Flip-Flop in LTSpice, set/reset assertion?

Post by Joerg
Today I found that a simulation wasn't working right because the set and
reset on the flip-flop (dflop) in LTSpice has the set and reset inputs
active high. Yet every modern flip-flop is active low. So I had to
invert everything on those inputs just for the simulation. The "Special
Functions" instructions in LTSPice are silent about this issue.
Does this date back to the days of Methusaleh or what could be the reason?

it's a behavioural model that does just what it says on the box,
active low inputs is more of a practical implementation detail

Piotr Wyderski

2019-09-22 00:03:44 UTC

Post by Lasse Langwadt Christensen
active low inputs is more of a practical implementation detail

Is there any practical reason behind the negative logic? Or just a
convention, as good as the opposite?

Best regards, Piotr

Lasse Langwadt Christensen

2019-09-22 00:11:52 UTC

Post by Lasse Langwadt Christensen
active low inputs is more of a practical implementation detail

Is there any practical reason behind the negative logic? Or just a
convention, as good as the opposite?

look at Gerhards post

Phil Hobbs

2019-09-22 00:51:38 UTC

Post by Lasse Langwadt Christensen
active low inputs is more of a practical implementation detail

Is there any practical reason behind the negative logic? Or just a
convention, as good as the opposite?
Best regards, Piotr

7400 TTL convention, perhaps to reduce power dissipation due to an
infrequently-used function. 4000-series CMOS usually used active high IIRC.

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

j***@highlandsniptechnology.com

2019-09-22 01:00:15 UTC

On Sun, 22 Sep 2019 02:03:44 +0200, Piotr Wyderski

Post by Lasse Langwadt Christensen
active low inputs is more of a practical implementation detail

Is there any practical reason behind the negative logic? Or just a
convention, as good as the opposite?
Best regards, Piotr

The basic gate in early DTL and TTL logic was the NAND. It was faster
and more economical to make a master-slave flipflop with additional
NAND inputs for preset and clear; active low over-rides into the main
bistable. Hence the 7474 type flipflop.

The basic RTL gate was a NOR, so RTL flops usually had active-high
preset and clear.

TTL killed DTL killed RTL, so TTL won. CMOS inherited the TTL
conventions.

Joerg

2019-09-22 14:26:15 UTC

Post by Lasse Langwadt Christensen
active low inputs is more of a practical implementation detail

Is there any practical reason behind the negative logic? Or just a
convention, as good as the opposite?

Like Gerhard wrote, better noise immunity. Old TTL pulled to within a
few hundred mV of the negative rail but not to the positive rail:

Loading Image...

If you got close to 4V in a 5V system you were lucky and the drive power
in positive direction was wimpy, meaning slow and potentially noise-prone.

--
Regards, Joerg

http://www.analogconsultants.com/

Helmut Sennewald

2019-09-21 15:15:22 UTC

Hello Joerg,

The basic digital primitives, (AND, OR, XOR, DFLOP, ...) are independent
of any device which you can buy.

Even the handling of not used inputs is different. Simply connect
nothing on the unused inputs. The LTspice circuit-compiler will then
"remove" these pins to make the circuit as simple as possible for the
simulation.

Don't forget to set a rise time or a delay in DFLOP.

Value: td=5n

If you want a Q=1 at the start, then somply add a IC=1.

Value: td=5n IC=1

Best regards,
Helmut

Joerg

2019-09-21 18:22:46 UTC

Post by Helmut Sennewald

Post by Joerg
Today I found that a simulation wasn't working right because the set
and reset on the flip-flop (dflop) in LTSpice has the set and reset
inputs active high. Yet every modern flip-flop is active low. So I had
to invert everything on those inputs just for the simulation. The
"Special Functions" instructions in LTSPice are silent about this issue.
Does this date back to the days of Methusaleh or what could be the reason?
Connecting unused AND gate inputs to terminal 8 as per instructions
also did not work but that was easy to figure out.

Hello Joerg,
The basic digital primitives, (AND, OR, XOR, DFLOP, ...) are independent
of any device which you can buy.

So they are always active high for control inputs?

Post by Helmut Sennewald
Even the handling of not used inputs is different. Simply connect
nothing on the unused inputs. The LTspice circuit-compiler will then
"remove" these pins to make the circuit as simple as possible for the
simulation.

In the instructions under "Special Functions" it says, quote "Unused
inputs and outputs are to be connected to terminal 8". I guess that's
not correct then. At least it fails in my simulation.

