Say for argument's sake you have an output stage with lots of local feedback. Something like a cathode follower, Schade style feedback, or something fancy. It drives the load (the OPT primary) with a low output impedance.
Supposedly this will also give increased bandwidth through the output transformer thanks to the parasitics being swamped by the low impedance.
Now, what I'm having trouble with is the following:
I guess the gist of my question is: despite the fact that you can apply lots more LOCAL feedback, is it actually worth doing in the real world and is stabilized global feedback just as good?
Supposedly this will also give increased bandwidth through the output transformer thanks to the parasitics being swamped by the low impedance.
Now, what I'm having trouble with is the following:
- Is there a benefit to using loads of local feedback instead of the limited amount you can from the OPT's secondary?
- Is global feedback essentially just as good as long as you take the phase shift into account?
- Does global feedback straighten out the phase shift the same way local feedback would?
I guess the gist of my question is: despite the fact that you can apply lots more LOCAL feedback, is it actually worth doing in the real world and is stabilized global feedback just as good?
Limited amount decreases output resistance by limited amount, but for the same amount turns low order distortions into high order ones that are more audible. According to my experience, feedback should be either zero, or as much as possible limited by terms of stability and phase shifts caused by compensation.
Global feedback decreases distortions caused by phase splitter and driver. In tube amps I usually use both; in hybrid amps with error-corrected source followers I use separate feedbacks, one over driver, another over an output stage.
All depends on the case: topology and components used. No single remedy exist. And of course, the more stages you surround by feedback the higher are phase shifts, the less stable is the amp, the more prone to overload of faster stages preceding slower ones.
Global feedback decreases distortions caused by phase splitter and driver. In tube amps I usually use both; in hybrid amps with error-corrected source followers I use separate feedbacks, one over driver, another over an output stage.
All depends on the case: topology and components used. No single remedy exist. And of course, the more stages you surround by feedback the higher are phase shifts, the less stable is the amp, the more prone to overload of faster stages preceding slower ones.
Local feedback is good for fixing up the output tube stage distortion and tube Zout. Since the output stage is where the worst and most objection-able distortion occurs (particularly for class AB), this is quite helpful. Local fb will also fix the xfmr magnetizing current distortion, and the shunt winding capacitance current problem too, but not the OT leakage inductance problem or winding resistance problem. Leakage inductance will cause some HF falloff. Winding resistance will cause a limit to the damping factor achievable.
Using global fb will in addition fix the earlier stage distortions and the OT leakage L and resistance issues. Giving a higher damping factor and better HF response. Global is more problematic to get much feedback without stability problems though (requiring a better, more BW, low leakage L, OT). And may cause harder clipping (depending on how much fb).
The best combination is to use some of both, this is called nested fb. The local fb reduces phase shift and distortion in the output stage, allowing the global fb to be more easily accomplished (stability), but the OT is still the limiting factor (so only worthwhile with a good high BW OT). Best to just split the fb dBs between local and global, rather than trying to push the total dB fb up further.
If you are trying to preserve "tube sound" though, global is not you baby. It will make the amplifier "clinical" sounding (ie, no euphonic distortion) if much more the 6 or 12 dB global is applied apparently. Local fb has some of this same effect, but since the earlier stages are still uncorrected tubes, not as big a deal. Then again, the OT magnetizing current distortion is often correllated with that "tube sound", and local fb suppresses this. Toroidal OT's (the ones made for VT OTs) don't have much magnetizing current distortion to begin with, so not much effect there. (on the other hand, using power toroidal xfmrs for OTs will usually have gobs of magnetizing current due to insufficient turns)
My $.02
Using global fb will in addition fix the earlier stage distortions and the OT leakage L and resistance issues. Giving a higher damping factor and better HF response. Global is more problematic to get much feedback without stability problems though (requiring a better, more BW, low leakage L, OT). And may cause harder clipping (depending on how much fb).
The best combination is to use some of both, this is called nested fb. The local fb reduces phase shift and distortion in the output stage, allowing the global fb to be more easily accomplished (stability), but the OT is still the limiting factor (so only worthwhile with a good high BW OT). Best to just split the fb dBs between local and global, rather than trying to push the total dB fb up further.
