Discrete Opamp Open Design

[stinius]

Originally Posted by bcarso

Richard, in your Linear Audio article, Table 1 shows an entry for signal-to-noise ratio with the input shorted of - [sic] 135dB. Is this for a reference level of 1V rms at the output, into 30 ohms?

Unweighted, I suppose.
Amazing.

[RNMarsh]

Originally Posted by scott wurcer

Folks here know me well enough I have nothing to sell and could care less what you think of me. The truth is you do have something to sell unless your web site is just for show.

I dont have a web site. Those products are long ago gone and company dissolved. The brand name (MARSH) is still used by Asian company (Magnet Technology) who made them originally. They have my blessing to sell the designs with my name in Asia. I havent recieved any money in years. And, they are not making or selling headphone amps that i know of. -RNM

[bcarso]

Originally Posted by stinius

Unweighted, I suppose.
Amazing.

Also highly unlikely, since e sub n from the 1k feedback divider alone will produce 13.4nV/sqrtHz at the output, which if flat with frequency will amount to 1.89uV rms. For 135dB S/N the signal will need to be 10.65V rms, more than 12V rails will permit. Hence my question.

[kevinkr]

I think a few of you guys need to take the tone down a notch or two...

[john curl]

I have a website too: Vendetta Research.com Yet there is NO product. Same as Richard.
And EVEN if we made product and advertised it, so what? You calling us, crooks? '-)

[RNMarsh]

The same table indicates the s/n in open loop measured -85dbv ref 1 v with input shorted and BW of 30KHz and No weighting applied. You can calc the closed loop noise from that. [Weighting is the norm. It can also make things look a lot better.]

Thx - Dick marsh

[bcarso]

Originally Posted by RNMarsh

The same table indicates the s/n in open loop measured -85dbv ref 1 v with input shorted and BW of 30KHz and No weighting applied. You can calc the closed loop noise from that. [Weighting is the norm. It can also make things look a lot better.]

Thx - Dick marsh

So are we to suppose that you calculated the closed loop noise from the open loop noise, rather than measuring it? I can't reconcile the stated results of 135dB S/N closed loop, especially if referred to 1V rms, with the analyzed, and as well, simulated behavior of the circuit.

Again, at best, the 1k feedback divider resistor alone limits S/N referred to 1V to a far smaller value than 135dB. I get at least 13.2nV/sqrtHz at the output, in a 30kHz BW, thus 2.29uV rms, which below 1V is an S/N 112.8dB. Not shabby and would trouble no one, but not the stupendous 135dB. And this is ignoring the JFET noise, and the noise of the feedback R, and the noise of the load resistors in the JFET drains, etc.

If one does herculean efforts with much quieter JFETs (including the nearly unobtainable 2SJ74) with higher gm, higher transconductances, a much smaller inter-source resistance trim, and a much lower impedance divider, it is feasible to push the S/N ratio up a lot. But it is not trivial, and "costs" a lot of voltage and (if operating the output stage in class A) a lot of power.

EDIT: OK I see where, as a computation , and based on the measurement of output noise of -85dBV, with an open-loop gain of 62dB and supposing an open-loop bandwidth of 25kHz, you could approximate the noise with closed-loop gain of 12dB as 135dB below 1V. However, this cannot be reconciled with the noise as described above. Etwas ist nicht in Ordnung


A free pdf of the Chaldean numerological correspondence for the alphabet for the first person to satisfactorily explain the discrepancy

[Wavebourn]

http://www.google.com/url?sa=t&rct=j...rVeJl6ZImBRIAA

[kevinkr]

I read and reread bcarso's comments on the theoretical snr of the RN Marsh discrete op-amp and after thinking about it worked the same analysis, and conclude he got it correct - this is at least the way I analyze my own designs when not being too lazy... I understand the argument, but the actual feedback components are not improved by feedback so adding the feedback margin to the calculated open loop noise performance doesn't look right to me either. Scaling the resistor values in a downward direction which looks possible and reasonable here would help a little, ultimately I suspect the device source resistance, external source resistors and balance pot contribute significantly as well, the actual noise is probably a couple of dB higher than calculated.

IMO the SNR is probably more than good enough, and I don't get me wrong I think this is a pretty cool design.

[bcarso]

Originally Posted by Wavebourn

http://www.google.com/url?sa=t&rct=j...rVeJl6ZImBRIAA

Hey no fair Wave!