Post by Helmut Sennewald
Don't forget to set a rise time or a delay in DFLOP.
Value: td=5n

Also, one must set Vhigh=3.3V or to whichever logic voltage it is to be.
Unfortunately that can't be fed in from any real rail which makes
unorthodox stuff such as controlled back-feeding difficult.

Post by Helmut Sennewald
If you want a Q=1 at the start, then somply add a IC=1.
Value: td=5n IC=1

In this case I needed to used the CLR pin in a more controlled fashion,
where reset happens a few msec after an event.

--
Regards, Joerg

http://www.analogconsultants.com/

j***@highlandsniptechnology.com

2019-09-21 15:46:00 UTC

Not the ECL parts, like 10EP51. Reset and clear are active high.

Winfield Hill

2019-09-21 18:59:15 UTC

Not the ECL parts, like 10EP51. Reset and clear are active high.

The same is true for the Ancient CD4000 series high-voltage CMOS.

--
Thanks,
- Win

Joerg

2019-09-21 19:43:42 UTC

Not the ECL parts, like 10EP51. Reset and clear are active high.

That's essentially RF stuff and can't easily be simulated with the
LTSpice behavioral models. Most of it needs to be breadboarded anyhow.

Post by Winfield Hill
The same is true for the Ancient CD4000 series high-voltage CMOS.

True but those are really long in the tooth. I still use them in designs
but figured that a modern simulator would use more modern conventions.

--
Regards, Joerg

http://www.analogconsultants.com/

j***@highlandsniptechnology.com

2019-09-21 20:06:45 UTC

Not the ECL parts, like 10EP51. Reset and clear are active high.

That's essentially RF stuff and can't easily be simulated with the
LTSpice behavioral models. Most of it needs to be breadboarded anyhow.

Post by Winfield Hill
The same is true for the Ancient CD4000 series high-voltage CMOS.

True but those are really long in the tooth. I still use them in designs
but figured that a modern simulator would use more modern conventions.

LT Spice is not very digital. I suspect that the library flops and
gates are not behavioral models. More than suspect.

I've had shift registers fail because I didn't set a non-zero prop
delay.

The EL and EP logic behaves very well, just about what the data sheets
say. If you need to divide down a 600 MHz oscillator, they are the
choice.

Just trekked to Safeway. It's brutally hot here, must be pushing 80F.
How is it out there?

Rick C

2019-09-21 20:22:40 UTC

Not the ECL parts, like 10EP51. Reset and clear are active high.

That's essentially RF stuff and can't easily be simulated with the
LTSpice behavioral models. Most of it needs to be breadboarded anyhow.

Post by Winfield Hill
The same is true for the Ancient CD4000 series high-voltage CMOS.

True but those are really long in the tooth. I still use them in designs
but figured that a modern simulator would use more modern conventions.

LT Spice is not very digital. I suspect that the library flops and
gates are not behavioral models. More than suspect.
I've had shift registers fail because I didn't set a non-zero prop
delay.

Same is true in any simulation. The individual devices are evaluated in an arbitrary order. Clock input causes an output change in one FF and the next FF is evaluated after the first, the change will ripple through when it would not in a real part with a finite delay. Kind of a Duh! That's why VHDL has delta delays which still won't protect you if you have any buffers, or other things in the clock path to some FFs (which add a delta delay).

Post by j***@highlandsniptechnology.com
The EL and EP logic behaves very well, just about what the data sheets
say. If you need to divide down a 600 MHz oscillator, they are the
choice.
Just trekked to Safeway. It's brutally hot here, must be pushing 80F.
How is it out there?

Here it is a very nice 90°F. Great day to be outside.

Why is Joerge ranting about LTSpice? Why isn't he at least thanking Helmut for the advice? Helmut is one of the truly knowledgeable people around on LTSpice. He has been a tremendous aid to the community. He also never insults anyone that I've ever seen.

--
Rick C.

- Get 2,000 miles of free Supercharging
- Tesla referral code - https://ts.la/richard11209

Joerg

2019-09-22 14:00:11 UTC

Not the ECL parts, like 10EP51. Reset and clear are active high.

That's essentially RF stuff and can't easily be simulated with the
LTSpice behavioral models. Most of it needs to be breadboarded anyhow.

Post by Winfield Hill
The same is true for the Ancient CD4000 series high-voltage CMOS.

True but those are really long in the tooth. I still use them in designs
but figured that a modern simulator would use more modern conventions.

LT Spice is not very digital. I suspect that the library flops and
gates are not behavioral models. More than suspect.

They work, at least for the most part. The unused input treatment seems
iffy but I just tie them to a logic level.