If you are trying to preserve "tube sound" though, global is not you baby. It will make the amplifier "clinical" sounding (ie, no euphonic distortion) if much more the 6 or 12 dB global is applied apparently. Local fb has some of this same effect, but since the earlier stages are still uncorrected tubes, not as big a deal. Then again, the OT magnetizing current distortion is often correllated with that "tube sound", and local fb suppresses this. Toroidal OT's (the ones made for VT OTs) don't have much magnetizing current distortion to begin with, so not much effect there. (on the other hand, using power toroidal xfmrs for OTs will usually have gobs of magnetizing current due to insufficient turns)
My $.02
Last edited:
I should clarify my last paragraph above. I often hear 6 dB of feedback stated as the max for "tube" sound, and anything above 12dB or 20 dB FB as clinical (no euphonics) sounding. In between seems to be the no-mans land that Wavebourn mentions, where noticeable higher harmonics are generated by the FB loop, but not suppressed sufficiently to be un-noticeable.
I think this may need to be put in perspective though, since the feedback is just straightening out the nonlinear characteristics smoothly into straighter curves with more feedback. It's just that 3/2 or square law devices mostly produce solely (significant) 2nd harmonic distortion (which seems to be acceptable to the ear), so as the curves do straighten out, other harmonics naturally show up until an absolutely straight line results. (and the harmonics are totally suppressed)
I think this may need to be put in perspective though, since the feedback is just straightening out the nonlinear characteristics smoothly into straighter curves with more feedback. It's just that 3/2 or square law devices mostly produce solely (significant) 2nd harmonic distortion (which seems to be acceptable to the ear), so as the curves do straighten out, other harmonics naturally show up until an absolutely straight line results. (and the harmonics are totally suppressed)
According to my experience "clinical" and "pristine clean" sounds are very different. The term "Clinical" is mostly used for amps that have very precise measurement results, but doe not sound clean. "Pristine clean" amps are those that fool imagination as if they don't exist. They may measure worse than "Clinical" amps, but sound is different. For example, my Pyramid amps have -80 dB of 2'nd harmonic on half (40W) of max power, the rest is invisible. With lower power that 2'nd order product goes even further down. Some SS amp I compared it with have lower 2'nd order product, but it's THD goes up with lower power, and higher order harmonics dominate. No wonder, it sounds "Sterile", while Pyramid fools imagination, as if sounds are real. However, not so real like in Tower amps that don't have output transformers, instead of them they have error-corrected source followers.
My experimental class A+C amps with nested feedbacks (quite complex arrangement) sound better than similar amps in class AB. I gave up with class AB SS outputs for now, it does not work better than class A+C. For class AB I use tube outputs.
My point is, it is not a feedback that determines whether the amp sounds "Sterile" or "Pristine clean", it is it's arrangement. I.e. topology and components used.
My experimental class A+C amps with nested feedbacks (quite complex arrangement) sound better than similar amps in class AB. I gave up with class AB SS outputs for now, it does not work better than class A+C. For class AB I use tube outputs.
My point is, it is not a feedback that determines whether the amp sounds "Sterile" or "Pristine clean", it is it's arrangement. I.e. topology and components used.
Another potential pitfall of local feedback I haven't seen mentioned yet is that if you are not careful, the amp can end up producing more distortion than open loop. That is because stages with local feedback place more demand on the preceding stage(s). A cathode follower requires much more voltage swing. Schade style feedback forces the driver to work into a much steeper AC loadline.
It is typically only useful up to a point. Beyond that you need to add global.
It is typically only useful up to a point. Beyond that you need to add global.
Last edited:
An amplifier which gets worse distortion as the signal decreases would be "defective" in my opinion, unfortunately this is commonly the case for many commercial SS bipolar amps with poor bias control. One of the reasons I like a Lateral Mosfet amplifier over a bipolar one, much better/reliable bias and crossover control, even though a bipolar SS one can surpass it in theory - when it's correctly adjusted and correctly thermally tracking (not too often). ThermalTrak like bipolar outputs may finally solve that problem.
The pitfall of local Schade feedback (as usually implemented back to the driver plates, presenting a low Z input) is worth noting. Significant design hurdles have to be met to get linear current output from the driver. The less commonly (at least lately) local feedback to the driver grids (Citation II) or cathodes (RCA SP-20 or the RC-19 handbook 6V6 or RC-30 7027A amps come to mind.) is a simpler approach to get right.
The pitfall of local Schade feedback (as usually implemented back to the driver plates, presenting a low Z input) is worth noting. Significant design hurdles have to be met to get linear current output from the driver. The less commonly (at least lately) local feedback to the driver grids (Citation II) or cathodes (RCA SP-20 or the RC-19 handbook 6V6 or RC-30 7027A amps come to mind.) is a simpler approach to get right.
