Hello Thibaud, Think this in the frequency domain.
A typical decimator 1. applies a low-pass filter to the incoming samples, and then 2. downsamples the data by extracting every n th sample. The low-pass filter does not only filter out non-desired signals, but also filters out noise. So, you must observe reduced noise. In your case, the thermal noise becomes -174 + 10*log(400kHz) when you decimate the signal. Your actual noise floor would be somewhat higher than the thermal noise (because real-world devices never achieve the thermal noise bound), however, and the actual noise level would depend on various factors, including the hardware itself and the low-pass filter design of the decimator. Regards, Kyeong Su Shin ________________________________ 보낸 사람: Thibaud Vial via USRP-users <[email protected]> 대신 USRP-users <[email protected]> 보낸 날짜: 2019년 12월 7일 토요일 오전 12:34 받는 사람: [email protected] <[email protected]> 제목: [USRP-users] Noise vs Sample rate Hi, I'm currently working on USRP sensitivity and noise. Noise power depends on receiver bandwidth (https://en.wikipedia.org/wiki/Johnson%E2%80%93Nyquist_noise#Noise_power_in_decibels). But what if : - I sample a signal at 4MHz with USRP + GNURadio (Thermal noise power = -174 + 10*log(4MHz) ) - I add a decimation by 10 in GNURadio What is the new thermal noise power ? The same, or -174 + 10*log(400kHz) ? I've made some tests with USRP + GNURadio, and the noise seems to be lower after decimation, but not of 20 dB (more of 10 dB). Thanks for your help,
_______________________________________________ USRP-users mailing list [email protected] http://lists.ettus.com/mailman/listinfo/usrp-users_lists.ettus.com
