Write to file at exact ms intervals

[Ray.R]

0.1ms is usually considered pretty good.  1 sec delta for every 10000 samples.

Have you considered the two loops and set each to a different core?

OR 

Trigger writing to file at every 10 second interval and see if that causes some improvement. 

Also, did you post your entire code?  We could look at it and see if there is some performance improvements that could be suggested.

[Ben]

---

@Ray.R wrote:

0.1ms is usually considered pretty good.  1 sec delta for every 10000 samples.

Have you considered the two loops and set each to a different core?

OR 

Trigger writing to file at every 10 second interval and see if that causes some improvement. 

Also, did you post your entire code?  We could look at it and see if there is some performance improvements that could be suggested.

---

Agreed (to both postings).

The best I ever got with a timed loop (under Windows) was to keep multiple loops (about 20-30) running at 1KHz without missing any cylces. I was using a hardware clock to do that.

For time stamps with delta that small I would use the hardware to time the acquisition and let the hardware keep up. I then gather what was collected afterwards.

Paraphrasing from on Old Memorex Disk drive service Manual "Attempting to (get zero read errors) get anything to run deterministically under Windows is an exercise in futility".

If you need tighter software time stamps, then RT is the way to go.

Ben

[Ben]

looking back over this thread...

The original error was 0.1 seconds.

TBob's version is showing 0.0001 jitter.

I suspect the change from Tbob was required to fix the 0.1 s error and what we have left over is due to the OS.

So are we done here?

Ben