I failed to find the way to cure "pure bias control" in class AB SS amps. Class AC (i.e. combination of class A with stable bias and class C with no bias at all) is much better. I believe it is not popular because of Walker's (Quad) patent that scared designers of it.
You may laugh, but in my previous post I mentioned comparison of Pyramid with such an amp. 😀
...and we discussed it thoroughly in several threads. Again, local feedbacks (long tails of driver tubes) helps to cure this problem as well.
Edit: I agree that amps with feedback clip worse than amps with no feedback. The cure I've found is to use fast attack optical compressors. For example, in Pyramid amps I sense screen grid currents of GU-50 output tubes in order to find when the amp approaches clipping.
Last edited:
The bipolar output stage (EF) has an optimum bias/current for minimum crossover distortion that is perilously close to notch-over. The Lateral Mosfet (SF) stage does not, it just gets better the higher the idle current. So it's just a matter of adjusting the idle current up enough until the FFT looks good enough (or it sounds good enough). This will definitely require more idle power than the optimum bipolar setup, but then it is no longer critical to adjust or subject to thermal drift off of the optimum. This does require more heatsink area, which is a cost issue, particularly for a commercial design.
Broskie mentioned a rather interesting fix for crossover distortion some while back (2nd diagram up from the bottom):
European Triode Festival and Crossover Notch Distortion and New OTL Design
A number of old patents use variations on it. Notice how the odd tapped off feedbacks transition voltage-wise during crossover. Broskie mentioned a 20 dB dist. improvement in simulation.
He then gave a tube version for a totem pole output configuration (not too useful). But it is actually quite easy to re-arrange this for a conventional P-P output using a center tapped OT with Schade feedbacks to the driver cathodes or crossed grids (diffl driver). I will leave this as an exercise to derive it. (I wouldn't want to spoil an interesting puzzle 🙂 , hint: only takes two additional resistors somewhere )
Broskie mentioned a rather interesting fix for crossover distortion some while back (2nd diagram up from the bottom):
European Triode Festival and Crossover Notch Distortion and New OTL Design
A number of old patents use variations on it. Notice how the odd tapped off feedbacks transition voltage-wise during crossover. Broskie mentioned a 20 dB dist. improvement in simulation.
He then gave a tube version for a totem pole output configuration (not too useful). But it is actually quite easy to re-arrange this for a conventional P-P output using a center tapped OT with Schade feedbacks to the driver cathodes or crossed grids (diffl driver). I will leave this as an exercise to derive it. (I wouldn't want to spoil an interesting puzzle 🙂 , hint: only takes two additional resistors somewhere )
Theoretically optimal bias exists, but practically you should apply sinusoidal signal, wait while temperatures settle up, then adjust bias... But don't turn of your sinusoidal signal, or you loose balance! 😀
Sure, local feedback across output transistors helps a lot. There were many solutions, some of them patented. But still, the best I could come up with, I presented some time ago (complementary 2-stage opamp with unity gain).
Sure, local feedback across output transistors helps a lot. There were many solutions, some of them patented. But still, the best I could come up with, I presented some time ago (complementary 2-stage opamp with unity gain).
It's not hard to prove that local NFB works better:
Consider an amplifier with a given gain-bandwidth product. It might be a perfect integrator (i.e., infinite DC gain, dropping to 1 at some frequency), or it might have finite low-pass gain. Op-amps are reasonable integrators in the audio band, thanks to the prevalence of dominant-pole compensation. The typical op-amp has a gain of perhaps 100dB (i.e., 10^5) at DC-50Hz, then drops at -20dB/decade until fT is reached at 50Hz * 10^5 = 5MHz.
With a feedback loop, you can divide the GBW however you want. If you want at least 50kHz bandwidth, you can get only 5MHz/50kHz = 100 gain.
If you chain two such amplifiers, you get twice the gain (i.e., 200dB at DC), and the same fT, but twice the phase shift. You can't even put NFB around this loop, it will oscillate (in principle, at *any* frequency)! The only useful range for this amplifier is before the cutoff frequency (i.e., 50Hz), because the phase shift is low there. If you apply enough NFB and compensation so that gain < 1 before phase shift hits 180 degrees (which is somewhere past cutoff, since the angle is exactly 90 degrees at cutoff, which is still okay), you can get it stabilized, but now your bandwidth is clearly so much less than expected.
This serves as an extreme example: chained stages, transinfinite gain and monstrous phase shifts are not useful, but you already knew that.
