Hi John,
In hindsight, that is exactly what I mean. The complimentary differential pair makes perfect sense. I had earlier tried to say that no matter how it came about or who thought of it, it was really inspired work. There is nothing there intended to diminish what you had done.
In other words, don't blame the inventor if the execution of a design falls short of the goal.
-Chris
In hindsight, that is exactly what I mean. The complimentary differential pair makes perfect sense. I had earlier tried to say that no matter how it came about or who thought of it, it was really inspired work. There is nothing there intended to diminish what you had done.
This has been one of my points all along. (my underline) It's very easy for a person to look at an example of your circuit configuration as executed by another company and conclude that it doesn't sound very good. This is because the manufacturer failed to satisfy the requirements, creating high DC offset currents at the input and also output. This also raises the distortion a great deal. The same thing happens with a normal long tailed pair, except the mis-match doesn't seem to create as much distortion. That only means that the simple normal long tailed pair is the better choice for mass production where parts can not be matched by hand.
In other words, don't blame the inventor if the execution of a design falls short of the goal.
There were many odd things designed and built back then. But I wouldn't call that overly complicated as far as a circuit design goes. The Marantz 500 (as an example) was designed in 1968 as well I think, and released in 1969. It used and input buffer amp, followed by three differential pairs. Certainly more complicated in design. You had one diff pair for overall control, then each polarity had it's own diff pair. That is a substantially more complicated approach. Something more like a brute force method while the complimentary differential pair is more elegant.
-Chris
The Marantz 500 was released in 1971. Before this, there were a few other power amps, including the 250, the Models 14, 15, and 16. These were VERY different from the Marantz 500, which I considered a kludge, and one of the slowest amplifiers in the world at the time for its power rating. We had one in our Swiss lab in 1974. We used it from 20-400Hz. It worked OK, there. I personally owned a 14, which I purchased in 1969. Measured OK, but gave me listening fatigue over time on the K-horn.
OK, just got in and I'm a little jetlagged. I'll bring it into the lab and even measure it.
Getting away from the bipolar complementary differential, from the late '60's, I might talk about what we built in our Swiss Lab, back in 1974-5. The first project that John Meyer and I attempted, was a 3 way all horn loaded, time-aligned PA speaker system. We used a 'derived' 3 way electronic crossover at 400 and 4000 Hz and I made an electronic time delay with analog shift registers to effectively move the tweeter 'back' about 15 inches. To implement this design quickly as possible, I selected the Marantz 500 for the bass, the Marantz 250 for the midrange, and I made a custom amplifier, ultimately to become the JC-3, to drive above 4KHz. I only needed about 15W with a 107 SPL/W Electrovoice T-350 horn loaded tweeter that we used. This is where I connected the Grateful Dead line driver to a clean complementary Darlington output stage to make a really successful amplifier, the best that I had ever done up to that time. This amp, ultimately became the fundamental building block for the Levinson ML-2, that got world wide acceptance.
The speaker was VERY successful, in that it was VERY loud, yet clean sounding.
Unfortunately, politics beyond our control, caused us to close the lab after making this design and a number of others, and move on.
The speaker was VERY successful, in that it was VERY loud, yet clean sounding.
Unfortunately, politics beyond our control, caused us to close the lab after making this design and a number of others, and move on.
Last edited:
This explanation of the complementary differential stage is very interesting.
Does it also have lower distortions than the simple differential stage ? Or other advantages beside the one mentioned ?
Fireworks, a test done, long ago, by making a complementary differential input stage and then forcing it to be a single differential with a current source load for the second stage can be implemented by just putting a large capacitor between the base emitter junction of one of the second stage transistors. Then this circuit reverts to a single differential with a single ended second stage drive and a current source load. Certainly, everyone is familiar with this topology. Well the distortion in my test went up 5 times. 2.5 times because of the extra distortion, and 2 times because the complementary differential added 2 times forward gain compared to the single differential. Someone making a JC-2 line amp could actually try this.
