"Three Chip" opamp preamp concept

So - as far as i've seen, guite a few classic preamps contain roughly three building blocks: Unbalancing via a transformer into a single op-amp gain stage into rebalancing via a transformer.

The transformers and op-amps all contribute with their own sound, so clearly we're not aiming for transparent "wire with gain" territory here.

Just out of curiosity: If unbalancing and rebalancing was to be done via op-amps instead - given the concept outlined in the pseudo schem below, how would you think a circuit like this would behave in terms of noise and coloration?

I'm well aware of THAT corps range of preamp and balancing chips, but the fun in this project is constructing a low-budget, small-footprint preamp using standard components only.

I suspect the transformer would be quieter at unbalancing at unity gain and would provide a balanced floating input. Similarly a transformer output also provides a balanced floating output which op amps do not.

Cheers

Ian

grrrunge said:

Just out of curiosity: If unbalancing and rebalancing was to be done via op-amps instead - given the concept outlined in the pseudo schem below, how would you think a circuit like this would behave in terms of noise and coloration?

Click to expand...

The best I've done with transformers is Noise Factor 1.6dB using a large Lundahl transformer & 5534 in my Calrec days. (ie 1.6dB from theoretical perfection)

Your circuit is essentially what the Cohen and its lesser copies do with one caveat.  For low noise, you want to have most of the gain in the very first stage.  That's why they have the gain switch on the emitters of the input transistors.

Crude versions like AMEK 2500 will have less colouration than a good transformer ... but worse noise.  SOTA like a properly implemented Cohen (or even THAT 1510/12, Eden etc) will have less noise too.

ruffrecords said:

I suspect the transformer would be quieter at unbalancing at unity gain and would provide a balanced floating input. Similarly a transformer output also provides a balanced floating output which op amps do not.

Cheers

Ian

Click to expand...

By floating, do you mean they're galvanically isolated from both chassis and circuit?

ricardo said:

Your circuit is essentially what the Cohen and its lesser copies do with one caveat.  For low noise, you want to have most of the gain in the very first stage.  That's why they have the gain switch on the emitters of the input transistors.

Crude versions like AMEK 2500 will have less colouration than a good transformer ... but worse noise.  SOTA like a properly implemented Cohen (or even THAT 1510/12, Eden etc) will have less noise too.

Click to expand...

That makes sense, I get that

Anyway, me being a first semester EE student, I'd much rather build a ton of cheap fun stuff, poke it with a stick and see how it works, than spend my limited cash on expensive audio transformers.

Am i correct in assuming (theoretically) that the noise in the input buffers will be the same, and cancelled out in the following differential amplifier - thus only carrying the noise from the differential amplifier to the next stage?

grrrunge said:

ruffrecords said:

I suspect the transformer would be quieter at unbalancing at unity gain and would provide a balanced floating input. Similarly a transformer output also provides a balanced floating output which op amps do not.

Cheers

Ian

Click to expand...

By floating, do you mean they're galvanically isolated from both chassis and circuit?

Click to expand...

Exactly.

Cheers

Ian

grrrunge said:

Am i correct in assuming (theoretically) that the noise in the input buffers will be the same, and cancelled out in the following differential amplifier - thus only carrying the noise from the differential amplifier to the next stage?

Click to expand...

No.  Noise is uncorrelated and will add.  Single OPA differential amps are also very noisy.  All that will be amplified by your 3rd 'Gain' stage.

There's good reason why a 3 OPA instrumentation amp (of which Cohen, THAT1510 etc are sophisticated examples) have the majority of their gain in their 1st stages.  All the complexity is directed at making that 1st stage as quiet as possible.

ricardo said:

grrrunge said:

Am i correct in assuming (theoretically) that the noise in the input buffers will be the same, and cancelled out in the following differential amplifier - thus only carrying the noise from the differential amplifier to the next stage?

Click to expand...

No.  Noise is uncorrelated and will add.  Single OPA differential amps are also very noisy.  All that will be amplified by your 3rd 'Gain' stage.

There's good reason why a 3 OPA instrumentation amp (of which Cohen, THAT1510 etc are sophisticated examples) have the majority of their gain in their 1st stages.  All the complexity is directed at making that 1st stage as quiet as possible.

Click to expand...

Alright. Bear with me here, I'm trying to get this stuff right in my head
The noise added before the differential stage consist of noise imposed on the signal during the cable run and noise from the buffer stage. Am i right?
Are there more sources of noise I'm not getting yet?

Is there any reason why this noise would not be similar?
How come all that noise that gets picked up doesn't get cancelled out in the differential stage, when the one signal gets flipped 180?

Also, how noisy is that "very noisy" you're talking about? Ballpark value

Understanding all this stuff is why I decided to begin studying electrical engineering in the first place. I'm just one curious bugger, and I'm not afraid of quite a bit of complex math if that's what it takes I hope you understand...

grrrunge said:

The noise added before the differential stage consist of noise imposed on the signal during the cable run and noise from the buffer stage. Am i right?
Are there more sources of noise I'm not getting yet?

Is there any reason why this noise would not be similar?
How come all that noise that gets picked up doesn't get cancelled out in the differential stage, when the one signal gets flipped 180?

Click to expand...

The noise picked up by the lines and the noise in your 'buffer' stage are 2 separate issues.

With a proper balanced line system, EMI picked up by the lines is cancelled because it is EXACTLY the same on each line. eg http://sound.westhost.com/articles/balanced-2.htm
Noise in your buffer is uncorrelated so it ADDS in the differential stage.

Also, how noisy is that "very noisy" you're talking about? Ballpark value

Click to expand...

The "Simple Single Opamp in Fig4 of the above link has 4x10k resistors.  Each introduces 12nV/rtHz of uncorrelated noise.  Because uncorrelated, you do an rms sum

Vn = rt(4 x 12^2) = 24nV/rtHz which is amplified by your 'Gain' stage.

Cohen has 0.5nV/rtHz at the equivalent point and THAT1510/12 about 1nV/rtHz

Understanding all this stuff is why I decided to begin studying electrical engineering in the first place. I'm just one curious bugger, and I'm not afraid of quite a bit of complex math if that's what it takes I hope you understand...

Click to expand...

Go out and buy The Art of Electronics - Horowitz & Hill

It will help you now and you will still be referring to it when you are a low noise guru in your old age.

Also download http://www.leonaudio.com.au/microphone.engineering.handbook..chapter.8.pdf which has Great Guru Baxandall's last thoughts on LN design.  You won't understand it all now but try.  Cohen's original paper and some later thoughts are also on that website.