I have a few questions and issues that I have not been able to resolve with forum searches, I will try to keep them concise here. #1 is the most important one to me.
LabVIEW 7.1, Windows XP.
Buffered Reads / Lost Data
When error 10846 (AI Buffered Read, app unable to retrieve data from background acquisition buffer fast enough) occurs, I receive no data from the AI Buffered Read function for a period of time. The error states data may be lost but it seems like I get NO data. See the attached VI and JPG for examples. The bottom graph shows a normal read while the top graph shows one with this problem. I realize the AI Read error cluster is not handled in the acquisition loop. From watching the backlog while the program runs, it seems the “dead spots” occur after the backlog reaches the buffer size. When stopping acquisition, sometimes the error handler outside the loop reports 10846, sometimes it doesn’t (when the dead spots are occurring). I can exaggerate dead spots with setups as follows:
Buffer size 5, Scan rate 1000, # Scans to read 100
Buffer size 10000, Scan rate 5500, # Scans to read 100
The size of the dead spots and continuous data are proportional to the buffer size.
Questions....
1. Shouldn’t the AI Read function wait until it has # scans to read of valid data before returning? For instance, scan rate of 100 and # scans to read of 1000 slows loop iteration speed to 0.1 Hz. It seems like it is returning nothing for a period of time after the backlog reaches the buffer size.
2. Error 10846 refers to the “data acquisition buffer.” Is this the software buffer? If so, what would LV refer to the hardware buffer as?
3. I know the PCI-MIO-16E-1 has a hardware buffer, does this mean I can accurately acquire data over a period of time at a specified frequency (within limitations of the card of course) without having to worry about Windows bogging down due to delayed writes and such?
4. Will hardware acquisition buffers generate an error if the data in them is lost/overwritten?
5. I tried loading the acquisition VI (attached) onto a LV computer that has only a new USB-6008 connected but the AI read/config/clear functions were not available on that system. Does this device not have hardware buffers (either I’m looking in the wrong places to find the answer this question and it uses different VIs OR it doesn’t have them).
6. The buffer/scan configuration listed above was used to exaggerate error 10846 and the accompanying dead spots but this problem has been intermittently plaguing a system I have been trying to understand/fix. I have logged the backlog value as the program runs and see that it increases during certain processor-heavy state machine states and am hoping upgrading from 256MB to 2GB RAM will improve this. Is there any reason I should not look at this as a solution? (I am also planning on adding an error handler for AI Read into each acquisition loop so the problem can not continue to go on without detection!!!!!!!!!)
7. Shouldn’t the AI Clear function report an error from the AI Read error cluster, even if it happened several reads earlier? It seems like it does not do this; if I put the error handler in the acquisition loop I receive an error as soon as the backlog reaches the buffer size, EVERY time. I thought I figured this out and changed the acquisition loop tunnels to shift registers... the error is reported each time this way; if AI Read will not acquire data if an error is fed into it in the first place then I guess I understand this one so I’m moving it to the bottom of the list.
Thank you all for your time, I have been struggling to get up to speed with DAQ and LV and appreciate all the support. I look forward to giving back to NI’s community in the future.
Regards,
David.
