Voltage spikes using USB-6009

[cevert73]

Hi,

I've been having some trouble eliminating voltage spikes from the output of a constant temperature hot-wire anemometer (55P15 probes with 54T30 multi purpose anemometer) measured, usually at 10kHz, using the analogue differential channel AI0 on a USB-6009 DAQ, with the negative tied to ground via a resistor respectively.  I have the wind tunnel motor set to a fixed speed with the probe mounted in the freestream where yes I would expect to see some voltage (and therefore velocity) fluctuations due to a finite freestream turbulence intensity but the voltage spikes are anomolous so I'm trying to work out where they are coming from.

I've tried many hardware permutations by eliminating superflous equipment and switching to an alternative PC without success, furthermore whilst I'm attaining good matches for my means with an oscilloscope the rms values don't appear to be closely correlated, particularly when I know I'm adding more noise to the system which doesn't seem to be felt by the 6009.  I wondered if anyone had any suggestions from the USB-6009 perspective based on the attached graph from Labview 8.5.

Thank you.

[richjoh]

You need to take your own average, it may clean up your spikes. Put the measurement vi in a For Loop, then divide the For Loop output by the...(you know number of measurements). Start with 25 or 50.

Take the usual precautions to minimize hardware noise, twisted leads, differential probes etc.

[cevert73]

Thanks for your reply, perhaps I should have provided more detail as to what it is that I'm trying to acheive.  I'm looking to use this hot-wire anemometry technique to capture the transition of boundary layers from laminar to turbulent flow.  Once I've acquired the signal (say 100k samples) it will then be post-processed for mean, stdev and intermittency (the fraction of the flow time deemed as turbulent).  Now whilst the mean values look good already this seemingly spurious noise (which I've been unable to associate to any other device) will falsely increase my rms and provide an inaccurate description of the evolution of the transition process.

I feel as though I've taken every reasonable precaution with my experimental rig but nevertheless this problem is persisting.  My application is somewhat specific so really I'm just left wondering what viable solutions others have implemented for the 6009 that might be able to help me or a suggestion of alternative hardware with some rationale as to why it could provide an improvement.

Thanks.

[johnsold]

Do you see those spikes on the oscilloscope?

How are grounds connected?

Lynn

[cevert73]

-ve to usb ground on the 6009 via a 10k resistor (I've tried various resistances).  The 54T30 CTA which provides the differential output voltage via BNC isn't grounded, it was designed to be floating but it is (currently) powered via a 12VDC car battery.

Currently I've being using quite an old 8-bit DS oscilloscope and haven't been able to export the stored data to PC as I've been having trouble interfacing through serial.  I'm looking to borrow someone elses higher-end scope in the near future.  It's difficult to see the spikes on DC-coupling with this scope, owing to the resolution but I would say that the spikes aren't present to the extent they are in labview.  What I can say though is that I've used AC coupling on the scope and can quantify the rms in various situations, e.g. 5mVrms with zero flow, which rises to around 40mVrms when I switch on a stepper drive/motor but labview and the 6009 don't 'see' this increase.

Incidentally the output impedance of the CTA is 50 Ohms and I'm using the ±2.5V range.

[Coq_Rouge]

Are you sure the differential setup is the best in your case. Unless your input signal is true differential. I would try a single ended input configuration. As a starter I would have connected a 1.5 volt fresh battery to the daq unit. Using as short wires as possible. A battery gives very little noise. So this will give you an Idea about the the daq+PC system noise floor. You should see very little noise in this setup. Unless something is wrong. If you are using a laptop try to run in on batteries only to see if it is significant difference.

[richjoh]

Hate to answer a question with a question but how bad are your spikes? In other words, your waveform graph shows y-axis average ~2.08 and x-axis is 1 unit of time. Take 25 or 50 of these and average them but I think you need to use the differential connections on the USB-6009 (OK I see you are already :-). Your plot looks normal from what I can see.

I used the 6009 to take mV measurement. Only after averaging did the problem subside. I believe it was due to the inherent noise of the 6009.

[cevert73]

The spikes have a detrimental effect on the rms and even more so on intermittency.

Can I just ask, having noticed that the input impedance on the 14-bit 6009 is 144k, could moving to a DAQ with higher impedance / better input bias current improve this situation (specifically) for me?  If you (anyone in this conversation that is) were looking at a DAQ (USB interface or otherwise) for a job of this nature what would be a sensible threshold for acceptable impedance?  I understand that resolution, noise, accuracy, hardware clock performance etc. are additional considerations but I'm familiar enough with these already.

@Coq Rouge It needs to be differential I'm afraid.  As for the battery idea yes I could try something like that or a signal generator but with zero air flow on the hot-wire I do obtain a reasonably stable voltage which is absent of these spikes.  Once there's flow though the system becomes more dynamic and things get a little crazy.

Thanks.

[Coq_Rouge]

Have you tried it in single-ended mode? If not you should do it. As I said your signal is not truly diferential. If you have some other daq card it could not hurt to try. The input circutry on the daq 6009 series is far from optional.

[cevert73]

Yes I've tried single-ended of the CTA BNC output.  I've attached the positive and negative single-ended voltage plots.  Differential mode is certainly required as stated in the CTA manual.

Did you mean optimal rather than optional?  I appreciate that the 6009 is a basic device but I would certainly like to be able to look at the hardware specifications and decipher what it is that makes higher-end devices better, hence why I'm asking about the input impedance.

Thanks.