Discussion:
PCB Layout (for ADC and DAC)
(too old to reply)
Jon Slaughter
18 years ago
Permalink
Reading over some layout issues involving data conversion it seems that most
of the stuff is well covered such as using seperate supplies, grounds,
bypass caps on power pins, etc...

Some issues I'm not sure about are when going from analog to digital and
vice versa.

In a datasheet involving a DAC(DAD1793 from TI) they show the recommended
layout but I'm a bit confused. The DAC is on the analog side but its half
digital. Its digital ground is grounded to the analog side. Would it not be
better to have its digital ground go to the digital ground plane? Of course
since the IC is over the analog ground plane it means that the there would
have to be a little routing but I'm thinking that maybe the digital ground
plane could be extendend to half between the chip and vias would go directly
to it?

Also, is there any reason for the ground plane to be on a seperate layer? Is
it for just convenience and to maximize area or is there some other concept
involved. Since I don't have any double sided pcb boards and if I did(which
I suppose I could make but sticking two single sided ones together) I'm not
sure how to make the vias(drill and solder/maybe use a wire connector). I'm
sure I can get away with using a single side for prototyping though? (As
noise issues won't be the ultimate factor unless its pretty bad)


What I was thinking about was splitting the ground plane/pcb board into two
halfs where the ic's that were partially digital and partially analog would
straddle both sides. This would allow me to sorta seperate the two sides to
some degree but not sure if it would cause other issues.

Also, in this case would using one digital supply for all digital elements
work? In the two layouts I have saw for ADC's and DAC's they use one supply
for the digital and then one supply for the the DAC and/or ADC.

i.e., one supply is completely digital but the one supplying analog power
also supplies the digital for the ADC's and DAC's. Is this necessary(in
that, is it better or worse than having only two supplies one for digital
and one for analog)?

One more question. I didn't realize that the DAC's and ADC's where
synchronous devices and need external clocks(thought they had them build in
and I could use interrupts to transfer data). Can I use the same clock for
both/all conversion ic's? I need a pretty stable and fixed clock because I'm
using it for audio so the pitch needs to be right(I don't think this will be
an issue but I'd like not to introduce more complications by having many
different clocks it isn't necessary).


Thanks,
Jon
MooseFET
18 years ago
Permalink
...
ADCs and DACs always have the issue of the digital circuit messing
with the analog workings. Internally, the chip has a path from the
digital logic ground to the analog ground. You don't want any current
to flow in this path. To prevent this, the digital ground must have a
very low impedance path. You want the digital ground plane to provide
all of the return currents for the digital lines so it needs to be
under them.
Post by Jon Slaughter
Also, is there any reason for the ground plane to be on a seperate layer?
Making a layer for a ground makes it work as a continuous sheet of low
impedance ground. When you combine it with a signal layer, you have
to plow traces through it. This increases the impedance.

And ...
Post by Jon Slaughter
Is
it for just convenience and to maximize area or is there some other concept
involved.
At RF frequencies slots and gaps in planes look like little tuned
circuits and antennas. If you circuit can drive an RF current into
the structure, it will radiate.
Post by Jon Slaughter
Since I don't have any double sided pcb boards and if I did(which
I suppose I could make but sticking two single sided ones together) I'm not
sure how to make the vias(drill and solder/maybe use a wire connector). I'm
sure I can get away with using a single side for prototyping though? (As
noise issues won't be the ultimate factor unless its pretty bad)
Use your single sided PCB with the traces making the top side.
Scatter largish holes (0.05 or more) for the ground connection. Take
an unprocessed sheet of copper clad and place it under your PCB and
match drill smaller holes (0.035) in it for the ground.

Solder long lengths of wire into the holes of the copper clad. Don't
use too much solder on the top surface. Bend the bottom side of the
wires over and not the top.

Feed the wires through the holes in the PCB and slide the PCB along
the wires down onto the copperclad. Bend over the wires and solder
and snip.
Post by Jon Slaughter
What I was thinking about was splitting the ground plane/pcb board into two
halfs where the ic's that were partially digital and partially analog would
straddle both sides. This would allow me to sorta seperate the two sides to
some degree but not sure if it would cause other issues.
Beware of increasing the impedance of the digital ground paths.
Post by Jon Slaughter
Also, in this case would using one digital supply for all digital elements
work? In the two layouts I have saw for ADC's and DAC's they use one supply
for the digital and then one supply for the the DAC and/or ADC.
Filter the heck out of your supplies. An RF bead, inductor or
resistor and a couple of 47u capacitors can make quite a difference.


Consider this bit of ASCII art:

---+------[ZT]-------+-------[ZT]--------+-----[ZT]-------+---------
! ! ! !
--/\/--+---A --/\/--+---B --/\/--+---C !
! ! ! ---
--- --- --- ---
--- C1 --- ---C3 !
! ! ! GND
GND GND GND

[ZT} = the impedance of the trace

Notice how if the reistors were zero ohms, some of an AC current
applied to B would flow through C1 and C3 into the local grounds at
those locations. It doesn't take much resistance to reduce this.
...
John Larkin
18 years ago
Permalink
On Sun, 22 Jul 2007 03:06:48 -0500, "Jon Slaughter"
...
Splitting planes is almost always a bad idea.

