How to avoid aliasing with DAQmx

[Basjong53]

I have a multifunction DAQ device. As an analog output I generate a 2kHz pulse train connected to an LED. The analog input is connected to a photodiode to measure the output of the LED. I'm not interested in the temporal reponse of the LED, I just want to measure the average photodiode voltage every 100 ms.

The code I have:

Though it's working, at certain frequencies, I get aliasing of the signal. Meaning that I measure, for example, 2 points in high, and 1 point in low, meaning that the average is too high. I just want to measure the average voltage over 100 ms. How can I make sure that I don't accidentely measure more high points than low points (or vice versa?). Is there an analog aperture time I can set? I found this https://www.ni.com/docs/en-US/bundle/ni-daqmx-properties/page/daqmxprop/attr2fee.html, but this property is not available to me in the DAQmx channel node.

Another options I guess, is to synchronize the input with the output, to ensure that I measure 1 point at every high and low output. But this only works because I happen to have all NI devices. What if an external device is generating the PWM signal? How can I avoid aliasing in that case?

[ZYOng]

Nyquist theorem states that you need to sample at a rate of at least twice the signal frequency. Since your pulse train is 2kHz, your AI sample clock should be at least 4k, although you should use 5 to 10 times to capture the waveform correctly. Is there any reason why you cannot use a higher sample rate on AI?

[Basjong53]

---

@ZYOng wrote:

Nyquist theorem states that you need to sample at a rate of at least twice the signal frequency. Since your pulse train is 2kHz, your AI sample clock should be at least 4k, although you should use 5 to 10 times to capture the waveform correctly. Is there any reason why you cannot use a higher sample rate on AI?

---

Yes, our system has a PCI-6225 which has a sampling rate of 250kS/s. We want to measure 32 LEDs simultaneously. Measuring boths voltage and current we need 64 channels, allowing a max sample rate of 250.000 / 32 / 2 = ~3900 S/s/ch. 

But like I said, I don't actually care about the shape or frequency of the LED output, I just want to measure the average intensity.

[Sebastian.Weber]

I think I would use a hardware low pass filter like a simple RC circuit. A time constant of 10ms would flatten the pulses, but a stable reading would be avaliable after about 50ms. Check if your circuit is able to source/sink the filter correctly! 

[Henrik_Volkers]

Some point that came into my mind:

You need a scan over all LED within 100 ms, thats about 31 ms per LED ..

You need to sync your 2 kHz (is that fixed?) with your aquisition !

(see manual , since you only need a square , you can also consider to use a timer for that)

I would try to configure the analog capture in a way that you sample one channel every 0.5 ms , I don't know the detailed possebilities of that DAQ but with some properties (you should be able   to configure settle time and capture time ) it should be possible to always hit the LED in active mode (better close t the end)

Another point is the settle time error at higher impedances. See figure 1 in the spec! We don't know your photodiode detector output impedance, so if it is not buffered..

uups I had a peek ... that card is even worse than I thought...     depending on your needed accuracy , forget the sync idea, go with a RC lowpass and choose a bigger C .. and choose the lowest samplerate (10 Hz ) to scan your channels.

As test try 20 Hz SR and use a channel list with every channel twice in the list (CH0,CH0,CH1,CH1, ...) and compare the results.