Hi,

I need to build in some gain compression in one of my amps so that the gain gradually drops off as the signal amplitude increases.

I could use a piecewise-linear gain characteristic but this gives too much distortion around the transition points.  I could also reduce the gain of the amp in general by factor 4-5x but this is hardly a good solution for small-signal operation.

Could someone suggest a good method to realize "soft" limiting?

Thanks
Vivek

Could you give me some references to AGCs? I was thinking along similar lines (AGC/VGA) but I don't know many good references.

Regards
Vivek

One attachement was lost; here is it.

Attachment deleted as it contained copyrighted material.
John M. Khoury. On the Design of Constant Settling Time AGC Circuits. IEEE Transactions on Circuits and Systems - II: Analog and Digital Signal Processing, vol. 45, no. 3, March 1998.
http://ieeexplore.ieee.org/Xplore/login.jsp?url=/iel4/82/14571/00664234.pdf?arnu...

Hi vivek,

Did you consider the possibility of using linear clampers? A comparator and eventually a feedback loop when such comparator is triggered might also help you. When the clamp circuit is on, the effective gain of your amplifier is virtually zero, keepìng the signal at the level driven by the clamper.

Just giving you some other ideas that might work for you.

Rgds
Tosei

Hi, vivek !

A very simple solution to your problem has come just into my mind - sometimes the most simple ideas are the last:

What about a differential stage with a tanh-characteristic, which performs soft limiting above app. 50 mV ?
Regards

Hi Tosei,

the principle you have described sounds interesting to me; can you please describe it in more detail ? (opamp with resistors and a comparator in the fedback loop ?). Thank you in advance.

Hi Buddypoor,

The approach I proposed is basically a closed loop system intended to control the clamp level
for the output voltage of an amplifier.In principle it could be used for either a closed or open
loop main amplifier (the one whose signal is going to be clamped or "soft limited").

The system would consist of two comparators (one for the positive or one for the negative clamp
thresholds) that would be connected to the output of the main amplifier.
If the amplifier's output signal is within the clamp thresholds, then the comparators are "off".
When the amplifier's output signal reaches either theshold, the corresponding comparator
will try to switch state and will be engaged. However, such comparator will be controlling either a current or voltage
that will be fed back to - for instance - the input of the main amplifier. In this way a linear nega
tive feedback loop will take place which will force the comparator to work in a linear fashion:
the comparator will feed back a signal that will try to compensate for the amplifier's input one in such a
way that the output level stays at the preset threshold.
Therefore such loop will comprise the active (linearly working) comparator and the main amplifier.
The larger the loop gain, the harder the clipping on the output signal. Therefore, by controlling
the gain of such loop (in particular the gain of the comparator) the limiting could be done in a more
soft or hard way.

When analyzing stability  for this loop system obviously the comparator has to
be biased at the proper (linear) operating point for performing such analysis.

Also, the tanh method could be combined with my proposed approach, since the in-loop comparator transfer function
characteristic could be done in such way.

Something to particularly consider is the case when the main amplifier is a closed loop one. This is another case
of a multi-loop system: the main amplifier loop and the main amplifier+comparator loop. IF the limiting signal
is fed back to the input of the main amplifier the multi-loop system will have a common break point
and the stability analysis might be simplified.


I hope this better describes what I proposed.

Vivek: I have also some experience with in-loop comparators too and I admit I had a hard time stabilizing those systems,
but this might be relatively simple depending on the type of amplifier and loop characteristics you are dealing with.

Regards
Tosei

Hi Buddypoor,

Unfortunately I´m not the one who invented this although I implemented it a couple of times  :). But certainly it existed before I used although I could not provide you with a reference right now (I´m not sure what exactly could have been published about this, but there has to be some paper or publication around I´d have to look for).

As for stability, as I tried to explain before, the main issue is to get the proper comparator bias point such that the analysis provides valid information. One way to do it is to run a transient analysis and stop it at some point in time when the comparator is limiting the amplifier´s output signal. Then you could run an ac analysis using the operating point provided by the transient final state.

I cannot think right now of special requirements as for the slope of the linear region except for possibly the amount limiting you are looking for given a certain overdrive : for example, the more overdrive on the comparator (the larger the signal if it was not clamped) the higher the gain you need in order to keep the clamp level as accurate (and constant) as possible.

Hope this helps.
Regards
Tosei