suggestions for high-resolution tiltmeter (inclinometer) sensor?

[rhb]

I make high resolution inclinometers mainly for marine use https://www.jmsensors.fi

Resolution is 0.001 degrees (RMS at 10 Hz, much lower after averaging). Accuracy 0.005 degrees + 0.5%. Long term stability (years) better than the accuracy. Measuring range +-17 degrees. The basic model is 1 k€ (USB connection, Windows software and IP68 aluminium enclosure). I don't think you can find cheaper ones with similar specs.

I also make an OEM model, which can be temperature calibrated for very good offset stability (typical +-0.01 and max +-0.02 degrees -20C - 80C).

I use a MEMS sensor and calibrate each sensors at 1 degree intervals.

I'd have to rate this as blatant advertising. It shows no knowledge of the thread.   The discussion in this thread is at least two orders of magnitude (100x!) more sensitive than your product.

Please behave better.  A tutorial series on applications for your product are, I'm confident, most welcome.  But, you are in nerdland.  You do not want to discover what nerds will do to sales puppies.  It's really quite appalling.

So, either step up to the plate and blow us away with what you know about measuring small angles, or please, go away quietly and don't do this again.

Or to put it another way:

If you can quote chapter and verse, I'll listen.  If not, I think you should move one.

Nobody in particular,
Reg

[rhb]

Who was the UK supplier of vials?    I need at least four 2"/2 mm vials.   @ez24 gave me a couple (I broke one :-(

I may get some more Chinese, but I want to compare with UK for my application.

But I *do* have my CA3039s! So once I recover from my gratuitous folly, I'll get a bridge going.

Many thanks to those who have been tracking down the theory via the patents.

When I calculate the theoretical possibilities and look at the cost I am blown away by what a very low BoM cost design can do.

[jmaja]

I make high resolution inclinometers mainly for marine use https://www.jmsensors.fi

Resolution is 0.001 degrees (RMS at 10 Hz, much lower after averaging). Accuracy 0.005 degrees + 0.5%. Long term stability (years) better than the accuracy. Measuring range +-17 degrees. The basic model is 1 k€ (USB connection, Windows software and IP68 aluminium enclosure). I don't think you can find cheaper ones with similar specs.

I also make an OEM model, which can be temperature calibrated for very good offset stability (typical +-0.01 and max +-0.02 degrees -20C - 80C).

I use a MEMS sensor and calibrate each sensors at 1 degree intervals.

I'd have to rate this as blatant advertising. It shows no knowledge of the thread.   The discussion in this thread is at least two orders of magnitude (100x!) more sensitive than your product.

Please behave better.  A tutorial series on applications for your product are, I'm confident, most welcome.  But, you are in nerdland.  You do not want to discover what nerds will do to sales puppies.  It's really quite appalling.

So, either step up to the plate and blow us away with what you know about measuring small angles, or please, go away quietly and don't do this again.

Or to put it another way:

If you can quote chapter and verse, I'll listen.  If not, I think you should move one.

Nobody in particular,
Reg

I'm sorry you feel that way. I just noticed this thread a few days ago. I read all of it before replying. The original question was about long term measurement and 0.001 or better resolution for which just a module was said to cost more than 1 k. I felt my inclinometer might be a solution for the original problem. Then the conversation went to very high resolution reading of a bubble in a vial. But was decent long term stability or temperature stability achieved? Not much point doing long term measurements without good stability. Of course very high resolution may have some other applications even without stability.

I can see my post regarded as marketing or even spam. I don't really do that. It must have been more than a year since the last time I linked my homepage to any forum.

I don't really know what is the resolution limit of my inclinometer with averaging, since it has not been used for applications needing sub 0.001 degrees resolution. It's dynamic response is quite fast, about 10 Hz. So it's quite different from the vial bubble one. As I said I get 0.001 RMS at 10 Hz and averaging reduces it with SQRT(n) as long as the stability effects come into play or noise can no longer be averaged out. I do know it's clearly better than 0.001 degrees, but I haven't really had any need to measure very long term. With one minute averaging you would get 600 samples and 0.001/SQRT(600)= 4e-5 degrees RMS in theory, but then you don't need much of a temperature change to cause more even with temperature calibrated ones.

