While loop deterministic timing

[djs_at_eaton]

I have a program operating a series of tests.  The main while loop period is set at 30mS.  The picture shows the configuration.  The wire to dt carries the value 30.

 

The loop contains a call to a DAQ running at 20KHz.  During the 30mS loop time, the program acquires waveform data, condenses it, and gets the timestamps.  During tests, the condensed data and timestamps are saved in a buffer.  When the test completes, the buffered data is formatted and written to a file.

When the file is written, timestamps are subtracted from t0 and the differences saved to the file.  A sample of the file output is below, with some data removed to get under the 10,000 character limit.

Notice that for most lines the time difference is 31mS, but every 11th or 12th line it appears that the loop time was 15 or 16mS.  Why would this be happening?

Sec	Pos (mm)	Amps	Brake V	Time	Cycle
0	-0.105	0.014	0.676
0.031	-0.105	0.014	19.083	0.031
0.062				0.031
0.094	-0.105	0.542	26.713	0.032
0.125				0.031
0.156	-0.101	1.015	26.751	0.031
0.187				0.031
0.219	-0.008	1.462	24.335	0.032
0.234				0.015	0
0.266	0.24	1.859	13.519	0.032	1
0.297				0.031	2
0.328	0.503	2.253	13.526	0.031	3
0.359				0.031	4
0.391	0.769	2.113	13.529	0.032	5
0.422				0.031	6
0.453	1.048	2.022	13.531	0.031	7
0.484				0.031	8
0.516	1.323	1.962	13.534	0.032	9
0.547				0.031	10
0.578	1.604	1.947	13.536	0.031	11
0.609				0.031	12
0.625	1.875	1.94	13.536	0.016	0
0.656				0.031	1
0.687	2.165	1.93	13.537	0.031	2
0.719				0.032	3
0.75	2.444	1.934	13.539	0.031	4
0.781				0.031	5
0.812	2.73	1.936	13.54	0.031	6
0.844				0.032	7
0.875	3.001	1.937	13.541	0.031	8
0.906				0.031	9
0.937	3.288	1.937	13.541	0.031	10
0.969				0.032	11
0.984	3.569	1.939	13.542	0.015	0
1.016				0.032	1
1.047	3.856	1.931	13.542	0.031	2
1.078				0.031	3
1.109	4.13	1.918	13.542	0.031	4
1.141				0.032	5
1.172	4.418	1.909	13.542	0.031	6
1.203				0.031	7
1.234	4.698	1.897	13.543	0.031	8
1.266				0.032	9
1.297	4.978	1.89	13.544	0.031	10
1.328				0.031	11
1.359	5.264	1.89	13.545	0.031	12
1.375				0.016	0
1.406	5.543	1.89	13.545	0.031	1

[crossrulz]

I do not understand the fine details for the timed loop, but I'm willing to bet that your problem here has something to do with one of these settings:

[dkfire]

It proberbly also has something to done with the resolution of the timestamp that the OS can supply.

Windows XP has only a resolution of around 15ms in it's timestamp.

You should use the Tick Count (ms) vi or some of the time info from the while loops Left Data node for this type for timing measurement.

[Mark_Yedinak]

---

@dkfire wrote:

It proberbly also has something to done with the resolution of the timestamp that the OS can supply.

Windows XP has only a resolution of around 15ms in it's timestamp.

You should use the Tick Count (ms) vi or some of the time info from the while loops Left Data node for this type for timing measurement.

---

Not to mention that if you are running under Windows the OS itself is not a realtime OS. That is, Windows is nondeterministic.

[Kevin_Price]

My bet's that dkfire nailed it.  If so, add your kudos to mine.  The stuff mentioned by crossrulz may also prove relevant after you handle the more fundamental issue of limited resolution.   I'm not actually sure whether the resolution problem kicks in natively with the timestamp or whether it comes the subtraction which converts to floating point "seconds since 1904".  Hang on a sec (no pun intended).

Hmm, looks like it may be inherent to the timestamp.  I tried converting to a time record cluster and saw the same quantization.  Go with dkfire's suggestions.

-Kevin P

[GregFreeman]

This does not answer your question, but just so you know, you can extend the terminal where the "error" input is and select previous iteration time so you don't have to calculate it yourself.

[Dennis_Knutson]

Personally, I'm not sure of the usefulness of a timed loop on windows and if you have an NI DAQ card, you can get much better timing just by using hardware timed input and the timestamp that you get from DAQmx. You really should post your VI.