suggestions for high-resolution tiltmeter (inclinometer) sensor?

[Martinn]

Jon, do you have a pitcture of its innards? Would be interesting!

But well, the whole point of this exercise was to find out if simply placing a $5 sensor on a PCB could achieve the same accuracy as a Niveltronic or Talyvel, which both (to my knowledge) use differential inductive transducers (LVDT). Looking back, figuring out how the pcap device works took such an effort (due to the miserable documentation) that at least design wise there's not much saving.

[branadic]

The only info I know of is given in their compenium:

https://www.wylerag.com/fileadmin/pdf/compendium/Kompendium%20deutsch%202018.pdf

The theory of operation is given on page 21 and 38. Looking into it, the sensitivity is specified to 10 µm/m aka 2 arcsec, while repeatability is specified to be 1 μm/m.

I agree, that the documentation of the PCap is hard and most of the really interesting details took me years to figure out myself, which eventually made it into my PhD thesis. However, since then we made huge progress with capacitive sensors in general, but also our capacitive two-axis inclination sensor. We recently finished a project, in which a process was developed to improve its linearity.

-branadic-

[Martinn]

I agree, that the documentation of the PCap is hard and most of the really interesting details took me years to figure out myself, which eventually made it into my PhD thesis. However, since then we made huge progress with capacitive sensors in general, but also our capacitive two-axis inclination sensor. We recently finished a project, in which a process was developed to improve its linearity.

-branadic-

The PCap devices are somewhat strange. I guess initially developed by acam in Studensee, then acquired by ams and then going to Sciosense in the Netherlands? Does not really feel like a successful product. Maybe they gave it a final chance before discontinuing the whole line. At least they are available at Mouser or Digikey, but I would not trust those in a product development. Also, there are near zero other projects to be found on the internet using the PCap devices. Little wonder given how time consuming it is to reverse-engineer them in order to compensate for missing docs.

Found this interesting paper most likely using a PCap device https://www.hahn-schickard.de/fileadmin/media/Hauptnavigation/Anwendungsberichte/AB_2-axis_inclination_sensor_en_R0.pdf

I would assume your PhD thesis is public? Would you share a link? Thanks!

- Martin

[branadic]

The PCap devices are somewhat strange. I guess initially developed by acam in Studensee, then acquired by ams and then going to Sciosense in the Netherlands? Does not really feel like a successful product. Maybe they gave it a final chance before discontinuing the whole line. At least they are available at Mouser or Digikey, but I would not trust those in a product development. Also, there are near zero other projects to be found on the internet using the PCap devices. Little wonder given how time consuming it is to reverse-engineer them in order to compensate for missing docs.

Well, that all happened right after the founder of ACAM, Mr. Braun, passed away. It is and all their other ICs are however a successful product(s), although you obviously don't see them in everyday devices.

Found this interesting paper most likely using a PCap device https://www.hahn-schickard.de/fileadmin/media/Hauptnavigation/Anwendungsberichte/AB_2-axis_inclination_sensor_en_R0.pdf

Yes, in the latest versions of this sensor we are using the PCap devices

I would assume your PhD thesis is public? Would you share a link? Thanks!

No, it's not public, hence why I can't share a link nor the IP how to properly use the PCaps at their full capability, which I already wrote took me years to figure out. But I can tell you, once understood how to write your own firmware and how to deal with the raw measurement values the PCap is an amazing powerful device. No, I'm not getting any money for that statement, it's just my experience from multiple different and current sensor projects using this/these device(s).

-branadic-

[Martinn]

No, it's not public, hence why I can't share a link nor the IP how to properly use the PCaps at their full capability, which I already wrote took me years to figure out. But I can tell you, once understood how to write your own firmware and how to deal with the raw measurement values the PCap is an amazing powerful device. No, I'm not getting any money for that statement, it's just my experience from multiple different and current sensor projects using this/these device(s).

I am not sure why they invented their own DSP in that chip. Sure, you can do all kinds of processing, but why not simply expose the CDC/RDC fully via registers and let some external microcontroller do it? This way the user would not have to mess with barely documented assembly innards.

