John Curl's Blowtorch preamplifier part II

[jan.didden]

Originally Posted by john curl

While on this topic, I would like to (once again) relate some audio history, where we got started, where we went wrong, and then got right again.
The history of TIM comes from 2 separate inputs. First, an accident where a power amp was put together wrong and it sounded BETTER than the correct design, based on op amp theory. This amp apparently had two changes: Higher slew rate and higher open loop bandwidth. In any case, this stuck with a young Matti Otala.
Second, in 1966 (a long time ago) a young graduate student Daughority(sp) and Dr. Greiner (well known, even today) published in the IEEE 'Audio' publication that HIGH OPEN LOOP was necessary for low distortion audio, even class A audio. I read this at the time and was impressed by the mathematical effort to show this. Still, I did not really understand it.
This IEEE article was cited by Matti Otala in his first papers, and initial efforts on his part showed that a high open loop bandwidth that tended, as well to create a high slew rate, as well, power amp sounded better.
However, before Jan L. and Matti O. built their amplifier design, I was designing complementary differential, pp throughout, complementary output power amps for the Grateful Dead group. I, then in 1971, 5 years after the D&G paper, was asked by a colleague then working at LLL, how to make a faster slew rate power amplifier. He needed on for a project. I thought about it, and even with my experience of making amplifiers from 10W-2000W in both 2 quadrant and 4 quadrant symmetry, I could NOT think of a better way to speed up my amps, except to use faster output devices, which would increase the gain-bandwidth, the open loop bandwidth, and the slew rate in proportion. He came back to me later, that somebody HAD made a really fast amplifier that overcame all his efforts, but they were not saying how. NOW, we can presume that they JUST degenerated the Gm of the input stage in order to increase the slew rate, without increasing the G-B, or the open loop bandwidth. What a breakthrough! At least it had not been OBVIOUS to me, at the time. Within the year, attending advanced engineering classes at UCB, the equations were shown, and all was clear. However, please remember, my engineering colleague at LLL did not have access to this knowledge in early 1971, or else he would have not have had to ask me what to do. It was just not that obvious in those days, what is so obvious now. (more later)

Well that's a nice story, but I have a 1954 electronics book that explains the concept of slew rate in amplifiers and how to design for high SR. Written by a US Naval Captain who taught at the Naval Tech University. What were you guys doing 20 years later???

So if you guys say things were not obvious probably means you read the wrong books

Anyway, we know that 'just degenerating the input stage' does NOT increase SR. Increasing the Iq in the input stage does. However, that might require a larger comp cap due to the increased transconductance, undoing the rise in SR you try to achieve. Fortunately, you could degenerate the input stage to lower transconductance again so you got your increased SR.

Agreed?

jan didden

[jan.didden]

Originally Posted by scott wurcer

NO, NO, and NO they will have the same frequencies. There are people here I assume that are interested in some of the basic principles of engineering. John just ask directly THIS question to Ron Quan and his advisor Tom Lee (an old friend) they will set you straight.

Volterra analysis extends to arbitrary non-linear systems even those with memory. The output with input at f1 and f2 can ONLY be at n*f1+-m*f2. Simple amplifiers do not fall into the non causal non time invariant category.

I'm not an expert on PIM, but I have always wondered how something that modulates the PHASE of a freq component could lead to additional harmonics?

jan didden

[john curl]

Jan, please look at Barrie Gilbert's paper for this info.

[john curl]

Moving on, Otala had published some theoretical papers starting in 1969 that promoted high open loop bandwidth and this tended to give high slew rate. BUT slew rate limiting is NOT TIM, it is the clipping that is caused by the TIM producing cap charging, the cap current being non-linear. In fact, a high slew rate op amp can have lots of TIM at much lower levels, IF a class B input stage is used. This is what Matti Otala wanted to include: The exceptions to the general rule, for example the uA741S, a high slew rate 741. Find one and measure it. You will find high TIM(SID) at moderate operating levels. So much for slew rate being the ultimate and complete measurement necessary. (more later)

[Bob Cordell]

Originally Posted by janneman

Well that's a nice story, but I have a 1954 electronics book that explains the concept of slew rate in amplifiers and how to design for high SR. Written by a US Naval Captain who taught at the Naval Tech University. What were you guys doing 20 years later???

So if you guys say things were not obvious probably means you read the wrong books

Anyway, we know that 'just degenerating the input stage' does NOT increase SR. Increasing the Iq in the input stage does. However, that might require a larger comp cap due to the increased transconductance, undoing the rise in SR you try to achieve. Fortunately, you could degenerate the input stage to lower transconductance again so you got your increased SR.

Agreed?

jan didden

Hi Jan,

Yes, slew rate and how to control it was fairly well-known for a long time. Unfortuantely, it was not considered adequately in the early power amplifier designs. Matti deserves credit for helping to shine the light on the need for high slew rate and good input stage linearity. As you point out, degenerating the input stage is a very effective and popular way to increase slew rate. The degeneration decreases the input stage transconductance without reduce input stage current capability. With reduced input stage transconductance, a smaller compensation capacitor is needed, which results in a higher slew rate. Input stage degeneration also linearises the input stage, greatly reducing "soft-TIM" (sub-slewing HF distortion).