[branadic]

MEMS based inclinometers suffer from drift due to a mixture of different materials with different t.c. forming the sensor element. Even burn-in can only reduce initial drift, but not fully eliminate it. Thus, MEMS-based inclinometers with capacitive principle are limited by physical constrains. Discussions with companies like murata confirmed that. In most cases their noise figure is defined by FFT, since their main use is vibration detection.
The only way to come across with this limitation, which is not given by the silicon itself, is to use a different methode of measuring the displacment of the seismic mass instead of capacitive measurement. The capacitive measurement also adds an additional portion of displacement due to electrostatic forces. Thus, longterm stability of current MEMS inclinometers is not that great, limiting the use of them for arcsec accuracy and resolution. But I can tell you, there is work in progress to overcome this limitation.

Vial based or fluid based inclinometers are a very simple approach with very good longterm stability. In most cases the readout circuitry or the mounting of the sensing element inside a package are the limiting factors on longterm stability, not the sensing element itself. The fluid defines the dynamic of the sensor. With such elements resolution in the arcsec and sub-arcsec range can be achieved, while having very big measurement range of several angular degree.

-branadic-

[jmaja]

I have used this MEMS chip since 2007 and had a lot of discussion with chip designers way before the company was bought by Murata. They still have the factory in Finland and all the designers I have been contact with are Finnish.

All the data I have given is based on my own measurements. I haven't really measured very long time stability, but my sensors have been used years in applications were 0.01 degrees long term drift would be noticed and I haven't heard any problems with that. Humidity and temperature will have an effect on long term stability rate. I still haven't found one needing recalibration to achive the specs I gave earlier. What kind of stability has been achieved with vial reading discussed in this thread?

Unfortunately they are discontinuing the chip I'm using due to subcontractor ASIC process change. They are developing a new one with about the same specs. I'm just ordering the current one hopefully lasting long enough for the new chip to be available. The discontinued chip seems to be totally unique. I haven't seen any other with comparable specs.

[smithnerd]

Who was the UK supplier of vials?    I need at least four 2"/2 mm vials.   @ez24 gave me a couple (I broke one :-(

Level Developments:

https://www.leveldevelopments.com/products/vials/

For the UK manufactured vials.

The chap on ebay.co.uk trading as 'caterpiller_red' is (I believe) selling imported Indian manufactured vials. The photos show them marked 'PIE', which seems to stand for 'Paragon Instrumentation Engineers'.

[RoGeorge]

My question is, why do a bubble vial is even needed, in the first place?

A bubble vial makes sense for a passive device, read by reading the marks with naked eye, but when reading the tilt electronically, why?

A simple bowl (or some other type of container) with some electrodes around it will do it just fine, maybe even better then the vial, with it's hysteresis in bubble's move because of the not so well polished interior of the glass.  The fluid itself can be the ground electrode, and two piece of Al adhesive tape on top would make the two measuring capacitor electrodes.

 

Later edit:
To go a little further, maybe having some liquid is also not a good idea, because of its inertia.  But capacitors are the simplest possible component to manufacture in a DIY regime.  Literally everything is a capacitor.  Easy to make 2 capacitors and connect them to that measuring bridge.

Maybe a simple ball bearing, suspended with a metallic wire inside a jar will be enough as a tiltmeter or even as a seismometer.  The jar is fixed to the earth, the ballbearing ball's inertia will act as a reference.  Two Al foil stick on the jar will make the capacitors.  Eventually 4 foils to make X-Y axes measurements and find the direction were the shake/wave is coming from.  A 3rd sensor for the Z axis would be harder to imagine, but not impossible.  Maybe a ball bearing suspended with a metallic resort.  I bet the exact waveform in 3D can be deduced by knowing the elasticity constants of the spring and the weight of the inertial ball.

