Precision Electronic Levels - The Germans Arrive

As mentioned before, I was able to read the German-language book on electronic levels well enough to tell that the book wasn't that helpful, and that the authors did not understand how these levels worked.

However, the book did have two or three references, also in German. So I googled them down, which led me to an article by a German engineer (now Diplom Ing) who had in effect combined the Taylvel and Wyler approaches, yielding a far simpler unit.

The relevant article is "Hochpräzise Neigungsmessung mit dem elektronischen Einachspendelsystem HRTM" by Timo Kahlmann (ETH Zürich), Christian Hirt (Universität Hannover), and Hilmar Ingensand (ETH Zürich), Ingenieurvermessung 2004, 14th International Conference on Engineering Surveying Zürich, 15. ­ 19. März 2004.

Even if you don't read any German, the figures and photos tell the story.

Some details:

The pendulum hangs on two 50-micron thick (0.002") by 3mm (0.118") wide bits of metal foil. The figure implies that the foil is made of spring steel (Federstahl) leafs (Blattfedern) but the text says that they are made of beryllium-copper foil. Brass shim stock or stainless steel foil should work, or some feeler gage stock.

Pendulum swing is limited by two nylon screws (Arrierschraube, Kunststoff).

While a Talyvel-like pendulum is used, the motion of the pendulum (Pendel) is instead sensed by means of a 3-plate (differential) capacitor, where the two outer capacitor plates (Kondensatorplatten) are fixed, and the pendulum mass is the moving third plate. The fixed plates appear to be made of brass, while the pendulum is aluminum.

Joe Gwinn

Reply to
Joseph Gwinn
Loading thread data ...

Joe:

Damping of the pendulum is accomplished by an eddy current brake. A powerful permanent magnet is placed below the pendulum in the base.

According to the article this reduces the damping time from about 20 seconds to 2 seconds.

Interesting concept, and buildable by any competent toolmaker.

If one of the electron-savvy members here is willing to prepare a circuit diagram for the electronics, this could be a very worthwhile project for some of us. The read-out could be arranged such that a commercial digital voltmeter could be utilized.

Thanks for pointing out this article, and the previous discussions.

Wolfgang

Reply to
wfhabicher

Yes. I didn't have the energy to translate the entire article, given my schoolboy German.

Yes. I think that Kahlmann and colleagues were worrying about the same things as I had been. But they started in 2000, and it was more than a home project for them.

I will publish the circuit when I learn it or develop it. It will be a traditional transformer capacitance bridge with synchronous detector, which is what Kahlmann probably uses as well.

See "Three-Plate Micrometer" (section 5.1.2) in "Capacitive Sensors" by Larry K. Baxter for the general theory. The usual implementation of the two voltage sources is a centertapped transformer. I built one of these in 1975 by winding three strands of #30 wire to fill the bobbin of a

10mm by 18mm ferrite cup core. One winding became the primary, the other two were connected together to make a center-tapped secondary. The transformer primary was driven with 20 volts at 500 KHz.

Welcome. I don't know why this so drew my attention.

Joe Gwinn

Reply to
Joseph Gwinn

That was a very interesting try to read.

I was looking at the drawing and noticed that magnet. A few winters ago one of the regulars mentioned how neat it is to drop a neodymium magnet in a length of copper pipe and watch it slither rather than drop through.

My school boy German was one year when I wasn't paying attention during the first grade. Dad was stationed in Germany at the time. I should have paid attention.

If you hit a section that you really want to understand (need translation), I suspect an email to Nick Mueller who posts on u.r.m.e. would get a reply.

If I'm understanding this correctly, the distance between the plates has a linear effect on capacitance so if the pendulum leans to one side, using what is essentially two capacitors in series will cancel since one side will decrease, the other side increase.

You are doing an excellent job of researching this. Your threads on this has been the high point of R.C.M. for the last week or so.

Thanks,

Wes

Reply to
Wes

Eddy-current dashpots work very well, and no messy oil.

Taylvel uses oil, originally in a little dashpot, lated as drops of oil between limiter screws and the pendulum.

Wyler uses either air going through small holes in the diaphragm, or active electronic feedback.

I had a few years in High School and then a year in College. And five years later spent a year in Sweden, 30 years ago. So I can get the gist from reading.

That's a thought. What is u.r.m.e. ?

