Determining Phase and Gain Margins in LTSpice

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Also, it might be a good idea to have an extra zobel added in the speaker box near the speaker itself, which would be at the far end of the speaker cable. This would make it 2 zobels, one at each speaker cable end. That should also play in favor of fighting against ingress of whatever interference might be around.

As discussed by Bop Cordell;s "Designing Audio Power Amplifiers", 2nd edition "Interfacing the real world". Your speaker cable may have a nominal impedance of 8R at audio frequencies, but at radio frequencies is going to vary all over the place, reflecting energy back to the amplifier. Bob suggests 100R in series with 0.01uF termination at the speaker.

Getting back to stability:
 

Attachments

  • Loop Gain and Q.png
    Loop Gain and Q.png
    247.3 KB · Views: 137
As discussed by Bop Cordell;s "Designing Audio Power Amplifiers", 2nd edition "Interfacing the real world". Your speaker cable may have a nominal impedance of 8R at audio frequencies, but at radio frequencies is going to vary all over the place, reflecting energy back to the amplifier. Bob suggests 100R in series with 0.01uF termination at the speaker.

Getting back to stability:

Unfortunately I don't have his new book. Hopefully some day..

So that's basically what I mentioned, an other zobel at the far end of the cable, at the speaker end. Except maybe for the values a bit different, as suggested by Bob. Not too different from the common values used at the amp's output.

In my case, at least for the low end channel, that speaker cable might not even go beyond 1m or so... Not too much of an antenna, and limited opportunity for much capacitance to build up..
 
Hi,
I located the following on the Tian probe :
Plotting Loop Gain Using the Tian Method - Spring 2011
In the text, it states you must zero out the AC Magnitude parameters for any voltage or current sources.
In examining the attached LTSpice .asc files, the voltage sources are not set to zero.

Therefore, is there another method of using the Tian probe and is there a suitable description of this alternative method ?
Thanks and regards,

Shadders.
 
I can see how this can definitely be confusing, but after reading this stuff many times, at least some of it starts making some sense.

Zeroing out the AC magnitude doesn't pertain to any of the supply sources, just whatever gives an input signal, so the input signal source's voltage should be set to zero. Not a problem, that's how we do this anyway. And for an AC analysis, the AC parameter just can't be set to zero, as this is what gets its value stepped through during the simulation. The tian probe gets activated by those stepped values, to 1 and -1, which in turns gives 2 curves superimposed on the plot.

So carefully consider what is meant by "zero out the AC magnitude parameters".
 
I can see how this can definitely be confusing, but after reading this stuff many times, at least some of it starts making some sense.

Zeroing out the AC magnitude doesn't pertain to any of the supply sources, just whatever gives an input signal, so the input signal source's voltage should be set to zero. Not a problem, that's how we do this anyway. And for an AC analysis, the AC parameter just can't be set to zero, as this is what gets its value stepped through during the simulation. The tian probe gets activated by those stepped values, to 1 and -1, which in turns gives 2 curves superimposed on the plot.

So carefully consider what is meant by "zero out the AC magnitude parameters".
Hi,
Thanks - i realised it meant the signal sources, not voltage rails etc.

This is why i asked - one of the asc. files has the input signal voltage source set to 1.08volts, as per :
GainWire-TIS-baxandall-TPC-OIC-LMOSnocomp-200W-2-DCservo-overVAS-15V-betterCCS.asc
Located at :
https://www.diyaudio.com/forums/solid-state/261973-middlebrook-gft-probe-11.html#post4098014
Regards,
Shadders.
 
It is confusing for sure.

But one thing would make no sense at all if all sources had to be zeroed out, if there is no voltage for supplies, then no DC op point could be determined, so then I don't think an AC analysis would work or give anything coherent.

But it also depends on how the Tian probe method is implemented I suppose. There is no way it would work with the AC source zeroed out if used like I use it with that sub-circuit. As the stepped analysis wouldn't give back anything useful.
 
It is confusing for sure.

