Liquid Nitrogen and Accelerometer Advice


(Pamela Hauff) #1

Has anybody thought about using liquid nitrogen to cool Accelerometers to eliminate noise?
Has anyone used liquid nitrogen at the space?
Any advice would be vastly appreciated.

Thanks

Pamela


(Nathan Beveridge) #2

I see at a quick google there are specialised accelerometers for use at very low temperatures such as on rocket boosters with liquid helium, etc. I wonder what temperatures a ‘regular’ accelerometer works reliably at? Could dry ice be enough of a temperature drop to intially (and cheaply and relatively safely) experiment with? It looks like it’s $6 per kilo. Are you dealing with an application that is getting wonky because of noisy signal/data? Or just pondering the possibilities? :slight_smile:


(Pamela Hauff) #3

All of the above. Yes I see that dry ice causes water to condensate on the electronics but maybe use CO2 could be the way to go gas small cylinder for smaller airborne sensors. But I would like to experiment with Nitrogen. I was watching a YouTube video about Voyager1 & Voyager2 the radio telescopes use liquid nitrogen on the amplifiers to detect the signals from them.
Pamela


(Svenska) #4

Not sure cooling is the way here.
Dealing with noise in sensors is a common thing to deal with. There is rolling averages, filters, EKF and other stuff to get rid of it.
If you’re interested in going that way give me a yell and I throw some reading material your way.


(Pamela Hauff) #5

Yes I was thinking of using tensorflow in python to filter the noise…But it isn’t really noise it’s random highly…chaos from a double oscillating pendulums.
Given a random seed or initial conditions a double oscillator will behave chaoticly. If tensor flow can identify the set of initial conditions from the error signal. Then remove the error from the data. The sensor produces no error.

This liquid nitrogen experiment is really me just trying to isolate that error signal.

Pamela


(Svenska) #6

I would stay away from any machine learning solutions. That is real hard to get right, has only been finding its way into control theory for a few years. There is no good papers or anything on it yet.

Have a look at the below. These are proven ways to deal with it.
In increasing complexity:

http://www.analog.com/media/en/technical-documentation/dsp-book/dsp_book_Ch15.pdf
http://greg.czerniak.info/guides/kalman1/


(Pamela Hauff) #7

Yes I understand the accerometer would not move when cooled. It could be possible to see the source of the error conditions more easily in this state.

Pamela


(Buzz) #8

The sensors in the ‘pixhawk2’ include multiple sets of accelerometers, gyroscopes, barometric sensors, etc, and they all can be calibrated for “drift” etc . We also find a stable temperature over time for the sensors is much more important than a low temperature, so we heat the entire sensor suite to a temp that is guaranteed to be above any ambient temp experienced, and get great results. The temp we use is 60degrees, but it’s not the value that’s important, it’s just the fact that it’s repeatable, and consistent over time etc.


(Pamela Hauff) #9

The error signal slows down when cooled? That’s what I would like to know just out of my own scientific curiousity. I would personally like to measure this with just a simple circuit. I am going to test with CO2 that maybe cool enough for my purposes.