[jmaja]

A bubble vial makes sense for a passive device, read by reading the marks with naked eye, but when reading the tilt electronically, why?

I guess it is because it is possible to manufacture a bubble vial, which amplifies the extremely small movement of the vial into much bigger movement of the bubble and the fluid. 0.001 degrees is less than 2 um on a 10 cm vial length. How would you detect that small movements? Any bearing will cause problems. One problem of a vial is very limited scale. But it's quite a task to get a wide scale together with very high resolution.

[tomato]

I'd have to rate this as blatant advertising. It shows no knowledge of the thread.   The discussion in this thread is at least two orders of magnitude (100x!) more sensitive than your product.

Please behave better.  A tutorial series on applications for your product are, I'm confident, most welcome.  But, you are in nerdland.  You do not want to discover what nerds will do to sales puppies.  It's really quite appalling.

So, either step up to the plate and blow us away with what you know about measuring small angles, or please, go away quietly and don't do this again.

Or to put it another way:

If you can quote chapter and verse, I'll listen.  If not, I think you should move one.

Nobody in particular,
Reg

Maybe the moderator(s) will let you preview all posts in the future to ensure they are of sufficient interest to you and meet your standards.

[rhb]

The bubble vial has the virtue that high sensitivity is easily attained.  It's pretty hard to beat for simplicity and cost.

A pendulum and LVDT as used in a Talyvan is a lot more complex.

The NASA Tech Brief noted that the application did not require the sensitivity of a commercial inclinometer.  Given that the NASA design was reading 0.05 seconds and a change of vial would result in 0.01 arc seconds should give some idea of how sensitive the commercial tiltmeters are.

[IDEngineer]

A bubble vial makes sense for a passive device, read by reading the marks with naked eye, but when reading the tilt electronically, why?

I guess it is because it is possible to manufacture a bubble vial, which amplifies the extremely small movement of the vial into much bigger movement of the bubble and the fluid. 0.001 degrees is less than 2 um on a 10 cm vial length. How would you detect that small movements? Any bearing will cause problems. One problem of a vial is very limited scale. But it's quite a task to get a wide scale together with very high resolution.

And capacitance is straightforward with fluids. Water has an incredibly high dielectric constant, way more than traditional "high DE" materials like tantalum.  I haven't looked into the DE of oils such as are commonly used in vials like these but the point is that basing an inclinometer on the movement of a bubble in fluid has several advantages that are hard to replicate in other ways. It's hard to imagine a mechanism simpler than a bubble in a sealed glass vessel.

[rhb]

The "spirit" in a spirit vial is *not* alcohol.  It feels like acetone, but smells different. 

I'd like to know what it is. 

In the meantime I suggest that if you break one, or think you might have, you place it outdoors until the contents evaporate.

[branadic]

My question is, why do a bubble vial is even needed, in the first place?
A bubble vial makes sense for a passive device, read by reading the marks with naked eye, but when reading the tilt electronically, why?

I can tell that 1µm/m stability over 24h is achivable with a capacitive, temperature compensated readout approach of a spirit vial. It's even harder to separate stability of the sensor from movements of the ground, even with high resolution inclinometers such as Leica Nivel or Wyler Zerotronic as reference.

I'd like to know what it is.

Well, some companies claim it's alcohol or hydrocarbonate.

-branadic-

[ballen]

I found this a very interesting thread.  I own a two-head Talyvel 4 which I repaired and calibrated, and used to map my surface plate.  I am going to experiment with making my own capacitive-readout bubble level as described here and in the NASA Tech Brief document from 1981.

From what I could tell, three people in this thread (JBeale, branadic, and rhb) have built these devices.

branadic read his using an MPT PCap01 capacitance-to-digital chip, and also characterized and removed the effects of temperature dependence.

JBeale and rhb interfaced with the NASA "circulating diode bridge" circuit.  JBeale started with 1N34 diodes, then shifted to using On Semi MMBD352LT1G diode arrays, then also increased the clock frequency to 1.8 MHz.  Rhb used old-stock RCA CA3039 diodes and some other, and also rolled his own copper foil.  With some help from another member here, he also located the patents that explained the operation of the "circulating diode bridge", which is clever and not trivial.