If you are restricted to etching single-sided boards, consider using
double-side copperclad and leaving the back side unetched, a solid
copper ground plane. Drill topside ground pads or, better, islands,
and solder via wires through to the ground plane. This will give you
far better signal integrity than a 1-sided, split-ground layout is
likely able to do.

The big performance cut in pc boards is between boards with true
ground planes, and boards without.

John
Eeyore
18 years ago
Permalink
Post by John Larkin
Splitting planes is almost always a bad idea.
And almost universally done in audio or you'll hear low-level 'birdies' in the
signal.

Graham
Joerg
18 years ago
Permalink
Post by Eeyore
Post by John Larkin
Splitting planes is almost always a bad idea.
And almost universally done in audio or you'll hear low-level 'birdies' in the
signal.
Until the audio gear is used near a strong transmitter and then any
split will cause some real birdies. Or worst case a fzzzzt ... poof.
--
Regards, Joerg

http://www.analogconsultants.com
Vladimir Vassilevsky
18 years ago
Permalink
Post by Joerg
Post by Eeyore
Post by John Larkin
Splitting planes is almost always a bad idea.
And almost universally done in audio or you'll hear low-level
'birdies' in the
signal.
Until the audio gear is used near a strong transmitter and then any
split will cause some real birdies. Or worst case a fzzzzt ... poof.
Well. Indeed the star grounding is very common in audio. As for the EMC
susceptibility problem, Graham is aggressively unaware of it.

Here are some practical considerations for the split ground:

1. A separate island in the ground plane can be usefull if you want to
isolate something very noisy or very sensitive from the rest of schematics.

2. Using common ground plane as the universal signal/power return path
may not be a good idea if the high currents flow through it.

3. There may be other connections to the ground. It may not always be
possible to connect everything to the ground at your board. So you have
to spit the grounds to avoid or alleviate loop currents.


Vladimir Vassilevsky

DSP and Mixed Signal Design Consultant

http://www.abvolt.com
Joerg
18 years ago
Permalink
...
It can help but a good design wouldn't need such islands :-)
Post by Vladimir Vassilevsky
2. Using common ground plane as the universal signal/power return path
may not be a good idea if the high currents flow through it.
It's the only way to avoid RF interference. Star grounds don't work well
in such situations because a connection to that star point will act as a
loop antenna and inject noise currents into wherever it ends. When I was
a young ham radio operator I had to remedy lots of those cases, mostly
stuff like electronic organs. To the point where I almost got sick of
it. Luckily copper stock was cheap in those days and I was often
rewarded with a nice tune played on the instrument (I can't play myself).
Post by Vladimir Vassilevsky
3. There may be other connections to the ground. It may not always be
possible to connect everything to the ground at your board. So you have
to spit the grounds to avoid or alleviate loop currents.
#3 is usually a sure sign that somthing went wrong. You brought up
exactly the point that is often misunderstood: Systems do not operate in
isolation, there is stuff connected to them or people wouldn't need the
system. A really good system will not have a problem with ground loops.

Of course you might not know the other end of the cable and sometimes
things like isolation transformers are needed, especially in audio.
Splitting grounds inside a system is, however, IMHO rarely a good idea.

The best class to learn ground structures would be to open up some
surplus military stuff. They know how to do it right because they have
to pass much stricter EMI tests than we do. OTOH if everyone would do
that a third of my business would shrivel up ;-)

The last star architecture I ripped out was about a couple of weeks ago.
Lower end of the audio range, noise up to kazoo. Now there is only the
normal background hiss left.
--
Regards, Joerg

http://www.analogconsultants.com
Vladimir Vassilevsky
18 years ago
Permalink
...
Perfect systems are made by perfect people in the perfect world. I haven't
met any perfect people, although I have seen many who claimed themselves to
be perfect. Just look into this newsgroup. What is scary some of them
really believe that they are :-) Most often you have to deal with what you
got, and there is not much that you can change.
Post by Joerg
Of course you might not know the other end of the cable and sometimes
things like isolation transformers are needed, especially in audio.
You are kidding. Audio is very cost sensitive. It is also the traditional
domain for fanatics and ignorants :-)
Post by Joerg
Splitting grounds inside a system is, however, IMHO rarely a good idea.
No good. But it is something that you may have to do.
Post by Joerg
The best class to learn ground structures would be to open up some
surplus military stuff.
Yes. I ripped apart some old stuff. Here is what is interesting: all
currents which belong to one stage are connected at one point. So there is
no AC current flow through the grounds. Then this point is connected to the
common ground rail. Only then this rail is connected to the chassis. The
position of the points of connection seem to be optimized.
Post by Joerg
They know how to do it right because they have
to pass much stricter EMI tests than we do. OTOH if everyone would do
that a third of my business would shrivel up ;-)
I develop the OEM stuff for automotive. They have hell of EMC requirements,
and there are the hard constraints on the cost also. Perhaps, the medical
equipment that you work on have the strict regulations, too. FCC class A/B
is just a joke.
Post by Joerg
The last star architecture I ripped out was about a couple of weeks ago.
Lower end of the audio range, noise up to kazoo. Now there is only the
normal background hiss left.
"There is no and there can't be any replacement for the intelligence,
experience, common sense and good taste" (Stroustrup)
Post by Joerg
--
Regards, Joerg
http://www.analogconsultants.com
Vladimir Vassilevsky
DSP and Mixed Signal Consultant
www.abvolt.com
Eeyore
18 years ago
Permalink
...
What is really scary is that you think YOU have a clue.