But one thing would make no sense at all if all sources had to be zeroed out, if there is no voltage for supplies, then no DC op point could be determined, so then I don't think an AC analysis would work or give anything coherent.

But it also depends on how the Tian probe method is implemented I suppose. There is no way it would work with the AC source zeroed out if used like I use it with that sub-circuit. As the stepped analysis wouldn't give back anything useful.
Hi,
I did state i realise the reference meant signal source, and NOT voltage supplies/rails.

From reading the webpage from Spectrum Soft, the signal sources are the probe itself, a voltage source and current source.

Regards,
Shadders.
 
This is why i asked - one of the asc. files has the input signal voltage source set to 1.08volts, as per :
GainWire-TIS-baxandall-TPC-OIC-LMOSnocomp-200W-2-DCservo-overVAS-15V-betterCCS.asc
Located at :
https://www.diyaudio.com/forums/solid-state/261973-middlebrook-gft-probe-11.html#post4098014

This is just a leftover from other tests, Damir would not have actually run the Tian probe with this turned on
If left on the result is obviously incorrect and easy to spot but it is an annoyance to have to redo the run.
So I usually replace the input amplitude with an expression that turns it off automatically when the probe variable is set.
More tomorrow- I am still a bit busy.

Best wishes
David
 
...you didn't notice are associated, such as the supplies and the load, with the zobel, which is also not visualized but put into text instead...

I did miss the supply definitions, probably because I can think of no reason why anyone would do it this way.
Why use a text definition when you can just place a source on the circuit?
Why put the definitions for the supplies far from the supplies?
Why not put the specifications of an object next to the object?

make a zip.. for that simulation...
But this isn't as easy as it seems...

Yes, exactly my point, it's a weakness of Spice and LTSpice hasn't improved this.
But your screenshot layout doesn't help.

...likely to screw it up. Then it would be a pain to track down my mistakes.

That tian probe in a sub-circuit has been used by many...so it's pretty much assumed to be valid...

...it's in a sub-circuit, it works just the same as if it wasn't, so why not treat its results the same?

Because, as your first quote says, it's easy to screw it up when people cut and paste stuff, especially if they don't really understand it.
So I tried to check your circuit was correct, my mistake.

...interpreted is an other story.
...From all the discussions I've read about this stuff, this seems like a pretty darn good amount, of both...

What discussions were those?
Both are borderline.
Do a Tian plot as the amp is close to clip, this is often the worst-case that sets the limits.
Maybe it won't be for your amp, I would be curious to see the plot.

Best wishes
David
 
I did miss the supply definitions, probably because I can think of no reason why anyone would do it this way.

I used to do it the other way, but like it better this way, because it cleans up the schematic and makes it smaller.

Why put the definitions for the supplies far from the supplies?
Why not put the specifications of an object next to the object?

I do usually place this close to where the supplies go in, with the labels. I just moved things around when I added other things and I will eventually move that back closer to the actual supplies "entry point". I prefer seeing the supplies info in one place, easy and quicker to change, both can be changed at the same time in one shot...

I do the same thing with other things, like load, zobel, things like that. Doing it this way allows having more than one set of those things, which can be switched easily by remming out one and un-remming an other...

What discussions were those?

All over spread out across the forums. Lots discussed about margins, mixed up with other stuff.

Both are borderline.

Under what criteria? This is where I find most confusion. The various opinions mention lots of different values, no actual consensus.

And that's why I ask the question again, but I guess that will vary depending on who answers.
What is the correct amount? What criteria is used to arrive at those values?

Do a Tian plot as the amp is close to clip, this is often the worst-case that sets the limits.
Maybe it won't be for your amp, I would be curious to see the plot.

That's confusing to me, as doing a sim with a Tian probe is an AC analysis, not a transient, and I don't see how referring to clipping in an AC analysis context can work.

How would you do that? There is no signal at the amp's input in an AC analysis, and that's been mentioned lately, that it should be "zeroed out".
And the Tian probe sim is stepped with values of 1 and -1, so how to relate any of this with a clipping condition?
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.