I can't easily contribute within the topic, but it might be useful how circuit topology, slew rate, and TIM evolved over the decades, and why we had to fix it.
Transistor amplifiers started to become very popular about 1965 or so. Tubes were still around, mature, and worked OK, BUT they, of course, had to be replaced at intervals, and Silicon was paralleling and even replacing Germanium transistors. However, back in those days, a silicon low power transistor like the 2N697, still cost extra, and all bipolar transistors were relatively expensive. This meant that we tried to use as FEW transistors in a circuit that we could get away with. For example, the differential pair input with transistors, while possible, was excluded from many audio circuits, both consumer and professional. It just added another part, and increased the potential noise 3dB. It must be remembered, that in those days, input and output transformers were popular with pro equipment, and consumer audio was all single ended input. For example, Ampex tape recorders used an input transformer (optional) and output transformer with cap coupled transistor circuits, and a single voltage supply. If you looked at Dolby at that time, they were doing much the same thing.
It was normal to use a single input transistor, hi beta, as an input device, emitter degenerated with a resistor to ground that also became a feedback resistor. This approach is more ideal than many people would realize, today. First, there is degeneration in the input device due to the RE(F) resistor, that does both functions. Secondly, one dominant noise source was the base resistivity or Rbb' that could be several hundred ohms typically with normal 'low noise', high beta transistors, so an extra 200 ohms or so was not so bad. Also, the original operating currents might be 100ua, or less, NOT 1ma or more, due to the noise current contribution, that dominated with high inductance sources used. (more later)
Transistor amplifiers started to become very popular about 1965 or so. Tubes were still around, mature, and worked OK, BUT they, of course, had to be replaced at intervals, and Silicon was paralleling and even replacing Germanium transistors. However, back in those days, a silicon low power transistor like the 2N697, still cost extra, and all bipolar transistors were relatively expensive. This meant that we tried to use as FEW transistors in a circuit that we could get away with. For example, the differential pair input with transistors, while possible, was excluded from many audio circuits, both consumer and professional. It just added another part, and increased the potential noise 3dB. It must be remembered, that in those days, input and output transformers were popular with pro equipment, and consumer audio was all single ended input. For example, Ampex tape recorders used an input transformer (optional) and output transformer with cap coupled transistor circuits, and a single voltage supply. If you looked at Dolby at that time, they were doing much the same thing.
It was normal to use a single input transistor, hi beta, as an input device, emitter degenerated with a resistor to ground that also became a feedback resistor. This approach is more ideal than many people would realize, today. First, there is degeneration in the input device due to the RE(F) resistor, that does both functions. Secondly, one dominant noise source was the base resistivity or Rbb' that could be several hundred ohms typically with normal 'low noise', high beta transistors, so an extra 200 ohms or so was not so bad. Also, the original operating currents might be 100ua, or less, NOT 1ma or more, due to the noise current contribution, that dominated with high inductance sources used. (more later)
Without understanding the fundamental aspects of single device vs differential device topolologies, it is pointless to compare the 'old' and the 'new', except the 'old' was not as bad as it first seemed, and it had a few intrinsic advantages. I would not go back, because understanding how to get the same 'advantages' with more sophisticated topologies gives us the best of both worlds.
I agree.
Maybe, as you say, some of these old designs did not actually sound that awful, but a diff input designs have their advantages and I personally think are a step up in sound quality.
Now what about the differential pair, or 'long tailed pair' as the British called it who appear to have invented it? This two similar device input stage either, emitter, cathode or source connected, is a natural for DC amplifiers. However, back in the old days, we did not always think that DC amplification was necessary or useful. This was true with analog tape recorders, as well as many other audio products. Therefore, cap or transformer input (output) was natural and customary. It saved active devices that were relatively expensive and delicate, and protected loudspeakers, etc from the power supply, if something shorted. This could potentially save a lot of connected equipment to the circuit that is having a problem.
