right-angle Gbit PoE compatible RJ45 jack that has multiple
drop-in sources, two LEDs on the high side, lots of stock from
non-Chinese sources, at a good price. They have to mount on my PCB and
ground to a cutout in my panel. That's an easy one.
Post by Fred Bloggs
https://www.electronicdesign.com/technologies/embedded/article/21272567/electronic-design-ai-lends-a-helping-hand-with-analog-and-custom-ic-design
https://semiengineering.com/ai-for-circuit-design-quality-productivity-and-advanced-node-mapping/
http://ieeexplore.ieee.org/document/31523/
The list is endless. Humans are not as unique and special as they make themselves out to be. They'll all be replaced by AI before long.
Has AI ever invented anything?
I check up on Flux.ai now and then. I wonder when they will run out of
money.
Fred Bloggs
2024-01-24 18:59:47 UTC
Permalink
Post by John Larkin
On Wed, 24 Jan 2024 08:00:15 -0800 (PST), Fred Bloggs
Post by Fred Bloggs
Post by Anthony William Sloman
I don't think so. Just a few words are not enough to specify and generate a specific, reliable design.
They are enough to specify a patentable idea. Reducing it to practice takes a lot more work, and documenting a complete system takes a lot of words (and pictures). Oddly enough, software can do a lot of the documentation.
There's not much I in AI. It's mostly a silly fad.
For particular problems it can already find solutions that humans can't
https://en.wikipedia.org/wiki/Protein_folding
AlphaFold is well ahead of any expert, and some theorem-proving programs operate without making mistakes to a degree that allows them to outperform human mathematicians on specific complex problems. It really isn't any kid of silly fad.
But brainstorming isn't so-called. Done right, it really works.
Mainly by discouraging status-seeking creeps from insisting on concentrating on their own ideas.
To be replaced by group concentration on no ideas at all.
The people have to be right for the process to be productive of ideas.
Some people will poison a brainstorming session, and too much general
sociability in the room will reinforce conventional thinking.
Post by Fred Bloggs
https://resources.pcb.cadence.com/blog/2022-the-role-of-machine-learning-in-analog-circuit-design
That's absurd. Sounds like they are trying to sell cad options to
beginners.
Post by Fred Bloggs
https://www.synopsys.com/blogs/chip-design/ai-analog-design-migration-samsung-safe-forum-2023.html
Certainly a lot of computing helps design digital ICs, but I wouldn't
call that intelligence. Smart people wrote very specialized software.
I sometimes write software to solve circuit problems, but the software
just does what I told it to do.
Post by Fred Bloggs
https://www.planetanalog.com/what-can-ai-do-for-analog-design/
I'd love to have a good component selection tool. The intelligence
would be in inferring things from bad data sheets that have no
standards. It would of course have to read and understand application
schematics and mechanical drawings and find gotchas buried in
footnotes and graphs.
Find me a right-angle Gbit PoE compatible RJ45 jack that has multiple
drop-in sources, two LEDs on the high side, lots of stock from
non-Chinese sources, at a good price. They have to mount on my PCB and
ground to a cutout in my panel. That's an easy one.
Lots of sources for that, but if you want a good price, it will be made in Asia. The shielded ones will ground to the panel. High demand parts like that will have a long lead time:

https://www.digikey.com/en/products/detail/w%C3%BCrth-elektronik/615008137421/2060608
Post by John Larkin
Post by Fred Bloggs
https://www.electronicdesign.com/technologies/embedded/article/21272567/electronic-design-ai-lends-a-helping-hand-with-analog-and-custom-ic-design
https://semiengineering.com/ai-for-circuit-design-quality-productivity-and-advanced-node-mapping/
http://ieeexplore.ieee.org/document/31523/
The list is endless. Humans are not as unique and special as they make themselves out to be. They'll all be replaced by AI before long.
Has AI ever invented anything?
It's doing things like running through impossibly large numbers of permutations to find something useful, as with drug discovery. It's more the case creative people are using AI to enable inventive ideas.
Post by John Larkin
I check up on Flux.ai now and then. I wonder when they will run out of
money.
john larkin
2024-01-24 19:15:19 UTC
Permalink
On Wed, 24 Jan 2024 10:59:47 -0800 (PST), Fred Bloggs
Post by Fred Bloggs
Post by John Larkin
On Wed, 24 Jan 2024 08:00:15 -0800 (PST), Fred Bloggs
Post by Fred Bloggs
Post by Anthony William Sloman
I don't think so. Just a few words are not enough to specify and generate a specific, reliable design.
They are enough to specify a patentable idea. Reducing it to practice takes a lot more work, and documenting a complete system takes a lot of words (and pictures). Oddly enough, software can do a lot of the documentation.
There's not much I in AI. It's mostly a silly fad.
For particular problems it can already find solutions that humans can't
https://en.wikipedia.org/wiki/Protein_folding
AlphaFold is well ahead of any expert, and some theorem-proving programs operate without making mistakes to a degree that allows them to outperform human mathematicians on specific complex problems. It really isn't any kid of silly fad.
But brainstorming isn't so-called. Done right, it really works.
Mainly by discouraging status-seeking creeps from insisting on concentrating on their own ideas.
To be replaced by group concentration on no ideas at all.
The people have to be right for the process to be productive of ideas.
Some people will poison a brainstorming session, and too much general
sociability in the room will reinforce conventional thinking.
Post by Fred Bloggs
https://resources.pcb.cadence.com/blog/2022-the-role-of-machine-learning-in-analog-circuit-design
That's absurd. Sounds like they are trying to sell cad options to
beginners.
Post by Fred Bloggs
https://www.synopsys.com/blogs/chip-design/ai-analog-design-migration-samsung-safe-forum-2023.html
Certainly a lot of computing helps design digital ICs, but I wouldn't
call that intelligence. Smart people wrote very specialized software.
I sometimes write software to solve circuit problems, but the software
just does what I told it to do.
Post by Fred Bloggs
https://www.planetanalog.com/what-can-ai-do-for-analog-design/
I'd love to have a good component selection tool. The intelligence
would be in inferring things from bad data sheets that have no
standards. It would of course have to read and understand application
schematics and mechanical drawings and find gotchas buried in
footnotes and graphs.
Find me a right-angle Gbit PoE compatible RJ45 jack that has multiple
drop-in sources, two LEDs on the high side, lots of stock from
non-Chinese sources, at a good price. They have to mount on my PCB and
ground to a cutout in my panel. That's an easy one.
https://www.digikey.com/en/products/detail/w%C3%BCrth-elektronik/615008137421/2060608
Post by John Larkin
Post by Fred Bloggs
https://www.electronicdesign.com/technologies/embedded/article/21272567/electronic-design-ai-lends-a-helping-hand-with-analog-and-custom-ic-design
https://semiengineering.com/ai-for-circuit-design-quality-productivity-and-advanced-node-mapping/
http://ieeexplore.ieee.org/document/31523/
The list is endless. Humans are not as unique and special as they make themselves out to be. They'll all be replaced by AI before long.
Has AI ever invented anything?
It's doing things like running through impossibly large numbers of permutations to find something useful, as with drug discovery. It's more the case creative people are using AI to enable inventive ideas.
That's not intelligent. It's just automating a lot of grunt work, as
programmed. Line monte carlo simulation. The person who set it up is
the intelligence. All that's new is having more compute power than we
had in the past.

