DAQmxReadAnalgog slow to transfer data / no effect of time out

[gavril]

NIDAQmx 7.5 on XP SP 3 with PCI 6034 E.

I want to read the most N most recent samples repeatedly in a 1kHz loop even if zero or less than N samples

have been acquired. I use NIDAQmx C API.

My problem is that the DAQmxReadAnalogF64 stops execution of the program for a few ms even though I have

I have set the property the RelativeTo property to Most_Recent_Samples and the Offset to -N.  The problem is

particularly pronounced when the sampling frequency is low. To illustrate the problem, I have timed

the function calls to DAQmxReadAnalogF64

  100 Hz, N = 2  => 8.5 ms

  100 Hz, N = 10  =>8.5 ms

 1'000 Hz, N = 10 => 1.45 ms

 1'000 Hz, N = 100 => 1.45 ms

10'000 Hz, N = 10 => 1.06 ms

10'000 Hz, N = 100 => 0.93 ms

 20'000 Hz, N = 10 => 0.30 ms

20'000 Hz,  N = 100 => 0.21 ms

20'000 Hz,  N = 1000 => 0.15 ms

In short, increasing the sampling frequency decrease the latency of the function call and,

to a lesser extent, increasing the number of desired samples also decreases the latency !

have tried setting timeout to -1, zero and 1 ms with no effect. Note that I always get the desired

number of samples and the function call never return an error.

 Any help appreciated.

Gabriel Baud-Bovy

Vita-Salute San Raffaele University

[Cla_Cup]

Hi Gabriel,

DAQmxReadAnalogF64 is a blocking call. One way to avoid blocking on a DAQmxRead call is to only execute the read operation when necessary.  The best way to do this is to make use of NI-DAQmx Events to monitor the transition of samples to the PC buffer (i.e. the area that is available to DAQmxRead).  The example "C:\Program Files\National Instruments\NI-DAQ\Examples\DAQmx ANSI C\Analog In\Measure Voltage\Cont Acq-Int Clk" shows how to call the DAQmxRead from a seperate callback using NI-DAQmx Events.

Hopefully this helps.

Best Regards.

Claudio

[gavril]

Hi Claudio,

Thank you for your answer. I will check this NI-DAQmx Events . I think that my problem was that I would an offset of N-1 samples instead of N samples because I expected that the current position designaged a sample that was alread acquired while it seems it is not the case (I did not find a clear statement in the documentation but I might have missed it). Anyways, DAQmxReadAnalogF64 waited to acquire one more sample before returning and that explains also why it was slower at lower sampling rates.