The idea you suggested that digital audio equipment will fail EMC if it doesn't
employ 'brick-wall' anti-aliasing filter is particularly telling.

It tells me you're clueless about what's tested in EMC tests, how those tests
are applied, how the signal path may be affected and what constitutes a failure
for starters.

Graham
Vladimir Vassilevsky
18 years ago
Permalink
Post by Eeyore
What is really scary is that you think YOU have a clue.
The idea you suggested that digital audio equipment will fail EMC if it doesn't
employ 'brick-wall' anti-aliasing filter is particularly telling.
It tells me you're clueless about what's tested in EMC tests, how those tests
are applied, how the signal path may be affected and what constitutes a failure
for starters.
My dear Graham,

What is tested, how applied and etc. is very clearly stated in the test
specifications. You can speculate and bitch about whatever, but if the
spec requires "no audible effects at BCI 240mA 1MHz.. ..1GHz", it should
work exactly like that. It is "yes/no" question, not "if/why" question.

I want to thank you for your valuable comments and assure that when I
will need your professional opinion, I will certainly ask for it. Until
then please refrain from sending messages addressed to me; I won't be
able to read it anyway.

With Best Regards,

Vladimir Vassilevsky

DSP and Mixed Signal Design Consultant

http://www.abvolt.com
Eeyore
18 years ago
Permalink
...
There is no IEC standard that ever specifies "no audible effects".

And the performance levels required are anything but yes/no. As you ought to know.

Graham
Joerg
18 years ago
Permalink
...
Yep. Only one ever was and got hanged on a cross.
Post by Vladimir Vassilevsky
Post by Joerg
Of course you might not know the other end of the cable and sometimes
things like isolation transformers are needed, especially in audio.
You are kidding. Audio is very cost sensitive. It is also the traditional
domain for fanatics and ignorants :-)
Cost sensitive? I find a lot of audio gear highly overpriced and some of
it over-engineered. But then, as you wrote, there are people who pay
whatever it takes.
Post by Vladimir Vassilevsky
Post by Joerg
Splitting grounds inside a system is, however, IMHO rarely a good idea.
No good. But it is something that you may have to do.
Not in my whole career so far. Except where required for isolation, for
example in patient interfaces.
Post by Vladimir Vassilevsky
Post by Joerg
The best class to learn ground structures would be to open up some
surplus military stuff.
Yes. I ripped apart some old stuff. Here is what is interesting: all
currents which belong to one stage are connected at one point. So there is
no AC current flow through the grounds. Then this point is connected to the
common ground rail. Only then this rail is connected to the chassis. The
position of the points of connection seem to be optimized.
Take some RF gear and check again. Inside you typically find a solid
ground plane with the perimeter exposed and properly plated. This is
mounted into a properly plated milled enclosure by a gazillion bolts.

A split ground can cause stuff to fry upon an EMP. For a pilot on a
mission that could create a major "Oh s..t!" experience, especially if
he has to land on a carrier at night.
Post by Vladimir Vassilevsky
Post by Joerg
They know how to do it right because they have
to pass much stricter EMI tests than we do. OTOH if everyone would do
that a third of my business would shrivel up ;-)
I develop the OEM stuff for automotive. They have hell of EMC requirements,
and there are the hard constraints on the cost also. Perhaps, the medical
equipment that you work on have the strict regulations, too. FCC class A/B
is just a joke.
Ours must not cause patient currents beyond the low uA range when hit
with a 5kV defibrillator pulse. In nearly all my cases there is the
added requirement that the system must not be permanently damaged and
has to come back to fully operational mode within very few seconds.
Post by Vladimir Vassilevsky
Post by Joerg
The last star architecture I ripped out was about a couple of weeks ago.
Lower end of the audio range, noise up to kazoo. Now there is only the
normal background hiss left.
"There is no and there can't be any replacement for the intelligence,
experience, common sense and good taste" (Stroustrup)
:-)
--
Regards, Joerg

http://www.analogconsultants.com
MooseFET
18 years ago
Permalink
On Jul 22, 4:39 pm, Vladimir Vassilevsky <***@hotmail.com>
wrote:
[... grounding ...]
Post by Vladimir Vassilevsky
1. A separate island in the ground plane can be usefull if you want to
isolate something very noisy or very sensitive from the rest of schematics.
The sensitive bit is usually analog and the noisiest is usually a DC-
DC converter or the like. Neither of these need as many routing
layers as the digital part so you can use up a layer as a local ground
plane above the overall one. In crosss section it may look lile this:

Power Logic Analog
Parts -> -[==]- -[==]- -[==]-
Top -> TTTTTTTT TTTTTTTT TTTTTTTT
Inner1 -> GGGGGGGG VVVVVVVV GGGGGGGG
Inner2 -> TTTTTTTT TTTTTTTT TTTTTTTT
Inner3 -> GGGGGGGGGGGGGGGGGGGGGGGGGGGG
Inner4 -> TTTTTTTTTTTTTTTTTTTTTTTTTTTT
Bot -> TTTTTTTTTTTTTTTTTTTTTTTTTTTT


T = Traces
G = Ground
V = Voltage

If you are careful, you can put the noisy nodes of the power supply
all above its local ground. This allows the rest of the layers to get
by without picking up too much noise.
Post by Vladimir Vassilevsky
2. Using common ground plane as the universal signal/power return path
may not be a good idea if the high currents flow through it.
3. There may be other connections to the ground. It may not always be
possible to connect everything to the ground at your board. So you have
to spit the grounds to avoid or alleviate loop currents.
Vladimir Vassilevsky
DSP and Mixed Signal Design Consultant
http://www.abvolt.com
Vladimir Vassilevsky
18 years ago
Permalink
Post by MooseFET
[... grounding ...]
Post by Vladimir Vassilevsky
1. A separate island in the ground plane can be usefull if you want to
isolate something very noisy or very sensitive from the rest of schematics.
The sensitive bit is usually analog and the noisiest is usually a DC-
DC converter or the like.
You will laugh, but one of the most noisy things are the delta-sigma AD
converters. They draw high ripple current with their clock rate; that
creates spikes up to hundreds of MHz. The DC-DC convertors are not too bad;
it's mainly low frequency conducted stuff plus magnetic fields.
Post by MooseFET
Neither of these need as many routing
layers as the digital part so you can use up a layer as a local ground
plane above the overall one.
Some books recommend to cut a hole in the overall ground plane under the
local ground to reduce the capacitive coupling. The idea is to lock the
noise on the island and don't let it spread around. All communication to the
island is done by one bridge, so there is no way for the noise currents to
the main board.
Post by MooseFET
If you are careful, you can put the noisy nodes of the power supply
all above its local ground. This allows the rest of the layers to get
by without picking up too much noise.
That certainly makes sense.


Vladimir Vassilevsky
DSP and Mixed Signal Design Consultant
http://www.abvolt.com
Eeyore
18 years ago
Permalink
Post by Vladimir Vassilevsky
Post by Joerg
Post by Eeyore
Post by John Larkin
Splitting planes is almost always a bad idea.
And almost universally done in audio or you'll hear low-level
'birdies' in the signal.
Until the audio gear is used near a strong transmitter and then any
split will cause some real birdies. Or worst case a fzzzzt ... poof.
Well. Indeed the star grounding is very common in audio. As for the EMC
susceptibility problem, Graham is aggressively unaware of it.
FYI you ignorant prick, I've been taking an active interest in EMC since 1987
and been the internal 'expert' on the subject for 3 companies.

I also have considerable experience designing products that sail through their
EMC tests. I regularly supervise such tests in fact, not least since the test
labs themsleves often don't understand how to appropriately test complex
non-consumer apparatus. And yes, those products include ones that have separate
analog and digital '0Vs' in them. There may of course be some care involved as
how those circuits interface to the outside world.
Post by Vladimir Vassilevsky
1. A separate island in the ground plane can be usefull if you want to
isolate something very noisy or very sensitive from the rest of schematics.
Doesn't get round the need for currents to have return paths though.
Post by Vladimir Vassilevsky
2. Using common ground plane as the universal signal/power return path
may not be a good idea if the high currents flow through it.
Well fancy that !
Post by Vladimir Vassilevsky
3. There may be other connections to the ground. It may not always be
possible to connect everything to the ground at your board. So you have
to spit the grounds to avoid or alleviate loop currents.
What did you say ? Split the grounds - Oh My !

Graahm
Nico Coesel
18 years ago
Permalink
...
I once worked with such an expert. When he had to build a 150W audio
amplifier he found the bolt which held the toroid mains transformer in
place an excellent grounding point :-)
--
Reply to ***@nctdevpuntnl (punt=.)
Bedrijven en winkels vindt U op www.adresboekje.nl
Eeyore
18 years ago
Permalink
...
As long as he didn't use it as a safety earth connection......

Graham
Eeyore
18 years ago
Permalink
...
For starters that's unlikely to meet safety regs.

Graham

Eeyore
18 years ago
Permalink
Post by Joerg
Post by Eeyore
Post by John Larkin
Splitting planes is almost always a bad idea.
And almost universally done in audio or you'll hear low-level 'birdies' in the
signal.
Until the audio gear is used near a strong transmitter and then any
split will cause some real birdies. Or worst case a fzzzzt ... poof.
You haven't heard of metallic enclosures ?