Computers automate grunt work and let us work faster and better and
move up the abstraction stack. Nonlinear differential equations were
never much fun.

I would like a Spice that was, say, 500x as fast as mine is now,
nvidia or something. And I'd love some way to specify results and have
a program juggle values and even library parts for a best solution.

Past attempts at such optimizations have tended to diverge. Even most
interns are smarter than that.
Post by Fred Bloggs
Post by John Larkin
I check up on Flux.ai now and then. I wonder when they will run out of
money.
Take a look at flux. It's funny.
whit3rd
2024-01-25 01:29:57 UTC
Permalink
Post by john larkin
I would like a Spice that was, say, 500x as fast as mine is now,
nvidia or something. And I'd love some way to specify results and have
a program juggle values and even library parts for a best solution.
Past attempts at such optimizations have tended to diverge. Even most
interns are smarter than that.
That's why we have math, that tells us that linear equations (like SPICE solves)
have multiple ways to generate large numbers. It's catastrophe theory,
to be precise.

For a stable sine wave oscillator, you can't use ideal C, R, L, and amplifier
components; there won't ever be any solutions that don't diverge, because
the linear-differential-equation solutions all have a matrix raised to
a power (and the power goes up with time). It'll always exponentially
decay or explode, because NO available component tolerances
are negligible effects.

The bugaboo of ALL multivariate optimizers is the fact that any solution
that's not unique is associated with regions in the parameter space
that have no optimum-direction sensitivity. Also, the parameter space
is huge. A dozen filter components means a ten-percent grid on
available values ranging over three decades has 72^12 = 1.9 *10^22
points to test, when the flat regions don't allow for gradient directed progress.
Anthony William Sloman
2024-01-25 01:50:53 UTC
Permalink
<snip>
Post by whit3rd
The bugaboo of ALL multivariate optimizers is the fact that any solution
that's not unique is associated with regions in the parameter space
that have no optimum-direction sensitivity. Also, the parameter space
is huge. A dozen filter components means a ten-percent grid on
available values ranging over three decades has 72^12 = 1.9 *10^22
points to test, when the flat regions don't allow for gradient directed progress.
I programmed one to fit my reaction data results when I was a graduate student - essentially the Fletcher-Powell procedure - which builds up a matrix of the second derivatives of the slope of the parameter space in the region it explored.

When the data was noisy it didn't work very well, but that was fine - those reactions runs got chucked out because the best fit reaction rate wouldn't have been all that well defined. You did have to start the process with reasonable values for the parameters being fitted (initial concentration, final concentration and reaction rate) but that wasn't difficult.\

As you say, if I'd had to start from scratch it wouldn't have worked. A dozen filter components wouldn't be selected at random - you'd start off with something classical (and probably fix the capacitors at values you could buy of the shelf).
--
Bill Sloman, Sydney
John Larkin
2024-01-25 02:07:24 UTC
Permalink
Post by whit3rd
Post by john larkin
I would like a Spice that was, say, 500x as fast as mine is now,
nvidia or something. And I'd love some way to specify results and have
a program juggle values and even library parts for a best solution.
Past attempts at such optimizations have tended to diverge. Even most
interns are smarter than that.
That's why we have math, that tells us that linear equations (like SPICE solves)
have multiple ways to generate large numbers. It's catastrophe theory,
to be precise.
For a stable sine wave oscillator, you can't use ideal C, R, L, and amplifier
components; there won't ever be any solutions that don't diverge, because
the linear-differential-equation solutions all have a matrix raised to
a power (and the power goes up with time). It'll always exponentially
decay or explode, because NO available component tolerances
are negligible effects.
The bugaboo of ALL multivariate optimizers is the fact that any solution
that's not unique is associated with regions in the parameter space
that have no optimum-direction sensitivity. Also, the parameter space
is huge. A dozen filter components means a ten-percent grid on
available values ranging over three decades has 72^12 = 1.9 *10^22
points to test, when the flat regions don't allow for gradient directed progress.
One of my specialities is designing instant-start super low jitter LC
oscillators. The Spice sims are dead on, except for tempco
compensation, which has to be done experimentally. I'd hate to design
such oscillators using differential equations.

If you think about it, all design is a mysterious mental process that
is follwed up by analysis, which can be equations, simulation, or
hardware prototyping. Whatever works. The real magic is step 1,
inventing things.

