Re: Re: VAS buffer
Hi Terry,
Do you think that also the harmless looking circuit below might have stability issues?
My latest design will be published on my website, probably in a couple of weeks. It don't use Cherry's NDFL or Hawksford EC, rather a lot of TMC stuff (and NFB of course).
Cheers, Edmond.
Terry Demol said:Bootstrapped buffer - a nice move, but can play games with
stability
What is your latest design Edmond?
cheers
Terry
Hi Terry,
Do you think that also the harmless looking circuit below might have stability issues?
My latest design will be published on my website, probably in a couple of weeks. It don't use Cherry's NDFL or Hawksford EC, rather a lot of TMC stuff (and NFB of course).
Cheers, Edmond.
Attachments
Edmond Stuart said:
You win. My amp did only 0.01%, but that was in 1973.
That's what my monster 20 pair output stage with 5A bias sims at 100W rms into 8 ohms open loop
My front end THD-20 with an ideal output stage sims a little higher than 700ppb with the complementary EF VAS buffer added.
It should be noted that many competent engineers can meet almost any measurement spec demanded of them.
My 'rule of thumb' for Parasound is: 100V/us or more, and .1% distortion at full output, if possible.
We established 100V/us, many decades ago, and most 'low TIM' amps will do this. Mine have done this for the last 35 years. The Otala designed 'Electrocompaniet' circuit also did it more than 35 years ago.
In fact, for a 100W amp, only 50V/us is deemed adequate by most experts on TIM.
In the early '80's with the introduction of ring emitter transistors, the efforts that we extended to make 100V/us virtually gave us 200-1000V/us without too much change. However, we still were using output coils, and without them, we found that 100V/us was still easily possible and the output coil could be removed, improving high frequency damping and potential ringing.
As far as distortion is concerned, that is usually related to the amount of feedback used, all else being equal, or extensive use of cascode circuitry to remove nonlinear capacitance effects. This is possible, but to what end? Better bragging specs? Better sound?
Many designers have found that high feedback sounds worse, all else being equal, than low feedback. Why, is still an open question. Therefore, I use as little feedback as possible to meet THX specs and no more. Sound quality is more important to me than better measurement.
My 'rule of thumb' for Parasound is: 100V/us or more, and .1% distortion at full output, if possible.
We established 100V/us, many decades ago, and most 'low TIM' amps will do this. Mine have done this for the last 35 years. The Otala designed 'Electrocompaniet' circuit also did it more than 35 years ago.
In fact, for a 100W amp, only 50V/us is deemed adequate by most experts on TIM.
In the early '80's with the introduction of ring emitter transistors, the efforts that we extended to make 100V/us virtually gave us 200-1000V/us without too much change. However, we still were using output coils, and without them, we found that 100V/us was still easily possible and the output coil could be removed, improving high frequency damping and potential ringing.
As far as distortion is concerned, that is usually related to the amount of feedback used, all else being equal, or extensive use of cascode circuitry to remove nonlinear capacitance effects. This is possible, but to what end? Better bragging specs? Better sound?
Many designers have found that high feedback sounds worse, all else being equal, than low feedback. Why, is still an open question. Therefore, I use as little feedback as possible to meet THX specs and no more. Sound quality is more important to me than better measurement.
G.Kleinschmidt said:
Keep working on it. My complete open loop amplifier (real, not simulated) has 0.015% into 8 ohms at 100 watts. And that is with only 4 paralleled devices each running at 80 mA bias. (Sixteen total output devices in a bridged configuration.) So basically equivalent performance but with 60% fewer output devices and 15x less idle current....
Re: Re: Re: VAS buffer
It depends where and how it is used.
Inside a feedback loop there could be issues, as I can confirm I have
used this circuit myself and other variations of it.
Another more stable variation is join the collectors Q1/Q2 together
and from this point connect a damping resistor to 'out'. This will
resistively isolate collectors of Q1/2 from emitter of Q3/4 which
can be a good thing.
