The Tian probe is not informative where you have placed it.
The servo loop bypasses it so your probe results will not reflect the actual feedback.
That would be a possible explanation but you will need to include your models if you want me to actually run the simulation and confirm this.
Best wishes
David
So if I disconnect the servo during AC simulations i'll get the desired results..??
Just tried without the servo, No difference in the results.
Just tried without the servo, No difference in the results.
Last edited:
I don't understand what results you desire or what you think the problem is.
A few pictures to explain would help.
Best wishes
David
If you look at the picture in post 358,
there you see that the curve is rising again above 3 MHz, with the tian probe in place the rise is above 0 dB and the phase-shift more than 180 degrees this should indicate problems.
If I remove the tian probe an look at the OLG plot, then the phase-shift at 0 dB is "only" 160 degrees and the circuit seems stable.
there you see that the curve is rising again above 3 MHz, with the tian probe in place the rise is above 0 dB and the phase-shift more than 180 degrees this should indicate problems.
If I remove the tian probe an look at the OLG plot, then the phase-shift at 0 dB is "only" 160 degrees and the circuit seems stable.
Last edited:
The Tian probe should never alter the stability, I wanted to make sure I had not misunderstood you.
If you post the models then I can run the simulation and I will try to determine where the problem lies.
Best wishes
David
Is it pure cherry though. I'm sure there's some shunt loading on the VAS output. Or have I misunderstood the meaning of pure cherry?
Hi kgrlee, I see that you follow CFA thread but you never commented after I tried to explain my use of "pure Cherry". Of course there is some shunt compensation as is usual in CFA amps.
I tried to use Cherry compensation in some of mine VFAs but never with much successes, it looks that is better suited for CFA.
So I am back from the dark side, Jedi now??
I know you write facetiously, but doesn't "pure" Cherry mean just one capacitor component?
What do you consider to make it second order, the inevitable shunt load of the OPS on the VAS mentioned above?
Internal component capacitances, trace PCB reactance?
Best wishes
David
Last edited:
Zan/Tian Probe
My favoured topology for 1 ppzillion THD20k is Enhanced VAS + EF2. Though this has 4 devices in the forward path, the system usually appears 2nd order. ie the rate of closure in the Return Ratio is 12db/8ve.
The extra orders in a well designed amp should appear above the 'ULGF'.
The usual tricks to stabilize a '2nd order' system, while looking at Bode/Nyquist bla bla, are appropriate .. though the practical methods are somewhat different from tweaking TPC Miller. You want Bode to be sorta 6dB/8ve as you approach the ULGF.
__________________
This is easily shown by a Tian probe on the 'inner loop' as in #262.
I shall dub this the Zan/Tian Probe; ZTP for yus TLA lovers. But I shall stick with the full name out of respect.
I know Guru Zan wasn't the first to suggest a Tian Probe at this position .. but he was the first to highlight & explain its importance.
My ambiguous references to 'inner' & 'outer' were obfuscating, confusing and beginning to irk me.
The VAS+OPS, in its simplest practical version of Single device VAS + EF2, is 3rd order; 3 devices in the forward path. The Cherry cap around this turns it into a 2nd order system cos the 'closed loop' gain demanded is differentiating.
My favoured topology for 1 ppzillion THD20k is Enhanced VAS + EF2. Though this has 4 devices in the forward path, the system usually appears 2nd order. ie the rate of closure in the Return Ratio is 12db/8ve.
The extra orders in a well designed amp should appear above the 'ULGF'.
The usual tricks to stabilize a '2nd order' system, while looking at Bode/Nyquist bla bla, are appropriate .. though the practical methods are somewhat different from tweaking TPC Miller. You want Bode to be sorta 6dB/8ve as you approach the ULGF.
__________________
This is easily shown by a Tian probe on the 'inner loop' as in #262.
I shall dub this the Zan/Tian Probe; ZTP for yus TLA lovers. But I shall stick with the full name out of respect.
I know Guru Zan wasn't the first to suggest a Tian Probe at this position .. but he was the first to highlight & explain its importance.
My ambiguous references to 'inner' & 'outer' were obfuscating, confusing and beginning to irk me.
