Passband gain variation - how good is good enough?

[Darktrax]

Choose for example a X-Band LNA followed by some transmission line, and maybe a down-converter to L-Band.
There are sometimes also bandpass filter to reject nearby uplink transmit signals.

I can find fine LNA kit out there claiming +/- 1dB gain variation across a passband around 700MHz wide.
Some have a second specification, typically 0.5dB over any 40MHz wide part within.

The down-converter specifications similarly might claim typically +/- 1dB in well respected kit.

So putting together the chain, it is possible that some drop in one place might be compensated by a lift elsewhere, but in the unlucky scenario, the variations might add in the worst way.

When a system is asked for, having less than (say) 0.5dB or 0.6dB gain variation, how can this be reasonable?
So what is the norm here? How good is good enough?

 

[jiripolivka]

Choose for example a X-Band LNA followed by some transmission line, and maybe a down-converter to L-Band.
There are sometimes also bandpass filter to reject nearby uplink transmit signals.

I can find fine LNA kit out there claiming +/- 1dB gain variation across a passband around 700MHz wide.
Some have a second specification, typically 0.5dB over any 40MHz wide part within.

The down-converter specifications similarly might claim typically +/- 1dB in well respected kit.

So putting together the chain, it is possible that some drop in one place might be compensated by a lift elsewhere, but in the unlucky scenario, the variations might add in the worst way.

When a system is asked for, having less than (say) 0.5dB or 0.6dB gain variation, how can this be reasonable?
So what is the norm here? How good is good enough?

Everything depends on the application. If your amp is a part of a communication system, you should know the signal spectrum and how the passband flatness affects the BER or signal distortion.Also, with more components cascaded, passband ripple can become worse if poles and zeros coincide, and reverse. For a single good amplifier, passband gain variation can be < 1 dB. If the system uses FM or FSK it is less sensitive to ripples than if QAM is used.

In radiometric systems where noise power is amplified, 1 dB variation is >10% and not acceptable.

 

[Darktrax]

My thanks jiripolivka.

Yes - I can accept that any radiometric measurement system has to be a lot better.
In this case, the application is indeed a communication system.

I have thought that the various types of PSK modulation, like 8QPSK, BPSK, DQPSK, like any FM, would be relatively insensitive to amplitude wobbles in the passband unless getting near the noise. The Eb/N0 for the kit is high enough to make the BER go far below the limit. I do not expect errors to be a problem.

I have specified equipment all claiming <1dB flatness, and I have to hope that LNA and X-Band down-converter taken together with some low-loss line will not show a 2dB departure. I know the real performance of delivered kit is often better than the advertised specification, but I cannot take it for certain.

Usually I am OK if I simply specify quality kit from vendors of good reputation. Attempting to model these effects is something I have not so far had to attempt.