Compared to many other 'tweaks' the complementary differential input stage 'improvement' over a single differential stage and a single driver with current source load, should be measurable by either nominal distortion test equipment, or Spice circuit analysis. I did it with my modified Heathkit IMA, back in 1973.
The improved alternative to the single differential stage with a single driver, is the approach first used by Richard Burwin in 1965 or so, for ADI. This consists of a single differential first stage, BUT then another differential stage for the second stage and then, some sort of current mirror, or similar circuit, to create a push-pull second stage.
This was used also in the Marantz 14 and 15 power amps in 1968, the Lohstroh-Otala power amp design, the Levinson JC-2 phono stage, the 5534 IC and Cordell's power amp, etc. This is a VERY popular design and it works about as well as the complementary differential. A serious student or designer might find this a circuit, problematic in some ways, because of thermal and voltage differences in the second stage.
The improved alternative to the single differential stage with a single driver, is the approach first used by Richard Burwin in 1965 or so, for ADI. This consists of a single differential first stage, BUT then another differential stage for the second stage and then, some sort of current mirror, or similar circuit, to create a push-pull second stage.
This was used also in the Marantz 14 and 15 power amps in 1968, the Lohstroh-Otala power amp design, the Levinson JC-2 phono stage, the 5534 IC and Cordell's power amp, etc. This is a VERY popular design and it works about as well as the complementary differential. A serious student or designer might find this a circuit, problematic in some ways, because of thermal and voltage differences in the second stage.
Last edited:
The example shown by Pooge, is a more MODERN version of this sort of single diff pair design, than the examples that we had made in the late '60's and the early '70's. This is where the differential folded cascode is used, rather than two stages. This sort of topology was first created (to the best of my knowledge) by Harris associates, some time in the 1970's in an IC. I don't have the part numbers available at the moment, but I have seen the data sheets.
Last edited:
As most of the common solid state circuit topologies have been discussed, I would like to add to the commentary about feedback resistors.
It is an known fact that capacitors actually have bigger potential problems and differences in materials than resistors. Since people can scoff at capacitors being very important, what about resistors?
There are specific locations that really make resistors more important than capacitors, in general. The first is the FEEDBACK RESISTOR between the output and the feedback input. This resistor is one of the most important passive components in an amp or preamp, because IF we are depending on feedback to LINEARIZE the circuit, the RESISTOR is the TEMPLATE that the feedback uses to get it right. Also, this is one of the hardest working resistors in the entire amp, not so much as average power dissipation, but CHANGES in power dissipation.
Now, this is where we go back to an amp that 'smoked' its feedback resistor under test.
This is an example of complete indifference to the feedback resistor. After all, the Maximum power dissipation can be calculated by E squared (power supply DC voltage) divided by the feedback Resistance. For example: a +/- 100V power supply and a 10K feedback resistor would be 10K/10K or 1W Kind of high, isn't it? But if a ground lifted, and full power square waves were accidently generated this would be the dissipation of the feedback resistor. Now, it might be possible to get away with a 1/2W resistor, but good engineering practice would probably be to increase the value of the feedback to 20K or more. That is what we do at Parasound, as we usually use 47K resistors for the main feedback resistor, for this very reason, primarily.
Now, there are other advantages of higher resistance or a larger power rating of the main feedback resistor. This is because of the TEMPCO of the resistor. This is where the resistor continually changing its actual resistance as the amp plays louder and softer. A larger resistor prevents as large a temperature change, because of its THERMAL CAPACITANCE. More later.
It is an known fact that capacitors actually have bigger potential problems and differences in materials than resistors. Since people can scoff at capacitors being very important, what about resistors?