In any case, we did NOT normally use the differential pair for AC amplifiers, but of course they were used for Analog Computers and DC amplifiers, where we wanted to process very low frequency or DC information. This made these circuits relatively expensive. The first time that I saw a lot of differential inputs was when the uA702, uA709, and finally the uA741 became normal circuit components, first in military, then in industrial, and finally in audio designs. Of course, for an analog IC, the differential pair was a 'natural' because it allowed direct coupling to the input and subsequent stages, with good DC stablity. However, unlike what Bob Cordell originally suggested, they OMITTED any emitter degeneration resistors in most of the designs, because they just added noise, and REDUCED the gain of the input stage. Just look at the design of the Ampzilla power amp. You will find resistors, BUT they are bypassed with capacitors to improve the open loop gain and reduce noise. Do you think that JB would actually add these parts, if he could have gotten away without them? No, he needed the resistors for better input matching, but he had the add the extra caps to get back the gain that he lost, to increase the open loop gain, to give the loop feedback, in order to meet spec. And so it goes. (more later)
In any case, we did NOT normally use the differential pair for AC amplifiers, but of course they were used for Analog Computers and DC amplifiers, where we wanted to process very low frequency or DC information. This made these circuits relatively expensive. The first time that I saw a lot of differential inputs was when the uA702, uA709, and finally the uA741 became normal circuit components, first in military, then in industrial, and finally in audio designs. Of course, for an analog IC, the differential pair was a 'natural' because it allowed direct coupling to the input and subsequent stages, with good DC stablity. However, unlike what Bob Cordell originally suggested, they OMITTED any emitter degeneration resistors in most of the designs, because they just added noise, and REDUCED the gain of the input stage. Just look at the design of the Ampzilla power amp. You will find resistors, BUT they are bypassed with capacitors to improve the open loop gain and reduce noise. Do you think that JB would actually add these parts, if he could have gotten away without them? No, he needed the resistors for better input matching, but he had the add the extra caps to get back the gain that he lost, to increase the open loop gain, to give the loop feedback, in order to meet spec. And so it goes. (more later)
Here it is. I do not know how old this is but by looking how the schematic is drawn it looks old enough to me.
Ampzilla Schematic
Ampzilla Schematic
So, as 1970 or so rolls in, use of the differential input stage, both discrete and integrated, becomes more prevalent, along with +/- supplies, and a reduction in interstage coupling capacitors and transformers.
This set the minimum input noise with the differential pair (with its 3dB increased voltage noise over a single device), and the feedback resistor to either ground or input as the added extra noise source. Usually, emitter degeneration resistors were left out, which gave a significant increase of first stage Gm, over the single device, with its necessary Re which was also the feedback resistor, and-or an added resistor for stability with temperature. In any case, this opened the way from the older single device input devices with limited Gm, to a much higher Gm input stage, which forced the lowering of the slew rate of the design for a given gain bandwidth.
With IC op amps, that were not really designed for audio, originally, this was an acceptable compromise, but when IC's were first introduced into audio, this became a problem. This problem also extended to discrete power amps, that need even a higher slew-rate than the preamp driving it. Typically, in the early 1970's, power amp slew rates of 10V/us, were typical.
This is where the Lohstroh-Otala amplifier departed from the typical amp, doing 100V/us or so. I followed suit with jfet input input stages that also lowered the input Gm and also gave 100V/us in the JC-3 prototype, back in 1974. Along with Electro-Research, this was seemingly design 'overkill', but it worked and it got both Electrocompaniet and Levinson into the mainstream of audio design, because of the subjective response to these designs at the time.
However, if we had followed the 'typical engineers path' at the time, we would have tried for even lower distortion by using more feedback, and dropped the high slew rate, high open loop bandwidth, and probably then tried for more rated power, by running more toward class B. Trust me, plenty of designers did, at the time, but these designs are long gone, and all you had to do was to live with one of them, especially with a pair of electrostatic headphones, to find out why.
This set the minimum input noise with the differential pair (with its 3dB increased voltage noise over a single device), and the feedback resistor to either ground or input as the added extra noise source. Usually, emitter degeneration resistors were left out, which gave a significant increase of first stage Gm, over the single device, with its necessary Re which was also the feedback resistor, and-or an added resistor for stability with temperature. In any case, this opened the way from the older single device input devices with limited Gm, to a much higher Gm input stage, which forced the lowering of the slew rate of the design for a given gain bandwidth.
With IC op amps, that were not really designed for audio, originally, this was an acceptable compromise, but when IC's were first introduced into audio, this became a problem. This problem also extended to discrete power amps, that need even a higher slew-rate than the preamp driving it. Typically, in the early 1970's, power amp slew rates of 10V/us, were typical.
This is where the Lohstroh-Otala amplifier departed from the typical amp, doing 100V/us or so. I followed suit with jfet input input stages that also lowered the input Gm and also gave 100V/us in the JC-3 prototype, back in 1974. Along with Electro-Research, this was seemingly design 'overkill', but it worked and it got both Electrocompaniet and Levinson into the mainstream of audio design, because of the subjective response to these designs at the time.
