Building differential electronic levels?

[Pete F]

Mount a small vibrator on the case to do the automatic thumping for you.

Well may you laugh, but that's what is fitted in analogue aircraft instruments, exactly for that purpose. I kid you not.

13.5 V is near enough to 13.8 V for warfies' work! 13.8V power supplies are a dime a dozen.

 

[craigS]

Nate,
Great post -- always like to know how fundamental things are measured. I can't quite make out in the photos what f
Federal used to hang the pendulum.
Is it some sort of Mylar film tape like thin recording tape or is it metal. Someone mentioned it was hung by 3 bands? That would make sense since it would make sure the pendulum always had constant force on the bands no matter any thermal twisting etc.

I noticed some surface bubbles on the oil, I think they could cause repeatability problem from variations in surface tension between the support bands.

Please describe the pendulum hanging system, width and thickness etc.

I saw a gravity experiment once that demonstrated the force between two masses one of the masses was hung from about 10' with audio recording tape. The mass was two steel balls with a shaft between hung in the center by the tape. So any unbalanced force pulling on one ball would rotate the assembly and twist the tape. A mirror was glued to the tape just above the balls and alight beam used to detect any rotation of the mass. The entire thing was in a draft proof housing. The second mass was a large box of sand on a dolly. As you moved the box close to one of the balls the light beam would move as the mass in the box rotated. I must have been 10 at the time but it impressed me. The resisting force needed to twist 10' of mag tape must have been so very very small.

Cool stuff

Craig

 

[NateA2]

It's been a while.

I have been working a bit more on my testing of LVDT as a sensor for an electronic level.

I have found (trial and error) that a LVDT that I have salvaged from a federal unit, can be interfaced into a gauge amplifier.

The LVDT works with the setup I have made for testing. With ~ .030" of lateral movement (.015" either way from center) I can get the Amp to register a total of three hash marks when set to the least resolution setting. Oddly, the higher res settings give me less movement of the needle. Regardless, the sensor works with the amp, and now its time to mount the sensor in a level body of some type for testing.

 

[NateA2]

well, playing with the connections a bit, I am able to get a much more drastic spread on the amplifier dial. This is great, but I need to figure out what the gage amplifiers outputs and inputs really are.

Anyone good at reading schematics? if so, could you look at the last page in this PDF, and let me know what the input/output connections are for this amp? I have put my meter to them, but have not been able to make heads or tails of the readings.

Since LVDT's require AC voltage to the center coil, I believe that the federal amp must be outputting a very low voltage AC. Sadly my meter cannot read this. 2 of the connections on the amp must be return from the other two coils, and one ground coming out of the amp. Any thoughts?

 

[toolnuts]

How the LVDT works

well, playing with the connections a bit, I am able to get a much more drastic spread on the amplifier dial. This is great, but I need to figure out what the gage amplifiers outputs and inputs really are.

Anyone good at reading schematics? if so, could you look at the last page in this PDF, and let me know what the input/output connections are for this amp? I have put my meter to them, but have not been able to make heads or tails of the readings.

Since LVDT's require AC voltage to the center coil, I believe that the federal amp must be outputting a very low voltage AC. Sadly my meter cannot read this. 2 of the connections on the amp must be return from the other two coils, and one ground coming out of the amp. Any thoughts?

An LVDT is really just 2 transformers hooked up to oppose each other. The center transformer coil is used to excite the
other two coils. The input is AC, and like any transformer, the output is AC. When the slug, the moving iron part, is
centered between the two output coils, each coil receives the same amount of excitation and they cancel each other
out. When the the slug has been move to a position away from center, one coil effectively gains some windings and the
other coil effectively looses windings, thus the coil with more windings has more output than the other and the output
is the difference between the two coils.

A transformer works by magnetic induction. If there is no iron, there is no magnetic induction. If you have a coil with
ten windings, but there is iron only on/around five of those windings, then effectively five of those winding do not get
get any magnetic induction. The amount of signal you get is proportional to the winding ratio. If a transformer has two
windings in the primary coil, and only one winding in the secondary coil, you get a step-down transformation IE:
10VAC on the input gives you 5VAC on the output.

