AW: Re: Couple EE questions...

Magnus Danielson e93_mda at drum.it.kth.se
Tue Oct 22 15:50:46 CEST 1996


> 
> >
> > It does make a diffrence, not only from the obvious heat problems but also
> > from
> > the noise level.
> 
> Noise is an important point indeed. I underestimated the noise of large 
> resistors
> for a long time. Think of this: Using Low noise opamps, but 500k feedback
> resistors all the time.
> I was surprised when I looked into an old EMS octave filter bank: 741 
> opamps,
> but resistors in the 1.5k range for the filters ! These guys knew exactly 
> what
> they did !

It does sound like a correct range...

> >  By makeing wise selection of the resistors you can balance
> > the
> > termal noise and the current noise. You can teoretically find a optimal 
> source
> > resistance to each input of an opamp, if you stick close to this value 
> (which
> > is op-amp type dependent) you will minimize the noise figure. You migth 
> want
> > to
> > stick in a resistor between the (+) input of the opamp and the ground when
> > doing an inverter for this reason among others.
> 
> Please correct me if I am wrong, but I think the smaller the resistor, the 
> better.
> The lower limit would be set by other criteria like power consumption and
> output drive capability of the opamp, and of course at a certain point
> improovement would become smaller and smaller, but
> does the noise floor really rise again at some point? I thought this whole
> "noise figure" stuff works the *other* way round: select an *opamp* for a
> *given* source impedance, and calculate the loss in SNR compared to
> an ideal resistor of given impedance.  But decreasing the impedance
> should always improove the SNR, shouldn't it ?

Well, you actually balance two sources of noise with each other, one noise 
source is the op-amp input itself and the other is the resistor noise. The 
trick
is to select the resistor value which results in the lowest noise figure.

You can go several ways really, for some applications you can just pick and
choose a resistor value, for some you migth need to pick opamp instead but what
you *really* should do is find the resistor value and opamp which makes the
whole circuit as a total work well. There is other issues within a circuit 
which
could govern the values more... say you are locked to use the CA3094, then
you must select the resistors for it... but if that makes you go out of good
caps you migth need to select less optimal resistors anyway...

I'd say there is no specific way which is correct, so when designing you should
be aware of the problem, have a noise budget and ensure that your design 
becomes
satisfactory in not only the noise sense....

There is however a important thing here... just replaceing the opamps in a 
circuit to better ones (noise wise) can _maybe_ improve the noise figure of the
curcuit.

Cheers,
Magnus




More information about the Synth-diy mailing list