I looked mainly at the return ratio because I was concerned about stability. The emitter capacitor does help.
Maybe no real dispute here.
Yes. I am not too worried by the RFI either, and the phase lead capacitor does not have any effect on THD by itself. I would point out that the extra lead does mean that it is possible to push the ULGF up and that would improve THD. As Andrew (Bonsai) noted, the phase lead capacitor can be overdone and become counter-productive.
But my uncertainty was whether it is possible to create a useful zero with the VAS/TIS emitter capacitor.
Yes also. This is rather simple for Miller compensation. Less clear for TMC and similar more advanced compensation schemes.
Best wishes
David
Last edited:
As noted Dave, the cap across the feedback resistor, if selected correctly, only affects gain and phase response beyond the ULGF, where it provides gain and phase margin benefits. On my e-Amp design, I use 5.1k feedback resistor and IIRC 150 Ohm in the lower feedback leg. I found 5-10 pF was all that was needed. Thus, when I see 22 k feedback resistors and 100 pF caps, I cannot help but wonder about the impact on stability, to say nothing of the possible RFI - although an output inductor and a Zobel can help here ( keep the output inductor damping resistor as high as practicable, use low inductance components for the Zobel)
Yes, I think we are in accord. This is a sensitive area and the capacitor must be optimised, not lucky-dip.
An output inductor and low inductance Zobel makes sense to me.
Probably a distributed Zobel directly on the output transistors and an RFI capacitor directly at the speaker terminal. All standard Cherry.
So we have a useful zero in the feedback network.
I have started to consider a zero from small inductors in the LTP emitters.
Done in the famous JE 990 discrete op-amp with ferrite cores.
Small air core toroidal inductors look practical.
The VAS/TIS still undecided.
Best wishes
David
Things have changed since the first version of Grey and Hurst
Designers are now using structural design with analog building blocks.
Complex mixed signal chips are using these blocks and the OTA is one of the main block.
Depending on the requirements ( speed, power, rail to rail, immunity , technology ....) you will need to make a choice between different OTA's topologies ( single stage, folded cascode, symetrical, two stages miller )
A traditional opamp is an OTA+ a buffer but can be a single OTA if output impedance is OK. The design of an OTA is more focused on open loop. An opamp is used in a feedback environment.
These concepts are structuring analog design and the total analog community is talking like that, there are no chapels on this subject.
See attached the slides of a typical design of a Wireless sensor ( sorry in french but self speaking). The design uses a Miller OTA. Very familiar design equations are given. The same approach is used in amplifier design.
I hope this clarifies the matter.
Excess phase is very well defined as the phase in excess of minimum phase and we can all call it that. Other uses are poorly defined.
This is one area that I do think could be improved in the Cordell book.
It is used loosely there and the intended sense is not always clear.
Best wishes
David.
I mean pole and zero in conjunction with emitter bypass capacitors and the usual LTP resistors of course. Should have been more explicit.
David
Sorry Davada, I don't understand what your point is. In fact, I had the non-inverting mode in mind when I posted your quoted post, but I believe there is no conceptual difference with the inverting mode.
In all cases, there must be an effective input signal between the two inputs that is exactly Vout/OLG.
jan
Well, it's not mine
and it's not very satisfactory, but 47 degrees is hardly pathological. And part of the reason for the poor phase margin is precisely because the RHP zero causes non-minimum phase behaviour. And if you really want then this same conventional amp can be stabilised with 60 degree PM and the non-minimum phase behaviour is still similar. Yes...and? I posted a Bode plot that clearly shows non minimum phase behaviour. I linked to the schematic and LTspice file so anyone who doubts the result can check the circuit and models. What's the problem?
Yes. I had not "worked out the details" of why the non-minimum phase behaviour was so conspicuous. Doesn't alter the fact it is a clear case of non-minimum phase behaviour. I added a capacitor across the VAS emitter resistor. The models are all in the linked post.
