Post History
Doing peak-detection in firmware is a valid method. I have done similar things. At the bare minimum, you want at least one sample guaranteed within the 100 µs pulse width. That means 100 µs samp...
#2: Post edited
by
Olin Lathrop
·
2026-07-08T22:29:49Z (2 months ago)
- Doing peak-detection in firmware is a valid method. I have done similar things.
At the bare minimum, you want at least one sample guaranteed within the 100 µs pulse width. That means 100 µs conversion period, which is 10 kHz conversion rate.- However, you also have to take the acquisition time before each conversion into account. The signal is not point-sampled, but acquired over a finite interval. The overall sample rate needs to be fast enough so that there is always at least one whole acquisition interval in every 100 µs.
For example, let's say the sample period is 10 µs. That would require a sample period of 90 µs absolute maximum, which is a sample rate of 11.1 kHz.- In reality, I'd want two whole guaranteed samples over the minimum 100 µs period. That way at least one of them isn't near an edge. Using the 10 µs acquisition time example again, that leaves 45 µs for each sample, for a sample rate of 22 kHz. Fortunately, that seems to be well within the capability of your A/D, so the digital peak detection method should work fine given what you have told us.
- Doing peak-detection in firmware is a valid method. I have done similar things.
- At the bare minimum, you want at least one sample guaranteed within the 100 µs pulse width. That means 100 µs sample period, which is 10 kHz sample rate.
- However, you also have to take the acquisition time before each conversion into account. The signal is not point-sampled, but acquired over a finite interval. The overall sample rate needs to be fast enough so that there is always at least one whole acquisition interval in every 100 µs.
- For example, let's say the acquisition time is 10 µs. That would require a sample period of 90 µs absolute maximum, which is a sample rate of 11.1 kHz.
- In reality, I'd want two whole guaranteed samples over the minimum 100 µs period. That way at least one of them isn't near an edge. Using the 10 µs acquisition time example again, that leaves 45 µs for each sample, for a sample rate of 22 kHz. Fortunately, that seems to be well within the capability of your A/D, so the digital peak detection method should work fine given what you have told us.
#1: Initial revision
by
Olin Lathrop
·
2026-07-08T22:26:08Z (2 months ago)
Doing peak-detection in firmware is a valid method. I have done similar things. At the bare minimum, you want at least one sample guaranteed within the 100 µs pulse width. That means 100 µs conversion period, which is 10 kHz conversion rate. However, you also have to take the acquisition time before each conversion into account. The signal is not point-sampled, but acquired over a finite interval. The overall sample rate needs to be fast enough so that there is always at least one whole acquisition interval in every 100 µs. For example, let's say the sample period is 10 µs. That would require a sample period of 90 µs absolute maximum, which is a sample rate of 11.1 kHz. In reality, I'd want two whole guaranteed samples over the minimum 100 µs period. That way at least one of them isn't near an edge. Using the 10 µs acquisition time example again, that leaves 45 µs for each sample, for a sample rate of 22 kHz. Fortunately, that seems to be well within the capability of your A/D, so the digital peak detection method should work fine given what you have told us.