How do I design TPC for a power amplifier?
Is there any way to do it?
Can I calculate or is it trial and error?
Is there any way to do it?
Can I calculate or is it trial and error?
I use a simulator for this, but you can calculate it as well. Start off by Miller compensating the amp for 70-90 degrees phase margin at the unity loop gain frequency (ULGF). This will usually be between 1MHz and 2 MHz in a bipolar amp using modern sustained beta output devices EF2 or EF3. Once that is done, split the comp cap into two parts. The cap going to the base of the VAS stage keeps the same value. Make the second cap going to the VAS output 2x that of the first cap. Now you have to calculate the resistor part. Use R = f/2*Pi*2c where c is the value of the Miller cap and F is 50 kHz. You may have to experiment a bit in the sim with the f value. What you want is a loop gain slope of 40 dB/decade initially that at around 500-800 kHz reverts back to 20 dB/decade.
Typical values you will end up with are 50pF,100pF and 1k to 2k for the resistor.
The phase margin will dip to c 45 degrees where the slope is 40 dB/decade before increasing again such that at the ULGF it will be back at 70-90 degrees as would be the case with a conservatively Miller compensated amp.
BTW, I stopped using TPC and now use TMC because TPC does show some closed loop peaking whereas TMC does not. For TMC, return the resistor not to ground but to the amplifier output.
Using either TPC or TMC will allow you to extract an additional 20-25 dB of loop gain at HF, dramatically lowering HF distortion. You will not see much if any improvement at 1KHz since the loop gain there is unaffected.
The technique works with VFA and CFA amplifiers. In both cases you should have well controlled square wave responses (I use 2-3 kHz with 100 ns rise/fall times to test) with no overshoot of ringing. The amplifier input must be BW limited to 300-500 kHz using a simple RC filter- good practice in any event no matter what comp method is used.
Typical values you will end up with are 50pF,100pF and 1k to 2k for the resistor.
The phase margin will dip to c 45 degrees where the slope is 40 dB/decade before increasing again such that at the ULGF it will be back at 70-90 degrees as would be the case with a conservatively Miller compensated amp.
BTW, I stopped using TPC and now use TMC because TPC does show some closed loop peaking whereas TMC does not. For TMC, return the resistor not to ground but to the amplifier output.
Using either TPC or TMC will allow you to extract an additional 20-25 dB of loop gain at HF, dramatically lowering HF distortion. You will not see much if any improvement at 1KHz since the loop gain there is unaffected.
The technique works with VFA and CFA amplifiers. In both cases you should have well controlled square wave responses (I use 2-3 kHz with 100 ns rise/fall times to test) with no overshoot of ringing. The amplifier input must be BW limited to 300-500 kHz using a simple RC filter- good practice in any event no matter what comp method is used.
Last edited:
Maybe you will find this thread interesting: https://www.diyaudio.com/community/threads/oitpc-output-inclusive-tpc-not-tmc.317335/
You forgot about this thread you visited some years ago. https://www.diyaudio.com/community/threads/tmc-or-tpc-my-dilemma-resolved.188133/
By the way this is TMC.
By the way this is TMC.
Yes - that's TMC. You should get 20-25 dB reduction in HF distortion. I'm wondering why your R12 is so high? Usually, 1-2k is about right.
Here is how I do it.
(I made an error in post 2 - I should have said 500 kHz and not 50 kHz because it is the TMC cross-over frequency that we need to calculate the resistor value for - apologies for that)
(I made an error in post 2 - I should have said 500 kHz and not 50 kHz because it is the TMC cross-over frequency that we need to calculate the resistor value for - apologies for that)
Apologies for nit-picking but I believe it should it be
,
giving 2.1kOhm in your example. (It is the sum of the two caps that defines the time constant, and you use Cdom + 2 Cdom).
In practice, one can simply trim the resistor in hardware to the minimum that still provides a sufficient phase margin. The latter can be gauged by e.g. ringing on a low level square wave, with the input LPF disconnected.
giving 2.1kOhm in your example. (It is the sum of the two caps that defines the time constant, and you use Cdom + 2 Cdom).
In practice, one can simply trim the resistor in hardware to the minimum that still provides a sufficient phase margin. The latter can be gauged by e.g. ringing on a low level square wave, with the input LPF disconnected.
H
HAYK
This circuit cannot be analyzed, the VAS is 40gm this means it is voltage input but what is its input impedance to get the current of the IPS into voltage?
Post in thread 'Unusual amp from 1987' https://www.diyaudio.com/community/threads/unusual-amp-from-1987.357369/post-7493597
Post in thread 'Unusual amp from 1987' https://www.diyaudio.com/community/threads/unusual-amp-from-1987.357369/post-7493597
Last edited by a moderator:
I usually look at it from the POV that the cap from the VAS output is being bypassed at LF by Rcomp. As F increases, Rcomp is bypassed by the VAS output cap and the OPS is transitioned out of the VAS inner loop.
I will take another look at your proposal - I think the front end and VAS impedances may make the analysis a little more complicated than simply 3Cdom, but I am open to correction.
In any event, once the values are in the ball park, it’s always a good idea to tweek the final values in a simulator.
There are two feedback loops in an amplifier like that. With TPC, the outer (global) loop gain has two poles and one zero, which your graph shows.
With TMC, one of the poles and the zero hide in the inner (local) loop, so you don't see them in your analysis. The loop gain available to correct the output stage errors is virtually the same in both cases. However, the input stage in this case has to work with higher signal levels at the HF end of the audio band, which is a source of extra distortion.
More like two small sections in Chapter 8 (I am looking at his 5th edition). BTW that's basically a reprint of his original 1994 article (page 137). Unfortunately, he doesn't explain any design procedure.
An accurate analysis of TPC was presented by Harry Dymond and Phil Mellor at the 129th AES convention in 2010 - AES sells their paper for $33, so I refrain from posting it here. In brief, the location of the poles depend on many factors, but the zero (Fcross in Bonsai's paper above) is at:
The practical design procedure is quite simple: choose the capacitance of the 2nd cap (connected to the VAS output) ) 2x to 10x of the first (connected to VAS input), then trim the resistor to the minimum value that still gives you a sufficint phase margin. To gauge the phase margin, disconnect the input LPF of the amplifier (if using), then feed the amplifier with low level square wave and observe ringing at the output. More ringing = less phase margin. (There is a mathematical relation there, but I doubt anyone would be interested). You do not want to eliminate ringing altogether, just make it aperiodic. After that, make sure the amp is stable at high levels with load connected and when clipping.
Last edited:
Also Arto Kolinummi covers this subject in his book 'Audio Power Amplifiers' (page 100); but also without any practical design suggestions.
H