I am not wise in the ways of those circuits but here is my thought. Download the circuit path specs for the Onyx 800R and the True Precision and also several different kinds of Neve stuff. Some of the newer Neve stuff is fixed gain/attenuation and some is continuously variable. Go ahead and do the same with your other favorite preamp modules. Now compare the layouts and your subjective/objective opinions on the pros and cons of each aural signature.

Thanks for the Input! This is what I am after with the thread. Varying the gain in the feedback usally affects the stability of the amplifer, as it can move poles around. This would mean at the very least the amplifier would behave differently at different gain levels. So the sound would changes as you cranked the gain.

Some designers like to eliminate this variable, and some use it for an effect.

For example: John Hardy vaires the Ri resistor to adjust the gain of his MPC-1 pre-amp.
http://www.johnhardyco.com/pdf/990.pdf

Where as Seven Circle attenuates the input on there N72 mic pre:
http://www.seventhcircleaudio.com/N72/N72R31/docs/n72_sch.pdf

So what have other people found it their work??

It all depends on what you are trying to achieve. I must have designed many thousands of amplifiers of different types, the majority of them being for instrumentation, where it is usually required to have a fully-specified response down to d.c. and a minimum bandwidth maintained over the range of gains. Multiple stages of amplification, each with switchable or programmable gain can be used to achieve this. For example, the input stage may have gains of x1, x10 and x100, and the second stage have gains of x1, x2 and x5, the combination giving gains up to x500 in a 1,2,5,10 sequence. These are almost invariably implemented as positive-gain amplifiers to give non-inversion and a high input impedance to each stage. The gain resistors are in a string from the output of the stage to ground, and feedback taken from a switchable or continuously-variable tap on the resistors back to the inverting input. This configuration eliminates the effect of switch impedance on gain accuracy (important in programmable-gain designs).

For low-level audio designs such as microphone pre-amplifiers, noise figures are important, so it is not usual to operate the amplifier at fixed gain with an input attenuator ("oscilloscope configuration"), as that gives a variable signnal-to-noise ratio but fixed bandwidth over the gain range. The John Hardy design is a positive-gain configuration similar to the instrumentation amplifiers I mentioned above. The Seventh Circle design uses distributed-gain, with input attenuation at low gains (e.g. for line-level inputs), and positive-gain amplification for higher gains.

Continuously-variable gain in a simple audio op-amp design is easily done using a potentiometer from the op-amp output to a.c. ground (i.e. to ground via a large capacitor). The negative input of the op-amp then connects to the wiper of the pot. In this way, the negative input has a bias current path but the d.c. response is not part of the gain equation, and so the offset at the output does not vary with gain. The pot should be an inverse log-law.

The is no definitive answer to your question. If the design has to cope with large inputs, it would be common to have the lower-gain ranges apply an input attenuation, and then the higher-gain ranges achieved by variable feedback an no input attenuation.

Thanks Boswell, great response!

Continuously-variable gain in a simple audio op-amp design is easily done using a potentiometer from the op-amp output to a.c. ground (i.e. to ground via a large capacitor). The negative input of the op-amp then connects to the wiper of the pot. In this way, the negative input has a bias current path but the d.c. response is not part of the gain equation, and so the offset at the output does not vary with gain. The pot should be an inverse log-law.

I have seen that idea once before and wondered what the cap was for... Thanks!

Keep those examples coming....