There are specific locations that really make resistors more important than capacitors, in general. The first is the FEEDBACK RESISTOR between the output and the feedback input. This resistor is one of the most important passive components in an amp or preamp, because IF we are depending on feedback to LINEARIZE the circuit, the RESISTOR is the TEMPLATE that the feedback uses to get it right. Also, this is one of the hardest working resistors in the entire amp, not so much as average power dissipation, but CHANGES in power dissipation.
Now, this is where we go back to an amp that 'smoked' its feedback resistor under test.
This is an example of complete indifference to the feedback resistor. After all, the Maximum power dissipation can be calculated by E squared (power supply DC voltage) divided by the feedback Resistance. For example: a +/- 100V power supply and a 10K feedback resistor would be 10K/10K or 1W Kind of high, isn't it? But if a ground lifted, and full power square waves were accidently generated this would be the dissipation of the feedback resistor. Now, it might be possible to get away with a 1/2W resistor, but good engineering practice would probably be to increase the value of the feedback to 20K or more. That is what we do at Parasound, as we usually use 47K resistors for the main feedback resistor, for this very reason, primarily.
Now, there are other advantages of higher resistance or a larger power rating of the main feedback resistor. This is because of the TEMPCO of the resistor. This is where the resistor continually changing its actual resistance as the amp plays louder and softer. A larger resistor prevents as large a temperature change, because of its THERMAL CAPACITANCE. More later.
Second on the list, are two places in an amplifier module, for resistance to appear to be most important. This is the INPUT RESISTOR and the 2'd FEEDBACK RESISTOR, the one that goes from the feedback tie point to ground. It is interesting that both of these resistors have almost the same AC voltage levels, no usual DC curent flow, and are both attached to the audio signal, itself.
Now, this is where it gets subtle.
Recently, we have gotten a lot of criticism for over-concerning ourselves with EXCESS NOISE in resistors. This is most common in cheaper, usually carbon based resistors, but it can also happen even in wirewound resistors. The nature of this noise is multi-varied, and still under study in some instances, but it tends to imply two factors:
One, the resistor can well be unreliable over time, because its internal connecting junctions are marginal, sometimes just a press fit.
Second, excess noise is sometimes defined as a 'dynamic change in the resistance' which creates the noise that shows up especially with DC flowing through it.
That doesn't appear to be a very accurate resistor to me.
Better quality resistors that cost a few penny's each in reasonable quantity, like 100, or more, are usually free of significant excess noise.
The rest of the resistors in an amplifier gain module are usually for DC biasing, load resistors, emitter resistors, etc. These resistors can be VERY sensitive to excess noise, especially low Mu triode tube load resistors, because of the high DC current necessary due to the DC voltage drop. I have gathered extreme examples of perfectly good looking, BUT very noisy resistors, once you add 10V or so DC voltage across it. So, just looking at the resistor and its external finish is not enough. More later.
Now, this is where it gets subtle.
Recently, we have gotten a lot of criticism for over-concerning ourselves with EXCESS NOISE in resistors. This is most common in cheaper, usually carbon based resistors, but it can also happen even in wirewound resistors. The nature of this noise is multi-varied, and still under study in some instances, but it tends to imply two factors:
One, the resistor can well be unreliable over time, because its internal connecting junctions are marginal, sometimes just a press fit.
Second, excess noise is sometimes defined as a 'dynamic change in the resistance' which creates the noise that shows up especially with DC flowing through it.
That doesn't appear to be a very accurate resistor to me.
Better quality resistors that cost a few penny's each in reasonable quantity, like 100, or more, are usually free of significant excess noise.
The rest of the resistors in an amplifier gain module are usually for DC biasing, load resistors, emitter resistors, etc. These resistors can be VERY sensitive to excess noise, especially low Mu triode tube load resistors, because of the high DC current necessary due to the DC voltage drop. I have gathered extreme examples of perfectly good looking, BUT very noisy resistors, once you add 10V or so DC voltage across it. So, just looking at the resistor and its external finish is not enough. More later.