Yes, the connection scheme I described will loose some VCE on Q1/2
but usually Q1/2 in these configurations are run at very low currents
so it is not a big issue.
However as you have drawn, the simple stand alone follower should
be fine depending on bias current and load (reactive?).
Ok, so is this the simpler 'PMP' version of the high spec 'PGP'?
cheers
Terry
Edmond Stuart said:
It depends where and how it is used.
Inside a feedback loop there could be issues, as I can confirm I have
used this circuit myself and other variations of it.
Another more stable variation is join the collectors Q1/Q2 together
and from this point connect a damping resistor to 'out'. This will
resistively isolate collectors of Q1/2 from emitter of Q3/4 which
can be a good thing.
Yes, the connection scheme I described will loose some VCE on Q1/2
but usually Q1/2 in these configurations are run at very low currents
so it is not a big issue.
However as you have drawn, the simple stand alone follower should
be fine depending on bias current and load (reactive?).
Ok, so is this the simpler 'PMP' version of the high spec 'PGP'?
cheers
Terry
Charles Hansen said:
Agreed.
Unfortunately just trying to drive 20 x OP devices is placing
constraints on the design that will limit it's ultimate performance.
I also agree with your design in as far as 4 to 5 OP devices per rail /
phase of the type you have used seems to be the sweet spot for
getting transconductance up but keeping drive requirements to a
minimum.
T
Re: Re: Re: VAS buffer
I've used a version of this with some success. It is the circuit that made me concerned about the capacitance of the Baker clamp diodes, as the Ccb capacitance of the transistors loading the VAS in this circuit is largely bootstrapped out. In some cases I've put zeners in the emitters of the input pairs to get the desired amount of bias spread, and tied the input bases together. This makes a good circuit to drive the HEC circuit I use.
Cheers,
Bob
Edmond Stuart said:
I've used a version of this with some success. It is the circuit that made me concerned about the capacitance of the Baker clamp diodes, as the Ccb capacitance of the transistors loading the VAS in this circuit is largely bootstrapped out. In some cases I've put zeners in the emitters of the input pairs to get the desired amount of bias spread, and tied the input bases together. This makes a good circuit to drive the HEC circuit I use.
Cheers,
Bob
john curl said:
I think we would all agree with that, John.
I'm glad to see that you are at least no longer blaming TIM on large amounts of NFB. Good measurements are not for bragging rights, but in my view they do leave less room for bad stuff. Your 0.1% may happen to be very clean, but 0.1% can really also be VERY dirty. There is much less room for dirt in an amp that performs to 0.01% or 0.001%. Such performance may help us to get there, but I agree that it is neither necessary nor sufficient for a truly good-sounding amplifier. By that I mean that there are really great sounding amplifiers that have a clean 0.1% THD, and that there are also bad sounding amplifiers that have only 0.01% THD.
That does not, however mean that THD, as one of many metrics, is not useful. As you know, I do not particularly like single-number THD, but prefer THD-20 with spectral analysis out a decade (unfortuantely many THD analyzers stop at about 80 kHz), or CCIF with spectral analysis.
IF high feedback sounds worse (I don't think it does), all else remaining equal (it usually is not, unfortunately), then I do agree with you that we do not know why. It is not related to something we have been able to measure, including PIM, TIM and IIM. It has been demonstrated repeatedly that high amounts of NFB do not exacerbate these distortions.
The hand-waiving arguments that we sometimes see about negative feedback, in its delay in traveling around the loop, does not have time to correct distortion that has already occurred, are just some of the bad science that has given NFB an undeserved bad rap. To the best of my knowledge, Otala did not resort to such silly arguments. To his credit, Otala proposed tests for most of the distortions he described, and ultimately these tests showed that increased amounts of NFB, when properly applied, did not exacerbate these distortions.
Cheers,
Bob
Charles Hansen said:
Hi Charles,
BTW, I got to see your MX-R in Montreal last weekend for the first time in the flesh. Sweet!
