Noise with Kulite Pressure Sensor on NI USB-6255

[HickoryDAQeryDoc]

Sitting in an ambient, static environment, our system has electronic noise 60 Hz equivalent to an RMS P'/P of 0.8-1.4%. We aim to measure fluctuations roughly an order of magnitude lower than this. Indeed we are approaching the quoted stated static accuracy of the transducer, but I'm confident that  the unsteadiness I'm seeing is line interference that can be further reduced.

Here's the system breakdown:

- Transducer: Kulite XCE-093-50A, 1.8 kOhm impedence. Four-wire connection... floating signal.

- Power conditioning: I'm running AC power to our DAQ, DC power supply, and signal conditioners through a Tripp-Lite line conditioner, with isolated filter banks.

- 10VDC Power: OMEGA PST-4130 Power supply, directly connected to the lead wires from the Kulite. RMS fluctuation in the 10.00 VDC power supply is 300-400 microvolts.

- Signal Conditioning: Kemo/Stewart VBF-44 bandpass filter set up to bandpass from 5 Hz to 25 kHz. Input and output gains are applied to a total of 60 dB.

- Signal Cabling: Signal is carried across short runs (< 3 ft each leg) of RG-58 cabling from transducer to filter, and filter to DAQ.

- DAQ: NI USB-6255 connected to a laptop running LabView, sampling at 50 kHz. The input from the Kemo Filter is connected as a differential input, with 100 kOhm bias resistors from both Signal+ and Signal- to AI GND, as recommended by http://zone.ni.com/devzone/cda/tut/p/id/3344. In accordance with another knowledge base article, I have attempted to run the laptop on battery power, without effect. I have also attempted to use a PC-6036E PCMCIA card with a BNC terminal block in place of the USB-6255, also without effect.

Note that taking the Kemo filter/gain unit out of the system and directly reading the unfiltered millivolt output of the Kulite also yields an unacceptable amount of 60 Hz noise (higher, in fact).

I do fully understand how many components of my system are not NI products, and that this is a complicated thing to troubleshoot, but can you provide any guidance on how to procedurally identify the source of the noise and eliminate it? I have tried most of the recommendations in the Field Guide to Wiring, but I'm still lost.

Thanks! 

- Jerrod

[HickoryDAQeryDoc]

I have also wired the Kulite input to be in a RSE configuration -- same result.

[HickoryDAQeryDoc]

Typical screenshot attached.

[HickoryDAQeryDoc]

Update -

Although the output of the sensor and the input to the filter/amplifier are _floating_, the output from the filter/amp is [b]single-ended[/b].

I've tried using both DIFF and NRSE with this in mind, but there's very little change in the noise level.

[Dan_K]

Hello Jerrod,

I have a couple of suggestions and questions about your setup.

• First, it is pretty obvious that your noise is environmental. If it is at all possible try making a similar measurement in a different location. By doing this you can see if it is just the room you are in that is causing the interference or if it is the equipment you are using that is generating it.
• What kind of lighting do you have in the room where you are making your measurements? Fluorescent lighting is notorious for introducing a lot of 60 Hz noise into measurements. In many case just turning the lights off can reduce the noise in your signal.
• You could also try wrapping all of your cabling and your USB-6255 in aluminum foil. This would add an extra level of shielding to your setup, helping to block environmental electric fields.
• Another possibility, depending on the frequency of the signal you wish to measure would be to sample at 60 Hz. This would cause each measurement to read the same amplitude of the 60 Hz noise wave, effectively ignoring it.
• If processing time isn't an issue in your application you could also try to use a software bandstop filter with very sharp edges to remove just the 60 Hz noise signal.
• Lastly, when connecting your signal differentially, did you connect pulldown resistors from both of your input lines to ground? If not, try performing your measurement with this type of configuration.

Let me know if any of this works or if you are still unable to remove this unwanted noise so that we can continue to work towards getting your data acquisition up and running and working.

