With double pole compensation you never need more than 68p lead compensation in series with 2k2.
Disable your major loop by shorting out your feedback network at the diff pair with a very large capacitor (10KF) with the input also short circuited to AC.
Open minor loop about the transimpedance stage (TIS and not VAS) and insert loop gain probe, (i.e. use Middlebrook's or Rosenstark's probe) to determine loop transmission.
About PSRR in your paper, it is noteworthy that simple RC filtering (R=10, C=1000uF) of the negative supply rail without cascode bootstrapping or moving the compensation's resistor's connection to ground should give more than adequete PSRR.
Infact after RC filtering you should find that there is no need to move the resistor connection from the supply rail to ground.
Last edited:
Yes but ''TMC'' is demonstrably inferior to TPC.
Note that is wise to start with a correctly (analytically) designed TPC network before converting to ''TMC'', as this ensures that the zero cancelling one of the double poles in the output minor loop is correctly located to prevent instability in that minor loop.
Another source of minor loop instability unless you use a very high Ft driver, or compensate the minor loop itself by shunt capacitance to ground at the TIS (never ''VAS'') output.
Last edited:
So, if you allow some overshoot you can have some additional loop gain at higher frequencies that can be used for distortion reduction. Question is, what is your limit? How far can you / dare you go knowing that you have no control over whatever (reactive) load is thrown at the amp out there.
Do you have a Rule of Thumb you use?
jan didden
jan, the added gain would apply over the entire audio range - for example say an "extra" octave could be had by allowing lower phase margin, with 2-pole compensation this would give +12 dB loop gain at all frequencies well below the 2nd order breakpoint
I understand that; my question was, how much lower phase margin would you be willing to allow? Edmond's view appears to favor some robust safety-margin. I often have the tendency to go a bit too far in the quest for max feedback gain and minimum distortion (and pay for it with smoke 😉).
Is it a matter of personal preference only?
jan didden
how "robust"? when do you give up - unconditional load stability without decoupling network, overdamped response only - likely not
Control Industry practice does accept smaller phase margin than "flat step response" implies
I just find it odd to argue so vehemently over means to obtain such an apparently arbitrary "optimum"
Control Industry practice does accept smaller phase margin than "flat step response" implies
I just find it odd to argue so vehemently over means to obtain such an apparently arbitrary "optimum"
I'm trying to find some rationale for an acceptable minimum phase margin rather than some arbitrary/personal preference. But maybe there is no straight-forward rationale.
Control industry practise probably isn't faced with unpredictable, widely varying loads, so that may not be a good reference point.
There is a rational way to determine phase margin. attachment 1/2 shows the typical blameless (my real "man cave" amp) with 22p compensation driving a resistive load in simulation. This is the simulated "breaking point" at 30 degrees margin. The first hint of overshoot is at 33p/60dg. margin. Into a simulated capacitive load , damped ringing can be seen at 39p/70dg. margin.
I then tried 33p and 39p on the real amp , a "tinnyness" could be heard confirming the validity of LT's prediction. 47p/80-dg. phase margin is what I initially designed the amp for and this will produce a very highly damped ringing with a large square wave driving a capacitor. 39p will work , but the sound suffers. I am scared to actually try 22p,
even through LT shows no sustained oscillation.
I could get a little more speed with the smaller cap , but opt for better sound and safety. Andy C. once told me to shoot for between 70-80 degrees margin and this has been faithful to me ... no smoke. 🙂
Other designs will behave differently with these same margins as will different devices. I have found the fairchild models to be pretty close to the truth as far as compensation is concerned. I've come close , but have never built a burning amp to date. 😀
OS
I then tried 33p and 39p on the real amp , a "tinnyness" could be heard confirming the validity of LT's prediction. 47p/80-dg. phase margin is what I initially designed the amp for and this will produce a very highly damped ringing with a large square wave driving a capacitor. 39p will work , but the sound suffers. I am scared to actually try 22p,
even through LT shows no sustained oscillation.I could get a little more speed with the smaller cap , but opt for better sound and safety. Andy C. once told me to shoot for between 70-80 degrees margin and this has been faithful to me ... no smoke. 🙂
Other designs will behave differently with these same margins as will different devices. I have found the fairchild models to be pretty close to the truth as far as compensation is concerned. I've come close , but have never built a burning amp to date. 😀
OS
Attachments
Of course but that was not the point. The point was, is there a rationale for a minimum acceptable phase margin.