Pliz dun mek fun of mi wid poly-syllabicity jus cos i kunt reed en rite proper 😱
OK, so 2nd order roll-off.
Just wanted to be sure what you meant.
It's possible to have 2nd order systems without roll-off, for instance an all pass filter.
Waly's post on the subject is probably educational, unless you think it pedantic instead.
Stabilization is a tricky subject.
The various component capacitances are of the order of the actual capacitors.
Plus the PCB parasitic capacitances and inductance.
You probably noticed I have started to discuss this in another thread.
Any input welcome.
Shucks😱
Best wishes
David
Last edited:
Du..uh! Should be cos the 'closed loop' gain demanded is integrating.
Pedants are by definition educational. Unlike some, Guru Waly's stuff is usually educational and worth close study.
Unlike da zillion page rants over semantic issues beloved by da pseudo gurus who are of no use to man or beast.
I look forward to the day when Guru Waly applies his undoubted expertise to a practical (or even SPICE world) design. 🙂
Last edited:
Hi all,
only recently, I found the Cherry ETI article (1983) on NDFL, where he also explains a little bit how to apply this to a "standard architecture" (although with additional intermediate Rush stage).
When considering this, a couple of issues, also related to the "pure Cherry", pop up.
1. Might it be that the additional differentiating feedback loop via the intermediate stage, whichs signal is overlaying with the feedback signal from the "pure cherry cap", helps in fact to stabilize the critical loop around the OPS?
2. In his example, Cherry uses no darlington OPS stage. I assume that this also helps quite a bit.
3. Cherry does not give frequencies and component values.
This leads to the main question:
Are you aware of a disscussion of this Cherry article, addressing component values and gain/phase margins of the different loops (with the rigour we are applying in the forum)?
Thanks in advance,
Matthias
only recently, I found the Cherry ETI article (1983) on NDFL, where he also explains a little bit how to apply this to a "standard architecture" (although with additional intermediate Rush stage).
When considering this, a couple of issues, also related to the "pure Cherry", pop up.
1. Might it be that the additional differentiating feedback loop via the intermediate stage, whichs signal is overlaying with the feedback signal from the "pure cherry cap", helps in fact to stabilize the critical loop around the OPS?
2. In his example, Cherry uses no darlington OPS stage. I assume that this also helps quite a bit.
3. Cherry does not give frequencies and component values.
This leads to the main question:
Are you aware of a disscussion of this Cherry article, addressing component values and gain/phase margins of the different loops (with the rigour we are applying in the forum)?
Thanks in advance,
Matthias
Last edited:
Sorry, I did also find the follow-on article discussing the concrete 60W amplifier with component values.
Anyhow, are there places with in-depth discussion of stability margins, additionally to the thread "New Cherry NDFL amp"?
Matthias
Anyhow, are there places with in-depth discussion of stability margins, additionally to the thread "New Cherry NDFL amp"?
Matthias
JAES March 1991 had a related article by Scott and Spears.
Cherry responded and the issue includes his comments.
There were other JAES articles by Cherry but I already recommended them to you in your thread.
I am not aware of much else in the audio literature, I expect most of the "nested" discussions will be in the IC literature, for the reasons already mentioned.
Best wishes
David.
This from someone who's favourite words are "obfuscation" and "pontification"😉
Thanks for the Cherry/Huijsing correspondence.
I had already found that one, it was your reference to IEEE Trans. that I was curious about.
Best wishes
David
I'm also curious why the beach bum is "unwashed" - too many salties to chance the water?
doubt it will help many undertanding Cherry nested loops but Mitchell a Reading has written a little on them
Dr Richard Mitchell's Selected Research Papers and Presentations
doubt it will help many undertanding Cherry nested loops but Mitchell a Reading has written a little on them
Dr Richard Mitchell's Selected Research Papers and Presentations
Thank you for the pointers. Unfortunetaly, I only have access to the IEEE publications via our library. Do you have electronic versions of the papers you mentioned?
And do I interpret you right, that probably nobody has done/shared a rigorous simulation analysis of the different loops using the methods we all together have applied in the the forum for TMC, TPC, MIC and so on?
Best regards,
Matthias
Last edited: