Data filtering and Input noise

[juhala]

Hi

I've encountered a problem that I hope someone can help me out. 

Im working with hydrogen sulfide (H2S), which is administered by a peristaltic pump, and is set adjust concentrations over a period of 24 hours.  I use a microsensor to measure H2S, which provides me values in mV, which range from 1.5 mV-10. This output signal is obviously quite low, so I use a LabJack U3 LV and adjusted my H2S sensor offset to be at 100 mV.  

However, when applying my conversion based on y= Mx + b, I get large variations in Value, even -  values. I currently take the average H2S(0) [0,10], but my values still fluctuates, despite my sensor measurements are quite stable. 

My three point measurements for conversion of values with D-A input, offset at 100 mV gave be the following: 

µmol/L	Volt
0	0,088
3,5	0,113
6,3	0,13

with the equation for conversion of mV to the units of µmol/L as y = 149,25x - 13,201

How can I stabilize my input signal, and e.g. ignore negative values so the output my pump is relying on becomes more steady? 

Thanks in advance.

Sincerely, 

Julie 

 

[AzeoTech]

You should probably talk to LabJack, but my first take is that you shouldn't be using an U3 for such a small signal.  Your full range from 0->6,3 umol/L is 42mV.  The U3 is a 12 bit device with a range of 0-2.4V which yields a minimum step size of 0.5 mV.  The specs indicate one bit of noise, so expect the signal to fluctuate at least +/- 0.5mV, probably more like +/- 1mV.  Using your slope calculation of 149,25 per unit x, you should expect noise on the order of +/- 0.15 umol/L.  Given your calibration range of 0-6,3 your expected noise is 2.5% of the your full scale (6,3 umol/L) reading.  That is pretty significant and probably what you are seeing.

To fix that you need one of two things:

1) a T7 or UE9 that has 16 bit resolution and gain

2) a nice quiet amplifier on your signal

That all said, the folks at LabJack are the experts, so I will prompt them to add their reply.

[LabJackSupport]

We would recommend you start by stepping back and tracking down your noise source. Our troubleshooting AIN application note has some guidance that could be helpful:

https://labjack.com/support/app-notes/troubleshooting-AIN#TooMuchNoise

Other than that, we agree with the response from Azeotech.

1) The T7 or U6 have the best resolution of any of our devices. We do not recommend the UE9 for new systems since the T7 performs better and costs less. We also have pro versions of the T7 and U6, which feature 24-bit resolution. The 24-bit converters have great noise rejection as well.

2) We sell an amplifier accessory called the LJTick-InAmp that could help improve your effective measurement resolution if you decide to stick with the U3:
     https://labjack.com/accessories/ljtick-inamp

[juhala]

Thank you very much for the answer.

In the H2S sensor software there is actually an option to change the analog output signal intensity. Currently my output is set at 100 mV and output for mV/unit is also 100. I will try and increase the output (mV/Unit) to see if that will decrease the noise, by amplifying the output signal .

Otherwise I can see that I will need to get some more gear.

[juhala]

Hi again. 

I tried amplifying the signal in my output software, which unfortunately didn't work. 

Afterwards I connected you amplifier(LJTick-InAmp), which only enhances the offset as my software already did (+ 0.4mV), but the signal is just as weak and fluctuating as ever. 

I suspect I need to use the switches (1 -4 as my cable is attached to IN A+), but no matter which switch I turn on I don't see a difference.  

Do you have an idea what i could be doing wrong ? 

Picture of the setup is attachedd (switch 5 is turned on though it is difficult to see on the picture). 

[juhala]

Corrected: 

I suspect I need to use the switches (7-10 as my cable is attached to IN A+), but no matter which switch I turn on I don't see a difference.  

[LabJackSupport]

Did you try the tests described on our troubleshooting page? How does a DMM measurement compare to what the LabJack reports?

If you can verify and understand the limitations of the U3, you can better understand your issue. If the U3 is operating within specification, but you still see higher noise than specified when measuring with the sensor, you know the issue is due to your sensor output or the U3 connection to the sensor. If the U3 is operating within spec and a DMM produces similar readings to the U3, you can infer that the issue is to do with your sensor.

You could likely do some filtering or more specifically over-sample and average values in software to help account for some noise, but you do not want to try that before you know there is not some fundamental issue with your setup.