Spice simulation

[keantoken]

Well because open-loop gain decreases with frequency after the corner, the high-order harmonics are not corrected as much as the fundamental. To me this is a very good reason for open-loop BW to cover the audio range. Unless perhaps your feedback loop is so powerful that CMRR or something else dominates distortion.

Also, if you have a low open-loop BW, doesn't the amp turn into something of an integrator? It ends up making gradual adjustments for some mS after any signal. This is essentially stored energy dissipating, and if RF turns your nonlinear semiconductors into mixers, you can end up with a shifting operating point. This is one distinct advantage to high open-loop BW.

[jan.didden]

Originally Posted by keantoken

Well because open-loop gain decreases with frequency after the corner, the high-order harmonics are not corrected as much as the fundamental. To me this is a very good reason for open-loop BW to cover the audio range.

The problem is that when you extend the OL BW, you will find that the OL gain that is available for FB does not increase. With a given amp, you can only extend the BW by lowering the low freq OL gain!

So you end up with the same FB correction factor at high frequencies as before, but now you've also lowered the lower freq FB correction to the high freq one. A big step backwards!

jan

[keantoken]

Often you do not need so much feedback to gain reasonable distortion figures, and BW can be extended by using things other than plain old Miller compensation. Symasym is a good example!

If there is an amp where high OL gain won't cause the need for massive compensation, and therefore lots of intermodulation with RF signals, then I guess there is not much stopping you from maxing out OL gain. What concerns me is the loading of the input stage; too much will cause it to become an RF mixer. VAS Miller compensation is particularly bad at this.

[tvrgeek]

Originally Posted by Bob Cordell

Even 60 degrees of phase margin is OK, IF it is real (meaning SPICE is really capturing and modeling properly all of the sources of excess phase in the real world). Chapeter 4 in my book "Designing Audio Power Amplifiers" covers a lot of this and has a useful plot, 4.10, that shows gain peaking and overshoot as a function of phase margin for a typical arrangement.

Very importantly, gain margin must also be good in order to achieve an amplifier with robust stability. One should have at least 6 dB of gain margin, meaning that the loop gain has fallen to -6dB by the time the phase reaches 180 degrees. Sometimes a designer will inadvertantly do something to a circuit to improve phase margin that only lessens gain margin, and this can be dangerous.

Cheers,
Bob

I have to go back and re-read again and again. Each time something else sinks in.

[kenpeter]

Originally Posted by janneman

The problem is that when you extend the OL BW, you will find that the OL gain that is available for FB does not increase. With a given amp, you can only extend the BW by lowering the low freq OL gain!

So you end up with the same FB correction factor at high frequencies as before, but now you've also lowered the lower freq FB correction to the high freq one. A big step backwards!

jan

Your assumption is based upon slugging the slowest pole
to make it dominant. If you instead speed up (neutralize
Miller of) other poles, until only one is left: The dominant
corner can be above audio, and stable, without reducing
the open loop audio gain plateau.

Neutralization actually doubles capacitance, collector to base.
Since you must add equal Miller cap to an opposite collector.
But the cap voltage divider eliminates moving effects, so is
not multiplied by voltage gain. Multiplication usually much
worse than double, can be avoided...

The corner frequency goes up and phase shift goes down.
What you would expect of a double unboosted capacitance.
Getting rid that boost is pure fired magic...

Reducing open phase shift does not require reducing gain.
Yes, the "gain bandwidth product" has thus been improved.

[kenpeter]

A more common way of hiding Miller is Cascode,
But this can only straighten the phase, it can't
go over center and bend it backward. Safe and
sane, but perhaps less versatile...

[jan.didden]

Originally Posted by kenpeter

Your assumption is based upon slugging the slowest pole
to make it dominant. If you instead speed up (neutralize
Miller of) other poles, until only one is left: The dominant
corner can be above audio, and stable, without reducing
the open loop audio gain plateau.

Neutralization actually doubles capacitance, collector to base.
Since you must add equal Miller cap to an opposite collector.
But the cap voltage divider eliminates moving effects, so is
not multiplied by voltage gain. Multiplication usually much
worse than double, can be avoided...

The corner frequency goes up and phase shift goes down.
What you would expect of a double unboosted capacitance.
Getting rid that boost is pure fired magic...

Reducing open phase shift does not require reducing gain.
Yes, the "gain bandwidth product" has thus been improved.

Sure, give it a try! If you can increase the OL gain at say 20kHz you make really progress.....

jan

[tvrgeek]

Originally Posted by keantoken

Looking at "loopgain2.asc" in the Examples directory in your LTSpice folder will give you a method for plotting open-loop phase margin. Most people prefer a phase margin of 80 degrees, or an open-loop unity phase of 100 degrees. (I've heard of the 40 degree figure, but I think it is either a mistake or it was taken out of context)

Smaller phase margins tend to cause overshoot in the square wave response, and more risk of anomalous oscillation. However the phase margin changes with loading, so you will need to look at phase margin with a number of different loads.

