Sampling window in PCIe-6351

[groumpos]

My experimental aparatus outputs a pulse train of gaussian pulses. Each pulse is expected to have a different height, and I wish to measure the height of every pulse individually. The pulse train has a repetition rate of 380kHz, and the width of each pulse can be in the range 100ns - 1us (depending on my design of electronics, which I have not yet finished). PCIe-6351 sounds a good choice for this purpose, but I am confused with the terminology about analog input timing.

1) What is the sampling window? The PCIe-6351 board has a 1.25MHz max sampling rate. If I use a sampling rate of 1MHz, does this mean that the board averages the input signal over 1us? If this is true, then I should use an additional peak-hold circuit before the analog-to-digital converter, to keep the pulse height constant over the sampling window.

2) I have an analog trigger which outputs square pulses 1Vp-p, and I want to use it to trigger analog detection. In the analog trigger specs, I read "accuracy = 1% of range". What does this mean exactly? Is it that if my trigger rises from 0 to 1V within 20ns, then the board can synchronize with the trigger within 0.01*20ns =0.2ns?

[Stefo]

Hi Groumpos,

If I understand correctly, you want to measure amplitude of pulses, which can be from 100ns to 1us long.

100ns corresponds to 10MHz - therefore you will not be able to use 1.25MS/s to be able to properly measure the shape of the signal. I beliebe that if you are interested in measuring the amplitude direclty, then you will need device with sampling rate of at least 20MHz, but becaus it is rectangular signal (which has lot of odd harmonics) then to reconstruct samplied signal properly (to properly measure it) you might want  to use even higher sampling rate - as high as 100MHz.

To answer yours questions:

1) NO, the sampling rate is not averaging, sampling means that the board is able to do AD conversion as fast as 1.25MHz. Refering to sampling, it is well described at Analog Sampling Basics. With 1.25MHz, you should be able to measure signal with highest frequency componend of 1.25/2 MHz. Unles you want to build your own electronics, I would recomend you to use high speed digitizer (oscilloscope).

2) Analog trigger means, that you will trigger measurement with measured signal - if the measured signal exceeds certain value, then you trigger. If you speak ask about accuracy, then I wouls say that in this case it has more to do with amplitude, rather then timing. If you specify to trigger on 1V, then trigger might happen between 0.99V and 1.01V. If you have digital signal which you want to use as trigger, then use it as digital trigger.

Ayway, I would recomend you to contact NI engineer who could recomend you the proper ways which could be used for application.

regards,

stefo

[groumpos]

Hi stefo,

Thanks for your reply. Actually, I only want to know the height of every gaussian pulse, and not record its full form. So my question is how well I can sample this height. I just now found out about this manual

http://www.ni.com/pdf/manuals/370784c.pdf

which explains a few things. Is the "convert period" (see page 4-32) the same as the sampling window. Namely, does the board average the signal over the convert period and outputs the result? Since the board has a 100MHz timebase clock. Does this mean that the shortest sampling window is 10ns?

Now I understand about the trigger, thanks for your explanation.

NI has a strange policy about contacting an engineer. Although my group has purchased numerous of items from them, worth tens of thousands, I keep receiving the message that I don't have an active contract, therefore I cannot contact an engineer.

[Brad_K]

Hello groumpos,

The NI 6351 uses a successive approximation ADC. This type of ADC does not average the input signal. Instead, a sample and hold circuit tracks the input signal for part of the convert period, then holds it at a steady voltage during the actual conversion from analog to digital. As a result, the "sampling window" is effectively instantaneous, for most practical purposes.

In your first post, you mentioned the idea of connecting a peak-hold circuit. I assume that you would export the convert clock to a PFI terminal and use it to reset the peak detector for each conversion. This idea sounds promising. Without a peak detector, however, you would need to sample much faster to reliably characterize the pulse train, as Stefo pointed out.

Brad

[tongxu]

Hi Brad,

I have the exact same situation as groumpos . I am thinking of using PCI6143 to get the peak amplitude .  We don't need to sample the whole pulse shapes. We only got about 10k random peaks/second.

I assume that as long as we know the exact location of the peak, for example, 100 ns after the trigger signal,  we can always triger the ADC to sample the peak amplitude even with a slow speed DAQ card. 

The question is, as you mentioned, there is a "sample and hold circuit tracks the input", I wounder what is the bandwidth of this  "sample and hold circuit"? In another word, what is the sampling window of this sample and hold circuit?

Many thanks,

Tong

[Seth_B.]

Hello Tong,

The bandwidth for the analog front end of the PCI-6143 is 490 kHz, per page 4 of the specifications.  Additionally, the PCI-6143 has a 20MHz clock with an accuracy of ±0.01%.  This means your timing accuracy is probably not good enough to precisely catch a peak with nanosecond timing.

You'll likely need to investigate the use of external peak detection hardware to use between your signal and the PCI-6143. 

[tongxu]

Hi Seth,

Thank you so much for the reply.

Could you please sugguest an external peak detection hardware by NI?

Tong

[tongxu]

Or, Should I better off to use a DAQ card with much higher speed. For example NI PCI-6133. What is the bandwidht of 6133?

Thanks again,

Tong

[Seth_B.]

Hello Tong,

You can view the Bandwidth chart on page 5 of the specifications for the PCI-6133.  If you want the card itself to catch the peak, you'll likely need a high speed/high bandwidth digitizer.  Alternatively, you can use a less expensive DAQ device and use an external Peak Detector circuit to capture the peak value of each pulse.  For this, you'll need to research non-NI sources to find one with appropriate bandwidth.