I am borrowing a copy of Kahlmann's thesis. Probably 150 pages of technical German. We shall see.

It's arranged so that for small displacements, the change in capacitance is the same, except that one increases and the other decreases. With a suitable circuit, one does get linear response.

When I was in Sweden, one thing I did was to design a 3-plate capacitance sensor for tracking of the movements of small animals in pharmacological research. I found some of my notes from then, dated

1975, and later notes from 1977 (when I built a sample of such a sensor).

Thanks. There has to be a way to find some left-right politics in this. But I guess that the idea of building such a level has tickled more than just my fancy.

Joe Gwinn

Reply to
Joseph Gwinn

Solid state accelerometers are smaller. lighter, cheaper, simpler for that sort of application.

For example:-

Analog Devices ADXL322JCP

I've got some similar ones waiting for a two plane balancer project.

Regards Mark Rand RTFM

Reply to
Mark Rand

I need to ask if they've thrown out all the old Rosemount differential pressure transducers from work. They used a differential capacitor sensing head. Ok in its own right, but useless for our purposes due to sensitivity to static pressure (The silicone oil dielectric used was compressible at a couple of thousand psi and changed the transducer calibration).

Mark Rand RTFM

Reply to
Mark Rand

[snip]

There is no reason that this cannot be done, and there are variants of the 3-plate differential capacitor bridge that allow the moving plate to be grounded, so the moving plate could be the rotating shaft (if round enough) or the housing.

One also needs a way to sense rotor angular position, so one can tell where the heavy spots are. A simple display would use the rotor angle signal to trigger an oscilloscope, which scope would display the housing motion signal. This yields a stable display of displacement versus angle, regardless of rotation rate (although the horizontal scale will change with speed).

How big are these rotors, what are they made of, and how fast do they spin?

If there are patent numbers on the commercial sensor units, it would be worthwhile to read those patents, if only for the litany of practical effects to be considered. And to educate me.

There must be a way ... which way did you say those rotors turn?

Joe Gwinn

Reply to
Joseph Gwinn

Widdershins

Reply to
Jim Wilkins

ADI makes very nice MEMS accelerometers, but for turbine rotor balancing, I would ensure that the bandwidth is sufficient, as turbines spin pretty fast, especially small turbines.

I have to mention that MEMS accelerometers use differential capacitor bridges to sense the motion of the silicon proof mass. Silicon is a very good mechanical material, almost as good as quartz. The capacitances are in the femtofarads, which makes for a noisy output signal if wide bandwidth is needed.

So, there is a tradeoff study to be made.

The usual bolt-on accelerometer used on big machines is a proof mass glued to a piezoelectric crystal, often quartz, with bandwidth in the tens of kilohertz. Sensitivity is often low. The commercial sensors are expensive, and home manufacture is difficult. However, I bet they turn up on evilbay.

What are you balancing?

Joe Gwinn

Reply to
Joseph Gwinn

Joe et all,

There was a discussion on this topic a while ago here where I promised to post some pics of the wee beastie....

Basically it is a functioning 1/16" scale turbo generator for live steam locomotives, with the prototype for my own. The alternator rotor is 3/8" dia. x 3/8" long, with permanent magnet. The turbine rotor is 5/8" diameter x 3/16" wide with buckets around the perimeter as per Stumpf design. Turbine is overhung design with the ball bearings on each side of the generator rotor. Rotational speed is about 56,000 rpm loaded. We've run it on steam and it easily lights up 4 flashlight bulbs at 80 psig steam pressure. It'll also burn them all out if the pressure is increased due to malfunctioning steam pressure regulator but this problem has been resolved:-))

I've searched the 'net and there is much information on dynamic balancing; but because of the small size most of the methodology is difficult to implement. I did jury-rig a crude machine to measure frequency, amplitude, and rpm. I used an inductive pick-up (old magnetic headphone) with the steel diaphragm glued to the vibrating part. This produced a very good sine wave output signal, displayed on an oscilloscope, but...

My son rigged up a LED timing light that would illuminate an ink mark on the rotor, but...

What we observed was that the timing mark would rotate from its original position. In fact I was able to move this timing mark over

360 degrees by increasing the turbine speed from about 5,000 rpm (as low as it would run) to over 60,000 rpm where the old bearings would howl (I have new ones with high temp light grease, not yet installed).

