2024-05-05 07:24 PM
The maximum sampling rate of STM32H743 ADC1 and ADC2 is 3.6MSPS. According to the aroma sampling principle, the maximum sine wave frequency that can be collected is 1.8MHz. I don't know if I understand it this way, right? I want to collect higher-frequency sine waves. Is there any way to use the built-in ADC? Can reducing the number of bits improve the sampling rate and collect higher-frequency sine waves?
Solved! Go to Solution.
2024-05-05 11:50 PM - edited 2024-05-06 12:14 AM
>Can reducing the number of bits improve the sampling rate and collect higher-frequency sine waves?
Yes. 🙂
see ds :
So up to 8 msps in 8bit mode possible; with all 3 ADCs interleaved up to 24msps .
+ see :
https://www.eevblog.com/forum/microcontrollers/stm32-triple-interleaved-example-code-by-st/
2024-05-05 11:15 PM - edited 2024-05-06 12:17 AM
> Can reducing the number of bits improve the sampling rate and collect higher-frequency sine waves?
No.
Edit: see AScha.3's response below, seems I was wrong... I didn't know about that ADC feature.
You cannot trick Mr. Nyquist (okay, there are a very few exceptions under very special conditions).
PS: look for some better translator, or is "aroma sampling principle" something I don't know? 😉
2024-05-05 11:50 PM - edited 2024-05-06 12:14 AM
>Can reducing the number of bits improve the sampling rate and collect higher-frequency sine waves?
Yes. 🙂
see ds :
So up to 8 msps in 8bit mode possible; with all 3 ADCs interleaved up to 24msps .
+ see :
https://www.eevblog.com/forum/microcontrollers/stm32-triple-interleaved-example-code-by-st/
2024-05-06 01:04 AM - edited 2024-05-06 01:06 AM
You can sample any sine of any frequency. You just cannot reconstruct them from the samples if they are above the Nyquist frequency as you then have frequency folding (sampling 9MHz signal with 5MHz reconstructs to a 1MHz signal). However in some cases frequency folding can be beneficial such as using interleaved ADCs or undersampling. But in that case you need a higher analog bandwidth of your ADC, since a signal that's already filtered out can of course not be sampled. So you need to verify that by checking the datasheet.
One common misunderstanding about Nyquist is that you can sample signals (without folding) up to and including Fs/2. This is false as signals at Fs/2 appear as a constant signal (DC). So you can only sample below Fs/2. And in practice you want to go much lower because because your nyquist filter is not a perfect filter so it needs a lower cut off frequency then Fs/2 in order to filter out unwanted signals enough. Also filters distort amplitude and phase of signals in the passband close to the cut off frequency. For instance a 8MHz ADC with a 5th order low pass filter with a cutoff frequency of 2MHz. Signals at 4MHz are reduced in amplitude by 30dB. Signals between 2MHz and 4MHz have reduced amplitude (and therefore also reduced resolution). Signals below 2MHz can then be accurately captured. Don't forget to take into account the bandwidth of the ADC.
As AScha.3 pointed out you can achieve higher sampling rates at different resolutions and configurations for your ADC.