Your point here raises a very interesting question. Does Class-A have any sonic advantage over a well-excuted Class AB amp that has equally good measured performance?
Cheers,
Bob
john curl said:
If open loop bandwidth is the real subjective factor, and if low open loop bandwidth causes an amplifier to sound worse (again, very big IFs in dispute), then its effect on the sound is not something we have been able to measure, as far as I know.
Low open loop bandwidth does not exacerbate TIM, PIM or IIM.
I am not the only person who has shown this to be true.
It is also not that difficult to design an amplifier with 40 dB of NFB and open loop bandwidth out to 20 kHz, if that is what one likes.
Cheers,
Bob
Terry Demol said:
Yes, but my output stage is actually rated at 800W rms continuous into 1 ohm, 1600W onto 0.5 ohm and 100A peak for 10mS. We are comparing apples to oranges here.
That (and the 5A bias) is why it needs 20 pairs of output devices, and 20 pairs are significantly harder to drive than 4 or 5.
Also the ON semi spice models for the driver transistors are bloody awfull, so that doesn't help with the simmed THD figures either.
If I substitute the MJE150XX driver transistors for low power Fairchild devices, for which I have very good models (you can do this in spice becuse the transistors don't blow up
), this is what I get for THD-20 at 100Wrms into 8 ohms:Harmonic Frequency Fourier Normalized Phase Normalized
Number [Hz] Component Component [degree] Phase [deg]
1 2.000e+04 3.991e+01 1.000e+00 0.21° 0.00°
2 4.000e+04 5.343e-04 1.339e-05 9.13° 8.92°
3 6.000e+04 1.470e-03 3.682e-05 28.08° 27.87°
4 8.000e+04 8.016e-05 2.008e-06 -175.73° -175.94°
5 1.000e+05 1.419e-04 3.555e-06 -121.29° -121.50°
6 1.200e+05 1.725e-05 4.323e-07 3.54° 3.33°
7 1.400e+05 2.493e-05 6.246e-07 79.25° 79.03°
8 1.600e+05 3.585e-06 8.983e-08 -175.96° -176.17°
9 1.800e+05 5.305e-06 1.329e-07 -86.17° -86.38°
Total Harmonic Distortion: 0.003940%
So, I don't think I really have to try much harder. Combined with my 200V/us 720ppb simmed front end, this OPS is going to kick ***. I got my alumimum channel on Thursday too, so I will start drilling and tapping the mounting holes for the trannies tomorrow afternoon.
And WRT to bias current levels, with higher bias you trade large signal linearity for small singal linearity, as outlined by D. Self.
Cheers,
Glen
Bob Cordell said:
No dispute if you just listen.
All of our designs are open loop, so that's all I ever listen to. Our all-FET V-1 has a bandwidth of 100 kHz. The mostly bipolar V-5 has a bandwidth of 250 kHz. When you listen to them, they have a similar tonal balance.
Based on the preceding paragraph, one would logically conclude that once a certain threshold is exceeded that extending the bandwidth wouldn't make any audible difference. When we were developing the mostly bipolar MX-R I made a prototype that had a bandwidth of 150 kHz. To my surprise, it sound dull and rolled-off in the top octave. I extended the bandwidth to 250 kHz, and now it had a similar tonal balance to the other two amplifiers.
I have no explanation why this degree of bandwidth would be audible. I have no explanation of why a certain bandwidth with a bipolar design would sound duller than a lesser bandwidth in an all FET design.
But as usual, the amp doesn't care what I think or whether I understand it. It just sounds how it sounds.
Bob Cordell said:
Hmmm. That's not the conclusion that John Atkinson came to last week:
"I auditioned the amps with no negative feedback and just 1.2dB of negative feedback. You wouldn't have thought it would make a difference, but darned if switching in even this minimal amount of feedback—which, in theory, should make the amplifier perform better—didn't diminish the enormous sense of space on the recording."
http://blog.stereophile.com/fsi2008/040708lamm/