Post by j***@highlandsniptechnology.com
I've had shift registers fail because I didn't set a non-zero prop
delay.
The EL and EP logic behaves very well, just about what the data sheets
say. If you need to divide down a 600 MHz oscillator, they are the
choice.

Haven't used ECL-style in more than a decade, mostly because they are
too power-hungry and power consumption matters more and mre these days.
Even on airliners they cound every drop of kerosene and thus watts in
the electronics.

Post by j***@highlandsniptechnology.com
Just trekked to Safeway. It's brutally hot here, must be pushing 80F.
How is it out there?

I did yard work yesterday, maybe it was in the upper 80's. Having lived
here for so long plus some Arizona exposure, I don't notice heat much.
When I ride a bicycle full bore on a 105F day people think that's crazy.
You just need enough water and an empty yoghurt beker to dump water over
the T-shirt.

--
Regards, Joerg

http://www.analogconsultants.com/

Gerhard Hoffmann

2019-09-21 20:53:06 UTC

Not the ECL parts, like 10EP51. Reset and clear are active high.

That's essentially RF stuff and can't easily be simulated with the
LTSpice behavioral models. Most of it needs to be breadboarded anyhow.

That has nothing to do with RF can be perfectly simulated. It's just
that the delay numbers are small. In a previous life, when 10kH an 100K
were the newest thing, I have built a nice 100K library for the Simucad
Silos simulator. That included setup/hold violations etc. I had only a
200 MHz scope, so I had to make sure that at least the logic was OK
to start with.

The same is true for the Ancient CD4000 series high-voltage CMOS.

True but those are really long in the tooth. I still use them in designs
but figured that a modern simulator would use more modern conventions.

You've got that the wrong way. Low active inputs are a very retro 74xx
thing. TTL designers speculated that it took more signal energy to
produce a LOW input than a HIGH (correct for TTL) and that would give
better noise immunity since most of the time these inputs do nothing.

That has survived for some time only for some things like 74HCT that
allowed burning less power without requiring heavy re-thinking.

And no, modern digital design has nothing to do with deploying 74xxx.
You formulate your system in VHDL, Verilog or Matlab and that's it.
Nobody cares about flipflops, let alone their reset pin polarity.

As it happens, I needed more space in may parts store last week and I
decided to move most things with pins, 74xx, 10K, 100K etc into
plastic containers in the basement. I did not use them in years.
Even the rests of glue logic in my designs go into a Coolrunner2.

cheers, Gerhard

Joerg

2019-09-22 14:21:02 UTC

Not the ECL parts, like 10EP51. Reset and clear are active high.

That's essentially RF stuff and can't easily be simulated with the
LTSpice behavioral models. Most of it needs to be breadboarded anyhow.

"Perfectly" depends on how far one has to go with parameters such as
input capacitances. I don't think those are part of an LTSpice
simulation. Would need a full non-behavioral model.

Post by Winfield Hill
The same is true for the Ancient CD4000 series high-voltage CMOS.

True but those are really long in the tooth. I still use them in
designs but figured that a modern simulator would use more modern
conventions.

Long in the tooth - retro, same thing :-)

Old TTL didn't pull very high. 3.5V to 4V in some cases so noise
immunity was indeed worse, until CMOS came. That made things cumbersome
because you could not use OC structures to operate it from external.

Post by Gerhard Hoffmann
That has survived for some time only for some things like 74HCT that
allowed burning less power without requiring heavy re-thinking.
And no, modern digital design has nothing to do with deploying 74xxx.
You formulate your system in VHDL, Verilog or Matlab and that's it.
Nobody cares about flipflops, let alone their reset pin polarity.

Nobody? Really nobody? Way north in France, in the province of the
Gauloises ...

For example, right now I have to design a circuit for a device that
replaces a uC function because the uC can't be trusted to do the job
reliably enough. I think it could be made reliable but a client's wish
is a client's wish. It'll need several 74LVC chips.

Post by Gerhard Hoffmann
As it happens, I needed more space in may parts store last week and I
decided to move most things with pins, 74xx, 10K, 100K etc into
plastic containers in the basement. I did not use them in years.
Even the rests of glue logic in my designs go into a Coolrunner2.

My designs are often cost sensitive, mass production, where programmable
logic is too expensive. A 20c Coolrunner could work, 50c would not.

Storing parts farther away has health benefits. In my office the book
shelf is 5ft from my desk so I have to get up from the chair every time.
Transistors are in the room next door from my lab bench so a short walk
is required. It's healthy. I saw a neighbor yesterday who stopped
retirement for another engineering job. Despite diet cutbacks he has
developed a profound belly, not very healthy. It'll easily take more
than a year to get rid of that.