[HickoryDAQeryDoc]

Dan - Thanks for the reply!

1) I've tried this in two locations on opposite sides of the building. 

2) Yes, there are plenty of fluorescent lights in this room and building. Turned them _all_ off, no change.

3) I will try out aluminum foil today. Do you recommend wrapping everything (BNC cables), or only the unshielded 30AWG leads to the transducer?

4) Unfortunately sampling at 60 Hz is not feasible. Our ultimate goal is a PSD from 5 Hz - 25 kHz, and we'd later like to look at even higher frequencies... 150-200 kHz.

5) True, but this becomes less feasible when I have so many harmonics in the system as well --- check out the example screen capture posted above. 

6) Yes, in the differential configuration, I indeed used 100 kOhm resistors between AI+ and GND, and AI- and GND.

If this fully eliminates environment & cabling, I'm beginning to thing that the noise must be generated by the either the Omega power supply or Kemo filter.

[HickoryDAQeryDoc]

FYI, aluminum foil didn't do it, either.

[AnalogKid2DigitalMan]

Did you apply the tinfoil with the shiny side IN? 🙂

I jest, these kinds of noise issues can be a bear to deal with. Is there any heavy machinery or electric loads (fans, motors, etc) in the nearby vicinity? AC power conduits/panels in/on the walls?

Do you have access to an oscilloscope? It's best to probe starting out with a mininmal configuration, working from the load cell back up the chain, with and with any DAQ hardware connected.

-AK2DM

[Dan_K]

Hello Jerrod,

AnalogKid2DigitalMan is right, all of the things he mentioned are good(?) sources of environmental noise. In addition from working from a minimal configuration and working up the chain you could perform tests on each of your devices individually to narrow down which one is the source of the unwanted noise. This would require designing specific tests for each device that would use only the specific devices, but the effort should pay off in the long run.

[HickoryDAQeryDoc]

Thanks guys.

As far as the environment...

We can turn off the fluorescent lights in the entire building without effect. There's nothing else major running in our small building, but perhaps another building in the immediate area is on our power grid, with interfering items?

I'm going to run the whole setup in a different building today, to see if that rules anything out. We're in a metal hangar at a small airport -- it can't be the best power source here.

As far as independent component testing... 

Kulite Alone

I can remove the Kemo filters from the setup and look at the unamplified Kulite signal on an oscilloscope. Here, I see the nominal 30 mV output, but already with a significant amount of noise (1-2mV RMS!). 60 Hz and its harmonics are still involved, but their dominance appears less well-defined in general than it does after the Kemo filters.

On a percentage RMS/nominal basis, this amount of noise from the unamplified Kulite is *worse* than when the Kemo filter/amplifiers are in place. However, while the output from the Kemo filters is better on a percentage basis, the appearance of 60 Hz and harmonics is more pronounced with the Kemos in place.

Kemo Filters Alone 

So, noise exists in the Kulite output upstream of the Kemo filters, but the character of the noise changes so dramatically that I can't rule out the filter/amps as a contributing cause either.

To isolate the Kemos, I can feed a waveform-generator signal into the Kemo filters in place of the Kulite signal. Indeed, I do see my 60 Hz + harmonics, in similar FFT character and amplitude as when the whole system is setup.

Now, is this the fault of the Kemo electronics directly, or continued evidence of a cabling/environmental issue?

Power Supply Alone

Since I'm seeing noise even directly out of the Kulite, it's possible my power supply is not good enough? I'm using an OMEGA PST-4130, a little $170 unit intended for sensor bridge excitation. Viewed on the oscilloscope, the RMS fluctuation in its 10VDC output appears to be about 100-200 microvolts. This is actually a little better than I got trying a pure DC source (two 9V batteries through a DC-DC converter).

Looking online, I have found a Krohn-Hite MV-106 DC source which is a considerably more expensive unit at $2800. According to specs, this will supply 10VDC with an RMS ripple & noise of 50 microvolts. Maybe it's worth a shot?