Michael, have you any 'real life' evidence for your assertions?
Certainly for the single transistor TIS, Cherry has a lot of real life evidence to support his matrix analysis and other pontification.
For a Blameless type amp, (enhanced VAS + driver & output) I have a limited amount of 'real life' to support my naive sims.
Both the above 'real life' stuff supports the use of a VAS emitter resistor but other stuff may be needed in addition
My own limited experience suggests that your arrangement (enhanced VAS + triples - 5 active devices (5 poles) in a feedback loop) WILL be unstable and may not be amenable to 'pure Cherry' without more complication which might severely affect the inherent distortion performance .. mainly cos the Holy HiZ point is compromised.
It will DEFINITELY be UNSTABLE WITHOUT A VAS RESISTOR.
__________________
I'll point out yet again
that the ONLY reason we do this sim sh*t is cos it might have some resemblance to 'real life'.If sims, theory or pontification don't coincide with 'real life' .. one of them must be wrong.
__________________
I'm in the process of analysing astx's excellent 2stageef-high-performance-class-ab-power-amp-200w8r-400w4r to see if 'pure Cherry' can enhanced its already good performance.
This has helped me crystallise my thoughts about Loop Gain probes and also the use and abuse of 'pure Cherry'. Watch this space.
___________________
Toni (astx) shows that excellent performance (better than most amps with triples) is possible. 'Pure Cherry' can improve this even further. Certainly to well past the stage where even SOA instruments give up.
I'm goiing to this trouble cos he has 'real life' results for both THD & stability which allow us to refine our sims and also get a better idea of where they may be lying. This gives us the confidence to use the sims to 'improve' the design.
An important question to 'triple' fans is 'are you getting better performance than Toni's much simpler circuit?' If not, why use the extra complexity?
Hi Richard,
Admittedly, I haven't tested a real amplifier with Cherry's output inclusive compensation, but the SPICE results are pretty telling don't you think?
After all, they confirm Douglas Self's findings.
Note that Cherry did not just recommend OIC for single transistor TIS amplifiers: he did not, in fact, place any caveats in this regard on its use.
Now, can you provide us with a schematic of a OIC that is stable in all respects?
Last edited:
Err.rh! The circuits in #4, #58, #84, #85 closely resemble 'real life' circuits I have built and tested. They are stable according to SPICE and more importantly ... the 'real life' criteria in post #1. You may like to download the .ASC files from post #5 and have a play.
I mention single transistor VAS amps cos the 'real life' examples that Cherry shows are of this ilk.
That scheme gives the amp an output impedance like a damped inductor. IE you are getting the amp to perform the function of the output series L/R network. The problem is that when you emulate this function with the amp, you increase distortion because the virtual impedance is generated by nonlinear transistors. This is something I would only do when I could not use an output network.
Zo only causes distortion if the load is nonlinear. The Zo of that compensation scheme itself will be nonlinear because it is a virtual impedance derived from nonlinear transimpedances, and this will cause distortion even with a linear load. The implication is that using a discrete impedance network (output series L/R) instead of a virtual one (caused by compensation) will distort less in both cases.
You're absolutely right but I felt compelled to insert a distinction.
Also, the comment about monstrous output Cbe dominating at BW extremes for very fast amps is true. For this your frontend has to be much faster than the Fc of Rm*Cbe (Rm=transresistance). However it is important to note that at the frequencies in question your entire amp will be in the throes of slewing with any significant signal. So is this really healthy? Such an amp will be relatively highly susceptible to RF intermodulation effects.
You're absolutely right but I felt compelled to insert a distinction.
Also, the comment about monstrous output Cbe dominating at BW extremes for very fast amps is true. For this your frontend has to be much faster than the Fc of Rm*Cbe (Rm=transresistance). However it is important to note that at the frequencies in question your entire amp will be in the throes of slewing with any significant signal. So is this really healthy? Such an amp will be relatively highly susceptible to RF intermodulation effects.
ALL opamps and Power Amps have an output Z with is inductive at HF. Just have a look at OPA datasheets which show Zo. eg Fig 2B in Intersil AN1092 that jxdking quotes.
You can measure this on a PA but you'll have to remove the Zobels and output inductors. Some amps may not be stable without these.
The Zo is also affected by feedback and usually to the same degree as THD ... both for linearity and value. For the usual high FB amp, Zo will be both small and 'linear'.
One could model the complete THD behaviour of an amp by considering all the THD is due to 'non-linear Zo'. I'm not sure there is any advantage wrt to the usual ways of looking at things.
__________________
keantoken, many thanks for your analysis of current mirrors on IPS. It explains a lot of serendipity in my Jurassic experience.
Last edited:
Your welcome..!
As for Zo, in almost all advanced amps, it is anything but inductive at any place other than the ULGF. I define inductive as an impedance rising at 6db/oct with a 90 degree phase shift. Anything else is not inductance and behaves totally different although you may not notice if you don't pay attention to these things.
As for Zo, in almost all advanced amps, it is anything but inductive at any place other than the ULGF. I define inductive as an impedance rising at 6db/oct with a 90 degree phase shift. Anything else is not inductance and behaves totally different although you may not notice if you don't pay attention to these things.