Even having the DSP, I am not sure what the benefit of using it would be. As far as I can see if you use the standard firmware, all it does is calculate the Cx/C0 ratios and do the averaging. For me this is fine, whatever correction I will use (for the time being a linear scaling factor) will be calculated in a STM32 processor, in plain C using float variables, with proper debug support.

Of course, if you are forced by whatever reason (cost?) to have a single chip version, you could make a sensor with PDM output and linearization for whatever sensor you attach to it, but I somehow can't imagine that this chip ever will find its way in such a price sensitive application.

Do you know if there is still anyone left from the original developers who knows how this chip works? Or is all Sciosense inherited was a mask set for production?

BTW nice work with the dual axis fluid sensor! It reminded me of my first ever sensor project ages ago, where I attempted to build an electrolytic inclination sensor for cave surveying. I used a chamber glued together from acrylic glass, with some vodka as fluid. Leaked of course. I would guess for your sensor, main problems were fluid properties (wetting/capillary effect at boundaries), stability and probably topmost fluid staying inside the cavity and not evaporating.

[RoGeorge]

Just out of curiosity, how much does the PCap sensor costs, or at least its price range for a 1pcs order?

[branadic]

I am not sure why they invented their own DSP in that chip. Sure, you can do all kinds of processing, but why not simply expose the CDC/RDC fully via registers and let some external microcontroller do it? This way the user would not have to mess with barely documented assembly innards.

Reason is that you can implement your linearization algorithm into it, beside many other things you could do with it, that I can't comment. If you don't need that ability I highly recommend writing your own firmware that simply reads all the raw data from the measurement registers and shift them into the result registers. That's all I can say about it.

Do you know if there is still anyone left from the original developers who knows how this chip works? Or is all Sciosense inherited was a mask set for production?

Yes, all the developers are still there.

BTW nice work with the dual axis fluid sensor! It reminded me of my first ever sensor project ages ago, where I attempted to build an electrolytic inclination sensor for cave surveying. I used a chamber glued together from acrylic glass, with some vodka as fluid. Leaked of course. I would guess for your sensor, main problems were fluid properties (wetting/capillary effect at boundaries), stability and probably topmost fluid staying inside the cavity and not evaporating.

Thanks. Getting everything fluid tight and selecting the right fluid was indeed one of the challenges, beside many others. Making the sensor linear was another one for example.

Just out of curiosity, how much does the PCap sensor costs, or at least its price range for a 1pcs order?

It's not a commercial product just yet, there are still some challenges that we are working on together with partners, so I can't put a price tag on it.

-branadic-

[Martinn]

Finally getting somewhere.

Sensor PCB is working fine, I added the display with a very basic UI. It shows two pages, a first one for parallel adjusting the PCB (using the adjustment sensors on the PCB) to its nominal 100 um air gap.
On a second page it shows the inclination in um over the baseline (currently 170 mm, the width of my "calibration" setup, later 1000 mm) with an offset zero and calib factor adjustment knob.
With this I think the core functionality is implemented. The rest will be some UI polish, nonvolatile storage of calibration and maybe offests, finishing the mechanics (3D printed housing for display module) and such.
1 um/m stability will be tough. I'll have to implement like a 5 s window averaging for this to work.

Just out of curiosity, how much does the PCap sensor costs, or at least its price range for a 1pcs order?

In case you meant the sensor chip itself: Search on Mouser for example for pcap04. Current 1pc price is around CHF 5.90.

Thanks. Getting everything fluid tight and selecting the right fluid was indeed one of the challenges, beside many others. Making the sensor linear was another one for example.

Why bother with linearity? As long as it's somewhat monotonous, you can linearize that easily enough. I used a piecewise cubic hermite interpolator for this on a different capacitive sensor, https://ch.mathworks.com/help/matlab/ref/pchip.html - works very well.

[branadic]

Why bother with linearity? As long as it's somewhat monotonous, you can linearize that easily enough. I used a piecewise cubic hermite interpolator for this on a different capacitive sensor, https://ch.mathworks.com/help/matlab/ref/pchip.html - works very well.

I would agree for a one-axis sensor, but as it's a circular spirit level things are way more complicated than you think.

-branadic-