I am trying to measure the modulation quality of low-speed, low-bandwidth (4.8ksps/12kHz) 4-FSK modulation using the E4406A and VSA 89601A software. The 4-FSK baseband generator i am using does not have a very accurate clock. It is rated for 250ppm, enough for USB, and it is well within that range. When measuring the signal quality with the software, and changing only the symbol clock rate, I find the software reports better modulation characteristics (lower FSK error) when the clock is slightly slow than when it it slightly fast. In this case the difference is between running the clock at about +100ppm vs -100ppm from the reference frequency. The FSK error reported more than doubles when running faster. Even when running significantly slower (-500ppm) the FSK error does not really change from -100ppm.
This seems rather odd to me. I am hoping someone with more experience in this area can shed some light on what I am seeing, whether it would be considered normal, and, if so, maybe explain why.
The baseband generator, for those curious, uses an STM32L4 with a 20ppm 32.768kHz LSE. The system clock is based on the MSI at 48MHz running in PLL mode (slaved to the LSE).