Need some clear info on data cquis. with LV in linux

[evren]

i just need multiple chans at one rate so that parts fine for me.

i actually don't do analog triggering. i process the analog data online and if it meets a criteria then a digital chan triggers the AO through a PFI pin.

if the performance of this process is unaffected then i'm in good shape.

thanks again.!

[Tom_W_[DE]]

Hi All-

Boris' comments on two points are incorrect. 

First, the fact that NI-DAQmx Base does not support multi-threaded operation has little to do with running AI channels at different rates.  It literally means that you can only access the driver in a single thread at one time.  This means that parallel loops in LabVIEW are not possible and that all code must execute sequentially.  Passing error clusters between VIs is a good method to ensure that all VIs are executed sequentially and in the desired order.

Also, it is impossible to sample analog inputs on different channels at different rates under any circumstances; this was never possible with Traditional NI-DAQ, NI-DAQmx, or NI-DAQmx Base.  This stems from the fact that the DAQ board has a single ADC that all input channels are multiplexed through, so the inputs must be scanned in the same task and at the same rate to avoid hardware conflicts.

Hopefully this helps-

[evren]

hey tom,

so if i'm calling AOwrite and AIread in the same loop there will be a preformance degradation. meaning it will take longer to get data off the input buffer.

but if i setup a long AO, and start it then take AI inside a loop while the output buffer plays out, there will be no degradation of performance?

in essence I'm asking if the drivers need to be accessed after an operation has be setup and started. as the output buffer plays out of the input buffer fills up is there driver access, or only when the AI or AO vi's are called is there driver access.

thanks.

-evren

[tibobo]

Hi tom and evren,

I saw my response was incorrect. Well, Tom could take a look at :

Sampling Different Channels at Different Rates with NI-DAQmx
http://digital.ni.com/public.nsf/allkb/96FD2F4685065C7686256F25006EE8DE

I said "undirectly", this is what I was talking about. This method doesn't work with NI-DAQmx Base since running two AI tasks in parallel is impossible. It does work with NI-DAQmx, even if it's not beautiful.

Regards,

Boris

[Tom_W_[DE]]

Hi Boris-

Just as a point of clarification on that example- with versions of NI-DAQmx prior to 7.4 where "round robin" sampling was used this method emulated a higher sampling rate.  Still, sampling one channel three times in a row and another twice in a row as the text of that article suggests does not provide uniform sampling as most applications demand on either channel. 

With versions 7.4 and above the sampling behavior was changed to sample all channels in a scan with minimal delay between channels to provide near-simultaneous sampling, so this method will really only provide for 3 extremely close samples on one channel and 2 extremely close samples on the second and then a comparatively long period of time before the pattern is repeated.

Regardless, I hadn't seen that method before and it is certainly interesting if the loss of uniform sampling rates is not a concern for your application.  Thanks a lot for your input!

[Lorne H.]

evren,

In response to your question:

" but if i setup a long AO, and start it then take AI inside a loop while the output buffer plays out, there will be no degradation of performance?"

That is correct, there will be no degredation in performance caused by the inability of DAQmx Base VI to be run in parrallel.

Lorne Hengst
Application Engineer
National Instruments