In case we are that much concerned about distortion of FB resistors, input resistors etc., how about distortion in power amplifiers? Sometimes we are convinced that distortion of some 0.05% in power amplifier, mostly 2nd and 3rd order harmonic is just fine (though resistor or capacitor distortion of less than 0.001%, again low order, is considered bad). BUT, this 0.05% amplifier distortion is perfectly audible, and no one seems to complain. How about that? Why it is not important, though assumed resistor distortion is important? I do not get it somehow.
Resistor distortion can be small, but everything is additive, and who knows, maybe some resistors might have nearly as much 7th harmonic as 3'rd harmonic. Just one bad contact could do this. Is saving a few pennies worth it?
I use some of the most expensive resistors in the world, because I know they will perform well, AND they have a really small 'footprint' on the circuit board. If you saw my latest circuit board, you would understand that this is really, really, important. However, for less stressful points, we might use DALE metal film, as measured by Ed Simon, in smaller dissipation than 1/2W, for a fairly small footprint. However, if I had the space, I might insist on 1/2W or more resistors, except for the servo or housekeeping circuits, just to keep the thermal stress as low as possible. I like to take care of each and every component that I have some control over. It has worked for me, over the decades. Overkill, sometimes? Yes.
I use some of the most expensive resistors in the world, because I know they will perform well, AND they have a really small 'footprint' on the circuit board. If you saw my latest circuit board, you would understand that this is really, really, important. However, for less stressful points, we might use DALE metal film, as measured by Ed Simon, in smaller dissipation than 1/2W, for a fairly small footprint. However, if I had the space, I might insist on 1/2W or more resistors, except for the servo or housekeeping circuits, just to keep the thermal stress as low as possible. I like to take care of each and every component that I have some control over. It has worked for me, over the decades. Overkill, sometimes? Yes.
I understand, but to me not enough care has been addressed to power amplifier distortion reduction. Yes, distortions superimpose, but -70dB of amplifier's 2nd order harmonic will always mask -120dB 2nd order harmonic of some resistor. And no evidence has been shown yet that resistor's 7th harmonic would be higher than that of the power amplifier. This is my point.
I do not mean poor contacts, that can result in horrible distortion. If there was a resistor distortion contributing to amplifier's distortion, I would like ta have a possibility to see an evidence. Let's talk real results, not guesses.
I do not mean poor contacts, that can result in horrible distortion. If there was a resistor distortion contributing to amplifier's distortion, I would like ta have a possibility to see an evidence. Let's talk real results, not guesses.
PMA, I think that Ed Simon's measurements show enough to at least get people to use the 'better' brands, and to take care with the really cheap stuff.
In production, it is VERY tempting to use the cheapest resistors, caps, etc that you can get away with. Trust me, many mid fi manufacturers ( mid hi end to most of you) use the cheapest resistors, etc that they can find. Each cheap resistor purchase gives savings enough for a couple of free lunches. ;-) I am always on guard for 'substitutes' used in my best products, but I sometimes just have to hope for the best in some of the cheaper designs that I am associated with.
In production, it is VERY tempting to use the cheapest resistors, caps, etc that you can get away with. Trust me, many mid fi manufacturers ( mid hi end to most of you) use the cheapest resistors, etc that they can find. Each cheap resistor purchase gives savings enough for a couple of free lunches. ;-) I am always on guard for 'substitutes' used in my best products, but I sometimes just have to hope for the best in some of the cheaper designs that I am associated with.
I would also like to point out that the exterior fit and finish does not necessarily prove excellence. I have hundreds of really beautiful resistors made for aerospace, that I suspect as to not sounding as good for audio as other resistors. Maybe I can sell some to people here and they can prove me wrong. Anybody in the SF Bay Area game?
The evidence is weak or non-existant. The argument that this resistor has -130dB and that has -150dB so this must be why they sound different is a logical fallacy. A test to isolate the contribution of the resistor has never been done (and never will be).
- Status
- Not open for further replies.