If, instead, you use local feedback around each amplifier, you get a series of amplifiers, each with moderate gain, and zero phase shift through the passband. The phase shift increases substantially past the passband, so you can still apply dominant-pole compensation to this system as long as gain drops below 1 at that point. For example, if you wanted 100kHz bandwidth, you can still get 2500 gain from two op-amps (each stage wired for 50), which is almost as much as you get from a single mediocre op-amp.
If you go with a *lot* of stages, you get a serious amount of phase shift at the end, at cutoff, making global feedback more difficult. This compromises the overall system bandwidth. Fortunately, the dominant element in a tube amp is always the output transformer, so you're looking at accommodating that, and driving it without causing distortion.
Note that the analysis works on both ends. If you have a bunch of stages with coupling capacitors, the phase shift and LF cutoffs will add up! This makes the Williamson topology rather comical: it seems to have been intentionally designed as a power phase-shift oscillator!
Tim
Consider an amplifier with a given gain-bandwidth product. It might be a perfect integrator (i.e., infinite DC gain, dropping to 1 at some frequency), or it might have finite low-pass gain. Op-amps are reasonable integrators in the audio band, thanks to the prevalence of dominant-pole compensation. The typical op-amp has a gain of perhaps 100dB (i.e., 10^5) at DC-50Hz, then drops at -20dB/decade until fT is reached at 50Hz * 10^5 = 5MHz.
With a feedback loop, you can divide the GBW however you want. If you want at least 50kHz bandwidth, you can get only 5MHz/50kHz = 100 gain.
If you chain two such amplifiers, you get twice the gain (i.e., 200dB at DC), and the same fT, but twice the phase shift. You can't even put NFB around this loop, it will oscillate (in principle, at *any* frequency)! The only useful range for this amplifier is before the cutoff frequency (i.e., 50Hz), because the phase shift is low there. If you apply enough NFB and compensation so that gain < 1 before phase shift hits 180 degrees (which is somewhere past cutoff, since the angle is exactly 90 degrees at cutoff, which is still okay), you can get it stabilized, but now your bandwidth is clearly so much less than expected.
This serves as an extreme example: chained stages, transinfinite gain and monstrous phase shifts are not useful, but you already knew that.
If, instead, you use local feedback around each amplifier, you get a series of amplifiers, each with moderate gain, and zero phase shift through the passband. The phase shift increases substantially past the passband, so you can still apply dominant-pole compensation to this system as long as gain drops below 1 at that point. For example, if you wanted 100kHz bandwidth, you can still get 2500 gain from two op-amps (each stage wired for 50), which is almost as much as you get from a single mediocre op-amp.
If you go with a *lot* of stages, you get a serious amount of phase shift at the end, at cutoff, making global feedback more difficult. This compromises the overall system bandwidth. Fortunately, the dominant element in a tube amp is always the output transformer, so you're looking at accommodating that, and driving it without causing distortion.
Note that the analysis works on both ends. If you have a bunch of stages with coupling capacitors, the phase shift and LF cutoffs will add up! This makes the Williamson topology rather comical: it seems to have been intentionally designed as a power phase-shift oscillator!
Tim
That depends. There are some drawbacks to local NFB, that have already been addressed: cathode follower types require larger than normal input drive voltages, and you could lose as much, if not more, linearity in the front end that you gain with the cathode follower finals.
If you use parallel local NFB ("Schade" NFB, as they call it here) you have the problem of causing the AC input impedance of the output stage to drop (same as Miller Effect that magnifies the reverse transfer capacitance). That can force the driver to work into a nastily steep loadline, so there are limits to how much you can apply.
In a design I did that used 807s, local NFB is needed, and was recommended by O. Schade himself, as the designer of this type. The 807 will make more of the objectionable higher order harmonics that sound really nasty if running 807s open loop. Schade recommended feeding back 10% of the output plate voltage back to the input. This recommendation turned out to be spot-on.
6.0db(v) of gNFB was all that was required to clean up the sound.
Other types (6BQ6, 6V6) make mainly h3. These types don't require any local NFB, since they sound fairly clean open loop. For a project that used 6BQ6s, ~6.0db(v) of gNFB made for a nice sound with hard rock and Techno. 13db(v) was definitely tending to a "solid state" sound that made hard rock sound "subdued".
If you do the open loop design right, you really don't need esssssssss-loads of NFB, and over doing the NFB (global or local) will make for a solid statey sound, and not good solid state either.
- Status
- Not open for further replies.