Hi everyone, as we know that amplifiers suffer from “gain compression” because of its nonlinearity. Assume the input of ampllifier is “vi=Vi*cos(wi*t)”, the amplitude of the output fundamental wave can be expressed by the following formula (see the following fig.).

When “k3>0”, it is called “gain expansion”; while “k3<0”, it is called “gain compression”. Well, we know that gain compression is the real case for amplifiers. But what are the physical reasons for gain compression of amplifiers? Or, anyone can physically explain why “gain expansion (k3>0)” cannot happen for amplifiers?  Thanks!!!
[img][/img]

jianke,
in a classical sense the total power in any compression is distributed among harmonics and distortion products. the reason the compression of desired tone is experienced is that power is distributed at non desired frequencies.  This is another way of looking at it and RFICDUDE is correct in example of some cases where this occurs, but ultimately the compression is a result of other harmonics playing a bigger part of the output spectrum, where the gain is still achieved until clipping but the desired frequency gain is competing against the introduction of new tones competing for the output power.  In other words total power gain deviating from desired frequency gain is what is observed in gain compression.

http://rfcooltools.com

aaron_do,

I believe what you are referring is clipping which is well beyond compression.
Consider 1dB compression point extrapolated from IM3 why does it happen around 10dB from the intercept?  Its not a coincidence that the power going into the IM3 products at 10dB backoff are equal to the power lost in the desired.  

http://rfcooltools.com

Hi rfcooltools,


I understood your point, but I think voltage headroom and current limiting are the root cause of the compression. IMO, the harmonics and intermodulation components are just a consequence of the compression. Also, your explanation doesn't cover gain expansion. When the gain expands, you get more power in both the fundamental and the harmonics.


regards,
Aaron

Hi, thanks sooo much for all your reply. I read the book <Microwave Engineering(2011_4th) David Pozar>.
It simply explain the physical reason for the "gain compression". But I still cannot understand it. Anyone can give me some comments?

At the risk of being heckled for my ignorance, let me ask a question which deviates a fair bit from the main topic of discussion.

Is gain compression a factor which limits performance of RF circuits critically? If so, is it at all possible/common practice to avoid the unpleasant effects of this by use of an on-chip 1:N transformer which would deliver significantly larger signal swings at the load side (Nx) while keeping swings low at amplifier output?

Or are on-chip transformers not yet efficient enough?

In some sense, the various load matching techniques used are doing something similar but while they all provide more signal at output (more power), they do not always avoid gain compression.

Vivek

Hi all,


I agree with loose-electron (i.e. what's important is what's really happening), but on the other hand its often useful to be able to put some numbers in. For instance, using what rfcooltools said, if you're at the 1dB compression point, you should be able to roughly calculate your third harmonic (I think it needs to be a well behaved system).

Vivek,

its a tradeoff (as it always is ). Maximizing the signal swing is necessary if you want good effiency. Also, reducing the load impedance (using your method or any other) reduces the power gain of the amplifier. On-chip transformers are used extensively in practice, but whether they are efficient enough depends on the application. Also, the method you described is exactly a load matching technique. Whether you choose LC matching or transformer matching depends on a few things, but transformation ratio is one factor. In theory, LC matching is better for small transformation ratios while transformers are better for large transformation ratios. However, in practice, it is difficult to realize large transformation ratios using on-chip transformers.

Correct me if I've misunderstood your question.


regards,
Aaron