Using Spice to verify a circuit works

[John Larkin]

On Mon, 24 Jun 2019 08:20:43 GMT, Steve Wilson <no@spam.com> wrote:

John Larkin <jjlarkin@highlandtechnology.com> wrote:

On Sun, 23 Jun 2019 21:57:05 -0700 (PDT), whit3rd <whit3rd@gmail.com
wrote:

On Friday, June 21, 2019 at 9:29:18 PM UTC-7, Steve Wilson wrote:
whit3rd <whit3rd@gmail.com> wrote:

On Friday, June 21, 2019 at 7:52:10 AM UTC-7, John Larkin wrote:

I don't model nonlinear chaotic systems in Spice, I model
electronic circuits that I know I can simulate usefully.

So, a self-starting oscillator, you don't model that?
Lame, like your other excuses and lies.

What's the big deal with self-starting oscillators? An HC14 is self-
starting.

That's a good example that Spice can't handle (well, not correctly); the
initial condition for a Schmitt trigger (or '555, or hysteretic gate)
with the right feedback is indeterminate (depends on tiny amounts of
noise or capacitor charge).

LT Spice will sim a Schmitt gate oscillator fine, if you set up the
gate params right.

What do you need to do for gate params?

Something like this:

https://www.dropbox.com/s/z5tetvvfoncsem1/Scmitt_Gate_Osc.jpg?dl=0

It really needs that nonzero prop delay param. LT Spice was apparently
not intended to be a digital simulator!

I haven't had much luck with PLLs in LT Spice either. We usually write
our own code for them. I've got one now that I want to simulate, but
I'd need better than double floats to do the math. PowerBasic supports
the 80 bit format, which just might do.


--

John Larkin Highland Technology, Inc

lunatic fringe electronics

 

[Steve Wilson]

John Larkin <jjlarkin@highlandtechnology.com> wrote:

Something like this:

https://www.dropbox.com/s/z5tetvvfoncsem1/Scmitt_Gate_Osc.jpg?dl=0

It really needs that nonzero prop delay param. LT Spice was apparently
not intended to be a digital simulator!

As long as you are not too fussy, Helmut's 74hc.lib includes that
automatically. Then you can use the whole range of 74HC devices without
worrying about gate parameters. This is very useful for a quick check on a
logic circuit.

Another alternaative if you just need a 74HC14 is Helmut's B-Source shown
in items 8 and 9 in my oscillator page at

https://drive.google.com/open?id=1ZsbpkV0aaKS5LURIb1dfu_ndshsSaYtf

I haven't had much luck with PLLs in LT Spice either. We usually write
our own code for them. I've got one now that I want to simulate, but
I'd need better than double floats to do the math. PowerBasic supports
the 80 bit format, which just might do.

I have come across simulations using Laplace Transforms. They are really,
really fast. However I never use them as I wrote my own software that gives
the values for the loop components.

 

[John Larkin]

On Mon, 24 Jun 2019 15:33:01 GMT, Steve Wilson <no@spam.com> wrote:

John Larkin <jjlarkin@highlandtechnology.com> wrote:

Something like this:

https://www.dropbox.com/s/z5tetvvfoncsem1/Scmitt_Gate_Osc.jpg?dl=0

It really needs that nonzero prop delay param. LT Spice was apparently
not intended to be a digital simulator!

As long as you are not too fussy, Helmut's 74hc.lib includes that
automatically. Then you can use the whole range of 74HC devices without
worrying about gate parameters. This is very useful for a quick check on a
logic circuit.

Another alternaative if you just need a 74HC14 is Helmut's B-Source shown
in items 8 and 9 in my oscillator page at

https://drive.google.com/open?id=1ZsbpkV0aaKS5LURIb1dfu_ndshsSaYtf

I haven't had much luck with PLLs in LT Spice either. We usually write
our own code for them. I've got one now that I want to simulate, but
I'd need better than double floats to do the math. PowerBasic supports
the 80 bit format, which just might do.

I have come across simulations using Laplace Transforms. They are really,
really fast. However I never use them as I wrote my own software that gives
the values for the loop components.

I want to resolve time-domain jitter to picoseconds, which gets
numerically ugly fast. We might just do some proto boards and skip
simulation.


--

John Larkin Highland Technology, Inc

lunatic fringe electronics

 

[whit3rd]

On Sunday, June 23, 2019 at 10:17:08 PM UTC-7, Steve Wilson wrote:

whit3rd <whit3rd@gmail.com> wrote:

That's a good example that Spice can't handle (well, not correctly); the
initial condition for a Schmitt trigger (or '555, or hysteretic gate)

There is nothing wrong with the HC14 in SPICE. It runs fine.

Whatever the initial conditions, the feedback will drive the input voltage
to the opposite side of the hysteresis voltage. It cannot fail to start.

Famous last words! The problem of metastability in flipflops, and of starting
conditions in hysteretic chips, is that there's an insoluble bit of math in
the center. There's TWO solutions at power-on, output HIGH and LOW
initial states are both allowed, unless you break the symmetry somehow.

Temperature, roundoff error, or a fudge constant in a model, will break
that symmetry, but those are all going to break it differently: the theory
just doesn't have the randomness of thermal fluctuations and thus
diverges from reality. The symptom, is that a SPICE model may always start
in some phase after power-on, but the chip might not. Absolute
phase usually doesn't matter (because we've learned not to depend on it),
but that isn't the same as being able to predict it.

A power-on sequence usually geta a few dozen clock cycles completed THEN
resets everything and starts the process. There's other good reasons for that,
but oscillator startup predictability plays a part.

 

[Steve Wilson]

whit3rd <whit3rd@gmail.com> wrote:

On Sunday, June 23, 2019 at 10:17:08 PM UTC-7, Steve Wilson wrote:
whit3rd <whit3rd@gmail.com> wrote:

That's a good example that Spice can't handle (well, not correctly);
the initial condition for a Schmitt trigger (or '555, or hysteretic
gate)

There is nothing wrong with the HC14 in SPICE. It runs fine.

Whatever the initial conditions, the feedback will drive the input
voltage to the opposite side of the hysteresis voltage. It cannot fail
to start.

Famous last words! The problem of metastability in flipflops, and of
starting conditions in hysteretic chips, is that there's an insoluble
bit of math in the center. There's TWO solutions at power-on, output
HIGH and LOW initial states are both allowed, unless you break the
symmetry somehow.

That does you no good. The output had to settle in one state or the other.
It cannot remain suspended between the two.

In the event that both states exist, feedback will quickly resolve the
issue and the circuit will settle to one state or the other.

Temperature, roundoff error, or a fudge constant in a model, will break
that symmetry, but those are all going to break it differently: the
theory just doesn't have the randomness of thermal fluctuations and thus
diverges from reality. The symptom, is that a SPICE model may always
start in some phase after power-on, but the chip might not. Absolute
phase usually doesn't matter (because we've learned not to depend on
it), but that isn't the same as being able to predict it.

Wordy nonsense. Metastability is resolved very fast. Many modern ic's in
the EcliPse class sho no measurable metastability at all.

A power-on sequence usually geta a few dozen clock cycles completed THEN
resets everything and starts the process. There's other good reasons
for that, but oscillator startup predictability plays a part.

The hysteresis is not a flip-flop circuit. There is no metastability in a
HC14. Once you hit on hysteresis voltage threshold, the output will switch
to the opposite side. There is no way to avoid it.