How can I stretch very fast pulses so PCI-6071E can see them?

[looper]

We put together our first LabVIEW measurement system
recently, and it seems we bought some incompatible
hardware. We need to measure pulses from 16 pixels
of a solid-state radiation detector, so we wired that
into a preamp and shaping amp that give us 16 channels
(differential) of 2-microsecond pulses. We then hook
those channels into a PCI-6071E card (via an SCB-100
box), and of course the pulses are too short for the
card to digitize. LabVIEW itself appears to set a
maximum sampling frequency of 6.7 MHz, but for 16
channels the PCI-6071E only supports about 600 kHz,
and LabVIEW warns of settling-time problems in the ADC
multiplexer if we go above about 60 kHz. This means
we need a pulse wid
th of several dozen microseconds
in order to get several ADC samples per pulse. (For
completeness, we are using a Mesytec MPR-32 preamp and
STM-16 shaping amp; they say they can modify the latter
to give us about 10 microseconds, but that's clearly
not enough.) Any suggestions as to a "pulse stretcher"
we could stick between the shaping amp and DAQ card?
Or is there an NI 16-channel peak-sensing DAQ card out
there that we've overlooked?

[chericks1]

Dear looper,

I would first like to apologize for the experience you've had with NI thus far. To address your questions, I don't know of a pulse stretcher, but I recommend looking into the simultaneously sampled input channels of our S Series devices on our web site.

If you have a PXI chassis, I recommend the NI 6132 or 6133. This PXI board has 8 differential analog inputs capable of transfering data at 2.5MS/s over every channel. You would benefit from this board in that you would only need to purchase two.

If restricted to the PCI bus, there is the PCI-6115 that has 4 simultaneously sampled analog inputs at 10 MS/s over every channel with 12-bit resolution. The PCI-6120 also has 4 simultaneously sampled analog inputs, but sacrifices the input rate of
800kS/s for 16-bit resolution. With both of these boards, you would need to purchase 4 to accomplish your task.

Let me know if you have any further questions or if this does not resolve your issue.

Thanks again and have a great day!

Chad AE
Applications Engineer - National Instruments

[looper]

Hello again--

Thanks for the suggestions. Before we drop three grand on one, or twelve grand on four, PCI-6110 or PCI-6115 cards, I'd better make darn sure in advance that we are going to be able to use them at full speed, though. In the course of investigating the sampling speed limits of the PCI-6071E we started with, I found that LabVIEW 7.0 itself, in the DAQ Assistant Express VI, refused to accept a sampling rate over 6 2/3 MHz; it gave an error -200077, saying that higher values than that are not supported values for the SampClk.Rate property (which I eventually figured out was related to the DAQmx Timing VI). In addition, I found that for N channels (N = 1 to 16) in the DAQ Assistant setup, the maximum sample rate per channel was

10 MHz / (N
+ 0.5)

and, if N > 1, the speed limit "to avoid possible ADC overruns" was

1.25 MHz / N

and "to avoid possible settling time violations" was

1 MHz / N

The last two presumably would not apply with the simultaneous sampling of the S-series boards, but are there other speed limits that would prevent us from getting the full 5M or 10M samples per channel per second from these two cards? If we got a PCI-6115 with the 10 MS/sec speed, is LabVIEW going to let us exceed the 6 2/3 MS/sec property limit I ran into, or the 10 MHz / (N + 0.5) limit for multiple channels? And, if we do get four of these cards, are there going to be multiple-card issues that limit the sampling speed? Our computer is pretty new, an HP Compaq D530 with a 3.2 GHz Pentium 4 and 2 GB of RAM, so I'd hope it would not be a bottleneck.

Again, my task is to do pulse-height analysis of 2-microsecond pulses in 16 channels; the pulses don't come in at a very rapid rate, but our choice of shaping amplifier
has caused them to be so short that we need a fast DAQ card to get multiple samples per pulse.

[chericks1]

looper,

I understand the hesitation. I'm 100 percent confident the PCI-6115 is able to transfer at 10MS/s over every channel. I'm concerned about the LabVIEW limitations you mention above and PCI limitations that apply.

The trouble with providing the maximum sampling rate for a board comes from the fact that often times reaching those upper bounds of the sampling rate becomes very system dependent. While the DAQ card will be able to meet the specifications published for that device, the board still relies on other components of the computer or system in order to get that data from the board to the application being run. Often times the performance of the devices and the ability to get the data from the boards will depend on what resource
s are available in the computer. For example, a machine using 4 DAQ cards in 4 PCI slots at the same time will perform slower than a machine with just a single PCI DAQ device due to the dependence on the system, in this case on the available bandwidth of the PCI bus.

I'm currently consulting with Research and Development on the issue, and will contact you as soon as I'm positive this solution will work.

Thanks for your patience,

Chad AE
Applications Engineer

[looper]

Hello again--

Since I posted my follow-up question, I have dug a little more into the detailed differences between the S-series cards. On the webpage

http://sine.ni.com/apps/we/nioc.vp?cid=11886〈=US

for the PCI-6115 card is a link to a Product Data Sheet in PDF form, and on the first page of that document (p. 197) in the lower right corner is a discussion of the card's "Large Onboard Memory," with the assertion that this enables a user to "capture high-bandwidth signals over long periods of time, even if multiple devices are sharing the same PCI or PXI bus." My understanding is that our computer is a fairly standard and fairly recent HP Compaq model D530 convertible minitower; however, I can't for the life of me find on the
HP website a spec for exactly how fast its PCI slots _are_. As you note, that doubtless enters into the equation... (Dammit, _Apple_ posts the specs on their PCI or PCI-X buses, why can't HP?)

Thanks!

[chericks1]

Hi looper,

I've been speaking with the Product Support Engineer about this issue, and I write to share what I've learned. The PCI bus of your computer can be implemented in two ways: (1) The capability to transfer 32 Bits / 33 MHz or (2) 64 Bits / 66 MHz. NI products are built to run on (1). Thus, the maximum transfer rate of the PCI bus is 132 MB/s.

32 bits/8 = 4 bytes * 33 MHz = 132 MB/s

Even though this is the maximum transfer rate, most busses only transfer at 80-110 MB/s. With 16 channels each transferring at 5 MSamples/s at 2 B/Sample, we get 160 MB/s.

16 * 5 MS/s * 2 B/S = 160 MB/s

This gives us a violation of PCI bus specifications.

I believe there still may be a pricey alternative for this appli
cation using PXI. Let me know if you're interested and I'll delve into the specifics.

Thanks,

Chad AE
Applications Engineer