0
\$\begingroup\$

A single-ended -10V to +10V 16-bit data acquisition system is composed of 16 single-ended channels where two of the channels are analog outputs and the rest are analog input channels. The transducers are outputting DC-like low freq analog signals such as temperature sensors ect. Only one channel(channel 7) is carrying pulses from different types of rotating instruments. Output channels output constant voltages during the data acquisition.

The daq board is sampling at 8kHz currently and multiplexing the channels.

I'm having random glitches sometimes. By random I mean 2 or 3 times in 1000 pulses in a pulse train.

Here are some glitchy readings from the pulse channel:

Assuming signal conditioners are not causing this, what could be the problem?

Can it be realted to BNC lengths or multiplexing speed is high due to 8kHz sampling rate? Or output impedance is high? But then I'm observing this kint of glitches when it comes to pulse channel not other channels.

What could be the reason?

edit: I came across this article: http://www.ni.com/white-paper/4494/en/

At the end of it it says:

"When scanning multiple channels at high sampling rates, be careful to notice the settling time and the impedance of the source of each channel. If the source impedance is too high, charges that accumulate on the input capacitance of the DAQ device are not dissipated by the time the signal is sampled. The result of this behavior is that the signal often appears to follow the signal of the previous channel. In this situation, either the sampling rate or the source impedance must be decreased. If the sampling rate cannot be decreased, the source impedance of the signal can be decreased by using a unity gain buffer or voltage follower. When adding a voltage follower to the measurement system, be mindful of the allowable measurement error and accuracy when selecting components and the input configuration for the DAQ device. However, keep source impedances below 1 kW when sampling multiple channels whenever possible"

But I dont know it would be ralated to my issue.

\$\endgroup\$
15
  • \$\begingroup\$ I would say it does not look like issue with impedance mismatch, or some interference on the lines. First and third picture shows that spike comes from nowhere, what I would do is check the device to see if it is not its glitch. For example, as you wrote at the end, use voltage follower at the device's side, and make two readings at the same time - after and before voltage follower. Where glitch will happen? \$\endgroup\$
    – Anonymous
    Oct 29, 2016 at 18:08
  • \$\begingroup\$ can it be related to multiplexer cannot keep up 8kHz sampling rate? \$\endgroup\$
    – floppy380
    Oct 29, 2016 at 18:40
  • \$\begingroup\$ Do I understand properly that this issue is only seen on one channel #7? \$\endgroup\$
    – Anonymous
    Oct 29, 2016 at 18:42
  • \$\begingroup\$ yes i observe it only on ch 7 \$\endgroup\$
    – floppy380
    Oct 29, 2016 at 18:44
  • \$\begingroup\$ What I say that you may check device first that its output is really free of these spikes. They look too legitimate to be good signal rather than product of some uncontrolled physical process. \$\endgroup\$
    – Anonymous
    Oct 29, 2016 at 19:05

1 Answer 1

Reset to default
1
\$\begingroup\$

Clearly you have a serious EMI problem on your MUX selector causing it to switch channels erroneously.

enter image description here

  • If you do not know how to measure signal integrity of each interface signal both input and output with a scope, then you have to blindly start filtering and shielding with CM chokes , shielded cable and Low Pass filters to lower the source impedance for rejecting Stray EMI either ground conducted or radiated.

    • this should be trivial to fix for an "analog" engineer with 1yr of RF experience.
\$\endgroup\$
11
  • \$\begingroup\$ Thanks. So the EMI is hitting to the MUX? But then doesnt that mean even-though I use filters in BNC side the EMI will still hit it though the air? I dont have experience in signal integrity (I know how to use scopes and el. instruments though). Cables are already shielded(BNC cables). Where should I place the choke filter considering my setup here: i.stack.imgur.com/8L7WX.png? At the entry of BNC box? would be very glad any kind of advice. this has been a nightmare for me. thanks in advance \$\endgroup\$
    – floppy380
    Nov 20, 2016 at 5:22
  • \$\begingroup\$ Very simple . USe very short gnd lead or two probes perfectly match in A-B mode and start verifying signal integrity. If you see noise, you should know how to deal with it by impedance ratios of load to source, that noise is rejected and signal is passed. In logic signals it often needs a termination resistor due to high dv/dt induced ringing from inductive cable not having twisted pairs etc. CM clamshell chokes around ribbon cable help raise CM impedance to reduce current injected by stray pulses while shunt caps lower differential impedance to stray crosstalk. \$\endgroup\$
    – Tony Stewart EE75
    Nov 20, 2016 at 6:33
  • \$\begingroup\$ If P2 cable is OEM it should have interleaved or twisted pairs with ground on logic. otherwise crosstalk.... I would examine effects of 500 Ohm termination for ringing and consider pull-up+down of 220R to 1K depending on length of cable. then add caps to high impedance sensors to shunt logic noise. There ought to be pads for these in BNC box or add R termination network chips.. Good luck . now you know what to look for. \$\endgroup\$
    – Tony Stewart EE75
    Nov 20, 2016 at 6:39
  • \$\begingroup\$ I think I should first check if the same noise is observed in all channels. Then it could be EMI on MUX. But if it is only at this particular channel, Im even doubting about the zener clamp for over voltage protection placed in the input of this channel. If the noise is only at this part cahnnel channel should I give a try to anti-aliasing filters? \$\endgroup\$
    – floppy380
    Nov 20, 2016 at 19:03
  • \$\begingroup\$ you must analyze the problem before guessing, because you don't have the experience. look for textbook quality waveforms and if not make it so \$\endgroup\$
    – Tony Stewart EE75
    Nov 20, 2016 at 21:31

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.