For myself and future readers, since the last posts in this thread were more than a year ago, I was hoping that the three of you might provide some brief summaries of the "end state" of your experimentation, and also say something about "lessons learned".  If any others reading this have also build these devices, it would be nice to hear from you too!

Cheers,
Bruce

[branadic]

We have a patent pending, so I can't say anything about the sensor.
Using the above mentioned circuit also involves a lot of IP build up over several years that I can't share, as this IP is what we earn money with when developing capacitive sensors for our customers.

-branadic-

[ballen]

We have a patent pending, so I can't say anything about the sensor.

OK, I understand.  Anyway, your posts, pictures and plots above are interesting and helpful.

Hopefully JBeale and rhb can provide a follow-up to their previous posts.

[rhb]

Sadly, I broke the vial :-(  That revealed that the Chinese vial was not fully ground making continuing seem a bit pointless.   And I didn't feel like buying UK made vials.   It also prompted a lot of reflection on methods for doing this.  The context can completely change the best choice.  In JBeale's case, the Chinese vials would be OK.  For me, trying to measure the lake level from my house not so usable.

There are a lot of options for detecting very small angles from vertical.  The TalyVel uses a pendulum and LDVTs and is accurate to a second.

At this point I'd use a wire flexure  pendulum with 3 or 4 capacitor plates and stops to prevent a short or other damage.  One then has the option of a huge number of bridge designs. 

One attractive option is to use the capacitors to shift the frequency of xtals.  Mix them to a frequency an Arduino can sample and measure the frequencies to determine the angular deviation from vertical.

An aluminum cylinder filled with fine lead shot with an 0.010" hole in the center of rotation and an 0.010" guitar string for a flexure should be robust and sensitive.  Calibration by reversal.  If you're making a portable device use three small carbide balls for contact points.

Have Fun!
Reg

[branadic]

Wyler in Swizerland uses a pendulum to build a differential capacitor. It is quite a fancy micro mechanical system.
Check Compendium "The secrets of inclination measurement" for further information.

Safran Colibrys has some quite expensive MEMS sensors based on a servo controller to keep the seismic mass in the middle position.

We currently have a project to improve our pcb based, capacitive, fluidic inclination sensors. Within the project we are improving linearity. Hopefully in about 1.5 years there will be a paper with results.

-branadic-

[ballen]

Dear Reg,

Thank you for the helpful reply!

Sadly, I broke the vial :-(  That revealed that the Chinese vial was not fully ground making continuing seem a bit pointless.   And I didn't feel like buying UK made vials.

Did the vial break because you soldered to the copper contacts while they were in place on the vial?  The ones I have are borosilicate glass (branded PIE, purchased from the UK) which has a low coefficient of expansion, so I thought that soldering directly to them might be OK. But I have not done so yet.

There are a lot of options for detecting very small angles from vertical.  The TalyVel uses a pendulum and LDVTs and is accurate to a second.

I own a Talyvel 4, which uses the same mechanism as the earlier Talyvels. It has a resolution of 0.1 arcsec, and according to the specs is accurate to 0.2 arcsec plus 2% of the (absolute value of the) reading.  I have calibrated it and can confirm that (at least for angles larger than a few arcsec) it meets these specs.

An aluminum cylinder filled with fine lead shot with an 0.010" hole in the center of rotation and an 0.010" guitar string for a flexure should be robust and sensitive.

I'm not sure that guitar string would be the best choice, as it might have a tendency to stretch. The Talyvel uses beryllium copper (BeCu), also known as copper beryllium (CuBe), beryllium bronze or spring copper, for its support wires. The diameter is fine enough that I suspect it is under a substantial fraction of its yield tension.

I am going to start by reproducing the NASA circuit, and then perhaps experimenting with a capacitance to digital converter.

