continous acq with PXIe5105

[Reuven.Stahl]

Hi all 🙂

i need to do a continous acq of 50MS/s of 4 channles with the NI PXIw 5105 .
1) can it be done?

2)how long can i acquire with either memory configouritons of the module (16M and 512M)

thanks in advance

Reuven

[Danny_Kaps]

Hi Reuven,

The answer to this question is going to depend on the chassis and controller you are using in this configuration. If we are streaming 50MS/s * 4 channels * 2 bytes/sample, we will be streaming 400MB/s out of this card.

PXIe can generally handle around 1GB/s streaming (this depends on which slot you are using in which chassis), so that would be fine (you can check this in the spec sheet of the chassis). However, this also depends on where you are streaming to. If you are writing to an embedded controller, host PC, or a RAID array, then there will be a separate spec for how quickly data received can be written to disk. This spec is further going to determine the answer to your second question: the memory configurations of the module refer to the size of the onboard buffer. For example, if you are outputting 400MB/s from the card, and your target disk is specced at writing 350MB/s, you would build up 50MB/s in the buffer, stopping the 16MB model almost immediately,  and the 512MB model after about 10 seconds. 

Here's a document that provides an overview of PXI Streaming: http://www.ni.com/white-paper/7700/en/

Best,

Daniel

[TimmTheEnchanter]

a) yes (with the right hardware).  I've streamed up to 6 channels continuously with that device using a MXIe-Express x4 and a rack mount controller.  In that test, I didn't do any processing with the data (like writing to disk), I just ensured I was getting the data.

b) depends on several factors, some of which were mentioned by Daniel.  Daniel mentioned that there are two bytes / sample (that is because the ADC 12 bit resolution is stored in 2 Bytes.

Another caveat applies if you are acquiring several records.  It doesn't sound like you're interested in that, but I just want to include it here for completeness.  There is head room that takes up additional memory that is associated with each record (see that in the specifications section about allocated onboard memory per record), and there is a re-arm time (or "dead time") in between records in which you will not be acquiring (but you may still be sending data back to the host.)