Part Number: TMS320F280049
I ADC RF received baseband single tone signal.
But the Signal to Noise Ratio after FFT on DSP is 10dB less than actual analog input signal.
I compared FFT length, It look's OK.
Signal level look's strong enough to, no need to fight with ADC base noise.
I got SNR from 20 * log10(RFFTmagBuff), subtract average value from peak max.
One thing I have to try is ADC auto-reload. I'm using timer interrupt for AD sample.
Hi Taeyeong,
Is the sampling coherent or are you using a windowing function? If you are using a windowing function, you may need to center the signal before passing it to the windowing function (so that the mean sampled value is 0). Note that even after windowing, the fundamental signal energy will still usually be spread over multiple FFT bins.
How do you know the SNR of the actual analog input after demodulation but before sampling with the ADC?
The SNR should be the ratio between the sum of all FFT bins in the fundamental divided by the sum of all other FFT bins (excluding harmonics). Depending on the fundamental frequency, the number of harmonics you want to consider, and the sampling rate, you may need to consider that some of the higher numbered harmonics will alias down from the 2nd or higher Nyquist zones.
You use 20*log(sum(fund) / sum(noise)) if the FFT bins are magnitudes and 10*log(sum(fund) / sum(noise)) if the FFT bins are power.
Note that the TI Precision Signal Injector EVM is a great tool if you want to easily get a very very clean sine wave for ADC evaluation.
)
Hi Taeyeong,
What is the source impedance and bandwidth of your source? You may need to use a high bandwidth and low noise op-amp to buffer the signal for the ADC and/or you may need to increase the ADC S+H duration (set by the ACQPS setting in the SOC configuration) to get good results.
You may or may not see the harmonics depending on the level of your noise floor and the distortion inherent in the signal and in the ADC, but you should attempt to identify them in any case. Attached is an example spectrum of a very clean 10kHz sine being sampled by the 12-bit ADC on F28388x; the noise floor in this case needs to be pretty low to see the harmonics.