I'm impressed by computer based high-order filter design, especially
lossy LC filters. That's been automated since the Fortran days. When I
twiddle a filter in Spice, it tends to diverge.
Anthony William Sloman
2024-01-25 04:32:05 UTC
Permalink
<snip>
Post by John Larkin
One of my specialities is designing instant-start super low jitter LC
oscillators. The Spice sims are dead on, except for tempco
compensation, which has to be done experimentally. I'd hate to design
such oscillators using differential equations.
Nobody in their right mind would design such oscillators anyway - they are never as low-jitter as continuously running oscillators.

The more sensible approach relies on free-running low jitter oscillators (which can run much faster) and a two stage timing procedure which digitises how long after the most recent clock edge the start pulse arrived and adds that to the desired delay (which is then realsied as a countablke number of clock edges plus a vernier delay to put it in exactly the right place.

That's what we did in 1988, and it would be a lot easier now. John Larkin is actually copying a Hewlett-Packard device from 1978 when there were fewer fast semiconductor parts stocked by broad-line distributors.
Post by John Larkin
If you think about it, all design is a mysterious mental process that
is followed up by analysis, which can be equations, simulation, or
hardware prototyping. Whatever works. The real magic is step 1,
inventing things.
If you are bad at it, it can look like that. Clowns that like to make a mystery of the process, in order to make their own efforts look more impressive. are a feature of every skilled profession. They don't get sent up as frequently as they should be.
Post by John Larkin
I'm impressed by computer based high-order filter design, especially
lossy LC filters. That's been automated since the Fortran days. When I
twiddle a filter in Spice, it tends to diverge.
Some people are better at twiddling that others. Mostly they have a clearer idea of what they are doing.
--
Bill Sloman, Sydney
whit3rd
2024-01-25 08:04:20 UTC
Permalink
Post by John Larkin
Post by whit3rd
Post by john larkin
I would like a Spice that was, say, 500x as fast as mine is now,
nvidia or something. And I'd love some way to specify results and have
a program juggle values and even library parts for a best solution.
Past attempts at such optimizations have tended to diverge. Even most
interns are smarter than that.
That's why we have math, that tells us that linear equations (like SPICE solves)
have multiple ways to generate large numbers. It's catastrophe theory,
to be precise.
For a stable sine wave oscillator, you can't use ideal C, R, L, and amplifier
components; there won't ever be any solutions that don't diverge, because
the linear-differential-equation solutions all have a matrix raised to
a power (and the power goes up with time). It'll always exponentially
decay or explode, because NO available component tolerances
are negligible effects.
One of my specialities is designing instant-start super low jitter LC
oscillators. The Spice sims are dead on, except for tempco
compensation, which has to be done experimentally. I'd hate to design
such oscillators using differential equations.
Oh, if you run SPICE, you ARE using differential equations... after
Laplace-transforming them to linear equations using lots of "j ⍵" bits.
The thing that cannot be done, is to do sines with all linear
components; there's no linear equation for a saturating logic
comparator, so it makes a fine oscillator (NE555)
but not a pure sine wave.

Bill Hewlett's classic sine wave oscillator design got around the
problem with a thermal-varying resistor (square law device,
NOT linear).

Doing an oscillator with LC instead of RC gets better jitter, I'm told.
Startup, though, is less linear if the inductor has any
kind of nonlinear character; you want to worry about things
(self-resonant frequencies, or remanent field) when component
selection time comes around.
John Larkin
2024-01-25 10:05:51 UTC
Permalink
Post by whit3rd
Post by John Larkin
Post by whit3rd
Post by john larkin
I would like a Spice that was, say, 500x as fast as mine is now,
nvidia or something. And I'd love some way to specify results and have
a program juggle values and even library parts for a best solution.
Past attempts at such optimizations have tended to diverge. Even most
interns are smarter than that.
That's why we have math, that tells us that linear equations (like SPICE solves)
have multiple ways to generate large numbers. It's catastrophe theory,
to be precise.
For a stable sine wave oscillator, you can't use ideal C, R, L, and amplifier
components; there won't ever be any solutions that don't diverge, because
the linear-differential-equation solutions all have a matrix raised to
a power (and the power goes up with time). It'll always exponentially
decay or explode, because NO available component tolerances
are negligible effects.
One of my specialities is designing instant-start super low jitter LC
oscillators. The Spice sims are dead on, except for tempco
compensation, which has to be done experimentally. I'd hate to design
such oscillators using differential equations.
Oh, if you run SPICE, you ARE using differential equations... after
Laplace-transforming them to linear equations using lots of "j ?" bits.
The thing that cannot be done, is to do sines with all linear
components; there's no linear equation for a saturating logic
comparator, so it makes a fine oscillator (NE555)
but not a pure sine wave.
If Spice runs differential equations and LaPlace transforms inside, at
least I don't need to know about it. The *concept* of the diff
equation for an LC resonator, energy sloshing around and initial
conditions, is of course basic.
Post by whit3rd
Bill Hewlett's classic sine wave oscillator design got around the
problem with a thermal-varying resistor (square law device,
NOT linear).
Yes, incandescent bulb filament. That's OK for a steady-state audio
oscillator. I recall that there's a bit of THD at low frequencies.
Post by whit3rd
Doing an oscillator with LC instead of RC gets better jitter, I'm told.
Yes, factor of a thousand maybe. Tempcos are much better too.
Post by whit3rd
Startup, though, is less linear if the inductor has any
kind of nonlinear character; you want to worry about things
(self-resonant frequencies, or remanent field) when component
selection time comes around.
Sure. But an air-core inductor is pretty linear. Startup is always
tricky, keeping the first few periods equal to within picoseconds.

