Oh man, now why'd you have to go and post that link? Who can look at that Marantz receiver and not be depressed at what our gear looks like today???
16 responses Add your response
I am not an EE but the part I found interesting is the point indicating feedback was not done well or completely with early SS gear and got a bad rap but improved greatly in the 90s on the back of A/D D/A technology and carried over to Class D amp design.
That rings true with my understanding of what feedback done right is supposed to sound like and the sound I have observed with various amps and designs over the years.
The best feedback was when you would turn the knob and it had that nice oily smooth feel with the little clunk. So much better than the click you get nowadays. Then again the big tuning knob, the mass that would let you really feel you were doing something, and if you got it going fast would keep spinning so that with a little practice you could give it a swipe and across the dial it goes gliding smoothly to right where you want it.
The knob or the dial, which provided the better feedback? I don't know. Both were great.
Nobody with a solid working knowledge of feedback theory chooses to avoid it. “I choose to avoid feedback” is an all too transparent code for “I’ve yet to figure it out”.The above is from the article. Its not exactly correct, although the gist of the article is.
We don't use feedback in our designs simply because so far we prefer to use tubes. Its almost impossible to add enough feedback in a tube amplifier due to all the frequency poles which limit phase margin and decrease Gain Bandwidth Product. And since not enough feedback is actually worse than none at all, being pragmatic, that's the direction we went. But if we could actually add enough (35dB being an absolute minimum to any design) then we would.
So actually what's been happening is most designs using feedback have not used enough, rather than 'too much'. When the feedback is insufficient, it is unable to cause the amplifier to compensate for the distortion added by the application of feedback itself. The distortion caused by the application of feedback (from Norman Crowhurst, so this has been known since the early 1960s) adds a host of higher ordered harmonics as well as intermodulations at the feedback node while suppressing some of the amplifier's innate distortion. This results in the noise floor of the amplifier being harmonic and inharmonic information rather than just a gentle hiss. The addition of the higher ordered harmonics causes the amplifier to sound bright and harder since the ear assigns a tonality to all forms of distortion and is keenly sensitive to higher orders since it uses them to discern sound pressure. I've seen a lot of people have problems with this fact over the last 40 years, it helps to understand the ear's sensitivity in that it has a 125-130dB range.
That is why nearly all solid state amps until very recently have been bright; certainly **all** solid state designs from Japan in the 1970s to possibly about 5 years ago. So in a way, the people criticizing the author's Pioneer for having too much feedback might have been right; if you can't add enough in your circuit you might be better off with none at al!
But if you can add enough feedback, and by that I mean in excess of 35dB then the circuit has the chance of being able to compensate for the distortion added by feedback: now you have a circuit that is actually really low in distortion. It has to be really low because the higher ordered harmonics are so easily heard by the human ear- 0.005%THD is easily heard if its all higher ordered harmonics!
While a lot of solid state advocates talk about neutrality, its hardly 'neutral' if its bright. That's a more heinous distortion than the 2nd or 3rd order (to which the ear is not that sensitive) by far!
Now some designers have taken a different approach, sculpting the distortion signature to include the 2nd and 3rd harmonics. Its been shown that if these lower orders are present (both treated by the ear as 'richness', 'bloom', etc.) in great enough degree, they can somewhat mask the presence of the higher orders. Nelson Pass takes this approach in his designs, which are some of the best sounding solid state amps out there. I've found certain early solid state designs that have this property as well although in the case of the latter it does not appear to be with intention.
Self-oscillating Class D amps appear to have the ability to run the level of feedback required. In fact such amps use the high amount of feedback as the means to set the switching speed. Its tricky- you have a lot of variables but in a nutshell the feedback loop is set with enough feedback that the amp exceeds its own phase margins and goes into oscillation. The loop then has timing constants within it that cause the oscillation to settle on a constant 'solution'; a constant switching speed. Since its easy to make a lot of gain in a class D amp, you can easily have over 40dB of gain and still have 25-30dB left over for actual amplification.
Now some designers have taken a different approach, sculpting the distortion signature to include the 2nd and 3rd harmonics. Its been shown that if these lower orders are present (both treated by the ear as 'richness', 'bloom', etc.) in great enough degree, they can somewhat mask the presence of the higher orders. Nelson Pass takes this approach in his designs, which are some of the best sounding solid state amps out there.https://www.passlabs.com/press/audio-distortion-and-feedback
This is an article Nelson Pass originally wrote and was published by 6Moons a while back which goes into more detail regarding Ralph's comments on Nelson's use of feedback. It also talks about implementation of feedback, both global and local.
There is another article elsewhere which I cannot find at the moment in which Nelson mentions that small amounts of feedback, the amount of which I cannot remember are fine, as are large amounts in excess of 30dB. So again, there seems to be some consistency in this line of thought that excessive amounts of feedback could be beneficial. I know that the last amplifier designed by Roger Modjeski used greater than 35dB of feedback, and as Nelson himself has mentioned regarding his First Watt designs the application of global negative is determined on whether or not the amplifier sounds better with it or not.
Properly implemented feedback is not a bad thing. I have some amplifiers that use it and some that don't and enjoy then all.
And Roger’s ( RIP ) masterwork RM-9 has switched 3 levels of feedback which aids one in discerning the impact on sonics.
certainly there are amplifiers that sound good with no global feedback as well: Atmasphere, Ayre ( RIP Charlie), Vandersteen, et...
and of course we should credit Dr Matti Otala for his discovery of TIM and relationship to negative F/B. I still have an Audionics CC2 wildly inspired by Otala. It is a very musical SS amp designed and produced in the late 70’s
fatigue free listening.....