So lets say you have an LVDT with 100 windings on the primary/input coil and 100 windings on each of the two
measurement/secondary coils - now you put a 10VAC signal into the primary coil, and move the slug to center
position between the two secondary coils. Each of the two coils have half of their coils windings able to support magnetic
induction because the iron slug only covers half of each coil (50 windings). So, each coil puts out 5VAC, and they cancel each other out. Now move the iron slug to the far end of it's travel, and now one coil has no magnetic coupling (no iron), and the other has all 100 windings magnetically coupled and so puts out 10VAC. If you move the slug to the opposite end the other
coil will now put out the same 10VAC but with opposite polarity.

A long explanation, but you should get how it works, and what the signals should be. To know exactly you
will need the windings information from the manufacturer.

Good luck - Paul

 

[Steelmind]

Talyvel 1 and 2 manual

Dropbox - 404

Hello; it is 5 years now, but I take my chances: I just bought an used Talyvel off Ebay and couldn't find the manual for it. Also, for some reasons, I couldn't access the link you posted here . Do you happen to have a copy of it in your computer ? Or maybe you uploaded it somewhere else too, where I could access it. Thank you very much for your time. Just in case, my email is : [email protected]

 

[RC99]

Here is an updated link. Dropbox - Talyvel 1& 2 Handbook 212-7.pdf - Simplify your life

 

[Jeanlucdemarc]

Bateri from TALYVEL 1

I got that device but 1 battery was inside skb830 6.75V.. Anyway as you know 2 should be inside which comes to 13.5V.
You can use 9 in series AA batteries (U=1.5V x 9=13.5V)

 

[ballen]

I have a Talyvel 4, which uses a 2x 9v NiCd battery internally. When they fail I plan to replace them with modern Li-ion batteries.

In the same vein, if you connect 4 standard 18650 cells to a 4-cell charge controller/balancer board it will put out 14.8V. You can drop that to 13.5 with two silicon diodes in series. It's an inexpensive rechargeable solution that was not available 50 years ago when your Talyvel was designed and built.

 

[Elitebook]

I very much liked the idea of a DIY differential leveling system.
Small core-less LVDT's are readily available together with cheap ADC and micro controllers for signal conditioning with infinite resolution (in theory anyway). Non-linearity is also possible to correct.

The level can be calibrated by placing it on a straightedge resting on slip gauges.
0,004848mm/1m of height difference equals to 1,0 arcsec.
Using 1031mm between blocks a 0,005mm height difference is needed to achieve the same angle.

Is it as simple as fixing the LVDT sensor to a central post and let the core move with the pendulum as proposed on the sketch?
Did a few FEA iterations and I am now displacing 0,800-1,0μm/arcsec using 0,05x5,0mm steel ligaments.


What it typical thickness of the ligaments holding the pendulum?
What kind of displacement seems to be "normal" for such equipment?
It seems like it's a critical spot when transitioning between directions - does the commercial levels do any magic here or is it just a theoretical problem?

 

[ballen]

You can find the Talyvel patents online. They make interesting reading.

The support is done with three fine copper-berillium wires. The design is clever. One of the support wires forms a "V" at the front of the pendulum, with the top ends of the V on either side and the bottom of the v in the middle of the moving carriage. A second identical support wire is at the back of the pendulum. Together these control four of the six degrees of freedom. The third wire is located at the side, and prevents rotation about the axis which passes through"tips" of the front and back Vs. The clever aspect of the design is that it controls but does not overconstrain the motion, and has zero backlash or hysteresis. When I did some adjustments to my Talyvel heads, I was terrified of breaking these wires. Later I had a phone conversation with the people at the UK home of Taylor Hobson, who do such work every day. The told me that they have seen broken suspension wires only a few times in the past decades. Apparently the designer, Richard Reason, would routinely test the read heads by knocking them from his workbench onto the floor, and wasn't satisfied with the design until they passed that test every time. You can read about him here:
https://homepages.abdn.ac.uk/npmuseum/article/Taly2010.pdf

 

[Elitebook]

You can find the Talyvel patents online. They make interesting reading.

The support is done with three fine copper-berillium wires. The design is clever. One of the support wires forms a "V" at the front of the pendulum, with the top ends of the V on either side and the bottom of the v in the middle of the moving carriage. A second identical support wire is at the back of the pendulum. Together these control four of the six degrees of freedom. The third wire is located at the side, and prevents rotation about the axis which passes through"tips" of the front and back Vs. The clever aspect of the design is that it controls but does not overconstrain the motion, and has zero backlash or hysteresis. When I did some adjustments to my Talyvel heads, I was terrified of breaking these wires. Later I had a phone conversation with the people at the UK home of Taylor Hobson, who do such work every day. The told me that they have seen broken suspension wires only a few times in the past decades. Apparently the designer, Richard Reason, would routinely test the read heads by knocking them from his workbench onto the floor, and wasn't satisfied with the design until they passed that test every time. You can read about him here:
https://homepages.abdn.ac.uk/npmuseum/article/Taly2010.pdf

Interesting.