Coilcraft has some great parts. Their 1812SMS is kind of magic. I cut
out maybe 4 layers of pcb copper plane below a part like that so the
field doesn't bounce off copper and especially so the pads don't see a
bunch of dreadful FR4 capacitance.
Phil Hobbs
2024-01-25 15:45:48 UTC
Permalink
Post by John Larkin
Post by whit3rd
Post by John Larkin
Post by whit3rd
Post by john larkin
I would like a Spice that was, say, 500x as fast as mine is now,
nvidia or something. And I'd love some way to specify results and have
a program juggle values and even library parts for a best solution.
Past attempts at such optimizations have tended to diverge. Even most
interns are smarter than that.
That's why we have math, that tells us that linear equations (like SPICE solves)
have multiple ways to generate large numbers. It's catastrophe theory,
to be precise.
For a stable sine wave oscillator, you can't use ideal C, R, L, and amplifier
components; there won't ever be any solutions that don't diverge, because
the linear-differential-equation solutions all have a matrix raised to
a power (and the power goes up with time). It'll always exponentially
decay or explode, because NO available component tolerances
are negligible effects.
Depends on what you mean by “ideal.“ For a pure LTI system, I agree.
However, even SPICE isn’t really LTI—besides roundoff and truncation error,
it’s implemented using floating point, which has magnitude limits.

You can easily use a nice noiseless behavioral amp whose gain is a weak
function of the time-averaged amplitude.

That’s an idealized model of the HP200-style ALC.
Post by John Larkin
Post by whit3rd
Post by John Larkin
One of my specialities is designing instant-start super low jitter LC
oscillators. The Spice sims are dead on, except for tempco
compensation, which has to be done experimentally. I'd hate to design
such oscillators using differential equations.
Oh, if you run SPICE, you ARE using differential equations... after
Laplace-transforming them to linear equations using lots of "j ?" bits.
The thing that cannot be done, is to do sines with all linear
components; there's no linear equation for a saturating logic
comparator, so it makes a fine oscillator (NE555)
but not a pure sine wave.
If Spice runs differential equations and LaPlace transforms inside, at
least I don't need to know about it. The *concept* of the diff
equation for an LC resonator, energy sloshing around and initial
conditions, is of course basic.
Post by whit3rd
Bill Hewlett's classic sine wave oscillator design got around the
problem with a thermal-varying resistor (square law device,
NOT linear).
Yes, incandescent bulb filament. That's OK for a steady-state audio
oscillator. I recall that there's a bit of THD at low frequencies.
Post by whit3rd
Doing an oscillator with LC instead of RC gets better jitter, I'm told.
Yes, factor of a thousand maybe. Tempcos are much better too.
Post by whit3rd
Startup, though, is less linear if the inductor has any
kind of nonlinear character; you want to worry about things
(self-resonant frequencies, or remanent field) when component
selection time comes around.
Sure. But an air-core inductor is pretty linear. Startup is always
tricky, keeping the first few periods equal to within picoseconds.
Coilcraft has some great parts. Their 1812SMS is kind of magic. I cut
out maybe 4 layers of pcb copper plane below a part like that so the
field doesn't bounce off copper and especially so the pads don't see a
bunch of dreadful FR4 capacitance.
Cheers

Phil Hobbs
(Off to a second day of deposition in a patent case.)
--
Dr Philip C D Hobbs Principal Consultant ElectroOptical Innovations LLC /
Hobbs ElectroOptics Optics, Electro-optics, Photonics, Analog Electronics
John Larkin
2024-01-25 16:35:32 UTC
Permalink
On Thu, 25 Jan 2024 15:45:48 -0000 (UTC), Phil Hobbs
Post by whit3rd
Post by john larkin
I would like a Spice that was, say, 500x as fast as mine is now,
nvidia or something. And I'd love some way to specify results and have
a program juggle values and even library parts for a best solution.
Past attempts at such optimizations have tended to diverge. Even most
interns are smarter than that.
That's why we have math, that tells us that linear equations (like SPICE solves)
have multiple ways to generate large numbers. It's catastrophe theory,
to be precise.
For a stable sine wave oscillator, you can't use ideal C, R, L, and amplifier
components; there won't ever be any solutions that don't diverge, because
the linear-differential-equation solutions all have a matrix raised to
a power (and the power goes up with time). It'll always exponentially
decay or explode, because NO available component tolerances
are negligible effects.
Depends on what you mean by “ideal.“ For a pure LTI system, I agree.
However, even SPICE isn’t really LTI—besides roundoff and truncation error,
it’s implemented using floating point, which has magnitude limits.
You can easily use a nice noiseless behavioral amp whose gain is a weak
function of the time-averaged amplitude.
That’s an idealized model of the HP200-style ALC.
I just soft clip the Colpitts LC oscillator to about 1 volt p-p, with
a voltage, a diode, and a resistor, with enough excess osc gain to be
sure it's reliable. I only care about the zero crossings - I'm making
a clock - so a little harmonic distortion doesn't matter. Done right,
one can make a sine wave that starts when triggered and is cycle-cycle
invariant starting with the first zero cross. It's a very simple
circuit that has taken us about 30 years to refine. I look back on my
early versions with horror.

One sales point in a digital delay generator is minimal and calibrated
insertion delay. An instant-start clock helps.

The other common DDG technique is to use a crystal-oscillator clock to
time out the coarse delays, and some sort of time measurement and
correction scheme to remove the inherent 1-clock p-p jitter from an
asynchronous trigger [1]. That's messy and adds a lot of insertion
delay. That's what the DG645 does. I rented one to see what's inside.
Turns out that all the "warranty void if sticker removed" stickers
(there were about 10) were the same as ones you can buy from Amazon.