Reading the rest of your post it seems that its basically personal preference based on good (or bad) experiences.
jan didden
Hello Jan
I think that when you are using simulation techniques to determin phase margin that 90 degrees phase margin is a good starting point. 90 degrees gives you a safety margin for when you go to build the hardware.
In my limited experience quite often you have to trim the compensation values to get a good pulse response.
Regards
Arthur
I think that when you are using simulation techniques to determin phase margin that 90 degrees phase margin is a good starting point. 90 degrees gives you a safety margin for when you go to build the hardware.
In my limited experience quite often you have to trim the compensation values to get a good pulse response.
Regards
Arthur
Hi Jan,
You're right, the designer/manufacture has no control over the (sometimes crazy) loudspeaker load. Therefore, you can't (mindlessly) apply general rules of thumb that are valid in an industrial process control environment, for example.
This being the case, I would say: stay on the safe side* and, whenever possible, always strive to a PM of 90 deg and a GM of >= 20dB.
As for a PM of 30 degrees, mentioned by JCX, syn08 reported that one of the channels of the PGP amp caused listening fatigue. It appeared the compensation cap in this channel was too small and caused ringing. After fixing it the amp sounded okay. I've simmed the amp with that too small cap and guess what? The phase margin was only 30 degrees !!!
Needless to say that a PM of 30 deg. is definitely too small.
* in particular when the additional cost is zero, as in the case of TPC vs TMC.
Cheers,
E.
So it seems 90 degr is conservative, and 30 degr is probably cutting it too close.
IIRC Bob Cordell aims for 45 degrees.
Any other opinions/experiences?
jan didden
IIRC Bob Cordell aims for 45 degrees.
Any other opinions/experiences?
jan didden
Hi Jan,
I aim for around 60 degrees, which I check in the lab by measuring the amount of overshoot with a 100Hz square wave at 80% of output clipping voltage with an 8R load. If I get <10% overshoot, I'm happy. Overshoot as a Function of Phase Margin
I remove the resistive load, and then use a capacitance box and increase cap load up to 4.3uF (the maximum with my cap box), to see if continuous oscillation can be provoked. I'll then try the cap box in parallel with a 4ohm (and sometimes 2ohm load).
Providing I use an output inductor (typically 1uH), the amps always seem to pass the capacitance box test with no issues.
I aim for around 60 degrees, which I check in the lab by measuring the amount of overshoot with a 100Hz square wave at 80% of output clipping voltage with an 8R load. If I get <10% overshoot, I'm happy. Overshoot as a Function of Phase Margin
I remove the resistive load, and then use a capacitance box and increase cap load up to 4.3uF (the maximum with my cap box), to see if continuous oscillation can be provoked. I'll then try the cap box in parallel with a 4ohm (and sometimes 2ohm load).
Providing I use an output inductor (typically 1uH), the amps always seem to pass the capacitance box test with no issues.
Well I think part of the discussion is what constitutes ' a good pulse response'. A pulse response with no overshoot appears to be needlessly conservative, more that 75 degr phase margin (see also previous post).
Intuitively I understand that a nice, non-overshoot pulse response looks good or safe. But a little overshoot still gives enough phase margin and gives you a bit more feedback and a little less distortion etc.
Don't forget that the square wave that shows the overshoot is not a practical musical signal.
jan didden
jan didden
Last edited:
Thanks for the link! That's a good paper to see the connection between phase margin, Q, overshoot etc. 60 degrees seems sensible.
jan didden