low frequency <1Hz Noise / Drift

[rsheets]

I am experiencing more than desired low frequency <1Hz noise/drift and can not seem to get rid of it.  I am takind STATIC measurememnets, we sample at 1 S/s

Hardware:

Cdaq 9172

NI 9205 card

Transducer: OMEGA PX02 0-10V transducer (0-20KPSI)

Power Supply: Sola HD 0-24V Regulated power supply

I am confident thsi is not related to wiring, our total runs are maybe 3m L  shielded wires, the entire control box has a dedicated line from our service panel run through an aluminum conduit.

 I have a 9209 on order, but NI really screwed up my order, and it is now 2 months late.  I undersand i am at the limit of the card (9205) as far as resolution.

But i dont understand why i am getting a drifting/oscilating signal.

I am attaching picttures of the pressure signals, first picture is an unfiltered chart.

Second picture is witha  point by point filter 1Hz

Amplitude is PSI  1 PSI corresponds to 0.0005 V

[Henrik_Volkers]

How do you measure the channels?

If you read a channel only once a second  line noise / hum can fool you ...  

How does the signal of one channel looks like if you read 100ms @ 100 kSPS ??

I expect some line noise 🙂      so if you read once a second the line phase is drifting giving you a drifting signal.

read faster, mean/filter decimate ..  better?

temperature ?   also a nice source of drift is a drifting temperature...

[rsheets]

The temperature i will rule out, how can you get a constant amplitude varying temperature inside of a boilding over a few seconds, maybee if they just drifted one way or abnother i would buy that.

We can try reading faster, but from my recolection we get the same problem.

We are reading 1S/s  and with a 9209 or 9208 thats about as good as we can do.

Why would line noise hum oscilate periodicaly?

[Henrik_Volkers]

Assume a constant value with some sinusodal power line AC hum added.

You have one oscillator in your DAQ, reading (exactly) every second and you have a hum with line frequency, that hits the same phase every second IF it would be constant 50 Hz or 60 Hz, BUT actually the power line frequnency is drifting (for europe see http://www.mainsfrequency.com/verlauf_en.htm )

Assume the power line frequency is off with .01 Hz: You would measure the constant value with a 100 second AC hum.

And the powerline frequency is changing all the time since the frequency is used to control the power plants 🙂

That why I asked for the higher sampled sensor signal 🙂  you have about 7mV 'drift' in your signal... BUT how does it look like when you sample it  as fast as you can ??  How, when you sample with 1 kHz? ?  If you know how your signal actually looks like, THAN you (or we 😉 ) can think about tricks or methods to extract the information you are looking for!

Usually, if your signal has an AC content (in the range of intrerest) you need to sample fast enough to use a software filter OR you need an analog filter before sampling.

[rsheets]

We are using a RC filter 3Hz low pass on one of the signals, it does not seem to help.  Let me do some sampling and get back to you.

[rsheets]

It apears higher sampling is helping some.

However when we sample more, now we have tons of data that we dont need, is there a convenient way to consolidate/average this data once per second?  Having data at more than 1Hz is a problem when you have tests lasting 10-20 hours.  

[Henrik_Volkers]

Data reduction is easy...

the vi's   called

sample Compression

or

decimate continuous

or

read the data in 1s blocks  and calculate the mean value (read the actual samplerate property and use that value for number of samples to read. LabVIEW (NIDAQ) will do the timing for you)

To analyse the noise ( seems more than line hum, some switched power supplies nearby? ) I would take a look at the power spectrum. 

If you want to post data, just copy and paste the graph with the data dispayed (100ms (or 1s) of  highest saplerate possible)  into a new empty vi, in the block diagram rigth click on the diagram icon and create- constant. save & post

[rsheets]

Thanks Henrik for the help!

I will give some of this compression a try, and get some pictures of the cabinet.

I dont believe any of the power supplies are switching If i remember the sola is not switching, i upgrade to this from a switching supply because i thought that was my issue with noise previously.

I will post some data soon too.

Power supplies:

24V

http://www.emersonindustrial.com/en-US/documentcenter/EGSElectricalGroup/products_documents/control_...

8-15V

http://www.omega.com/pptst/PST-4130.html

[rsheets]

Decimate signal seems to work well and is quite simple, but i am not able to get a time stamp out of this, I have tried different methods, but alas, no time stamp.

sample compression is still giving me tons of data, I am unsure what i am doing improperly with this. 

For your third example will you post a simple VI showing this, that would eb tremendously helpful.

[Henrik_Volkers]

here is a quick modification of an example shipped with LV

If you select browse for the Physical Channel you can (ctrl) select more than one channel, however the powerspectrum is shown only for the first channel listed