[1] RMS jitter is the clock period divided by sqrt(3) for some bizarre
reason.
boB
2024-01-25 23:02:45 UTC
Permalink
Post by john larkin
On Wed, 24 Jan 2024 10:59:47 -0800 (PST), Fred Bloggs
Post by Fred Bloggs
Post by John Larkin
On Wed, 24 Jan 2024 08:00:15 -0800 (PST), Fred Bloggs
Post by Fred Bloggs
Post by Anthony William Sloman
I don't think so. Just a few words are not enough to specify and generate a specific, reliable design.
They are enough to specify a patentable idea. Reducing it to practice takes a lot more work, and documenting a complete system takes a lot of words (and pictures). Oddly enough, software can do a lot of the documentation.
There's not much I in AI. It's mostly a silly fad.
For particular problems it can already find solutions that humans can't
https://en.wikipedia.org/wiki/Protein_folding
AlphaFold is well ahead of any expert, and some theorem-proving programs operate without making mistakes to a degree that allows them to outperform human mathematicians on specific complex problems. It really isn't any kid of silly fad.
But brainstorming isn't so-called. Done right, it really works.
Mainly by discouraging status-seeking creeps from insisting on concentrating on their own ideas.
To be replaced by group concentration on no ideas at all.
The people have to be right for the process to be productive of ideas.
Some people will poison a brainstorming session, and too much general
sociability in the room will reinforce conventional thinking.
Post by Fred Bloggs
https://resources.pcb.cadence.com/blog/2022-the-role-of-machine-learning-in-analog-circuit-design
That's absurd. Sounds like they are trying to sell cad options to
beginners.
Post by Fred Bloggs
https://www.synopsys.com/blogs/chip-design/ai-analog-design-migration-samsung-safe-forum-2023.html
Certainly a lot of computing helps design digital ICs, but I wouldn't
call that intelligence. Smart people wrote very specialized software.
I sometimes write software to solve circuit problems, but the software
just does what I told it to do.
Post by Fred Bloggs
https://www.planetanalog.com/what-can-ai-do-for-analog-design/
I'd love to have a good component selection tool. The intelligence
would be in inferring things from bad data sheets that have no
standards. It would of course have to read and understand application
schematics and mechanical drawings and find gotchas buried in
footnotes and graphs.
Find me a right-angle Gbit PoE compatible RJ45 jack that has multiple
drop-in sources, two LEDs on the high side, lots of stock from
non-Chinese sources, at a good price. They have to mount on my PCB and
ground to a cutout in my panel. That's an easy one.
https://www.digikey.com/en/products/detail/w%C3%BCrth-elektronik/615008137421/2060608
Post by John Larkin
Post by Fred Bloggs
https://www.electronicdesign.com/technologies/embedded/article/21272567/electronic-design-ai-lends-a-helping-hand-with-analog-and-custom-ic-design
https://semiengineering.com/ai-for-circuit-design-quality-productivity-and-advanced-node-mapping/
http://ieeexplore.ieee.org/document/31523/
The list is endless. Humans are not as unique and special as they make themselves out to be. They'll all be replaced by AI before long.
Has AI ever invented anything?
It's doing things like running through impossibly large numbers of permutations to find something useful, as with drug discovery. It's more the case creative people are using AI to enable inventive ideas.
That's not intelligent. It's just automating a lot of grunt work, as
programmed. Line monte carlo simulation. The person who set it up is
the intelligence. All that's new is having more compute power than we
had in the past.
John, AI is NOT just more computing power.

It is neural networks running on hardware that work well with NN

I see that with the Agent filter you put out that stops the Google
Groups postings stops the Fred Bloggs and other responders evidently.
But I haven't seen any sporge posts either so that is good.

boB
Post by john larkin
Computers automate grunt work and let us work faster and better and
move up the abstraction stack. Nonlinear differential equations were
never much fun.
I would like a Spice that was, say, 500x as fast as mine is now,
nvidia or something. And I'd love some way to specify results and have
a program juggle values and even library parts for a best solution.
Past attempts at such optimizations have tended to diverge. Even most
interns are smarter than that.
Post by Fred Bloggs
Post by John Larkin
I check up on Flux.ai now and then. I wonder when they will run out of
money.
Take a look at flux. It's funny.
Joe Gwinn
2024-01-25 23:38:28 UTC
Permalink
[snip]
Post by boB
Post by john larkin
That's not intelligent. It's just automating a lot of grunt work, as
programmed. Line monte carlo simulation. The person who set it up is
the intelligence. All that's new is having more compute power than we
had in the past.
John, AI is NOT just more computing power.
It is neural networks running on hardware that work well with NN
Yes and no. No computer hardware "works well with NN" (Neural
Networks), because nobody has invented a true direct associative
hardware memory yet, so the computer hardware roughly emulates the
desired NN kind and behaviors, at stunning expense in hardware and
electrical power (plus cooling systems to remove all that heat).

It is useful to note that the pacing task for implementing a neural
net in a digital computer is matrix inversion, where the matrix to be
inverted may be billions of lines by billions of rows, and is not at
all sparse. It's a long story, but well documented.

By contrast, the human brain has a volume of about 1.3 liters and
consumes about 20 watts, and contains something like 171 billion
cells, of which 86 billion are neurons. Computations are performed by
analog hardware, cells. The number of synapses per neuron is
something like a factor of ten thousand larger. Every synapse needs a
cell in the matrix holding at least an 8-bit value. And so on.

