DAQmxBase Acquiring at high frequency

[Atos]

Hi everyone,

Part of my job is to acquire continously on multiple channels using NIDAQmxBase.

As I faced problems in my application I decide to work with the provided examples but I doesn't succeed to make them work.

Here are my needs :

12 Channels

48000 Hertz

Here is my code (contAcquireNChan.c with a few modifications):

#include "NIDAQmxBase.h"
#include <stdio.h>
#include <time.h>

#define DAQmxErrChk(functionCall) { if( DAQmxFailed(error=(functionCall)) ) { goto Error; } }


int main(int argc, char *argv[])
{
    // Task parameters
    int32       error = 0;
    TaskHandle  taskHandle = 0;
    char        errBuff[2048]={'\0'};
    int32       i,j;
    time_t      startTime;
    bool32      done=0;

    // Channel parameters
//    char        chan[] = "Dev1/ai0, Dev1/ai1";
    char        chan[] = "Dev1/ai0, Dev1/ai1, Dev1/ai2, Dev1/ai3, Dev1/ai4, Dev1/ai5, Dev1/ai6, Dev1/ai7, Dev1/ai8, Dev1/ai9, Dev1/ai10, Dev1/ai11";
    float64     min = -10.0;
    float64     max = 10.0;

    // Timing parameters
    char        clockSource[] = "OnboardClock";
    uInt64      samplesPerChan = 1000;
    float64     sampleRate = 48000.0;

    // Data read parameters
    #define     bufferSize (uInt32)1000
    float64     data[bufferSize * 12];
    int32       pointsToRead = bufferSize;
    int32       pointsRead;
    float64     timeout = 10.0;
    int32       totalRead = 0;


    printf("Press Ctrl-C to exit\n");
    DAQmxErrChk (DAQmxBaseCreateTask("",&taskHandle));
    DAQmxErrChk (DAQmxBaseCreateAIVoltageChan(taskHandle,chan,"",DAQmx_Val_Cfg_Default,min,max,DAQmx_Val_Volts,NULL));
    DAQmxErrChk (DAQmxBaseCfgSampClkTiming(taskHandle,clockSource,sampleRate,DAQmx_Val_Rising,DAQmx_Val_ContSamps,samplesPerChan));
    DAQmxErrChk (DAQmxBaseCfgInputBuffer(taskHandle,144000));

    DAQmxErrChk (DAQmxBaseStartTask(taskHandle));

    // The loop will quit after 10 seconds
    printf ("Acquisition starting\n");

    startTime = time(NULL);
    while( time(NULL)<startTime+50 ) {
        DAQmxErrChk (DAQmxBaseReadAnalogF64(taskHandle,pointsToRead,timeout,DAQmx_Val_GroupByScanNumber,data,bufferSize*12,&pointsRead,NULL));
        totalRead += pointsRead;
        printf("Acquired %d samples. Total %d\n",pointsRead,totalRead);
        // Just print out the first 10 points of the last data read
        for (i = 0; i < 10; ++i)
            printf ("data0[%d] = %f\tdata1[%d] = %f\n",i,data[2*i],i,data[2*i+1]);

    }
    printf("\nAcquired %d total samples.\n",totalRead);

Error:
    if( DAQmxFailed(error) )
        DAQmxBaseGetExtendedErrorInfo(errBuff,2048);
    if(taskHandle != 0) {
        DAQmxBaseStopTask (taskHandle);
        DAQmxBaseClearTask (taskHandle);
    }
    if( DAQmxFailed(error) )
        printf("DAQmxBase Error: %s\n",errBuff);
    return 0;
}

With my  values (12 channels, 48000Hz) it doesn't seems to work; I get the following message when readingAnalogF64 :

DAQmxBase Error: RLP Invoke Node <err>There is <b>not enough data</b> in the DMA buffer to fill the VI's read buffer.  Either request less data, or wait until there is additional data in the DMA buffer.

With smallers values (10 channels or 30000Hz) it works better, even if sometimes I still get this error message.

After a few investigation it seem's that this not an error but just a warning (code 42), but I don't know if the values are correctly acquired.

