It would be helpful to learn the sequence of Cherry's ideas.
I understood that he expressed concern but then built the NDFL amp that used a lead capacitor, which implies he had found no real problems.
If it is the NDFL article that came first then that is obviously different.
Leach may also have been leery but I have not seen any evidence he may have had.
Some of his other concerns definitely turned out to be ill founded, so simple leeriness does not convince me.
Best wishes
David
A better RF filter makes sense to me.
Deal with RF with an RF filter and then optimise the feedback with the feedback network.
Best wishes
David
His concerns were with stability with capacitive loads, which lead to his split feedback scheme in linked paper and latest amp designs. While a Zobel/RFI filter scheme ideally designed for a given load would bring flexibility for various feedback parameters, Leach chose a system approach that made the output network have less affect on compensation. Whether or not you agree with his approach is beside the point. You made statements to the effect that there's no major problem with lead comp. I simply offer some consequences stated by the same Cherry you referenced as well as Leach's paper.
http://users.ece.gatech.edu/mleach/papers/Feedforward.pdf
Yes. "a...small capacitor in the overall feedback network B can be used to correct the group delay and improve the reproduction of transient waveforms"
Best wishes
David
Hi David,
Okay, that cap. This one is not connected to the output. Instead, it is connected to the Zobel cap. As a result, it doesn't provide any lead compensation. So no fear for extra HF ingress. Same story as outlined in post 2809. The two circuits from my website behaves exactly the same. (and please, ignore what Mr Troll said about it. He doesn't know what he's talking about)
Cheers,
E.
Okay, that cap. This one is not connected to the output. Instead, it is connected to the Zobel cap. As a result, it doesn't provide any lead compensation. So no fear for extra HF ingress. Same story as outlined in post 2809. The two circuits from my website behaves exactly the same. (and please, ignore what Mr Troll said about it. He doesn't know what he's talking about)
Cheers,
E.
Last edited:
Yes, I have his book but his approach does not seem to have convinced many other EEs.
Does any manufacturer use this?
No, I wrote that it seemed Cherry did not consider it a major problem to use a capacitor across the feedback resistor (or else he would not have done so in the NDFL amp). I still want data to make an informed decision
Best wishes
David
It is connected to Zobel cap but I think that Cherry's quote implies it still provides some lead. Not sure this is relevant to my question anyway since I am interested in the extent of EMI in a more conventionally connected amplifier.
No comments on that?
Best wishes
David
Actually loop gain remains the same in the audio band whether you use a real output stage or an ideal one.
I have tried this but it seems to introduce a pole in the major feedback loop which compromises stability margins. Perhaps someone knows different?
To be fair, one should compare the situation with lead network in the feedback loop to another situation with the same loop gain, e.g. shaped by a small resistor in series with the Miller cap (just as an example, maybe this brings another disadvantage).
Both variants will lead to roughly the same distortion residuals, but in the first variant the not exactly linear input stage will see more of the HF noise.
Matthias
Hi Mike,
It is important for you to understand that real output stages have finite bandwidth and contribute excess phase. This in turn influences the maximum achievable unity loop gain frequency for a given stability criteria. That in turn directly affects available loop gain in the audio band.
This concept is fundamental to experienced amplifier designers who must deal with real output stages, with all their warts, every day.
Cheers,
Bob
Cannot be Michael. You have a big fat pole in the o/p stage that I am afraid will just not go away - and especially if you are using bipolars. That's one of the main reasons all this MC, MIC, TPC, TMC etc stuff. Otherwise, you could probably stick a pole at 50 or 100 kHz and be done with it and get sub 1 ppm without even trying.
Hi Edmond ,
It is obvious that the feedback path going through a shunt RC network
will result in non linear frequency response past the said RC -3db frequency ,
and will result in a peaking at about 600Khz with the values in your schematic.
The 12pF cap will in principle act as a high pass network whose effect
should be reciprocal of the zobel network and hence provide a frequency
constant feedback ratio but it will also provide inherently a phase lead
compensation.
very awkward wording - a linear response may have a frequency dependency - "non linear" is not commonly used to describe linear filters - even in their transition bands
"non flat" frequency response however is assumed in linear fliters
discussing audio amplifier circuit design would I try to avoid ambiguous uses of "non linear"
OK! I meant "with phase lead capacitor (compared to without phase lead capacitor)"
and you meant "with phase lead cap. (compared to with the phase lead done some other way)".
I understand now and your point is reasonable.
I want to zero out as many as the amp poles as is practical so I already plan to use a resistor in series with the Miller capacitor (or TMC equivalent).
I know that was just your example, but I expect to optimise both the forward and feedback gain. The phase lead capacitor is one extra free parameter.
Best wishes
David