Hi

I was wondering if anyone has any experimentally observations or insight they can share on reducing the A/D module's clock of period=Tad below the minimum recommended. For example, yesterday I incorrectly calculated the divisor required to satisfy the min.Tad of 75ns for PIC24FV. I was clocking at ~20ns, yet the error was not evident from eyeballing the voltages I was getting back. Based on this, I suspect that only the precision of the lower few bits is compromised, and that there's probably quite a large safety margin building into the 75ns eg. 75ns will be derived from the maximum possible internal stray RCs whereas typical stray RCs may be much lower. Also for simplicity they may only quote the maximum of the minimum Tads for all the various application scenarios, but if you're only using 10bit precision rather than the full 12bit option then perhaps the real min. Tad could be ~35ns.

Any thoughts?

Robin

 

Based on this, I suspect that only the precision of the lower few bits is compromised, and that there's probably quite a large safety margin building into the 75ns eg. 75ns will be derived from the maximum possible internal stray RCs whereas typical stray RCs may be much lower. Also for simplicity they may only quote the maximum of the minimum Tads for all the various application scenarios, but if you're only using 10bit precision rather than the full 12bit option then perhaps the real min. Tad could be ~35ns.

Any thoughts?

Robin

Better would be to just run the A/D with fewer bits of conversion. I think that the PIC's allow this. I don't have time to verify this right now....

 

Better would be to just run the A/D with fewer bits of conversion. I think that the PIC's allow this. I don't have time to verify this right now....

Thanks Richard. I have the option of 10 or 12 bits which is 12 or 14 Tad per conversion respectively, so not much of a saving. My thinking is that cutting Tad by 50% knowing that the error introduced would only be comparable to the 12th bit's resolution would be much more worthwhile

 

Any thoughts?

Robin

Unless this is absolutely necessary (and checked for possible chip to chip variations) I wouldn't do it with a production device.

 

I would use an external a2d.
I designed a USB oscilloscope and found the external a2d about 10 times quicker.