Graham
Joerg
18 years ago
Permalink
Post by Eeyore
Post by Joerg
Post by Eeyore
Post by John Larkin
Splitting planes is almost always a bad idea.
And almost universally done in audio or you'll hear low-level 'birdies' in the
signal.
Until the audio gear is used near a strong transmitter and then any
split will cause some real birdies. Or worst case a fzzzzt ... poof.
You haven't heard of metallic enclosures ?
Sure. But keep in mind that the inter-winding capacitance of audio
baluns is not zero pF. That's how RF got in in my first debug case and
it fried some stuff. I also learned there that audio folks sometimes use
opamps with very limited diff input range (that's what had fried).
--
Regards, Joerg

http://www.analogconsultants.com
Eeyore
18 years ago
Permalink
Post by Joerg
Post by Eeyore
Post by Joerg
Post by Eeyore
Post by John Larkin
Splitting planes is almost always a bad idea.
And almost universally done in audio or you'll hear low-level 'birdies' in the
signal.
Until the audio gear is used near a strong transmitter and then any
split will cause some real birdies. Or worst case a fzzzzt ... poof.
You haven't heard of metallic enclosures ?
Sure. But keep in mind that the inter-winding capacitance of audio
baluns is not zero pF.
Very little audio equipment uses signal transformers these days.
Post by Joerg
That's how RF got in in my first debug case and
it fried some stuff. I also learned there that audio folks sometimes use
opamps with very limited diff input range (that's what had fried).
I'm curious as to what op-amp that might be.

Graham
Joerg
18 years ago
Permalink
...
Yes, and a diff amp imput can make it worse if there is no full low
impedance ground plane inside.
Post by Eeyore
Post by Joerg
That's how RF got in in my first debug case and
it fried some stuff. I also learned there that audio folks sometimes use
opamps with very limited diff input range (that's what had fried).
I'm curious as to what op-amp that might be.
It's been a long time ago but IIRC this one:
http://focus.ti.com/lit/ds/symlink/ne5534a.pdf
--
Regards, Joerg

http://www.analogconsultants.com
Eeyore
18 years ago
Permalink
...
I really take issue with the need for a ground *plane* per se. It is obviously good
practice to arrange proper EMC measures for all input and output signals for best
performance but I see no reason for a actual ground plane when a metallic enclosure is
used. In fact I can't think of any audio equipment I've ever seen that uses one, or
needs one.
Post by Joerg
Post by Eeyore
Post by Joerg
That's how RF got in in my first debug case and
it fried some stuff. I also learned there that audio folks sometimes use
opamps with very limited diff input range (that's what had fried).
I'm curious as to what op-amp that might be.
http://focus.ti.com/lit/ds/symlink/ne5534a.pdf
I wondered if you might mean that. Its differential input range is basically a couple
of diode drops.

How did you manage to *fry* one ?

Graham
Joerg
18 years ago
Permalink
...
Basically something like a simulated EMP. After that it was dead like a
door knob. Placed a balun there and the problem went away.
--
Regards, Joerg

http://www.analogconsultants.com
martin griffith
18 years ago
Permalink
On Mon, 23 Jul 2007 19:12:16 GMT, in sci.electronics.design Joerg
...
I suppose if you need a EMP proof audio system at home,

1) you should move
or
2) use EF86's
or
3) stop playing heavy metal



martin
Joerg
18 years ago
Permalink
...
It's not for here, it's for medical gear. If a doc has to do a marathon
shift, has become tired and then goofs a defib shot this cannot render
the machine useless. Because it might be urgently needed right then. A
5kV dump at tens of amps is nothing to sneeze at.
Post by martin griffith
2) use EF86's
I do use tubes but not the EF86.
Post by martin griffith
or
3) stop playing heavy metal
You mean, not even "Stairway to heaven"?
--
Regards, Joerg

http://www.analogconsultants.com
martin griffith
18 years ago
Permalink
On Mon, 23 Jul 2007 21:37:47 GMT, in sci.electronics.design Joerg
...
Why do doctors work such ridiculous hours? MBA management?
Post by Joerg
Post by martin griffith
2) use EF86's
I do use tubes but not the EF86.
Mullard I hope, for the British sound
Post by Joerg
Post by martin griffith
or
3) stop playing heavy metal
You mean, not even "Stairway to heaven"?
It has a dynamic range > 3dB so not heavy metal


martin
Eeyore
18 years ago
Permalink
Post by martin griffith
Post by Joerg
You mean, not even "Stairway to heaven"?
It has a dynamic range > 3dB so not heavy metal
More like >40dB dynamic range actually ! And deafinitely not heavy metal ! :-p

Graham
Joerg
18 years ago
Permalink
martin griffith wrote:


[...]
Post by martin griffith
Post by Joerg
Post by martin griffith
2) use EF86's
I do use tubes but not the EF86.
Mullard I hope, for the British sound
Nope. Sylvania 5654W. It's the ruggedized mil version of the 6AK5. Very
nice tiny pentode and good for low noise amps up to over 100MHz. Pretty
much EMP proof :-)
Post by martin griffith
Post by Joerg
Post by martin griffith
or
3) stop playing heavy metal
You mean, not even "Stairway to heaven"?
It has a dynamic range > 3dB so not heavy metal
Ok, then, rap music?
--
Regards, Joerg

http://www.analogconsultants.com
martin griffith
18 years ago
Permalink
On Mon, 23 Jul 2007 22:57:10 GMT, in sci.electronics.design Joerg
...
Nope, rap music is a commodity, disguised as music, to fund the IRA,
sorry, RIAA/MAFRIAA

