Melcor 1731 understanding

[Deepdark]

Hi guys

For personnal knowledge and brain training, I was looking at the 1731 opamp schematic linked and I want to fully understand it. Here is what I analysed but their is some place where I 'm not sure what is going on.

Looking at the input. We have a first differential pair (Q1 and Q2) tied in a common emitter (source) so we amplify the differential between the input + and input -. The output signal at the collector is then reverse. R1, R3 and R4 are our current source.

Then, exiting the first stage, we enter another pair of transistor, Q3 and Q4, which their emitter are tied together, as well as their collector and R5 is the current source. The signal at the base of Q3 comes from the output of Q2 and teh signal at the base of Q4 comes from the output of Q1. Is that it? So what is the goal of R2 and C1? Some sort of filtering, or stabilizing frequency? Now, can we say this stage is configured as a common collector (drain)? So the signal is not reverse and, by the same, we should talk about a common mode instead of a differential mode??

Next step is where I struggle to understand what is going on. If the output signal of the second stage is effectively taken at the source, then the signal from Q3 split at R6, which is tied to V-, and go to the base (gate) of Q5. Q5 looks to operate in Common source mode so the signal taken at the drains should be phase reverse. The output then go to the gate of Q6, which is also tied with the output of Q4. Here the signal shouldn't be the same polarity, isn't it?

Q6 seems to be tied in common drain and it's output feed Q7 and Q7 then go to output? the D2 diodes are there to form a voltage reference, isn't it?

Q8 looks to be feed by the In-, which also connect to the output or R8. This entire part is a little obscure. Is it there to form a loop?

Thanks for your time

 

[Samuel Groner]

These are all bipolar transistors, so we have collector, emitter and base. And don't count on such reverse-engineered schematics to be fully correct...

Q1/Q2, Q3/Q4 and Q5 are common-emitter configurations. Q7/Q8 is a complementary common-collector (AKA emitter follower) stage, and Q8 is driven by the collector of Q5. Q6 is slightly obscure--without having studied it in detail I'd say it is a common-collector stage which, together with Q5, forms an adventurous push-pull drive for Q7/Q8.

C1/R2 shapes the high-frequency response. C2 forms a local high-frequency feedback loop before the output stage; it probably improves stability into capacitive loads. C3 is probably part of the compensation as well (Miller feedback for Q5), but the detailed intention is a bit obscure as R9 is in series with it (which forms a strong zero).

From the positive input to the output we have two inverting (common-emitter) and two non-inverting (common-collector) stages; from the negative input to the output we have three inverting and one non-inverting stage. So polarity is correct in both cases.

Samuel

 

[Deepdark]

These are all bipolar transistors, so we have collector, emitter and base. And don't count on such reverse-engineered schematics to be fully correct...

Q1/Q2, Q3/Q4 and Q5 are common-emitter configurations. Q7/Q8 is a complementary common-collector (AKA emitter follower) stage, and Q8 is driven by the collector of Q5. Q6 is slightly obscure--without having studied it in detail I'd say it is a common-collector stage which, together with Q5, forms an adventurous push-pull drive for Q7/Q8.

C1/R2 shapes the high-frequency response. C2 forms a local high-frequency feedback loop before the output stage; it probably improves stability into capacitive loads. C3 is probably part of the compensation as well (Miller feedback for Q5), but the detailed intention is a bit obscure as R9 is in series with it (which forms a strong zero).

From the positive input to the output we have two inverting (common-emitter) and two non-inverting (common-collector) stages; from the negative input to the output we have three inverting and one non-inverting stage. So polarity is correct in both cases.

Samuel

Thanks Samuel

That's an interesting reading and analysis. I found opamp to be fascinating. There surely a way of doing a simpler opamp that can deliver enought gain in a preamp amplification stage, such as api 312, or so.

 

[PRR]

Learn your BJT vs FET terminology. I thought I knew some of what this plan does but now I am totally confused with all your sources and drains.

