Hi, how do group delay distortion degrade receiver sensitivity ?

In a digital receiver the group delay can create quadrature errors in the IQ demodulator, and in this way degrading the BER sensitivity.
http://my.ece.ucsb.edu/yorklab/Usefu...s%20AN1314.pdf

Everything depends on how you define "receiver sensitivity".

The general definition is the signal-to-noise ratio, or S/N. This parameter is defined at receiver input where the noise power, N, is defined in dBm as Pn = -174 + NF + 10 log B. The signal level, S, is the signal power coming from the antenna.

This definition is not dependent on signal modulation, so no effect of spectrum distortion caused by group delay. Group delay is only dependent upon the narrowest part of the receiver which is most often the IF band-pass filter. This filter must be designed to pass the modulated-signal spectrum without distortion while limiting the noise power as above.

At receiver output, after the demodulator, the "output S/N" is modulation-dependent. In digital systems, bit-error rate, BER, is a function of the input S/N (also expressed as C/N for wideband signals), and the type of modulation. This function is non-linear, and sometimes it shows a threshold, under which (for lower S/N than a limiting value) the BER grows to unacceptable numbers.

BER doesn't have a threshold as PER (Packet Error Rate), FER (Frame Error Rate), or BLER (Block Error Rate).
PER, FER, BLER can set a fake threshold in the software (making the receiver looking good), when BER doesn't allow doing this.
BER have a smooth characteristic and is direct proportional to S/N of the receiver or NF (thermal noise) of the receiver.

But going back to the initial question which is very specific to group delay, BER can be affected in the same time by the quadrature errors in the demodulator (due to linear distortions as group delay).

Thank you for the details! My field is rather low-noise microwave receivers and radiometers. In my company, after many non-realistic requests for a low GD, we used a VNA to measure various microwave BPFs and LPFs to learn the GD of various bandwidths. Good filters passing reasonably the spectra of different modulation types also had a GD with only small ripple beyond band limits.
No other components (if good) affect system GD.

That's fine. As you know Group Delay of the filters can generate Intersymbol Interference (ISI) in a digital system.
From my experience the front-end high frequency filters (800MHz and up) do not have a GD such bad to produce ISI (or at least in the systems that I worked).
But the problem appears at IF filters, with much narrower BW.

You mentioned radiometer. Isn't thermal noise is uncorrelated over frequency, so that group delay should have no effect on your radiometer measurement?

In any case, for fixed frequency (single tone) sensitivity measurement, group delay has no effect on sensitivity.

Is the group delay maybe a secondary effect, and what reall affects your measurement is filter insertion loss?
Narrow band or high order filters will have group delay issues as well as increased insertion loss.

Unfortunately CDMA and WCDMA are single tone (spread spectrum) systems and they suffer from ISI which might be caused by narrow filters group delay.
This is a common issue in Base Stations design which still use superheterodyne approach for the receiver.

You are right; radiometers evaluate thermal noise, often of natural sources. The band-pass filter is used to select the frequency range of interest, and reject human-made interference. For calibration pass-band response should be flat, not to -3 dB but rather to 1...5%. Filter skirts or reject bands are much less important.

Contrary to the above, communication receivers utilize filters to reduce the thermal noise out of the bandwidth used for intelligence spectrum transmission. With complex modulation schemes, the spectrum becomes not limited to a rectangle (like for AM) but rather trapezoidal. Filters do not match the spectrum shape as their response is more rectangular over frequency. The compromise between limiting the thermal noise and limiting the information spectrum leads to filters narrower than the complete spectrum. This leads also to GD distortion.

@vfone: With single tone I meant unmodulated carrier. We don't know what modulation scheme the original post referred to.