I did not find the patent however I found this link with Talyvel schematic diagram on page 19.
https://royalsocietypublishing.org/doi/pdf/10.1098/rsbm.1990.0039

You are describing three wires in total compared to the PDF which is describing five wires. I guess four of the five wires are the "V" in each end of the carriage?
Is it only one wire on one of the sides of the carriage?
How does it in that case control rotation in both directions?

I seem to to failing totally in understanding the concept here

If you have pictures to share from your Talyvel interventions it would be much helpful

 

[DeadMahoDude]

If anyone is curious about the schematic, I was bored lately and traced out the schematic and figured out a little on how the electronics/LVDT inside those Talyvels are working:

Schematic with explanation over here:
Taylor Hobson Talyvel mit Schaltplan&Erklarung - Zerspanungsbude

I'm playing a little with the thought about building a big one, like 300mm by 450 high, with about 300mm pendulum length, would be probably 5-10x more accurate, because of 10x increase in pendulum length.

Electronics shouldn't be an issue, could even be improved with new low-noise/offset opamps, basic principle of mechanics and LVDT with electronics in these Talyvel levels is great though, would be hard to improve...

 

[ballen]

If anyone is curious about the schematic, I was bored lately and traced out the schematic and figured out a little on how the electronics/LVDT inside those Talyvels are working:

Schematic with explanation over here:
Taylor Hobson Talyvel mit Schaltplan&Erklarung - Zerspanungsbude

That's fantastic! I have tried for ages to get a schematic for these Talyvel 1 units.

Well done!

Some more Talyvel information can be found in this thread and in this thread and in this thread.

Cheers,
Bruce

 

[fobyellow]

you dont have to make it that big. i already build one, pendulum length 100mm, total height about 150mm, it could easyly resolve 0.001um/m, if average 15 result per second, it could have a resolution of 0.0001um/m, but i'm still working on the repeativity

 

[fobyellow]

I hope to take the level unit apart again now that I have made it functional. It does not repeat as well as I wish it would at the .5AS resolution.

My suspicion is that the 3 brass shims, suspending the pendulum are not positionally correct. I suspect they need to be set very carefully to have the pendulm not be influenced by them.

Additionally, I don't understand why they used 3 instead of 4. was a triangle layout more stable rotationally than 4? I also want to mic them and get the OA measurements.

The transformer (LVDT) looks to be very similar to whats inside federal electronic indicators. Federal made LVDT's like these, they are pictured in the scan of the manual. One is also for sale on ebay at this time. Federal Electronic Gage Head Model Ehe 1049 Used | eBay

the LVDT is sealed with clear epoxy.

Nothing complex in these units folks, and I dont see any reason to (at this time) explore all sorts of other sensors. The LVDT's clearly work, are available and their interface with amplifiers is understood.

I wonder why the pendulums length was not made greater? it could give greater resoloution to the degree that .1AS could be repeatably obtained without too much trouble.

For instance: the 20AS resoloution repeats spectacularly. 1AS is significantly worse. .5AS is not good at all. Tapping the unit seems to bring it into compliance (most of the time)

If I built one that had a longer length of the pendulum suspension bands, I should be able to read .5AS more reliably. I believe...

i'm working on the repeativity now，i followed federal's idea, change the long reed spring to 1mm short one, the rest length of the pendulum using 6061-t6 , i use silicon oil as dumping fluid, now i have a single direction repeativity of 5um/m, while dual direction repeativity of 100um/m worst case, have you solved your problem?