Be able to write an approximate value for EVERY current and voltage. Fer example, ass-uming the inputs are near ground (as they will be for an audio inverter or a high-gain NI audio stage), and that Vbe is "small", the current in R1 is (V-)/130K; and we *may* get half of this in each of Q1 Q2. (Already a problem: supply voltages are not shown.) Q2 Q2 current leads to voltages across R3 R4, which leads to voltages at Q3 Q4 and a current in R5 (modified by R10 fed from Q6).

The D2 D2 Q7 Q8 stage idle current has no matchbook answer. Ideally we would consider the relative areas of the diodes and transistors, modify by the funny doping used in diodes, then consider R11 R12.... From experience we know it will run several mA. We can quickly show that 1mA or more is fine for the audio, though (depending on supply!) dozens of mA might be a lot of heat.

Heed Samuel's comment about flawed transcriptions. I suspect there *is* a flaw here but am not inclined to trace it tonight.

Pick a simpler plan. There are several major tricks here. R10 tied-back to Q3 Q4, for one. These may be above your present understanding, just headaches until you get further down your path.

There is an older mid-level textbook with a *complete* run-through of the '741 op-amp. I have it here somewhere but it is not leaping to my hand; maybe someone else knows what I mean. While the '741 is old and despised in audio, it works The Same as the Melcor, just different. (Honda, Subaru, and Napier Deltic engines all turn petrol into power but the parts are laid-out different.)

Doug Self's book on power amplifiers is VERY good food. (Power amps are mostly op-amps with fat transistors on the end.) He does not go into all the many different ways you can skin cats, he cuts to a good solid topology you should KNOW, his basic "blameless" amp.  Cordell's audio-amp book covers some of the clevererer ways to do the deed, but much of it is hard-chewing for a beginner.

> There surely a way of doing a simpler opamp

Yes, yes, yes. Maybe not fewer parts, but easier analysis.

 

[PRR]

> tricks here. R10 tied-back to Q3 Q4

I started doing your homework. I am tired and not getting a grade so there may be goofs.

ASS-uming +/-15V supply, and diff-pair symmetry(?), Q1 Q2 current is trivial. Red crayon.

From that I get ~~0.8mA into the second pair. But working through Q5 Q6 I get a voltage at R10 which *cuts-off* Q3 Q4. So that can't be right.

You could write the equation; it's not massive. However *knowing* that a plan like this was designed (even if hit/miss) and worked well in quantity, my cave-man hack is to assume that there is a balance and it is probably mid-way. So write 6mA instead of 13mA and work back through the 2nd stage (orange "maybe" crayon). The number I got is within 20% of self-agreement, and another iteration would give an answer correct for any practical purpose, and we do not need an exact answer.

Pondering why R10 in 3rd stage stabilizes the current in the 2nd stage is good "brain training".

I do think it would be "simpler" to put a '5534 chip inside a B-B DOA module and get on with making music.

Oh, the polarities do come around correct. And the push vs pull gain-difference is very close.

Another point: what is the MAXimum current which the output stage will pass? This is actually trivial but they make one part do two jobs so it may not be evident that there is current-limiting. Further point: what is the maximum dissipation in Q7 or Q8 for an AC signal centered around ground? (For further smoke-proofing, consider the case of Out pulling up hard and a load shorted to V-. However this is unlikely in a happy console.)

 

[jdbakker]

There is an older mid-level textbook with a *complete* run-through of the '741 op-amp. I have it here somewhere but it is not leaping to my hand; maybe someone else knows what I mean.

The Monolithic Operational Amplifier: A Tutorial Study.

JDB.
[there are more, but this one lists the basics and is generally sufficient to at least 'get' the anatomy of most op-amps, discrete or otherwise. The Self books PRR mentions are a good alternate approach, though]

 

[JohnRoberts]

I am too lazy to give you a point by point analysis but in passing, the output stage is not biased on very hard (or maybe it is?). I would be tempted to better define the class A current in that output stage (right now it depends on transistor Vbe and small signal diode drops) , but get it to work first, before tweaking. Be careful about substituting different devices there.

