Buffered event counting. Why can't I explicitly sequence generating the Sample Clock Pulse and reading the counters?

Question

At irregular occasions I need to grab counts from several counters, and buffering the counts must be done simultaneously for all counters. I'm modeling my approach after zone.ni.com/devzone/cda/tut/p/id/5404 which someone kindly pointed out in an earlier thread. However, that example only uses one counter, and you can't test the synchronization with only one counter, so I am using two counters configured the same way, and they're wired to a single benchtop signal generator (for example at 300 kHz).
What I want to do, I can test in a loop with a somewhat random wait in it. I want to drive a hardware digital output line high for a few ms and then low again. The hardware line is physically connected to terminals for my timing vi's Sample Clock Source and so will cause them to buffer their counts for later reading. After I pulse this line, when I know new good buffered counts await me, I want to read both my counters. If their bufferings are simultaneous, then each counter will have counted the same number of additional counts since the last loop iteration, which I can check by subtracting the last value sitting in a shift register and then subtracting the two "additional counts" values and displaying this difference as "Diff". It should always be 0, or occasionally +1 followed immediately by -1, or else the reverse, because buffering and a count could happen practically at the same moment.
When I do this using a flat sequence to control the relative timing of these steps, so the read happens after the pulse, the counters often time out and everything dies. The lengths of time before, during, and after the pulse, and the timeout value for the read vi, and the size of the buffer and various other things, don't seem to change this, even if I make things so long I could do the counting myself holding a clipboard as my buffer. I've attached AfterPulse.vi to illustrate this. If I get 3 or 10 or so iterations before it dies, I observe Diff = 0; at least that much is good.
When I use two flat sequences running in parallel inside my test loop, one to control the pulse timing, and the other to read the counters and do things with their results, it seems to work. In fact, Diff is always 0 or very occasionally the +/- 1 sequence. But in this case there is nothing controlling the relative timing such that the counters only get read after the pulse fires, though the results seem to show that this is true. I think the reads should be indeterminate with respect to the pulses, which would be unreliable. I don't know why it's working and can't expect it to work in other environments, can I? Moreover, if I set some of the pulse timing numbers to 1 or 2 or 5 ms, timeouts start happening again, too. So I think I have a workaround that I don't understand, shouldn't work, and shouldn't be trusted. See SeparateSequence.vi for this one.
I also tried other versions of the well-defined, single sequence vi, moving the counter reads to different sequence frames so that they occur with the Sample Clock Source's rising edge, or while it is high, or with the falling edge, and they also often time out. I'll post these if anyone likes but can't post now due to the attachment limit.
Here's an odd, unexpected observation: I have to sequence the reads of the counters to occur before I use the results I read, or else many of the cycles of this combine a new count from one counter with the one-back count from the other counter, and Diff takes on values like the number of counts in a loop. I though the dataflow principle would dictate that current values would get used, but apparently not so. Sequencing the calculations to happen after the reads fixes this. Any idea why?
So, why am I not succeeding in taking proper control of the sequence of these events?
Thanks!!!
Attachments:
AfterPulse.vi ‏51 KB
InSeparateSequence.vi ‏49 KB

Answer 1

Kevin, thanks for all the work.
>Have you run with the little execution highlighting lightbulb on? -Yes. In versions of this where there is no enforced timing between the counter and the digital line, and there's a delay inserted before the digital line, it works. There are nearly simultaneous starts on two tracks. Execution proceeds directly along the task wire to the counter. Meanwhile, the execution along the task wire to the digital high gets delayed. Then, when the digital high fires, the counter completes its task, and execution proceeds downstream from the counter. Note, I do have to set the timeout on the counter longer, because the vi runs so slowly when it's painting its progress along the wires. If there is any timing relationship enforced between the counter and the digital transition, it doesn't work. It appears to me that to read a counter, you have to ask it for a result, then drive the line high, and then receive the result, and execution inside the counter has to be ongoing during the rising line edge.
>from what I remember, there isn't much to it.  There really aren't many candidate places for trouble.  A pulse is generated with DIO, then a single sample is read from each counter.  -Yup, you got it. This should be trivial.
>A timeout means either that the pulse isn't generated or that the counter tasks don't receive it. - Or it could mean that the counter task must be in the middle of executing when the rising edge of the pulse arrives. Certainly the highlighted execution indicates that. Making a broken vi run by cutting the error wires that sequence the counter read relative to the pulse also seems to support that.
>Have you verified that the digital pulse happens using a scope? -Verified in some versions by running another loop watching a digital input, and lighting an indicator, or recording how many times the line goes high, etc. Also, in your vi, with highlighting, if I delete the error wire from the last digital output to the first counter to allow parallel execution, I see the counter execution start before the rising edge, and complete when the line high vi executes. Also, if I use separate loops to drive the line high and to read the counter, it works (see TwoLoops.vi or see the screenshot of the block diagram attached below so you don't need a LV box). I could go sign out a scope, but think it's obvious the line is pulsing given that all these things work.
>Wait!  I think that's it!  If I recall correctly, you're generating the digital pulse on port0/line0...  On a 6259, the lines of port 0 are only for correlated DIO and do not map to PFI. -But I'm not using internal connections, I actually physically wired P0L1 (pin 66) to PFI0 (pin 73). It was port0/line1, by the way. And when running some of these vi's, I also physically jumper this connection to port0/line2 as an analog input to watch it. And, again, the pulse does cause the counter to operate, so it clearly connects - it just doesn't operate the way I think it is described operating.
For what it's worth, there's another mystery. Some of the docs seem to say that the pulse has to be applied to the counter gate terminal, rather than to the line associated with the sample clock source on the timing vi. I have tried combinations of counter gate and or sample clock source and concluded it seems like the sample clock source is the terminal that matters, and it's what I'm using lately, but for example the document I cited, "Buffered Event Counting", from last September, says "It uses both the source and gate of a counter for its operation. The active edges on the gate of a counter is used to latch the current count register value in a hardware register which is then transferred via Direct Memory Access...". I may go a round of trying those combinations with the latest vi's we've discussed.
Attachments:
NestedSequences.png ‏26 KB