--
Regards, Joerg

http://www.analogconsultants.com/

Gerhard Hoffmann

2019-09-23 01:36:53 UTC

Post by Joerg
Old TTL didn't pull very high. 3.5V to 4V in some cases so noise
immunity was indeed worse, until CMOS came. That made things cumbersome
because you could not use OC structures to operate it from external.

TTL levels have nothing to do with 5V. That is just the supply voltage.
The switching threshold is somewhere near 1V8, anything below 0V6 is
definitely low and anything above 2V4 is definitely high.

CMOS has its threshold at 1/2 VCC, but the historic importance of
TTL required the extraneous 74HCT family with the lower TTL input
threshold, implemented by playing games with the width/length ratio
of the FETs.

Post by Gerhard Hoffmann
And no, modern digital design has nothing to do with deploying 74xxx.
You formulate your system in VHDL, Verilog or Matlab and that's it.
Nobody cares about flipflops, let alone their reset pin polarity.

Nobody? Really nobody? Way north in France, in the province of the
Gauloises ...

Deploying a few gates is not digital design.

Post by Joerg
For example, right now I have to design a circuit for a device that
replaces a uC function because the uC can't be trusted to do the job
reliably enough. I think it could be made reliable but a client's wish
is a client's wish. It'll need several 74LVC chips.

Consultant's creed:
It's our policy to give the customer what he wants.
That is very strong medicine, and usually only required once.

Gerhard

j***@highlandsniptechnology.com

2019-09-23 02:50:01 UTC

TTL levels have nothing to do with 5V. That is just the supply voltage.
The switching threshold is somewhere near 1V8, anything below 0V6 is
definitely low and anything above 2V4 is definitely high.

Classic TTL was about 1.2 volts threshold.

Post by Gerhard Hoffmann
CMOS has its threshold at 1/2 VCC, but the historic importance of
TTL required the extraneous 74HCT family with the lower TTL input
threshold, implemented by playing games with the width/length ratio
of the FETs.

74HCT is great for receiving inputs from FPGAs at 3.3 or 2.5 volts. It
will have much less shoot-through current than HC, too.

Nobody? Really nobody? Way north in France, in the province of the
Gauloises ...

Deploying a few gates is not digital design.

It's our policy to give the customer what he wants.
That is very strong medicine, and usually only required once.
Gerhard

I often use Tiny dflops to resync the outputs of an FPGA to a clock.
That really cleans up the jitter that FPGAs tend to make.

Joerg

2019-09-23 14:07:01 UTC

Post by Joerg
Old TTL didn't pull very high. 3.5V to 4V in some cases so noise
immunity was indeed worse, until CMOS came. That made things
cumbersome because you could not use OC structures to operate it from
external.

TTL levels have nothing to do with 5V. That is just the supply voltage.
The switching threshold is somewhere near 1V8, anything below 0V6 is
definitely low and anything above 2V4 is definitely high.

And sometimes the drive signal didn't quite get there. Or not all the
time. The threshold in TTL is lower though, but occasionally it wasn't
low enough. Unless you used 244 bus drivers this stuff just didn't have
any oomph.

That is one reason I was never much of a fan of HCT. In my youth I built
a lot of circuits with CD4000 logic because it didn't have such
problems. It had other problems but the main upside was a vastly lower
power consumption.

Nobody? Really nobody? Way north in France, in the province of the
Gauloises ...

Deploying a few gates is not digital design.

That part fulfills logic funtions. If this and that happens at the same
time tug on an alert rail, otherwise not. Preferably while consuming
less than 1uA. What's not digital about that?

It's our policy to give the customer what he wants.
That is very strong medicine, and usually only required once.

:-)

Once is the goal, of course. After that I (and the client) expect this
to run for the next few decades.

--
Regards, Joerg

http://www.analogconsultants.com/

Rick C

2019-09-23 14:17:04 UTC

Post by Joerg
Old TTL didn't pull very high. 3.5V to 4V in some cases so noise
immunity was indeed worse, until CMOS came. That made things
cumbersome because you could not use OC structures to operate it from
external.

TTL levels have nothing to do with 5V. That is just the supply voltage.
The switching threshold is somewhere near 1V8, anything below 0V6 is
definitely low and anything above 2V4 is definitely high.

Nobody? Really nobody? Way north in France, in the province of the
Gauloises ...

Deploying a few gates is not digital design.

That part fulfills logic funtions. If this and that happens at the same
time tug on an alert rail, otherwise not. Preferably while consuming
less than 1uA. What's not digital about that?