@cleo I sent you an email, need bit of help with bringing an RM-4 back to life
Great posts above; thanks to all.
Ralph (@atmasphere), what do you think of Mr. Putzeys’ statement that ...
... the idea that feedback causes TIM is probably most noteworthy for being not just wrong, but also the exact opposite of the truth. TIM happens in the input stage. An increase in global feedback makes the input stage work less hard. That causes a disproportional reduction in TIM.
Thanks. Best regards,
@tomic601 the idea of the feedback switch on the RM-9 was not about adjustable feedback, it was about adjustable gain (Nelson Pass talks about the feedback/gain relationship in the article as well), which is why it is referenced as a gain switch. Roger wanted users to have a means to use both passive and active preamps with the RM-9, so he put that feature in to allow the gain to be adjusted accordingly. However, after more questions from users about which position sounded better than he wanted to deal with (Roger's stock answer was, "whatever position sounds best to you") he removed the switch and the feedback/gain was factory set. The RM-10, while not having a switch, has instructions and a chart in the manual for adjusting the gain by swapping out some resistors and capacitors.
Here is another little snippet from Roger. This time on IMD which he felt was very important to limit in circuit designs:
"I intend to write a paper on the more complete story of 2nd harmonic distortion theory. I will say here that it must be minimized (but not at the expense of adding other distortions). What listeners don't seem to realize is that large amounts of 2nd is fine for a single note but not for a full orchestra. As I like to say "Please tell me what is the second harmonic of Beethoven's 9th? The amplifier is having a little trouble figuring it out". Once music becomes a complex signal, the second harmonic argument goes right out the window. The amplifier can only act upon the instantaneous voltage it is given and has no idea what notes are being played. At that point, it's Intermodulation Distortion we had better be taking care of. How many of us have heard our single-ended amps make a mess of a symphony yet play a solo voice beautifully?"
Tony I certainly appreciate the listen and pick which switch position sounds best.... and Lord knows, my thinking on any subject evolves a bit with time but....
this is from the RM-9 owners manual, which includes RM note back to Stereophile in the review of the RM-9
” Not wanting to play “ God” over the situation, I chose to give the user a three position feedback switch so he could change the damping by a factor of approximately 4:1 “... he then mention gain, passive etc
in my system with what I consider to be a very neutral tweeter, I prefer the high gain / low feedback setting.... it sounds glorious...
Modjeski got his hands on a Futterman OTL amp, looked into it's design, and discovered that Julius employed an extremely large amount of negative feedback. A search will bring up more details on Roger's thoughts on the Futterman, negative feedback in general, and their relationship in different amp designs, including of course OTL's (he designed and built a number of them).
@bdp24, the Futterman H3 which Roger referenced for his last OTL project used an incredible amount of feedback, but even more interesting were the large number of 1" ferrite beads inside the amp that Julius omitted from the schematic.
I wish Roger had written more about the use of feedback in amplifier design. His knowledge of the subject was in my opinion better than that of transformer design and tubes. Roger technically only manufactured one OTL amplifier, the OTL-1 which saw very limited production. The RM-6 design was purchased by Counterpoint and became the SA-4. Contrary to popular belief Roger did not design the Beveridge OTLs, although he did QA and test them. Here is a link to the story of the OTL-1: https://www.audiocircle.com/index.php?topic=126907.0
If you scroll down to the Circuit Description you can read some of Roger’s thoughts on feedback. A portion of which I will quote here:
In the OTL-1 the input tube literally rides on the output terminal. So as the input rises the output follows it perfectly in phase. Rather than being an injection point for feedback this point exists in the fundamental circuit. This allows the amplifier to react immediately. In conventional amplifiers the feedback comes through the output transformer (with considerable phase shift that is load dependent). The feedback then needs an injection point which is often the cathode of the input tube or grid of a differential amplifier. Internal delays in the loop (phase shifts at low or high frequencies) can cause the negative feedback to become positive and make the amp oscillate at low or high frequencies or both. Many amps are unstable without a load and many have low frequency instability that causes the woofer cone to wander about its rest position.
Ralph (@atmasphere), what do you think of Mr. Putzeys’ statement that ...
... the idea that feedback causes TIM is probably most noteworthy for being not just wrong, but also the exact opposite of the truth. TIM happens in the input stage. An increase in global feedback makes the input stage work less hard. That causes a disproportional reduction in TIM.@almarg Bruno is spot on here. The problem is that certain elements of the amplifier design can lie outside the feedback loop, but nonetheless be affected by the operation of the feedback. When you're dealing with something like this you get distortion.
Feedback has mostly been used in the last 40 years to decrease output impedance so that the amp can behave as a voltage source. That is why in most cases there is brightness/harshness that isn't part of the original signal since the feedback needed to get the amp to behave this way is always going to be less than what is needed for the feedback to have the ability to correct for distortion it adds.
One issue that's not seen a lot of attention is the simple fact that a speaker does not need more than about 20:1 for damping and many need considerably less. If overdamped, you get a phenomena often called 'tight bass' which is pleasurable but a coloration. So right here you can see that adding the feedback that is really needed for the amp to do its job right is overkill for the needs of the loudspeakers.
I've seen nothing to address this; audio is still a developing field.
Thanks, Ralph ( @atmasphere ). So would it be fair to say that feedback can cause TIM if the amount of feedback that is applied is insufficient to prevent it, but in many designs it is not practical to apply that much feedback "due to all the frequency poles which limit phase margin and decrease Gain Bandwidth Product." (Quoting from one of your earlier posts).
If so, I believe that would reconcile Mr. Putzeys' statement with the seminal work of Dr. Matti Otala, mainly in the 1970s, that was referred to by @tomic601 in an earlier post.
Thanks! Best regards,