RHP zeroes in the forward transfer function lead to non-minimum phase behaviour.
The Miller feedback can have the RHP zero eliminated with a series resistor.
But all amplifiers are ultimately non-minimum phase. The only question is whether that behaviour is of practical importance.
If it is not relevant for audio why is it "a relevant RHP zero"? Kind of a confused phrase really. But I completely concur if what you meant is that an RHP zero often has no practical importance.
This doesn't seem to have any point? We both know CMOS op-amps have issues with RHP zeros There is an extensive literature. I have a few theses with literally dozens of ways to deal with the problem...And?
I suspect you mean the minimum phase property
Otherwise it is perfectly possible to adjust the gain and phase separately. That is exactly what an all pass filter does. It adjusts the phase separately from the gain.None of this affects the fact that you stated
"Audio amplifiers are minimum phase, period."
And it's demonstrably not always true.
Best wishes
David
Last edited:
On what basis do you make the assumption that the miniscule variations in closed loop frequency response at HF are due to TIS loading?
Note that connecting a resistor in series with the Miller capacitor seems to worsen the stability margins of the minor loop.
Actually, I was refering to the fourth edition of Grey, Hurst, lewis and Meyer published in 2001. This can hardly be described as outdated.
I've also looked at the fifth edition (2009) and there is no reference therein of the two stage Miller compensated amplifier as an OTA. Nevertheless, please feel free to call it an "OTA" if you wish; I, on the other hand, will remain firmly in the camp of those authorities who don't, as I have indicated here:
http://www.diyaudio.com/forums/soli...lls-power-amplifier-book-387.html#post3551501
Last edited:
Ok, let's end this by amending: "for all practical purposes, audio amplifiers are minimum phase systems". I hope you realize that, in usual audio implementations, any property that goes beyond, say, 5MHz is usually swamped in parasitics, and I don't think you'll see an audio amplifier designed on low permittivity RF substrates anytime soon. OTOH, as I already said, if the stability of an audio amplifiers depends on a few dB of lost gain margin because of a RHP zero, that's a rotten design to start with.
I'm waiting (but not holding my breath) for the first audio amplifier design with independent phase correction using an all pass filter.
P.S. My example is relevant because a) is very simple and b) the RHP zero is under 100MHz (not in the 100's of MHz). Irrelevant for audio because the UGF is 200KHz and the 3dB bandwidth is about 200Hz.
Last edited:
I had not considered that, and I don't think I have seen it mentioned before.
You Spiced this?
Best wishes
David.
Yes. See below.
Minor loop gain phase shift (Green trace) at unity gain without series resistor=153 degrees
Minor loop gain phase shift (Blue trace) at unity gain with 100R series resistor=225 degrees
Attachments
Last edited:
In the example I posted the effect of the RHP is already quite visible at 5MHz.
And this was for an actual amp, no attempt was made to select component values that lowered the frequency
In the example post the GM improves by 6dB.
If you say a 6dB improvement makes no difference "for all practical purposes" then I won't waste time in a debate.
Any readers can decide for themselves.
Best wishes
David
Last edited:
Concern with Gm2 really isn't an issue with discrete amplifiers with a BJT TIS whose quiescent current is greatly exceeds 1mA. It's only an issue in monolithic op. amps. with sub-milliamp. TIS standing currents, with the resulting low Gm2 causing potential failure of or inadequate pole splitting.
Note, also, that if the emitter follower in the buffered TIS isn't biased with sufficient current, then pole splitting may be inadequate or fail altogether as other non-dominant poles become significant. See Solomon, pg. 327, below:
http://ece.wpi.edu/~mcneill/524/handouts/solomon.pdf
It's a good idea, I think, to bias the buffer to draw roughly the same current from the LTP as the current mirror to mitigate quiescent current imbalance in the LTP.
Last edited:
Yes, seems reasonable, but sometimes this non-minimum phase behaviour is not intuitive.
Thanks for that.
Best wishes
David