suggestions for high-resolution tiltmeter (inclinometer) sensor?

[branadic]

Received one V-block from Norelem as well as a 3D printed vial holder (a first draft for some testing). The vial is fixed inside the V-block using some rubber gasket ring elements between vial holder and vial. It works wonderful as expected. The vial holder as well as the recycled board needs some redesign and the V-block some machining (grinding on top and bottom), but sure enough I will have a full level vial based inclination sensor soon.

-branadic-

[StillTrying]

Why so much copper right to the very ends of the vial, doesn't the last 25-30% of it just reduce the sensitivity.

[metrologist]

I think because the effects can still be measured out to the very ends, it give the vial a much larger range.

[_Wim_]

I think because the effects can still be measured out to the very ends, it give the vial a much larger range.

If the bubble is completely under 1 side of the copper, can you still see the difference on where the bubble is? I also would have expected that you need a change in surface area, so when a bubble is completely under one copper piece, it does not matter where it is...

[branadic]

Each electrode is almost as long as the bubble, so yes you can measure a difference if the bubble is completely under one electrode and that is where the linear range of the capacitive difference measurement ends.

-branadic-

[_Wim_]

Each electrode is almost as long as the bubble

Ooh, that was not clear to me in the picture. From the picture above it seemed the electrode is almost twice as long as the bubble, hence the question of "StillTrying" and me if it would not be better to make the electrodes approx the same length as the bubble...

[branadic]

Ooh, that was not clear to me in the picture. From the picture above it seemed the electrode is almost twice as long as the bubble, hence the question of "StillTrying" and me if it would not be better to make the electrodes approx the same length as the bubble...

To be honest, the picture shown is not the latest one. The electrodes were shortend after some tests to a length a lil' bit longer than twice the bubble length.

Measured the behavior of the sensor today on a manuel test bed and I'm totally amazed by the performance. Without any noise generated by motors/pushers I crossed the limit of resolution of our reference, a Zerotronic ±1°. To be fair enough, the measurement range is in the order of ±0.35° / ±21' / ±1260" / ±6.10865mrad, but cost is far below 3.000 bucks.

But don't listen to me, say it with pictures. 

-branadic-

[JBeale]

Measured the behavior of the sensor today on a manual test bed and I'm totally amazed by the performance. Without any noise generated by motors/pushers I crossed the limit of resolution of our reference, a Zerotronic ±1°. To be fair enough, the measurement range is in the order of ±0.35° / ±21' / ±1260" / ±6.10865mrad, but cost is far below 3.000 bucks.

Very clean looking graph from your sensor, congratulations on that impressive performance!

[branadic]

Very clean looking graph from your sensor, congratulations on that impressive performance!

Thanks JBeale and yes you are totally right, this is an impressive performance, even if you compare repetition rates (1Hz vs. 6.1Hz).
I'm sure if you change to a more modern way of reading the capacity, you will get excellent results. Make the distance between both electrodes as small as possible, so you don't loose anything.
And not to forget temperature compensation. Your circuit will have excess temperature coefficient by rectification.

-branadic-

[JBeale]

I am interested to try the C2D circuit approach when I have more time. I couldn't do any more work on it lately, but here is 6 days of data from my old simple circuit, on the garage slab.  I added a slow 10-second-per-rev gearmotor with an offset weight on the shaft, all on a wood plank that is flexible enough to give the sensor a slow tilt wobble of several arcsec; that seems to have fixed the "sticking" steps I saw before. The one big step near the start is a manual readjustment to prevent going offscale, although it still did at the peaks over the next few days. With the wood plank, the whole thing may just be a humidity sensor, however all the smaller-scale features on the graph I checked do correspond exactly in time to real events, like a car driving into or out of the garage.

[branadic]

Mesaured the temperature stability of my setup together with the CDC today. Therefore the sensor was tilted by 90°, so that the vial is looking to the sky and each electrode is fully covered by the inner fluid. With the aquired data I did a temperature compensation comparison. Attached are the results. The large spikes seem to be my presence disturbing the measurement. However, with the temperature compensation coefficient I can now calibrate the sensor against a reference to deliver the final angle.

-branadic-

[LaserSteve]

I'm amazed that you are unaware of TI's FDC1004 EVM demo kit and the excellent data logging  software available for it.
Would like to know how the FDC compares to your NASA bridge.

I'm using the FDC1004EVM to monitor the position of  40 micron oil droplets inside an air  filter at work.. So far we're happy with it.

Steve 

[branadic]

Hi Steve,

if you refer to me, I use a PCap01 for my sensor not a NASA bridge. It's cheap, low power as per principle, has a DSP integrated that is sleeping while measuring so the complete calibration can be integrated writing some lines of assembler, needs only a few external components and has 3 additional resistive ports beside 8 capacitve ports. And since I fully understood how to configure it, how to programm it and how data has to be processed to get the best results, there is no need for me to use FDC1004. Nevertheless, we have evaluated FDC1004, but with worse results compared to PCap01. So PCap01 became the best sucessor for expensive and "dumb" AD7746 chasing atto farads, at least for me.