Just about to organise my hifi, to be shipped from the yUK, spanish
willing (Spain shuts down for August, beach time ! Cafe's stay open,
thats about it)

Quad electrostatic speakers, 57 version (2 pairs)
Quad II/303 poweramps
A couple of Studer tape decks, I have to archive loads of tapes

And a preamp, with 5534s, from Elektor 25 years ago

It should sound pretty good



martin
Joerg
18 years ago
Permalink
...
Well, as long as the pubs (bodegas?) stay open that should be fine :-)
Post by martin griffith
Quad electrostatic speakers, 57 version (2 pairs)
Quad II/303 poweramps
A couple of Studer tape decks, I have to archive loads of tapes
And a preamp, with 5534s, from Elektor 25 years ago
It should sound pretty good
Don't wait to long with tapes in a hot climate. Out here tapes lasted
about 10 years max, then degradation set in. The first VCR tapes lost
sync after a mere 6-7 years :-(
--
Regards, Joerg

http://www.analogconsultants.com
Eeyore
18 years ago
Permalink
...
Fair enough. Not actually a normal EMC test. I don't see a commercial case for EMP hardening
hi-fi gear.

Do you recall what the field stength was ?

Graham
Joerg
18 years ago
Permalink
...
EMP is just one example. Gear that hasn't been built to withstand mil
std susceptibility tests tends to keel over the instant a strong
transmitter starts next door. For example a ham radio station. Or in a
hospital a diathermia machine. So instead the usual 10V/m I often ask
the EMC lab whether they can go above 30V/m.
Post by Eeyore
Do you recall what the field stength was ?
IIRC well north of 30V/m but with leads attached and that's probably
what killed it. The other stuff that tends to kill inputs is the
"zapper" for ESD tests. No problem with an opamp that is diode-clamped
to the rails but that's nearly impossible if you must accept line level
and the opamp croaks above 1V.
--
Regards, Joerg

http://www.analogconsultants.com
John Larkin
18 years ago
Permalink
On Mon, 23 Jul 2007 17:01:33 -0700, Joerg
...
One of my customers tests stuff at 50 KV/m. We sold him some fiber
optic links.

John
Joerg
18 years ago
Permalink
...
Probably not a good idea to stick around when they begin that test ;-)

I remember my first defib-safe design where the guy from TUEV insisted
on witnessing the test. I had to build my own tester which in itself
looked like a machine from hell. Huge coils and all that stuff. Catheter
in a trough, water filled in, electrode dropped in, electrode connected
to the machine from hell. Always one hand in the pocket. "Ok, charging
up now, BACK UP THEM COMPUTERS". An evil buzz filled the room and the
TUEV guy slowly moved towards the far corner of the room ....

Afterwards I had to do a hard reset of the PBX system but the TUEV
inspector was thoroughly impressed. I think I might have seen some beads
of sweat.
--
Regards, Joerg

http://www.analogconsultants.com
Eeyore
18 years ago
Permalink
...
LOL.

I build my own 8kV 'zapper' for pre-compliance testing using an old EHT transformer I had lying
around. Not quite in the same league as your kit above for impressiveness though. Also, once having
got the measure of the required design strategies we never bothered with it any more. It's not
difficult to design for EMC once you know the ropes.

Graham
Joerg
18 years ago
Permalink
...
Yep. It's just surprising and sad how few of the young grads know the
ropes or are willing to really learn them.

BTW, the gear I used to charge the 32uF cap to 5000 volts wasn't much
bigger than your flyback transformer either. It took a while. Much
easier: Transformers from bug lights. Those can be had for $10 when on
sale and you can still use its tube as a spare for another bug zapper.
But when we did that test it was wintery weather so no bug zappers at
the stores.
--
Regards, Joerg

http://www.analogconsultants.com
Eeyore
18 years ago
Permalink
...
For your market I can fully understand that.

Audio isn't a life-critical application normally. The typical response to large RF fields is
graceful degradation of s/n ratio.

