Eeyore said:
I'm a great believer in local feedback. Linearises stages nicely and
improves HF and phase response.
But it does have limited value if such local feedback is in an overall loop.
In which case, it may well make no practical difference at all. Its a
topological thing, and the free lunch issue. The above circuit with a dual
loop cannot be stabilised at the same BW as the one with no local loop. The
local loop does not improve UG phase response, it makes it worse.
For example. Consider inserting emitter resistors in the input differential
pair of a power amp. The argument goes that well, "linearises stages
nicely...", unfortunately, it usually doesn't achieve much.
Inserting the resisters, reduces the loop gain. That is, consider that the
other stages have been *optimally* designed, such that they produce all the
gain that is practically achievable for that topology. In this case,
irespective that the input satge might be deemed to have lower distortion,
due to the local feedback, the resisters reduce the overall loop gain. This
reduction in gain, is less gain to reduce distortion. I have performed
experiments on this, and simulations, and by and large, I have found that
the net distortion is always lower or the same, without the resisters, all
things being equal
Regarding, the "HF and phase response." bit, it often doesn't. If the local
BW improvement is being achieved by feedback, when you actually do the full
sums of stability, you get back to where you were. I have already given the
2 stage, loop within a loop example, and the same analysis usually applies
here.
What emitter resistors can buy you is slew rate, i.e some speed. This is,
by arguments I have already given, at the expense of accuracy.
An issue here, is that one has to make sure apples are being compared with
apples. Its all too easy to not actually do a proper AB examination of the
circuit topologies.
For no overal loop feedback, local feedback is great. e.g the input stage of
the earlier studiomaster mixers, and does what you say.
If you need to, you can make up some
lost overall gain with jellybean transistors (or ICs) at the front
end.
If you are *optimally* designing an amplifier. That is, it is as fast as
possible with the devices available, then you can't add any extra gain
without having a detrimental effect on stability. If you find that you can,
then you haven't designed the amp as best as could have been done in the
first place. Again, the issue with additional i.cs to make up the gain, is
that they have another fundamental roll off to deal with, and fundamentally,
the goal is to have at most, only two major poles at most. Another stage,
just moves the stability problem to another place, it don't away.
When one is designing an i.c amplifier, one is generally (often?), always
trying to get the maximum BW possible, ie at whatever the Ft limits of the
process will allow. This means that one is always running out of steam for
that process. There is therefore simply no way to add any extra gain,
without the phase penalty of that gain.
If you use an amp inside the loop of an amp, if it is not the main dominant
roll off, or have a BW exceding the UGB that the amp is being designed to
achive, it will just cause more stabilty grief.
Kevin Aylward
www.kevinaylward.co.uk
