Inconsistent readings with HoneyWell Sensors on USB-6281

[NWMI]

I am using HoneyWell Sensors with my USB-6281 to measure pressure vs time.  While calibrating the sensors show no problem.  It was only by fluke that I noticed the sensors were behaving strangely.  I measure a device that regulates fluid pressure and have a system in which a constant flow of water is pumped through this device.  When the water is being pumped and the device is operating properly the problem is not noticeable.  The water pump has about 4 hours of steady flowing water being pumped through the device.  One day I left the pump and data running over night.  It was easy to see that roughly at about 4 hours all the pressure began to dip low indicating that the water had run out and there should be no more pressure stimulus.  But I started seeing peaks and valleys throughout all 10 of my sensors, sensors that had no active pressure being applied.  I began to consider the design of my water pressure system so I decided to run a test over night with absolutely no pressure at all, that is to say closing all channels to any stimulus.  I found the same kinds of peaks and valleys occurring through all 10 channels again.  These peaks and valleys occur similarly throughout all channels save for a scalar value differentiating from channel to channel.

I have spoken to NI Application Engineers over the phone and they have run a bunch of tests resulting in finding no problem.  There has been speculation of the occurrence of "ghosting" which doesn't seem correct since my sample rate is 1 Hz.  There was also the suggestion that this particular model has only 1 ADC and that may be the issue, also an alliteration to ghosting.  Maybe a DAQ card in which there is 1 ADC per channel would be the solution but it was recommended to keep that as a last resort as this problem should be fixable without purchasing a new card.  

I am using this DAQ card in conjunction with a different card to power the sensors.  The grounding is done through both cards and ofcourse the data is in the NI card.  There is no problem with the wiring configuration but the problem continues to persist when given static pressure, whether during the day or night.  

I am looking for recommendations of sensors that will work well in condjunction with the USB-6281 and will give consistent pressures given a period of static pressure.  Has anyone heard of this problem occurring with these sensors per say or just the problem in general?  Ideally I would like the accuracy of these sensors to be within ± 0.1 mmHG.  The HoneyWell Sensors I use are 1 PSI analog single ended.

Thank You

[Henrik_Volkers]

What range do you use?

What signal amplitudes do you see?  With pressure and without (but with the ghost-signal)?

Sounds like you have a sine like ghost signal?

How do you capture the signal , one single reading per second or more and mean value?

My first guess would be power line noise, 50/60Hz on your signal, but if you capture only one ms every second you will find a  sine like signal (folded, or beat signal) due to the power line frequency drift. (actally you migth see more power line frequency drift than a pure sine)

Using MAX , can you capture a signal with a higher samplerate ? Say 3000 points @ 10kHz or 100kHz samplerate?  How does this signal looks like, with and without pressure?

And pictures of the signals tell more than words 😉

[NWMI]

The sensor and power from the daq cards are both at 5v and the range I need to capture is from 0 to roughly 60 mmHG with special emphasis at the 5-15 mmHG range.  The calibration on these sensors shows a linear relationship between voltage and pressure as the equation y = 13.387x - 6.472 where y is mmHG and x is voltage.  I'm seeing noise with a range of roughly 100 mV which would correspond to 1.33 mmHG of variability, which I need to eliminate.

I'll attach some signal pictures to show the amplitudes.  All signals in the pictures are when there is no pressure being applied and the channel is closed to outside pressure.  It does seem like a very loose sine wave im not sure if it due to ghosting.  Capturing the signal is simply at 1 sample per second.  I havent tried to capture at a higher sample rate on NI Max since it wont let me capture a 4 hour period of data.  If im capturing 3k samples at 10kHz or 100kHz wouldnt my readings be done in 0.3s or 0.03s respectively?  The problem doesnt occurr in such a short time span.  I need to run atleast 2 hours of samples thats why I keep my sample rate as low as possible.

I apologize for the slow response I'm getting busier these days but thanks for the prompt reply.

Hope you are having a nice Fall.

[johnsold]

How will you be doing your actual test? You have only mentioned MAX. What language will you be using for the real data collection? 

You can certainly sample at much faster rates and then average or discard excess data.  I suggest running some tests at 1000 samples per second. At that rate you can easily detect power line frequency interference.

Try putting a simple voltage divider on one channel. Connect it to the 5 V source powering the sensor. Connect another voltage divider to the 5 V output of the DAQ device and wire it to a different channel. These signals will detect any shift in the power supply voltages and will separate DAQ effects from sensor effects.

You said "there is no pressure being applied and the channel is closed to outside pressure." If there is a small amount of air in an otherwise closed system, temperature variations or atmospheric pressure variations might change the pressure at the sensor. Can the sensor trap a small air bubble at the diaphragm? Does the heating or cooling system for the laboratory where the sensors are located operate on a time scale similar to the signal variations? Can you monitor temperature at the sensors?

Lynn

[Henrik_Volkers]

Reading 300ms of data  will help us to analyse your problem 🙂

Samplerate >= 1k SPS will be fine.

I would suggest that you do this with and without the pump running.

The trick is to get 1s (or at least one periode of power line frequency (that will sligthly change) ) of data with a higher samplerate (for example 1kSPS) and condense this data to one value. Usually you take simply the mean value. That will (usually) reduce the noise.  

BTW:

here is your data but I pasted into irfanview and saved it as a png file. Not everyone like to open xls files from the public 😉

And Lynn make a very good point about the static pressure changes if you have a closed system and the temperature is changing! In a water system it's even worse without air!

Do you have a temperature sensor in your system?

[NWMI]

The actual test is done through a custom software developed using labview.  

The wiring is set up as 5V power source daq to the sensor, data from the sensor to the NI daq, and ground from the sensor to both the NI daq and third party power source daq.  The 5V output of the NI DAQ device is not used.  Please tell me if I am misunderstanding anything.  Also if you can give me a recommended Voltage Divider for this application it would be much appreciated.   

The system is always purged for any air bubbles, visible or non visible as that would compromise the test.  I do not have any temperature sensors applied to this system but the room is kept at a constant temperature.