As someone still faced this problem and have more information.

Thank in advance

Sylvain 

[Atos]

A few more informations (forgotten) :

I run on linux (RHEL4 and RHEL5)

I tried on two differents machines (PentiumD 2.8Ghz, 1Go RAM  and C2D 3GHz, 2Go RAM)

The approximately same code works with NIDAQmx (but this last isn't supported on RHEL5 that's the reason why I try NIDAQmxBase)

[R_Duval]

Hi Sylvain.

Could you send us a screen of the error message?

Thanks in advance.

Regards.

[Atos]

Of course I can,

You can find enclosed a screen shot of my console.

I have tried many combination with no more results.

I'm thinking about a lack of optimisation in NIDAQmBase cause it works well with NIDAQmx.

Thanks for any help 

[R_Duval]

Hello Atos.

Did you solve your problem?

Could you tell me the version of DaqmxBase did you use?

Thanks.

Regards. 

[Atos]

Well, I didn't solve the problem cause it's currently impossible to do this with DaqmxBase.

NI support, reproduce the problem with their example given below.

They also confirm that today the DAQmxBase layer is unable to acquire at the samplerate given in the hardware specification.

It seems to be a problem with the DMA writing/reading at high rates.

For me it's not very straight that NI sell a card with hight hardware specifications and linux support, althought they do not provide the software that enable to achieve this.

A bit incomprehensible ... 

[Tom_W_[DE]]

Hi Sylvain-

Which version of NI-DAQmx Base are you using?  Which PCI/PXI hardware device are you using? 

The error you are seeing is an internal software error that should not appear to the user in most cases.  Some enhancements were made in NI-DAQmx Base 3.2 to help avoid this error, so I'm curious if you might be using an older version.

Thanks-

[Atos]

Hi Tom,

Thanks for replying.

Actually I've tried with 2 configurations

    + RHEL4, DAQmxBase 3.2 and 6259PCI Card

    + RHEL5, DAQmxBase 3.2 and 6254PCI Card

I've got the same problem. It seems that I try to acquire at too high rates.

Moreover there's not many configuration's functions provided with the DAQmxBase C API, so I can't try many configuration's combinations.

If someone can tell me more ...

[JoeFriedchicken]

Hi Atos,

One option you may want to evaluate is using the full NI-DAQmx driver on Linux. Both your 6254 and 6259 cards are supported and shouldn't experience the DMA problems you've been observing. The trade-off, however, is supported Linux versions. The only release of DAQmx for Linux is version 8.0 [1], and so it supports slightly older distributions. The latest RedHat release that DAQmx 8.0 works with is Enterprise WS 3 [2].

As Tom mentioned, the problems you're seeing shouldn't be happening in Base 3.2. We're looking into the cause and fixes, but I don't know when they will be added to the driver. The fastest resolution for you may be using DAQmx 8.0 and not Base 3.2

[1] DAQmx 8.0 for Linux
http://joule.ni.com/nidu/cds/view/p/id/374/lang/en

[2] DAQmx 8.0 for Linux Readme
http://ftp.ni.com/support/softlib/multifunction_daq/nidaqmx/8.0/Linux/Mandriva/readme.txt

[Atos]

Hi Jo,

Thanks for your answer,

Sorry for my being late on replying, I was a few day on holydays...

Actually when I start working on this project, I've began with NIDAQmx on RHEL4.

At that time, I was given instructions, drivers and the NI card and I didn't know that NIDAQmx doesn't support RHEL4.

Almost every things worked well  ...

But today we have upgrade on RHEL5 (we've got new computers unsupported on the previous versions of RedHat). 

That's the reason why we try NIDAQmxBase (Only this driver, supports the last versions of RedHat)

We faced the above problems ...

But as I have just said, every things run well with NIDAQmx on RHEL4/5 ... even thought it's not supported.

Does someone know in which circumstances the kernel stack overflow happens ?

This could help us make more test on the reliability of our application ...

Actually it's quite simple, it can be summarized as an infinite acquiring task.

Thans for any help,  

 Sylvain

Message Edité par Atos le 09-22-2008 07:57 AM