This moving of the timing mark threw me; I suspect I know why it moves, critical speed and shaft deflection spring to mind, as do mass, velocity, and acceleration, but my education is almost 35 years old when I learned about this. Professionally I deal with statics and make sure things don't fall on people or break unexpectedly:-))

Since the magnetic pick-up is really a velocity transducer, and the velocity of the vibrating mass lags the rotor position with the amount of lag probably a function of rotational velocity, I thought that a position transducer would make the problem more tractable by removing some variables from the system.

Hence my interest and queries on capacitive transducers plus circuitry.

And thanks for your efforts on this.

Wolfgang

Reply to
wfhabicher

I was thinking only 40,000 RPM, like the rotors in a jet engine. But

56,000 RPM is 933 RPS, call it 1,000 RPS. To see the details within a revolution, you need minimum five and better ten harmonics, which means that one must have bandwidth to 9,333 Hz, call it 10 KHz. And more is better.

Visible LEDs are quite slow, and the delay to max light could cause the rotation. As one spins faster, a constant delay in time becomes an increasing delay in angle terms. LEDs are also not that bright, forcing use of long pulses to get enough light to be visible, which causes blurring. At high speeds, this blurring will become severe.

The standard way to address this is to use a very short pulse xenon flash. This is very easy if powered from the AC line. The key trick is to use an optocoupler to deliver the Fire! signal to the flash. Pulse transformers won't work, for an odd reason. They fire the flash well enough, but too much flash energy comes backwards through the pulse transformer and discombobulates the trigger logic. Been there done that. Ended up ripping the pulse transformer circuit out and replacing it with an optocoupler circuit.

The pulse must be very short, with very low jitter. At 1,000 RPS, make the rotor appear to stand still, with resolution of 1/1000 of a rotation, requires light pulses about 1 microsecond wide, with jitter no larger.

There are flashtubes intended for this, but one can short-pulse almost any tube, probably well enough.

The only way this can work is a one-tooth gear and a magnetic pickup. Drill a hole under the one tooth to maintain dynamic balance.

What I use is an HP/Agilent (now Avago) optical encoder. Look for motion control encoders at . But these may be too big. For such a small rotor, some other approach may be needed. The signal from the generator might be used.

Well, the alternative is to integrate the velocity signal to get a position signal. This can be accomplished with an opamp with a capacitor (and parallel resistor) in the feedback path.

It's an interesting albeit tiny project, but with high power-to-weight ratio.

Joe Gwinn

Reply to
Joseph Gwinn

Damn leftists! Hmm. Damn rightists! Depends on which end you view.

Joe Gwinn

Reply to
Joseph Gwinn

What about using a record player( remember them) cartridge as a displacement sensor? A stereo one could supply a signal from two dimmensions.

Reply to
Grumpy

An explanation of how these things work can be seen at

formatting link
This cite says there is a linear region, which I think means linear enough for many purposes. The transfer function is of the form f(x) = k*x/(a^2 -x^2) which obviously is not linear anywhere but may be close enough to linear in regions where a >> x.

Indeed! I would not have thought it possible to sense microinch displacements this way but Jones (1973) demonstrated detectable displacement of 10^-10 mm.

Reply to
Don Foreman

Not Joe, but not bashful either...

Something to think about is that sensing of imbalance may be easier to do by strongly constraining motion and sensing vibratory force rather than velocity or accel. This greatly reduces phase shift between imbalance vector and observed response, which can be very significant and somewhat chaotic at speeds near a mechanical resonance.

Strain gages are quite inexpensive. Another possibility might be to make piezo sensors using the piezo elements in cigarette lighters or those long-neck grill and fireplace lighters. I would think that a couple of kilohertz bandwidth would be easy enough to get from those using a transimpedance amplifier made with a jellybean opamp. That's

120K RPM.
Reply to
Don Foreman

uk.rec.models.engineering

Wes

Reply to
Wes

If you can characterise the pickup, circuit and LED delay, perhaps with a temporary contact sensor on a second scope channel, you can make a calibration chart to compensate for it.

Reply to
Jim Wilkins

Of course. RCM with an accent. Thanks,

Joe Gwinn

Reply to
Joseph Gwinn

Toolpost grinder...

Mark Rand RTFM

Reply to
Mark Rand

PolyTech Forum website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.