S/N ratio is, especially for CW, a function of detection bandwidth, or in the case of an SDR, a function of sampling rate and fft size. For a regular analogue radio with real HW filters, the CW power is (hopefully!) considerably more than the noise, so the (peak) power measurement will fairly accurately depict the carrier energy level with even quite wide detection bandwidth. For noise however, we need to consider the noise bandwidth. The broadband, filter passband-filling noise power (which should be detected in RMS by the way) will be highly dependent on the bandwidth of it, i.e. how much noise energy is being integrated into a number. Increasing the detection bandwidth from e.g. a ~300Hz CW filter to a ~3kHz SSB filter will result in 10dB more (noise) power. In other words the CW carrier S/N ratio will appear 10dB worse with a 3kHz filter than a 300Hz filter
The point is, to make any kind of S/N ratio comparison, we need to know the modulation (CW is very different from 400BPS PSK) and we need to know the detection bandwidth. And in the case of SDR it is important to know the sample rate and FFT size (bin count). Only in this way can we compare apples with apples or make orange/apple conversion calculations to make comparison possible.
And for really meaningful S/N measurements, peak detection should be used for the (CW) carrier (S) and RMS detection for the noise (N).
It does not hurt to realize that the noise factor (N) is often noise + interference (I) and that a more accurate measurement would be S/(N+I), but fortunately for us, due to it's very high linearity, the QO-100 transponder noise floor has been and is pretty much random noise. In AO-7, AO-40 and the like, this was not so simple as HELAPS and other high efficiency RF amplification and other linearization and non-linearity processes resulted in noise-sounding (N+I.)
The result of all this is that 9dB SNR can easily be the same this as a 25 dB SNR. Only the measurement conditions are different.
73 - Michael, oh2aue