edit- in the version with gain did he use one or two feedback networks. /edit]
I think Ian's post makes that clear
Thanks, but that's a disappointment. Those - inputs are not connected together. That is a pretty common, well known, strategy to increase current output from unused op amp sections.
You're strictly correct in that it is the circuits that are paralleled, rather than the opamps themselves.
But tbf that's a common type of usage linguistically. eg the output of an opamp buffer is often described as "very low impedance".
It is, of course, the circuit that exhibits the low output impedance rather than the opamp device itself that might have a datasheet output impedance of, say, 200 Ohm. But this is multiplied up by the circuit. That's not an exact analogy
but I think the point is clear.
Regardless, it works to lower the noise as well as increasing the available current drive.
I can't verify the dB figures quoted. But professional respect would lead me to accept them as correct.
fwiw last time I used this technique was some decades past in a different application area (control signals for High Voltage work - typically 0 -10V to be multiplied up to 0 - 10kV. Different devices and dc accuracy required - but same principle and a useful benefit.
I have used that more frequently when I have a spare section of TL074 quad op amp left over and could benefit from lower impedance drive capability so I can use lower value, lower noise, feedback network resistors (bifets only drive 2k comfortably). I actually did that inside a DSP project I consulted on after leaving Peavey (this century but already a couple decades ago). When using it for other than unity gain buffers, I connect the + input of the second op amp to the output of the first, but still use the degeneration sharing resistors between the two op amp outputs.
If I'm translating that properly - increases current but not reduce noise ?
Correct me if I am misunderstanding - my thoughts are Malbec assisted tonight
In advance of a high voltage corona / ozone issue that I'll need to address in the morning (not an audio thing apart from thr HV "fizz" noise).
I never considered it to make a lower noise opamp. The Cohen topology with very low noise discrete front end active devices was my goto approach for low noise since the mid-late 1970s (but I didn't know it was called "Cohen" until decades later).
Yes. Am aware of your friction with the "Cohen" term
And I wasn't aware of the term until well after I'd seen the topology discussed.
But yes - it's impressive. I don't think the "parallel circuits" approach being considered here is meant to be regarded as an alternative for it.
But it's a relatively easy win where a Cohen type solution isn't practicable but you have a spare opamp stage.
I am a little uncertain about noise benefit from that simple Self parallel. Since the bottom of the input long tail pairs and both inputs are not connected, this is not the standard active devices in parallel topology known in the literature. I will ASSume that Self documented bench measurements of the noise reduction from his 5532 twofer.
In my time wasted searching the WWW I did find a post from the man himself (many years ago) saying the output noise was too low to measure practically. But I think he was referring to an oddball low power audio amp (headphone amp?) crafted from paralleling multiple opamps together using that same build out output series resistor topology across many devices.
Rings a bell.
Here's a hint, the technique used by uber low noise op amp makers to characterize extremely low noise opamps is to test them while running at some significant closed loop gain (like 100x) then dividing down the output noise by that closed loop gain factor. I recall being critical of that measurement technique (for measuring distortion) because the inputs are terminated with very low source impedances but for basic noise measurements it seems practical.
Yeah - that seems a common datasheet measurement approach.
)
It will work from the wiper of course but with lower performance. So why buy a "premium" part and then knowingly hamper it 