As an example. This is the LTSpice Zout of one of my ideas. Hardly any inductance there. Phase reaches all the way back to 270 degrees! This is an indication that feedback is more fully utilized. It is more common to have a phase of 180 degrees or less, but not usually less than 90 degrees.
I wish LTSpice had a polar+logarithmic Nyquist plot feature (with projected cartesian rulers). That would be so much more useful than the Cartesian nyquist plot.
I wish LTSpice had a polar+logarithmic Nyquist plot feature (with projected cartesian rulers). That would be so much more useful than the Cartesian nyquist plot.
Attachments
I'm going to call, 'Foul' and guess its not a PA or maybe has the evil Zobels & other stuff still connected.
Further cheats are the use of Linear axis for the impedance to astound & amaze us unwashed masses.
You gonna come clean?
Last edited:
I stand corrected. There appears to no part of the frequency spectrum where Zout of my poor efforts are even remotely Inductive. I grovel at Guru keantoken's feet.
The evil effects in fig 2D of Intersil AN1092 http://www.intersil.com/content/dam/...n10/an1092.pdf appear to be rife. My excuse is that I only pretended to read it.
Last edited:
'pure Cherry' on Toni's amp
I've posted a series of sims on Toni (astx) 's huge amp which has excellent performance using just EF2 outputs instead of evil triples. It starts from #177 2stageef-high-performance-class-ab-power-amp-200w8r-400w4r
Doing it has clarified some of my thougts & Jurassic memories of using 'pure Cherry'. More pontificating on the thinking .. later in this thread.
I've posted a series of sims on Toni (astx) 's huge amp which has excellent performance using just EF2 outputs instead of evil triples. It starts from #177 2stageef-high-performance-class-ab-power-amp-200w8r-400w4r
Doing it has clarified some of my thougts & Jurassic memories of using 'pure Cherry'. More pontificating on the thinking .. later in this thread.
You came to the Solid State thread quite late(loosing your time in other thread?) and, probably don't know about other very good projects, EF2 and triples(not so evil), using TPC or TMC, presented here. There was very good and quite long TPC-TMC discution with suggestion how to choose capacitance values(not just connect the resistor to the ground or the output with the same C values). You could try to search for TMC or TPC some years back and see.
BR Damir
There is absolutely nothing wrong with triples. They maximise forward path gain by reducing the loading of the output stage on the second stage (TIS), and they insulate the TIS from the non-linear loading of the output stage because the first member of a Loncanthi triple can be persuaded to operate in class A. Double emitter followers, even those in which the first pair are cross coupled, can not achieve this.
damir said:
Actually I HAVE been looking at some of these threads. But it is pretty obvious that this thread of mine is to champion 'pure Cherry'.Michael Kiwanuka said:
My other obsession is simplicity. I'm not really interested in supa dupa new topologies unless they show advantage (simplicity, stability, THD in that order) over my much simpler circuits. And of course, I am looking for stuff that has 'real life' measurements to back up theory. I like Toni's project cos it is simple and has better 'real life' performance than most 'triples + supa VAS + dupa IPS' projects.
I've posted sims & .ASC files and explain where these are close to what I found in 'real life' Jurassic times and also where they might not accurately reflect 'real life'.
But surely someone is going to try my .ASC files and show better performance with their favourite compensation .. even if only in SPICE world?
My claim is that the .ASC file is close to a 'real life' amp I designed and built in Jurassic times for production. I've also enumerated several differences ... the devices cos I don't have SPICE models for the 'real life' ones I used, and also where I think the model is furthest from 'real life' ... the PSU.
Michael, your .ASCs dun count cos Cherry, myself (& 'real life') are convinced they will be unstable. We agree that your circuits will be unstable .. because of the design choices .. evil triples + enhanced VAS, no VAS emitter resistor, other evil stuff .. . Why use an example that is known to be wonky (according to Cherry & others and 'real life') to criticise 'pure Cherry'?
How about showing my circuits are unstable?
Of course whatever criteria or 'real life' additions to my model that you use to damn it .. has to be applied to whatever competition you put up against it.
___________________
Damir, I'm aware of several ways to improve TMC and have worked on these (theoretically). I think Edmond Stuart's treatment on his website is good. But I haven't managed to get SPICE world results anywhere near a good 'pure Cherry' implemenatation.
But I don't believe in TMC so am not qualified to do optimum TMC variants. eg You would not ask Self or Cordell to design a good CFA cos they don't believe it is better than VFA.
I asked TMC fans for their expert recommendations and simulated Michael's in posts #56, 57, 58. They show the equivalent 'pure Cherry' version has nearly 10x less distortion.
____________________
Damir, I've just noticed you posted some recommendations too .. in post #55. They show much better performance. My apologies for not looking at them properly. I will play with your recommendations & report back. But you will have to allow me to use the good devices that you have on your recommendations
____________________
Lastly, I hope everyone knows that when I say 'evil', it is tongue-in-cheek (ie a joke). I don't really think triples evil (actually I do ... but not for the reasons discussed here
) ... but that you can get similar or better performance using simpler circuits. - Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.