Noise in high frequency signal generators from an amateur

Tl;dr: Don’t trust the noise performance of high-freq signal generators except at the signal frequency.


The large background current measured in our sample made me feel suspicious of the performance of the high frequency signal sources: HP 8753D Network Analyzer and E8257D PSG Analog Signal Generator, so I checked the frequency specturm of the source using an oscilliscope, and the result turned out to be very interesting ┑( ̄Д  ̄)┍


The PSG frequency was set at 51.511MHz and -20 dbM, and the detected frequency peak include:
-3dbM at 51.511MHz (expected)
-30dbM at 1.511MHz -35dbM at 11.36kHz -50dbM at 60Hz

Image - Expected frequency
Expected frequency
Image - Peak at 11.36kHz
Peak at 11.36kHz
Image - Peak at 55Hz
Peak at 55Hz

The 1.511Mhz, 11.36kHz and 60Hz frequency may be coming from the working principle of the generators: my guess is that they excite the signal with oscilliscope working in different ranges, like 100kHz to 10MHz. Another possiblity is just the integration artifact of the oscilliscope.

Another quick check on the <100Hz resonance also seem to be interesting, below are the results: When PSG signal is set at 60MHz, there is a peak at 40Hz;
40MHz shows 25Hz;
30MHz shows 19Hz;
10MHz shows 6.5Hz.

In summary: Don’t trust the generators except at the target frequency. Otherwise, use a bandpass filter.

Image - Our suspended graphene device
Our suspended graphene device