I like the approach to drive the output stage push-pull but there is an extra delay in the Q5 inverter that isn't in the other direction's path which may require extra compensation for stability.  C2 connecting back to one of the input LTP  bases could exhibit some variability with source impedance there. This extra path delay reminds me of the quasi-complementary power amp output stages used before we had robust complementary power transistors. Probably not much of an issue for low power, but more symmetry is always nice. Of course always a trade off between complexity and performance.

I recall reading the op amp tutorial Sam mentioned and JDB provided a link to, I think I have the hard copy around here somewhere (If i haven't given it away).  While most dismiss the 741 these days as a very low performance op amp, it was a remarkable accomplishment for what it was back in the day.  In the good old days they used to publish full schematics for the innards of most ICs and I loved studying them.

JR

 

[PRR]

JD's cite is excellent, but there is a college text-book where the middle chapter(s) go through the '741 point-by-point, deriving every current and frequency response corner.

 

[PRR]

> better define the class A current in that output

I suspect it runs rich-enough. The 8 ohm bias resistors are quite small.

Note that in the original, "D2" was a 2-diode device sold for the purpose of bias-setting and sorted for voltage-range. Two 1N4148 is a later hack.

With the later-hack part numbers, a run made from one batch may have >2:1 variability of idle current, with different batches >4:1. The parts specified are super-available and I agree it would not be wise to throw, say, '4001 for '4148 or TIP in place of BD, without limiting resistors on the smoke-test.

The idle current isn't super fussy. If under 4mA there may be crossover distortion. At 20mA the transistors run warm. Anywhere in there is OK. System power should not care because it is scaled to drive 150 ohm loads and at full signal level could draw 35+mA, so 4mA-20mA idle should be fine.

> extra delay in the Q5 inverter

I'll mildly argue this. Q6 is also "common emitter", not emitter-follower. It is driven from an infinite impedance (Q4 collector, naked). It finds its Base drive across R9 which ties to R8 to Q6 Emitter. As long as Q7 Q8 Base currents are "small" (I think they are), then Q6 gives similar speed to Q5 (somewhat different because 3:1 ratio of resistor values). And both Q6 and Q5 run quite rich and will be very fast. And tightly coupled. I suspect any asymmetry will be only at highest levels and frequencies. And at that it will still blow the pants off of '741 or really ANY opamp available in that day except 3 or 4 similar audio-specials.

Where it really "falls down" is input common-mode voltage action. If inputs come much away from zero the Q1 Q2 current changes, ill-effects down the chain. But in this day we didn't use an amp unless we NEEDED amplification and lots of it. (These things cost too much.) And designers had only recently moved away from consoles based on fixed-gain tube modules with typical gain of 35dB. Input to an amp was rarely over -30dBm, 24mV, which is "nothing" compared to the 14V across R1.

And the input shift for larger input levels may be why folks are still lusting for the 1731 when the later Jensen 990 fixed most of these quirks.

 

[Samuel Groner]

I've quickly put it into LTSpice and it's a surprisingly efficient design. The DC open-loop gain is above 100 dB, and the GBW product in excess of 100 MHz at audio frequencies. Unity-gain bandwidth is 5 MHz.

The connection of R9 is quite elegant--it is bootstrapped to the output voltage, and thus the collector currents of Q3/Q4 remain roughly balanced during large-signal output.

Samuel

 

[tv]

I played with this opamp in simulation (and a modded prototype) a couple of years ago and the R5/R10 is from memory a sneaky trick (forgot exactly what it does tho), R8 is there for peak output simmetry (but could be done differently), while R6 could be also done a bit different to provide similar "bootstraped" pseudo-constant current load for Q3 as it is done for Q4 with the R8.

 

[Whoops]

I have an original Non-Working Melcor 1731 that was replaced in an OSA Mic preamp.

I dont remember now what the problem was, if it stopped working or making noise. I can test it for sound when I go back to the shop. It was replaced by a gar1731.

If anyone around here is interested in having it for studying or research I will gladly send it to you for free anywhere in the world.

The components seem to be mounted upside down, and some components seem that were not mounted on the board but with wires, all the spaghetti and components are surrounded by some sort of transparent silicone rubber inside the black plastic case.