AW: improoving SNR of OTA ?

[Haible Juergen]

	>I know that emitter resistors in diff stages introduce feedback,
	>ie. lower gain, thus allowing higher input levels making SNR
	>better. The (terrible) schematic of my discrete
4-quadrant-multiplier
	>uses this method. The resistors won't introduce much noise.
	>Maybe I don't get the point, but why using diodes as "resistors"
	>with all disadvantages when you're allready going discrete and have
	>good resistor? I don't think that the diodes will linearize
anything
	>in your circuit. 
	>
	>m.c. 

Fixed resistors in the emitter path will introduce local feedback,
i.e. linearizing the circuit. But it will also kill control of gm by means
of the tail current. Good for fixed gain, bad for OTAs.

With diodes, you have current controlled, variable resistors.

The basic principle is the same as with the internal be diode:
increase current -> decrease dynamic resistance. Both have
basically the same (exponential) nonlinearity, so the "quality" of
distortion is the same as before, only that N devices in series allow
for N times the voltage, for a certain amount of THD.

So "Signal" could be increased by factor N, while "Noise" would
only grow with SQRT(N). Thus the improoved SNR.

_BUT_ this will only make sense if all of these N devices have a 
similar noise. This is true for N transistors (in parallel), or N
diodes (in series), but I don't know about the transistor's be
path in comparison with a cheap diode, in terms of noise voltage.

JH.