Cheers,
   Bruce

[Kleinstein]

Because of magnetic forces I would be a bit carefull with a LVDT. It may still show a stable reading, but may be off from the earths magnetic field, if one is not carefull with the shape  (a ball should be OK).

With the string it may be important to avoid anything magnetic, as from the shape the string would cause a magnetic force. So normal steel should be out and under high stress / after some cold working some stainless steel versions can show weak magnetism.

The shift in the resonance frequency of crystalls (especially low frequency tuninig forc ones)  can be uses to detect variations in gravity but it could be tricky with the direction. The difficulty is not in measuring the frequency, but in the mechanical mounting and getting a stable temperature.

[rhb]

I was smoothing out the Cu foil and hit the teat :-(  Had not got to the solder stage.  However, I've used the tape for Manhattan style prototyping on SS PCB and soldering is a non-issue.  When the *wire* is hot you're done.  Press the wire to the tip, place it in contact with the foil and solder.  The foil is such a small thermal mass and so much more conductive that the glass never sees the heat if done properly.  An excellent "belt and suspenders" dodge is to leave the foil sticking up where the connection is to be made, solder it and then stick it down.

Rather than speculate, I suggest looking up the  material mechanical properties and calculating the relevant stresses and strains of music wire and beryllium bronze wires.  As I understand, the TalyVel uses the suspension wires for signal as well.  That presents different requirements.

The dominant parameter of a flexure wire is the stiffness.  The stiffer the flexure wire, the longer it must be.  But it is the product of stiffness and length that matters.  Larger wire, longer pendulum.  Music wire is available as small as 0.003".  That's pretty fine.  Naturally, with a draw plate you can go finer.  The point of a wire flexure is that, properly designed, it matches the mathematics of a perfectly flexible string suspension.  Knife edges, etc not needed.

I got involved in this as far as I have for two reasons: I'm a professional geoscientist and I was given three 4"/div Chinese vials by a thread participant.  I was interested in observing the crustal deformation associated with the water level in the lake a mile from my house.  Spending $70+ for a pair of UK vials wasn't a high priority.  For workshop use I'm much more interested in a suspension based system. I'll build out an XY system using the Chinese vials just to see how they do.  But the crappy quality took the bloom off the project.

The reason that a turned solid aluminum cylinder with a small center hole (over a *short* distance, e.g. 0.1") is that made from round bar stock on a basic lathe it will be as true as you can afford.  And not hard to accomplish to 0.0001" with proper practice.

Have Fun!
Reg

[JBeale]

I have not really done much with the bubble-tiltmeter since my last post here. I discovered that the electronic readout has quite good resolution in theory, but the mechanical properties of my glass vial were less than ideal. The resolution of the overall system in my case was determined by some "stickyness" or hysteresis of the bubble, due to the roughness or texture on the inside of the glass vial.  Maybe this could be smoothed out by a fixed amount of vibration applied by a speaker or similar transducer, but that adds its own opportunity for uncontrolled offsets and I didn't investigate the possibility.

Separately I also tried measuring the left-right timing of a driven pendulum with optical center detector; timing with Raspberry Pi Pico at 200 MHz. Very high-resolution in theory (output being a frequency ratio avoids some analog pitfalls). It worked but was still somewhat noisy, perhaps due to the optical sensor, and background horizontal earth motion (microseism).

Separately I also tried DC analog optical position sensing of a suspended pendulum using a slit and a split-field photodiode. That works but the optical beam has some scattering by fine particles so it is also a dust meter.

I would imagine the pendulum with capacitive sensing would be very sensitive, although I have not tried it.

As far as I know, all tiltmeters are also horizontal seismometers, so you have to do a lot of low-pass filtering if the system is not already well damped, like a nearly-horizontal fluid bubble probably is.

-john

[ballen]

Hi Reg,

Thanks for your detailed reply.  There are certainly many possibilities, both for sensing and for readout.