Graham
Joerg
18 years ago
Permalink
...
In my cases as a young ham operator it was a lot worse than that. The
worst was an electronic organ from a company called Wersi. When I did CW
transmissions (morse code) the speaker was thumping so violently you'd
think it would come right out of its frame. During the on cycles the
organ sound itself completely vanished. I'd say the SNR dropped to zero
dB and was replaced by a "THWACK!". And this wasn't even next door.
--
Regards, Joerg

http://www.analogconsultants.com
John Devereux
18 years ago
Permalink
...
What I like to do is use double sided PCB material, but dedicate one
side as a ground plane. This side does not need to be etched, so it is
just like making a single-sided board from the processing point of
view. All ground signals go to a via (and all vias are to
ground). Then, for these "ground" vias just drill them and use tinned
copper wire (or resistor leads etc) to make the connection. This works
very well for surface mount boards, and is much easier than a full
double-sided layout (no alignment issues). Since there is no ground
routing on the "tracking" side (the component side) a single sided
layout is not too difficult. You could use the same artwork for
production boards too. (Actually I now only make my own boards for quick
prototypes of tricky bits of the circuit).
--
John Devereux
John Popelish
18 years ago
Permalink
...
I haven't tried this technique, but it sounds very useful.
For through hole boards, I would use one of those spur or
brad point bits (that cut a circle around a central point,
made for drilling wood) to clear the ground plane from a
little circle around the lead hole where components are
passed through the plane to connect to traces on the other
side. Do you think that is practical?

The biggest problem might be finding bits smaller than 1/8".
http://www.japanwoodworker.com/product.asp?s=JapanWoodworker&mimid=XSDUSPF0SMPN8KAE67JE0V0E3H0U7L26&pf_id=54%2E521%2E125&dept_id=12815

I'd probably have to grind my own low rake versions to cut
copper without diving through the board.
John Popelish
18 years ago
Permalink
...
I found a 2mm version, but look at the price!
http://www.mikestools.com/570-7020-Stern-Brad-Point-Drill-Bit-Metric-2mm-Hss.aspx
Glenn Gundlach
18 years ago
Permalink
...
components
are passed through the plane to connect to traces on the other side. Do
you think that is practical?
The biggest problem might be finding bits smaller than 1/8".
http://www.japanwoodworker.com/product.asp?
s=JapanWoodworker&mimid=XS...
I'd probably have to grind my own low rake versions to cut copper
without diving through the board.
I found a 2mm version, but look at the price!http://
www.mikestools.com/570-7020-Stern-Brad-Point-Drill-Bit-Metric-...

Not to pick on anybody, particularly John Popelish, but but when I
think of folks spending several hundred bucks on computers and cell
phones and $100+/month on cell service, internet and cable TV, I'm
always a bit taken aback when there is a comment on a 1 time expense
like a carbide drill bit being expensive.

As for the technique, I've seen low quantity broadcast gear with one
side ground planes manually drilled for clearance.

GG
linnix
18 years ago
Permalink
...
3/32" and 1/16" are quite common at your local hardware store.
Post by John Devereux
http://www.japanwoodworker.com/product.asp?
s=JapanWoodworker&mimid=XS...
I'd probably have to grind my own low rake versions to cut copper
without diving through the board.
I found a 2mm version, but look at the price!http://www.mikestools.com/570-7020-Stern-Brad-Point-Drill-Bit-Metric-...
That link is broken, can't see the price. I got a few 1.2mm and 1mm
steel bits if you want them.
Post by John Devereux
Not to pick on anybody, particularly John Popelish, but but when I
think of folks spending several hundred bucks on computers and cell
phones and $100+/month on cell service, internet and cable TV, I'm
always a bit taken aback when there is a comment on a 1 time expense
like a carbide drill bit being expensive.
Carbide bits are easy to break, especially for these sizes.
Post by John Devereux
As for the technique, I've seen low quantity broadcast gear with one
side ground planes manually drilled for clearance.
GG
John Popelish
18 years ago
Permalink
Post by Glenn Gundlach
Not to pick on anybody, particularly John Popelish, but but when I
think of folks spending several hundred bucks on computers and cell
phones and $100+/month on cell service, internet and cable TV, I'm
always a bit taken aback when there is a comment on a 1 time expense
like a carbide drill bit being expensive.
As for the technique, I've seen low quantity broadcast gear with one
side ground planes manually drilled for clearance.
$10 for a carbide bit might not be so bad, but that one:
http://www.mikestools.com/570-7020-Stern-Brad-Point-Drill-Bit-Metric-2mm-Hss.aspx
was high speed steel.
John Devereux
18 years ago
Permalink
...
Yes, in fact that is precisely what I used to do. You can buy special
bits from Farnell IIRC. (It has been so long since I made a through
hole board that I had genuinely forgotten all about this)!
The biggest problem might be finding bits smaller than 1/8".
http://www.japanwoodworker.com/product.asp?s=JapanWoodworker&mimid=XSDUSPF0SMPN8KAE67JE0V0E3H0U7L26&pf_id=54%2E521%2E125&dept_id=12815
I'd probably have to grind my own low rake versions to cut copper
without diving through the board.
--
John Devereux
John Devereux
18 years ago
Permalink
John Devereux <***@THISdevereux.me.uk> writes:

(replying to own post)
...
Thinking about this, it might be better to have the components on the
groundplane side, and the tracks on the "solder" side. Then you use a
normal ~3mm bit to countersink the area around non-ground pins. The
ground pins can then be soldered on the component/ground plane
side. Can be tricky when the pins are under the device.

It's much easier with surface mount.
--
John Devereux
Jon Slaughter
18 years ago
Permalink
Thanks for all the info guys.