Joe Gwinn
john larkin
2024-01-26 00:03:10 UTC
Permalink
Post by Joe Gwinn
[snip]
Post by boB
Post by john larkin
That's not intelligent. It's just automating a lot of grunt work, as
programmed. Line monte carlo simulation. The person who set it up is
the intelligence. All that's new is having more compute power than we
had in the past.
John, AI is NOT just more computing power.
If it runs on a compuer it sure is.
Post by Joe Gwinn
Post by boB
It is neural networks running on hardware that work well with NN
Do any NNs work well? NNs are cargo-cult crude cartoons of an actual
organism. Single- and few-cell organisms without a nervous system
learn and do complex stuff, and our brain has billions.
Post by Joe Gwinn
Yes and no. No computer hardware "works well with NN" (Neural
Networks), because nobody has invented a true direct associative
hardware memory yet, so the computer hardware roughly emulates the
desired NN kind and behaviors, at stunning expense in hardware and
electrical power (plus cooling systems to remove all that heat).
It is useful to note that the pacing task for implementing a neural
net in a digital computer is matrix inversion, where the matrix to be
inverted may be billions of lines by billions of rows, and is not at
all sparse. It's a long story, but well documented.
By contrast, the human brain has a volume of about 1.3 liters and
consumes about 20 watts, and contains something like 171 billion
cells, of which 86 billion are neurons. Computations are performed by
analog hardware, cells. The number of synapses per neuron is
something like a factor of ten thousand larger. Every synapse needs a
cell in the matrix holding at least an 8-bit value. And so on.
Joe Gwinn
And a human brain can play tennis, or recognize one face out of a
million, or design circuits, with wet chemistry gates that have
millisecond prop delays.
Anthony William Sloman
2024-01-26 03:59:10 UTC
Permalink
Post by Joe Gwinn
[snip]
Post by boB
Post by john larkin
That's not intelligent. It's just automating a lot of grunt work, as
programmed. Line monte carlo simulation. The person who set it up is
the intelligence. All that's new is having more compute power than we
had in the past.
John, AI is NOT just more computing power.
If it runs on a computer it sure is.
It runs on a different kind of computer, which works in a rather different way from the micropressors we are used to.
Post by Joe Gwinn
Post by boB
It is neural networks running on hardware that work well with NN
Do any NNs work well? NNs are cargo-cult crude cartoons of an actual
organism. Single- and few-cell organisms without a nervous system
learn and do complex stuff, and our brain has billions.
They are an abstractions of what biological neural networks do, and to that extent they are cruder than the parts they are designed to emulate.

As with all new and poorly understood technology, there are loads of confidence tricksters trying to sell people neural network solutions that don't work.
Post by Joe Gwinn
Yes and no. No computer hardware "works well with NN" (Neural
Networks), because nobody has invented a true direct associative
hardware memory yet, so the computer hardware roughly emulates the
desired NN kind and behaviors, at stunning expense in hardware and
electrical power (plus cooling systems to remove all that heat).
It is useful to note that the pacing task for implementing a neural
net in a digital computer is matrix inversion, where the matrix to be
inverted may be billions of lines by billions of rows, and is not at
all sparse. It's a long story, but well documented.
By contrast, the human brain has a volume of about 1.3 liters and
consumes about 20 watts, and contains something like 171 billion
cells, of which 86 billion are neurons. Computations are performed by
analog hardware, cells. The number of synapses per neuron is
something like a factor of ten thousand larger. Every synapse needs a
cell in the matrix holding at least an 8-bit value. And so on.
And a human brain can play tennis, or recognize one face out of a
million, or design circuits, with wet chemistry gates that have
millisecond prop delays.
So what?
--
Bill Sloman, Sydney
John Walliker
2024-01-26 12:17:06 UTC
Permalink
Post by Anthony William Sloman
Post by Joe Gwinn
[snip]
Post by boB
Post by john larkin
That's not intelligent. It's just automating a lot of grunt work, as
programmed. Line monte carlo simulation. The person who set it up is
the intelligence. All that's new is having more compute power than we
had in the past.
John, AI is NOT just more computing power.
If it runs on a computer it sure is.
It runs on a different kind of computer, which works in a rather different way from the micropressors we are used to.
Post by Joe Gwinn
Post by boB
It is neural networks running on hardware that work well with NN
Do any NNs work well? NNs are cargo-cult crude cartoons of an actual
organism. Single- and few-cell organisms without a nervous system
learn and do complex stuff, and our brain has billions.
They are an abstractions of what biological neural networks do, and to that extent they are cruder than the parts they are designed to emulate.
As with all new and poorly understood technology, there are loads of confidence tricksters trying to sell people neural network solutions that don't work.
Post by Joe Gwinn
Yes and no. No computer hardware "works well with NN" (Neural
Networks), because nobody has invented a true direct associative
hardware memory yet, so the computer hardware roughly emulates the
desired NN kind and behaviors, at stunning expense in hardware and
electrical power (plus cooling systems to remove all that heat).
It is useful to note that the pacing task for implementing a neural
net in a digital computer is matrix inversion, where the matrix to be
inverted may be billions of lines by billions of rows, and is not at
all sparse. It's a long story, but well documented.
By contrast, the human brain has a volume of about 1.3 liters and
consumes about 20 watts, and contains something like 171 billion
cells, of which 86 billion are neurons. Computations are performed by
analog hardware, cells. The number of synapses per neuron is
something like a factor of ten thousand larger. Every synapse needs a
cell in the matrix holding at least an 8-bit value. And so on.
And a human brain can play tennis, or recognize one face out of a
million, or design circuits, with wet chemistry gates that have
millisecond prop delays.
So what?
Some useful NN architectures such as the multi-layer perceptron can fit very well
and efficiently onto a DSP using fixed point arithmetic once training is complete.

John
whit3rd
2024-01-26 22:25:23 UTC
Permalink
Post by Anthony William Sloman
Post by john larkin
Do any NNs work well? NNs are cargo-cult crude cartoons of an actual
organism.
And a human brain can play tennis, or recognize one face out of a
million, or design circuits, with wet chemistry gates that have
millisecond prop delays.
So what?
The takeaway here, is that we social animals have evolved IFF (identification
friend-or-foe) firmware, because that's an important task. Thus,
it's no surprise that facial recognition can be precise.
So can food-aroma and flavor distinction, there's things your
ancestors did not want to eat.

Some of our other important tasks are also hard to relegate to logic gates,
or analog computers, or locks/keys/cams. The expansion of a part of
our technology into those important tasks is inevitable, and... just
in its infancy.