Regarded a suspended aluminium cylinder, one possible complication concerns the mechanical "stops" which prevent the cylinder from hitting the capacitive sensing pads or frame.  If these are "hard" then each impact will deform both the cylinder and the pads, changing their shape and hence their capacitance and the balance.  If they are "soft" then they will also be deformed by impact.  This won't change the shape/balance of the pendulum but by changing the distance to the pendulum, it will also influence its capacitance.  The fluid bubble sidesteps some of these issues, although as the discussion here has shown, it is not free of hysteresis.

A second issue is the orientation.  I am not sure if it is correct to assume that a single wire will return the cylinder to the same orientation, particularly if that wire is very thin.   (Even if the cylinder is round to 2.5 microns, that may not be good enough, as we are sensing significantly smaller motions.)  The time constant for this might also be very large.  And if you suspend the cylinder from multiple wires, that creates other potential issues.

For the record, I wanted to clarify one point:

As I understand, the TalyVel uses the suspension wires for signal as well.

This is incorrect.  The Talyvel armature is suspended from five BeCu wires.  One is a single strand, and the other four form two "V"s.  All five of those (conductive) wires are attached to the internal metal frame of the device at the top end, and to the metal armature at the bottom end. While it is not their primary purpose, they do provide a conductive path between the moving armature and the fixed frame.

The sensing is done via two coils, located on either side of the suspended armature.  The inductance of these coils is a function of their separation from the (steel) armature.  The wires from each of these coils are entirely independent from the suspension wires, and are connected to the readout via a shielded set of twisted pairs.  The two coils are part of a classic bridge circuit, driven by a 3kHz oscillator.

Cheers,
  Bruce

PS: for the NASA/spirit vial design I would be grateful for guidance about the size/shape of the two top electrodes.  For my PIE vial, the total internal length is about 95mm (3 3/4") and the bubble (when centered) has a length of about 32mm (1 1/8").  The NASA document suggests a rectangular electrode with a length comparable to the bubble, but if I do that, then the capacitance difference will not be a monotonic function of the displacement angle from horizontal.  This is because, as the displacement increases from zero, the bubble first moves completely under one electrode (decreasing its capacitance) but then exits the electrode at the far end, increasing its capacitance again.  If I make the electrode the same length as the bubble, then the capacitance will be a (roughly) linear function only for a total displacement of +- 16mm from the central position, corresponding to (nominally) +- 24 arcsec.

PPS: a second technical question.  Are the CA3039 "diodes" just Si NPN transistors with the base and collector tied together?  If so, can I just use (say) a transistor array such as CA3127 for this?  If I understand correctly, the matching is desirable so that the offset voltages and junction capacitances are the same (as a function of temperature & forward current, and temperature & reverse voltage respectively).  Is that correct?

[Kleinstein]

The wire will allso restrore the rotation. If needed one can measure the rotation (e.g. from thermal effecs) too. There is a torsional more of oscillation and the frequency can be relatively low, but for a small system (e.g. < 2 cm diameter) this would not be so extraordinary slow (e.g. < 1 Hz). The air would provide some and possibly enough damping of the oscillations. If really needed magnetic damping (with a permanent magnets on the outside, providing eddy current damping) could be added.

The mechanical stop needs to be taken into account, but I don't think it would be such a bad point. It dones not have to be the same surface as used for sensing. The sensing distance would need to be larger than the mechanical stops, which limits the resolution.  Still the resolution of capacitive readout is usually quite good, well OK to see the seismic motion.

One will need some filtering. Today this would be mainly in the digital domain. It would be not just a simple low pass, but also additional suppression of the main resonant modes of the system. I would even go beyound just a classic filter and consider fitting the resonant mode(s) to also take into account a decaying oscillation if needed.

[babysitter]

@branadic: Located at Stuttgart, not too far from artificial (Straßbourg) and natural (Schwäbische Alb) serial offenders regarding seismic activity, I wonder why the inclination measurement experts didn't recommend against getting sensitive measurement devices.  ;)You might want to set up one or three geophones (HX711 ADCs do well enough) to evaluate shaking floor.

BR
Hendrik