Where Matti went off the tracks was in blaming high TIM on low open-loop bandwidth. It is very easy to show that when things are compared on the proper basis, apples to apples, low open-loop bandwidth does not decrease slew rate and does not increase TIM.

The important thing to keep in mind is that for a given loop gain at 20kHz, a smaller open-loop bandwidth simply corresponds to a higher feedback factor at low frequencies.

Cheers,
Bob

[scott wurcer]

Originally Posted by john curl

Jan, please look at Barrie Gilbert's paper for this info.

Sorry it's not there either, frankly I don't see what the fuss is about.

[jan.didden]

Originally Posted by john curl

Jan, please look at Barrie Gilbert's paper for this info.

No I won't. SR is well understood and it has been explained again and again how Barrie viewed this, and your misunderstanding of it.

jan didden

[Bob Cordell]

Originally Posted by janneman

I'm not an expert on PIM, but I have always wondered how something that modulates the PHASE of a freq component could lead to additional harmonics?

jan didden

Hi Jan,

Every 6 months or so the PIM thing gets brought up again and everthing is thrashed out on both sides again. Those who want a good explanation of PIM, what causes it, and how NFB is involved, just need to go to my web site and read the JAES paper on PIM that I wrote. While Barrie Gilbert's long-subsequent paper is technically correct, it is really not a complete treatment. Moreover, John is putting words in Barrie's mouth when he asserts that Barrie's paper shows that low open-loop bandwidth exacerbates PIM.

The conclusions in my paper are:

1) PIM does indeed exist.

2) It is basically what has been known by video engineers forever as differential gain and phase.

3) One form of PIM is indeed "caused" by NFB and I show the math for predicting it.

4) the PIM generated by NFB depends on the closed loop bandwidth of the amplifier, NOT the open-loop bandwidth.

5) the PIM caused by negative feedback depends for its creation on the conversion of amplitude intermodulation distortion to phase intermadulation distortion. PIM from this mechanism cannot exist without substantial amounts of AIM (which is easily detectable).

6) Amplifiers without negative feedback have PIM as well, from numerous sources, only one of which is nonlinear junction capacitances.

7) Negative feedback actually REDUCES the PIM from #6.

In any case, those who want a fairly decent coverage of PIM should check out my paper, to at least understand the arguments. They certainly don't have to agree with me.


Cheers,
Bob

[scott wurcer]

Please folks get the BS straight. PIM somehow magically creates distortion at impossible in-harmonic frequencies. The stuff that uccurs at the normal predictable frequencies is no problem.

[john curl]

Moving on still further: It might be noted that a 'bible' of amplifier design was written by Dr's. Cherry and Hooper in about 1969. It is about 1000 pages, AND there is NO mention of slew-rate, rate of change, or any such term ANYWHERE, yet it was the best, most up-to-date book on the subject. Find a copy and see for yourself. Yet, while we all knew what slew rate was, mostly from IC op amp spec sheets, we did not have the insight yet to improve it significantly.
Now the years pass, remember, the problem first came up in 1966. By 1972, several papers were available from Otala, promoting high open loop bandwidth in audio design.
So in 1973, when working for the Grateful Dead, I designed what later came to be the Levinson JC-2 line stage, and the JC-3 power amp driver stage, which had about 100V/us slew rate, AND about 20KHz open loop bandwidth. The GD were happy, as was Mark Levinson, and I had found a SUCCESSFUL formula for making amps, and preamps.
These designs still sound good, just try them. Perhaps not perfect, but remember, they were designed 38 years ago. Perhaps we can do somewhat better, today, especially with faster output devices, and better jfets.
John Meyer and I worked together in making a super 3 way driven all horn loaded loudspeaker design in 1974. We tried all kinds of commercial power amps, Marantz, SAE, etc. and they worked pretty well up to about 5KHz or so but a better tweeter amp was necessary. That is when I coupled a JC-2 line amp to a class A complementary symmetry output stage, and got a satisfactory solution. Later, I gave this design to Mark Levinson to make a small power amp, and unfortunately Mark and I fell out, mostly over royalties for this amp, and he marketed it under the name of ML-2, with a few revisions of the circuitry.
However, about the same time (1973) or so, Matti and jan L. made an amp together at Philips Research and published a paper on it. This amp had the same overall specs of the JC-3 that I made later, including 100V/us and 20KHz open loop bandwidth. Both sounded great, and I actually bought the first prototype ever made by Electrocompaniet, and used it, off and on, for 15 years, until it was destroyed in a firestorm, in 1991. I have another, and it sounds good, even today. This amp is NOT special because of the amp topology, it is successful because of the high open loop bandwidth. Some similar designs of mine, even with a more linear topology, do not sound as good as this little Electrocompaniet, and I keep it in service as a constant reminder. (more later)