The wetware IFF works for facial recognition quite well. It doesn't
apply to distinguishing Donald Trump against a real leadership candidate
nearly so well. Some voters are virtually still infants, I suspect.
boB
2024-01-28 02:35:30 UTC
Permalink
Post by john larkin
Post by Joe Gwinn
[snip]
Post by boB
Post by john larkin
That's not intelligent. It's just automating a lot of grunt work, as
programmed. Line monte carlo simulation. The person who set it up is
the intelligence. All that's new is having more compute power than we
had in the past.
John, AI is NOT just more computing power.
If it runs on a compuer it sure is.
Different kind of computer.

Well, in this case, using multi-core graphics controllers is
organized way better for the die size allowed in the neural network
application. Its different than what you can get these days in a
particular die size for the particula application. A PC micro,
although as or more complicated than an Nvidea graphics processor,
it works better for that.

As for being THE processor for a NN like Joe better points out, these
processors are what we had (and have) at the time so they were used.

Just wait for a just as complicated, circuitry wise, neural network
chip somes out. It will be even better.

Anyway, NNs are different too in that they "learn" better than a
"program" running on a 100X faster PC chip. That's what I considered
being different than just more computing power. In a non-linear way
mabe.

boB
Post by john larkin
Post by Joe Gwinn
Post by boB
It is neural networks running on hardware that work well with NN
Do any NNs work well? NNs are cargo-cult crude cartoons of an actual
organism. Single- and few-cell organisms without a nervous system
learn and do complex stuff, and our brain has billions.
Post by Joe Gwinn
Yes and no. No computer hardware "works well with NN" (Neural
Networks), because nobody has invented a true direct associative
hardware memory yet, so the computer hardware roughly emulates the
desired NN kind and behaviors, at stunning expense in hardware and
electrical power (plus cooling systems to remove all that heat).
It is useful to note that the pacing task for implementing a neural
net in a digital computer is matrix inversion, where the matrix to be
inverted may be billions of lines by billions of rows, and is not at
all sparse. It's a long story, but well documented.
By contrast, the human brain has a volume of about 1.3 liters and
consumes about 20 watts, and contains something like 171 billion
cells, of which 86 billion are neurons. Computations are performed by
analog hardware, cells. The number of synapses per neuron is
something like a factor of ten thousand larger. Every synapse needs a
cell in the matrix holding at least an 8-bit value. And so on.
Joe Gwinn
And a human brain can play tennis, or recognize one face out of a
million, or design circuits, with wet chemistry gates that have
millisecond prop delays.
Anthony William Sloman
2024-01-25 01:17:08 UTC
Permalink
Post by Fred Bloggs
Post by Anthony William Sloman
I don't think so. Just a few words are not enough to specify and generate a specific, reliable design.
They are enough to specify a patentable idea. Reducing it to practice takes a lot more work, and documenting a complete system takes a lot of words (and pictures). Oddly enough, software can do a lot of the documentation.
There's not much I in AI. It's mostly a silly fad.
For particular problems it can already find solutions that humans can't
https://en.wikipedia.org/wiki/Protein_folding
AlphaFold is well ahead of any expert, and some theorem-proving programs operate without making mistakes to a degree that allows them to outperform human mathematicians on specific complex problems. It really isn't any kid of silly fad.
But brainstorming isn't so-called. Done right, it really works.
Mainly by discouraging status-seeking creeps from insisting on concentrating on their own ideas.
To be replaced by group concentration on no ideas at all.
You've not taken apart in a brain-storming session?

<snipped ads>
Post by Fred Bloggs
The list is endless. Humans are not as unique and special as they make themselves out to be. They'll all be replaced by AI before long.
And the sooner we can automate pessimism the faster we can replace Fred Bloggs.

Fred doesn't seem to realise how much of circuit design is tedious drudgery, which can and should be taken over by machines. He seems to be equally ignorant of the necessity to find the right point of view, which doesn't seen to be a task that anybody has automated yet. Quite a few circuit designers aren't aware that they need to find the right point of view, and soldier on, using the 555 (or whatever tool they've latched onto) to solve every problem that gets served up to them.
--
Bill Sloman, Sydney
RichD
2024-01-27 01:47:42 UTC
Permalink
Post by Fred Bloggs
Humans are not as unique and special as they make themselves out to be.
They'll all be replaced by AI before long.
This has already begun -
https://arxiv.org/abs/1810.10525

Who needs physicists?



--
Rich
Anthony William Sloman
2024-01-22 22:50:55 UTC
Permalink
Post by Fred Bloggs
Post by John Larkin
Post by Phil Hobbs
On 2024-01-21 10:12, John Larkin wrote:>
<snip>
Post by Fred Bloggs
Post by John Larkin
Post by Phil Hobbs
Hmm. I don't think that I agree in general, because you make it sound
as though the process were just intuitively plucking one idea out of
somewhere-or-other and cranking it out.
If an idea is new, where else would come from?
New inventions tend to show up in several different places at much the same time. New ideas spread around, and turn into patentable innovations in different people's brains. I have had original ideas that stayed original for nearly twenty years, but most of them turned out to have shown up elsewhere earlier, even if they weren't put into practice all that well, if at all.
Post by Fred Bloggs
Within 5 years, all this manual fiddling, and so-called brainstorming, will be reduced to an AI-app resident on a $ phone. It may not be optimum, but it will work.
Probably not. The trick is finding the right point of view, and software is written around a particular point of view.
--
Bill Sloman, Sydney
Anthony William Sloman
2024-01-22 00:07:29 UTC
Permalink
Post by Phil Hobbs
JL wrote an interesting post in the depths of the "better
microelectronics from coal" thread that I thought was worth pulling out
on its own.
On 2024-01-21 10:12, John Larkin wrote:>
"...what IS electronic
design, and what's the best way to do it? <snip>
Short answer, cobbling. When presented with a problem or an
opportunity to design electronics, the most efficient way to do that
is to grab a piece of paper and immediately sketch a circuit or an
assembly.
"Efficiency" is a ratio of a theoretically known best way of doing something with what actually happens.

