I am working with others to write a LV program to acquire data at 1.25 MHz on all 8 channels of the USB 6356. For the initial version the on-board 32 megasample per channel buffers are large enough to assure no data loss. Later I anticipate that runs longer than 25.6 seconds might be required. I would have probably not recommended the USB device for this application but was not asked. The program architecture will be Producer/Consumer with only acquisition and enqueue in the DAQ loop while acquiring data. The data will be graphed (after some reduction) and saved to file in the Consumer. It is not a part of a control system.
The users would like to see recent data on a graph so several reads per second would be nice, but this is not a hard requirement.
My question is: What is the optimum (or as near as we may be able to get with the non-deterministic Windows/USB interface) number of samples to read?
I looked at this thread where William Earle provided some very useful advice.
I cannot determine from the X-Series Users Manual or other documents I have found whether the analog inputs are transfered as raw binary (2 bytes per sample) or as double precision (8 bytes per sample). This will make a big difference in the throughput. Edit: Page 10-2 refers to USB Signal Stream (USB bulk) but does not offer guidance on read sizes. End Edit.
I do not have ready access to the device or the time for extensive experimentation, so I hope to benefit from the experiences of others.
Thanks,
Lynn