If you want a masterful example of stability compensation, look at Symasym.

I am getting about half of all this. Guess that's progress. Synasym does have some features that are interesting. ( cap in driver stage I have not seen before) At least two references I have suggest the poles on the output of the VAS to be not the best. I found ( was lead by the nose to) the big issue it caused in my DH-120 with highly asymmetric rise and fall. However my RB 951 uses the same method. Is this really a matter of sufficient current?

I will punch all this into spice so I can pay and see the direction changes make. I am getting a bit lost on the exact placement of parts for the canceling miller in the last few posts.

On " a number of different loads". Yup, good advice. What I have not seen is any set of suggested loads, easy to Apogee. I have been just using a 4 or 8 Ohm resistor and 1u cap. No idea what a ribbon transformer would do if I ever try some of them.

It would be quite a project but I guess I could take the plots from WT-2 for several of my speakers and see if I could come up with a network that has something like that. It would be limited to the pretty easy loads of my speakers. ( at least I try to build them so)

I take heed that my idea of just measuring phase across the input is not really valid. Experience with silicon fuses reinforces that. I am not completely sure why beyond the chicken and the egg problem.

With servo systems, one can make it unstable and force it into submission. That tends to make a very fast responding system. Or, one can make it as good as you can without the feedback and then just nudge it a bit. Failure modes differ for sure. Anyway, I see elements of both directions in amp feedback and compensation. Is there any relation with one or the other and sound? Preferences?

[keantoken]

By "various loads", I mean different size capacitors, from 1nF up to 1uF. No need for a load resistor, just hang a capacitor on the output and see if it oscillates. That is as diverse a test as I have needed so far. Generally inductive and capacitive loads are not a problem, and if they are it will show up in the capacitor test anyway.

What are your objectives for compensation? Slew rate? Distortion? Bandwidth? EMI immunity? How do you really tell when one scheme is better than another?

[tvrgeek]

kean,
My objectives are mostly learning. If at some time it leads me to why my Rotels pass my wife's super critical hearing and almost nothing else does, all the better. I just got fascinated with amp design and have been getting a crash course ( much thanks to many forum members) in the basics. While repairing my baby Hafler, I went through quite a bit of effort to get it stable with the Exicons instead of the Hitachies. Among other things, I am learning LTSpice and how to use it to gauge as best we can reality.

As far as what is in my system I am listening to, I have a sufficient supply of old Rotels and Parasounds. I am tempted to pick up a Forte or Nak Pa5 just to hear the results of Mr. Pass's affordable efforts. Different approach than EB, JC, DS, BC and others.

Which parameter is most important? Well that does seem to be the big question. We have been having quite a conversation over in the lounge about it. I learned a bit over in the solid state forum about ccs's. I have some suspicions, but in no way am I going to second guess the likes of many members here who have made a living at it for 50 years.

I suspect: The problem my wife hears is actually in the tweeter. My guess in high order distortions exciting breakup and resonance issues in the tweeters. I judge the Rotels to have less detail, say compared to one of my Parasounds or Haflers. If you look at the profile of the distortion, the Rotels fall off where the others seem to remain level relative to the background noise. The fix is then best done with better tweeters. ( Excel or Revelator, not decided yet) Very aggressive notch filters in the tweeter could also be the correct answer. If I can quit playing with amps, I can get back to my speakers and find out.

I suspect: consistency/linearity to be more important than some absolutes, like in a ccs for the IPS. A fast JFTE is consistent across the AF band, where a cascode is stiffer, but rapidly changes impedance above 10K. I plan on testing this by ear. Resistor, Jfet, and JFET cascode.

I suspect: multiple outputs to be superior to single higher power ones.

I suspect: some of the extremes in IPS design are not audible in a system as the output is an order of magnitude worse.

I suspect: why I never liked the passive preamp I built was I was using a BJT input power amp and it was too load sensitive. Evidence from the teaching by Mr. Curl and others as well as my Hafler does sound better with the Nak CA5 I bought than a passive. Hard to point to, just "better" If you have ever judged vintage cars, sometimes one is just better. You can't point out why, can't point a flaw on others, just better.

I suspect: symmetry in the square wave to be far more important than the absolute slew rate.

I suspect: More local feedback to be better than more global feedback, IF the output stage is included somewhere and is better to start with. I am not at all adverse to feedback. I just have some sneaking suspicions about a day late and dollar short. Probably totally unfounded.

I suspect: we are not measuring the right things. We need measurements that are more about dynamics, graceful recovery, and tolerance. Not the steady state easy to print in adds numbers. Otherwise the .0000 something amps would always sound better. Not to my ears.

I suspect: the transistors themselves have a much bigger impact than is obvious. They are even more variable than a cap, and caps are a disaster.

I KNOW: Darn near every transistor I see or look up I can't get.

I have observed: a difference in how a driver starts from rest with a pure pulse of several cycles in how with some amps the cone lags the input more than others until it catches up in a couple of cycles. I can not correlate it to any amp spec; DF, power, BW. No match. I believe this to be important.