However, if we had followed the 'typical engineers path' at the time, we would have tried for even lower distortion by using more feedback, and dropped the high slew rate, high open loop bandwidth, and probably then tried for more rated power, by running more toward class B. Trust me, plenty of designers did, at the time, but these designs are long gone, and all you had to do was to live with one of them, especially with a pair of electrostatic headphones, to find out why.
hi John
firstly i would like to thank you for your time and trouble relating your design experiences, i find it very interesting.
with regard to the 'long tail pair' have a look here
The Alan Blumlein Homepage
Alan D Blumlein, engineering genious and prolific inventor
firstly i would like to thank you for your time and trouble relating your design experiences, i find it very interesting.
with regard to the 'long tail pair' have a look here
The Alan Blumlein Homepage
Alan D Blumlein, engineering genious and prolific inventor
I agree.
Maybe, as you say, some of these old designs did not actually sound that awful, but a diff input designs have their advantages and I personally think are a step up in sound quality.
I recently acquired a vintage QUAD 33 preamp because I wanted to experiment with its particular tone control system.
But it is a nice surprise that this preamp, which uses just a couple of transistors, with a single-ended, zener-stabilised 12VDC supply, sounds so very good! Phenolic PCBs, tinned (no trace of gold!) plug-in PCB connectors, that sort of thing. Still pretty good!
I DID replace some of those unobtainium 5-pole DIN jacks with RCAs though...
jan
Attachments
Shortly I had this model for service and by perform a soundcheck I agree with you. BTW - great work concerning refitting the thick WBT's.
For my customer I want to find a 4-pin DIN male connector for thick cable - go to post #14 about
http://www.diyaudio.com/forums/solid-state/131318-quad-33-please-help-2.html
Last edited:
I agree.
Maybe, as you say, some of these old designs did not actually sound that awful, but a diff input designs have their advantages and I personally think are a step up in sound quality.
Maybe not if the singleton is a CFP, in this case I think the CFP might even be more linear than the LTP depending on how much degeneration is used on the LTP. Singleton input stages display favourable harmonic spectrums in acordance with Hiraga and can have as low distortion figures as LTP based designs.
Joachim is kinda right saying they are coming back, but it started already 7-8 years ago with commercial designs employing them in entry and mid level consumer amps. As more recent comercial schematics are not easily available most are unaware of this.
Hi John,
As usual, you are making up as you go along what you assert I said, and getting it wrong.What you are saying here is fairly silly. First of all, reasonable values of input LTP emitter degeneration do not cause a significant noise problem for power amps. Do the math. It is not difficult to get below 10nV/rt Hz in a properly designed power amp with 10:1 input LTP degeneration. I do it all the time. Please note that the voltage noise of even a 1k resistor is only about 4 nV/rt Hz.
Secondly, any properly designed audio amplifier has plenty of gain even with emitter degeneration. Maybe JB didn't get it right.
Stop attributing things to me when you can't get it right.
TIM is about getting adequate slew rate and decent open-loop linearity - it is not rocket science.
Bob
Telling me what TIM is, once again, Bob? I thought that we cured you of that back in 1980. '-)
In any case, the PRIMARY reason that JB added the capacitors was to get back the open loop gain that he lost. However, for completeness, it should be noted that cap bypassing lowers noise also.
IF JB had left the bypass caps out, he could have increased his slew rate to something more competitive with Otala or me, but like so many others, he needed the gain to reduce the xover distortion that is inherent in such a design. Of course, IF you, Bob, had been there to advise him, back in 1970 or so, all would be well today, and we would be building Ampzillas for fun and profit, even today.
In any case, the PRIMARY reason that JB added the capacitors was to get back the open loop gain that he lost. However, for completeness, it should be noted that cap bypassing lowers noise also.
IF JB had left the bypass caps out, he could have increased his slew rate to something more competitive with Otala or me, but like so many others, he needed the gain to reduce the xover distortion that is inherent in such a design. Of course, IF you, Bob, had been there to advise him, back in 1970 or so, all would be well today, and we would be building Ampzillas for fun and profit, even today.
Bob,
When the only tool you have is a hammer.... Lets not forget preamps....maybe your next book!
ES
I have one spare 4-pin DIN that takes max 6mm OD cable. If you want it, just send me a snail mail address.
Edit: I removed the remote main output jacks and put the output RCAs there.
Then I removed the superfluous output DIN jack, as well as the Radio 2 DIN jack, and put two Radio 2 input RCAs in those holes.
jan
Last edited:
Actually, Walt Jung reviewed the Ampzilla in The Audio Amateur at the time and found that the input pair was severely unbalanced - to the point that he called it 'a design oversight'. And so it goes, indeed.
jan
- Status
- Not open for further replies.