improoving SNR of OTA ?

[Haible Juergen]

Just another crazy idea:

We all know that one can improove the SNR of a transistor by connecting
several ones in parallel, or the SNR of a diode (as VC Resistor) by
connecting several ones in series.
Now what about this: Build a differential pair with variable tail current
(i.e. core of OTA), and connect 10 diodes in series with each emitter.
This would allow 11 times larger input levels, but the noise of each
element would only contribute in a geometric (and not arithmetic) law.
This would only work if the noise of a diode and the noise of a transistor
would be the same order of magnitude. No idea if this is the case or
not. Anybody knows for sure ?

JH.