I'm curious as to just how bad not having a ground plane is. Surely the
circuit will still function properly? How much noise will be introduced? I'm
willing to sacrifice quality for ease to make at this point. (since if the
only issue with the layout is going to be noise once I get everything
working through a prototype I can add a ground plane to the layout and then
send off and get it fabricated)

I think here there won't be any issues masking noise from other sources but
that might be a problem. (I think at this point I'm just trying to get the
system to work as a whole)

I assume that the issue is with the analog domain or is there a reverse
issue with digital? i.e., can the digital be affected by the analog domain
also when not using these special precautions such as a ground plane? (so
that the circuit could actually not function properly rather than just
having noisy conversions)

I think that I will end up doing the 2 sided boards though. I'm sure its
much easier to route/layout ;) Since there are vias that I can buy that look
like a snap to use and it will just involve a drilling a few holes.

(mainly just curious about the consequences of not using a ground plane. I
know its not as good but I don't think any mentions just how bad. (remember
its for prototyping so I can sacrifice some things))


Thanks,
Jon
Jon Slaughter
18 years ago
Permalink
...
Oh, not sure if I mentioned this but its for audio. I'm not doing really
high speed stuff and my sample rate will actually be 96khz or max 192khz but
definitely nothing past 100Mhz.

I suppose though that it is actually in the mhz range since these are
oversampling so I guess I do have to look at it as high speed?
MooseFET
18 years ago
Permalink
...
The important number is the gain * bandwidth value. Even with only
audio signals, having a gain of a million and no ground plane is
asking for it.
Vladimir Vassilevsky
18 years ago
Permalink
Post by Jon Slaughter
Thanks for all the info guys.
I'm curious as to just how bad not having a ground plane is.
It depends.
Post by Jon Slaughter
Surely the
circuit will still function properly?
BTW if you open a consumer electronics device, you will see single layer
PCBs. Works fine for them.
Post by Jon Slaughter
How much noise will be introduced?
It depends on the quality of the layout and on what exactly are you
trying to accomplish.

I'm
Post by Jon Slaughter
willing to sacrifice quality for ease to make at this point. (since if the
only issue with the layout is going to be noise once I get everything
working through a prototype I can add a ground plane to the layout and then
send off and get it fabricated)
I think here there won't be any issues masking noise from other sources but
that might be a problem. (I think at this point I'm just trying to get the
system to work as a whole)
Good plan.
Post by Jon Slaughter
I assume that the issue is with the analog domain or is there a reverse
issue with digital? i.e., can the digital be affected by the analog domain
also when not using these special precautions such as a ground plane? (so
that the circuit could actually not function properly rather than just
having noisy conversions)
Don't worry upfront. As the matter of fact, quite complicated systems
can be built using just two layer PCBs without a solid ground plane. Of
course, the layout of the power and ground requires careful attention.
Post by Jon Slaughter
I think that I will end up doing the 2 sided boards though. I'm sure its
much easier to route/layout ;) Since there are vias that I can buy that look
like a snap to use and it will just involve a drilling a few holes.
(mainly just curious about the consequences of not using a ground plane. I
know its not as good but I don't think any mentions just how bad. (remember
its for prototyping so I can sacrifice some things))
Vladimir Vassilevsky
DSP and Mixed Signal Design Consultant
http://www.abvolt.com
Michael A. Terrell
18 years ago
Permalink
Post by Vladimir Vassilevsky
BTW if you open a consumer electronics device, you will see single layer
PCBs. Works fine for them.
Sure, for a cheap radio, or other low density designs where they can
get away with a handful of jumpers. When was the last time you saw a
single layer computer motherboard? Most of the computer monitors I've
had apart have double sided PC boards. The highest layer count I've
worked on in industrial designs were 16 layer. They were in a DSP based
diversity telemetry receiver system.
--
Service to my country? Been there, Done that, and I've got my DD214 to
prove it.
Member of DAV #85.

Michael A. Terrell
Central Florida
Eeyore
18 years ago
Permalink
...
If you're making stuff in the west, the cost saving of single vs double sided is
probably marginal and not worth the trouble.

If you're manufacturing in Asia it's another story. I didn't even know you could
get 1/2 oz copper foil until we discovered our Bombay sub-contractor was using
it. I could toast my colleagues' asses for not specifying the copper thickness
on the materials specs (what materials spec?).

Also you'll have a choice of paper-phenolic/SRBP/paxolin/FR2 vs composite/CEM
(paper-glass-epoxy) or real epoxy-glass/FR4 as the base material.

Graham
b***@yahoo.com
18 years ago
Permalink
...
No, this question comes up often, follow the manufactures advice,
there is "noisy" digital and "quiet" digital, this is quiet digital
(noisy digital would be a DSP or high speed parallel buffers), see
link for an explanation of the ADC/DAC digital vs analog ground pin
issue

http://www.analog.com/analog_root/static/raq/raq_groundingADCs.html
Jon Slaughter
18 years ago
Permalink
...
Following some links I see

http://www.hottconsultants.com/pdf_files/june2001pcd_mixedsignal.pdf

and figure 8 shows exactly what I was talking about with straddling!
Loading...