LabVIEW

cancel
Showing results for 
Search instead for 
Did you mean: 

Sampling Bottlenecks PCI-MIO-16E/1kHz/6 channels/12-bit/1.25GB/1.79GHz

 

 

 

I am using a PCI-MIO-16E to sample 6 12-bit input channels at 1kHz.  I believe a bottleneck is giving me problems reading this data as my backlog does not remain at 0 with a buffer size of 300.

 

I am using a Athalon XP 2200+ (1.79GHz) processor, 1.25 GB RAM, VT8375 North Bridge, and VT8235 South Bridge.

 

The PCI-MIO-16E has a maximum sampling rate of 132 MB/s, close to the PCI bus speed.

 

Is there someplace this system might be bottlenecking the data transfer?  I can't imagine where a 72 kbps transfer would be bottlenecking here...

 

I have noticed that the backlog overflows more often when a sub-vi is opened.  Increasing the memory from 256 MB to 1.25 GB did not rectify this problem and I have only two things I see as the possible problem, the processor or the video card.

 

Does it seem likely that a 1.79 GHz processor would not be able to handle a 72 kbps transfer?  The only calculations that are performed while it is transferring is finding the maximum values in a buffer full of scans and storing the data in arrays (no disk writes).  The LabVIEW program takes up 100 MB of memory, if that is significant.

 

If anyone can point me to information that may help me resolve this problem, I'd appreciate it.  I found a on-demand video on the site that discusses eliminating bottlenecking problems but it would not play for me.

 

Thanks,

 

Dave
0 Kudos
Message 1 of 2
(2,089 Views)
Hello Dave,

I probably need to know little more about the specifics to answer more precisely but here is what I can see so far:
I assume that the sample rate you meant is 1.25 MS/s as in Mega Samples per second (and not 132 MB/s as in Mega Bytes per second). Given the actual system you have:
6 Channels at 1 kHz
Find the max and store into arrays

Does not seems to be anything very hard for the machine you have. However you have a large memory consumption (256 MB to 1.25 GB) and I do not believe this is only for the operations you mention. You are right, there should not be any bottleneck with those speeds.
Here is the kind of details I would like to see:
- Could you explain if you are using DAQmx driver? How many Samples per channel? What's the number of samples in the buffer?
- What other data/operations you do in the program?
- If is really small can you post it here?

Let us know if there is any extra update you can give us.
Gerardo O.
RF SW Engineering R&D
National Instruments
0 Kudos
Message 2 of 2
(2,071 Views)