There's no theory that can let you work out the "best" design for any particular application, so it's not a word that meas anything useful i this context.

In my experience the first step in any design is to work out exactly what you want to happen. Then you have to think of ways of making it happen.
Sometimes there's an obvious solution - if you are answering a frequently asked question there's often a frequently adopted solution.

New components can suggest solutions that weren't previously practical.
Post by Phil Hobbs
Sometimes one can do that instantly, without thinking, or
sometimes one can ignore the issue for a few days and then the design
pops up. Sometimes brainstorming and whiteboarding help. Sometimes
fiddling with Spice helps.
Thinking about the problem always helps. Conscious thought can often kick the sub-conscious into action, and solutions can pop up from there, but you have to put in the conscious effort to get the process started.
Post by Phil Hobbs
All that literature research and math analysis and simulation and
breadboarding and prototyping are just slow and expensive follow-up
chores for people who don't have 100% confidence in their instincts.
Sane people.

For me the hall-mark of actual design is the process of discarding one approach and trying another. John Larkin never talks about that.

It's not fun, and it doesn't make you look good, but it does clarify your thinking and let you hone in on what really matters.
Post by Phil Hobbs
Analysis, sometimes prudent to do, but not design.
Not John Larkin's idea of design, which is idiosycratic.
Post by Phil Hobbs
Design is subconsious and instinctive. And it's free! And to some
extent, it can be taught, but seldom is.
Design can go on in the subconscious, but it isn't free. Waking up in the middle of the night and feeling the urge to sketch a circuit diagram isn't something that makes your wife happy.
Post by Phil Hobbs
Most of us design things to sell, so do whatever works. We're selling
stuff, not publishing papers.
If you have got something unique to sell. writing a paper about it is useful publicity. The instrument literature exists to inform people about solutions that people have worked up to solve academically interesting problems, but most of them are commercially interesting problems as well.
Post by Phil Hobbs
Hmm. I don't think that I agree in general, because you make it sound
as though the process were just intuitively plucking one idea out of
somewhere-or-other and cranking it out.
I don't think that John Larkin actually does circuit design in the way that most people understand process. A great many circuits are produced by evolution rather than intelligent design. and quite a few of them work really well. Quite a few more can work better (and be built more cheaply) if you spend some time thinking about what they are doing. I spent about half my time industry doing just that.
Post by Phil Hobbs
You've often argued in favor of brainstorming, where you get a few smart
people in front of a white board and try out ideas to find the best one
and flesh it out. We've done that together, very fruitfully.
It's possible to do more or less the same thing by oneself, but it
requires the ability to tolerate uncertainty for extended periods.
(That's a skill well worth developing, which most people are really,
really bad at, IME.)
I sometimes need to do a family of designs, rather than just one.
Recently I've been working on some very fast, very cheap SPAD preamps,
intended to go in the guts of positron-emission scanners.
Designs with lots of real-world constraints are often the most fun, and
this one's specs include: 300-ps edges with 100-ps timing repeatability
from unit to unit; no magnetics allowed; and a BOM cost of $1 or less.
(You need a whole lot of channels, and PET and MRI machines are often
combined.)
I do a fair amount of analysis of circuits of that sort, to figure out
what actually limits their performance. It isn't super detailed--in
this case, just enough to figure out whether it'll be the base-emitter
time constant, the Miller effect, or the SPAD's series resistance that
will be the limiting factor.
Miller, I can deal with using circuit hacks. The BE time constant is
Rbb' * Cbe, which gets slightly worse at high current, but is mainly a
device parameter--to get a big improvement you have to change
transistors. The SPAD can be negotiable depending on whose process
you're making them on--when each machine needs thousands of them,
vendors tend to listen.
If you can buy a 100,000 parts you can go for an application specific device.
Post by Phil Hobbs
Eventually, of course, you have to pick one and go with it, but picking
a topology usually takes me an iteration or two.
My experience too.
--
Bill Sloman, Sydney
Dan Purgert
2024-01-22 23:04:57 UTC
Permalink
Post by Phil Hobbs
JL wrote an interesting post in the depths of the "better
microelectronics from coal" thread that I thought was worth pulling out
on its own.
On 2024-01-21 10:12, John Larkin wrote:>
"...what IS electronic
design, and what's the best way to do it? <snip>
Short answer, cobbling. When presented with a problem or an
opportunity to design electronics, the most efficient way to do that
is to grab a piece of paper and immediately sketch a circuit or an
assembly. Sometimes one can do that instantly, without thinking, or
sometimes one can ignore the issue for a few days and then the design
pops up. Sometimes brainstorming and whiteboarding help. Sometimes
fiddling with Spice helps.
All that literature research and math analysis and simulation and
breadboarding and prototyping are just slow and expensive follow-up
chores for people who don't have 100% confidence in their instincts.
Analysis, sometimes prudent to do, but not design.
Design is subconsious and instinctive. And it's free! And to some
extent, it can be taught, but seldom is.
Most of us design things to sell, so do whatever works. We're selling
stuff, not publishing papers.
Hmm. I don't think that I agree in general, because you make it sound
as though the process were just intuitively plucking one idea out of
somewhere-or-other and cranking it out.
You've often argued in favor of brainstorming, where you get a few smart
people in front of a white board and try out ideas to find the best one
and flesh it out. We've done that together, very fruitfully.
It's possible to do more or less the same thing by oneself, but it
requires the ability to tolerate uncertainty for extended periods.
(That's a skill well worth developing, which most people are really,
really bad at, IME.)
It helps if one has a rubber duckie (or maybe I'm just that bad at it!)
--
|_|O|_|
|_|_|O| Github: https://github.com/dpurgert
|O|O|O| PGP: DDAB 23FB 19FA 7D85 1CC1 E067 6D65 70E5 4CE7 2860
Loading...