-branadic-

[raptor1956]

I recently purchased a few 20"/2mm levels from Aliexpress and when they arrived they were all off by minutes. 

https://www.aliexpress.com/item/HACCURY-High-Precision-Metal-Long-Level-Bubble-Precision-Equipment-Installation-Bubble-level-Accuracy-0-1mm-m/32856559241.html


The best one was about 1' off while the other two were more like 3-5 minutes off.  So I did a little dremel/file job and now they are within about 1-5" which I about what I wanted.  We'll see if they drift over time as I have no idea how they affixed the vial to the Al enclosure.  I suspect that you'd want some compliant adhesive that kept it in place but offered some compliance with temp change or force application.


Brian

[branadic]

Tried to calibrate my sensor, but without success since the Zerotronic is much noisier, slower and of less resolution. Need to wait unit our Nivel 230 is back in house to see if it is any better. Meanwhile I can further optimize my setup and design the final circuit board.

-branadic-

[StillTrying]

Is anyone going to make an X,Y (N-S, E-W) one.

[branadic]

Easily combine two single axis sensors and align them to perfectly match 90° to realize a two axis sensor.
If you are asking for a circular spirit level based two axis sensor instead, there is a patented solution available.
You won't get better results with a DIY solution based on commercially available circular spirit levels. Don't ask me why, I can't speak about it as this is part of the secret of the sensor design.

-branadic-

[rhb]

Received one V-block from Norelem as well as a 3D printed vial holder (a first draft for some testing). The vial is fixed inside the V-block using some rubber gasket ring elements between vial holder and vial. It works wonderful as expected. The vial holder as well as the recycled board needs some redesign and the V-block some machining (grinding on top and bottom), but sure enough I will have a full level vial based inclination sensor soon.

-branadic-

Are you saying that you bought a v-block and then two point mounted the vial with O rings? If so, why bother with the v-block?

[branadic]

Are you saying that you bought a v-block and then two point mounted the vial with O rings? If so, why bother with the v-block?

No, you misunderstood me. I used a segment of an o ring on the upper side as some sort of spring to elastically fix the vial into the V-block.

-branadic-

[JBeale]

slightly off-topic: clearly not "high precision" but I found an android app ("clinometer + bubble level" by plaincode) which uses the phone's 3D accelerometer to report the phone's angle with 0.1 degree resolution. After doing its builtin 180 degree calibration step, the actual precision is reasonably good. It does averaging, so there is a settling time.  There is a 1-D mode with the phone on edge, and 2D with the phone flat facing up.  Best with phones that do not have rounded backs, and have one edge flush without protruding buttons.

[branadic]

Can say for now, that resolution of my sensor (sample rate: 3.05Hz) is even better compared to Leica Nivel230 (sample rate: 1Hz, range: ±1.1mrad, resolution: 1µrad, accuraccy 1σ: 5µrad, accuracy 2σ: 10µrad). So now the golden question: How to calibrate this sensor without having a better reference?
Both (Nivel and Zerotronic) are not accurate enough and even the Leica's measurement range is to small.

-branadic-

[rhb]

Would you please explain what the plots show?

What is your estimate of the resolution?  I seem to recall a small fraction of an arc second.

The only thing I can think of that *might* work is something I saw in a NASA tech brief.  They needed a precise angle at very high resolution (though less than this).  They used a turned disk with a steel tape pulled by a micrometer head to rotate the shaft.  In your case you'd need at least a pair of 10:1 levers constructed in that fashion.  But if the disks were turned between centers and then mounted in the same fashion it should be quite accurate and easy to construct.

Visualize a pair of stepped disks with a 1" diameter section and a 10" diameter section. Pulling on the 10" section pulls a tape on the 1" section which is attached to the 10" section of the other disk.  So the distance is divided by 100.  You might be able simply to interpolate between steps of 0.0005" using a single disk.  That would give you 10 second calibration points.  A good digital mike reading to the tenth would give you 2 second calibration points.

[branadic]

First picture shows the response in x-axis of Leica Nivel. As can be seen two steps on a manuel test bench have been performed during searching for zero on Leica Nivel. The unit of Leica Nivel is mrad, thus one division equals 2µrad (2µm/m).
Second picture is the response of my inclination sensor in counts on the same change of angle, which was measured in parallel. It hasn't been calibrated yet, thus counts.

The small steps in my measurements close to zero signal of Leica Nivel (the test bench needs a while to settle down after a change in angle which can be seen clearly) are due to colleagues entering and leaving the lab    With a lot of fantasy you can see them also in the signal of Leica Nivel.

-branadic-

[Harm314]

Would you mind to share the raw data for the 'calibration' graphs? I would like to have a go at some filtering/correlation of the data to see how close these graphs match.

[branadic]

Would you mind to share the raw data for the 'calibration' graphs? I would like to have a go at some filtering/correlation of the data to see how close these graphs match.

Sorry, but no. I'm currently on that topic by myself.

-branadic-