Benchmarks for Buffered Counter Input with X-Series Devices

[Eyal_S.]

Hi,

Does anyone know the maximum transfer rate (using dma) for the x-series cards? 

I'm looking for data simillar to the one in this link:

http://digital.ni.com/public.nsf/allkb/72A7E41EE5A8756A862571DA0076F1D7?OpenDocument 

More specifically I want to know the transfer rate for the PCIe 6343.

Thank you,

Eyal

[stilly32]

I've benchmarked 10MS/s on all four counters at the same time - this is for all X Series boards. I don't recall the CPU usage but it was relatively low on a single core machine. I mentioned this and more in this post - X Series is here – New Counter Features!

Cheers, 

Andrew S

[Eyal_S.]

Hi Andrew,

Do you happen to know the transfer rate for just 1 counter? I'm currently using the USB-6210 which has a bigger FIFO (1024 ) but if the transfer rate is substantially better for the X series it might work out alright for my application. For 1 counter I get 1.2MHz transfer rate on my current card.

I'm using 1 counter for period measurements. My signal has occasional high frequency bursts which is the reason I need a big FIFO and/or fast transfer rate.

Thanks for the information,

Eyal

[stilly32 · Accepted solution]

For a single counter, I've benchmarked up to 20MHz. You can't achieve that on all four counters because of PXIe/PCIe bandwidth limitations (4 counters X 20Ms/s X 4 B = 320MB/s) but for a single you can achieve those rates. I didn't run it for an extended period of time, but at rates like this if we're going to get an overflow we'll usually get it pretty quickly.

Cheers, 

Andrew S

[ceties]

Do you anybody know the benchmark results for one/four counters on NI USB-6356/6366. It is also X series but USB. Are the counters digitally triggerable for input operations?

Thanks.

[Kevin_Price]

Hi ceties,

I can't speak to benchmark figures but I actually just replied to another post of yours on the MIO board and mentioned that the X-series (among many other boards) DO indeed support triggering for counter input measurements.  I'm not sure whether the X-series handles it differently, but prior boards that supported triggering for counter input tasks required the use of an "Arm Start" trigger rather than the standard start trigger.

-Kevin P

[stilly32]

I benchmarked 8MS/s on a single channel on a USB-6353 (should have similar performance to the USB-6356/66) with a buffered counter input task. This was only a 10 minute test on my Dell E6400 laptop ( Windows 7 32 bit, Dual core P8600, 4GB memory) so a sustained 24/7 rate is likely less. Also, I wasn't doing much with the data, just displaying it in an array. This is pretty much topping out USB streaming on this system (32MB/s) so I wouldn't expect to get higher. I wasn't running much else in the back ground (Explorer, Powerpoint, Lotus Notes) and CPU was at ~20%.

Hope this helps,   

Andrew S

[ceties]

Hi Andrew and thanks for your time,

Just to be sure since I was not very clear - those 8MHz is the frequency of the input = how fast the TTL pulses were coming? And if you went higher with the freq you got a buffer overflow error, right?

Thanks

[stilly32]

Correct - I generated an ~8MHz counter output with one onboad counter and did a buffered edge count with that as the sample clock - same operation under the hood as a single counter frequency measurement. Anything higher I get buffer overflows. Notes, even at higher rates I was able to acquire some samples before getting an overflow.

Hope this helps,

Andrew 

[ceties]

Thanks for the clarification!