 

[fobyellow]

DIY electronic level

I followed federal's design(seems more tough than talyvel to me), using silicon oil as damping fluid. i change the long reed spring to as short as 1mm(the fixing aera not counted), the rest length of the pendulum is made by AL-6061-T6 or C5191, the total length of the pendulum is around 100mm, i can get better than 0.2um while static reading,the following chart is half a minute data logging(the sampling rate is 16sps, 16 data window averaging, final data display-rate is 1 sps)
View attachment 347377
the long term data logging is mystry to me, sometimes i can get better than 3um/m a day. some times worse than 30um/m a day.

the following pic are my work, hoping to get advices
the first time i designed 2 sensor, thats differential set, it could lower the temp-coeff therotically, but when i finished it , i found 2 LVDT coils interfere eachother. so i disconnect 1.
View attachment 347378
i think its because the distance between 2 coils. i tried to put iron in the middle to do shielding, no effect
View attachment 347379

View attachment 347380
the earlier long reed spring

i cut the long reed spring to short pieces

the test platform is a 1200*1000mm granite surface plate, i build a simple small-angle-generator using s micrometer and a 1000mm long guided-rail

the following chart is a 2.5hrs logging 1 data per second non averaging,showing about 0.3um/m drift


everything seems perfet till now. BUT!!!
the repeativity is very very bad, if i move the level single direction, i get about 5-10um/m repeat, if i move the level back and forward, the result could be 100um/m in worst case. and everytime i move the level, there will be a 10minutes level slow drift period

 

[fobyellow]

I hope to take the level unit apart again now that I have made it functional. It does not repeat as well as I wish it would at the .5AS resolution.

My suspicion is that the 3 brass shims, suspending the pendulum are not positionally correct. I suspect they need to be set very carefully to have the pendulm not be influenced by them.

Additionally, I don't understand why they used 3 instead of 4. was a triangle layout more stable rotationally than 4? I also want to mic them and get the OA measurements.

The transformer (LVDT) looks to be very similar to whats inside federal electronic indicators. Federal made LVDT's like these, they are pictured in the scan of the manual. One is also for sale on ebay at this time. Federal Electronic Gage Head Model Ehe 1049 Used | eBay

the LVDT is sealed with clear epoxy.

Nothing complex in these units folks, and I dont see any reason to (at this time) explore all sorts of other sensors. The LVDT's clearly work, are available and their interface with amplifiers is understood.

I wonder why the pendulums length was not made greater? it could give greater resoloution to the degree that .1AS could be repeatably obtained without too much trouble.

For instance: the 20AS resoloution repeats spectacularly. 1AS is significantly worse. .5AS is not good at all. Tapping the unit seems to bring it into compliance (most of the time)

If I built one that had a longer length of the pendulum suspension bands, I should be able to read .5AS more reliably. I believe...

have you solved your problem of repeat?

 

[fobyellow]

digital meter

your meter is too old, due to hystersis of the coil spring, the analog meter usually dont have very high resolution. and also they are limited in dynamic range. i suggest you connect a 5-digit voltage meter(which now have a reasonable price) to the analogue output of the meter.

i'm using the solartron orbit system, it could almost resolve 18-bit signal, 16bit noise free, and it has a powerful logging software for free!!

 

[fobyellow]

I hope to take the level unit apart again now that I have made it functional. It does not repeat as well as I wish it would at the .5AS resolution.

My suspicion is that the 3 brass shims, suspending the pendulum are not positionally correct. I suspect they need to be set very carefully to have the pendulm not be influenced by them.

Additionally, I don't understand why they used 3 instead of 4. was a triangle layout more stable rotationally than 4? I also want to mic them and get the OA measurements.

The transformer (LVDT) looks to be very similar to whats inside federal electronic indicators. Federal made LVDT's like these, they are pictured in the scan of the manual. One is also for sale on ebay at this time. Federal Electronic Gage Head Model Ehe 1049 Used | eBay

the LVDT is sealed with clear epoxy.

Nothing complex in these units folks, and I dont see any reason to (at this time) explore all sorts of other sensors. The LVDT's clearly work, are available and their interface with amplifiers is understood.

I wonder why the pendulums length was not made greater? it could give greater resoloution to the degree that .1AS could be repeatably obtained without too much trouble.

For instance: the 20AS resoloution repeats spectacularly. 1AS is significantly worse. .5AS is not good at all. Tapping the unit seems to bring it into compliance (most of the time)

If I built one that had a longer length of the pendulum suspension bands, I should be able to read .5AS more reliably. I believe...

have you solved your problem? this repeat problem has troubled me for months, i tried different way to adjust the suspending point, but have no result. i think thats the main difference between Talyvel and Federal. the talyvel use very tiny wires to suspend the pendulum, even if it deforms, it could only produce very tiny force compare to the weight of the pendulum. the federal design use wide and thick reed spring, if it deforms at the top of the pendulum, it could result a considerable force, and also the piece of spring could form a curved surface(microscope level, small one),like a dome-switch, so i'm preparing to do a heat-treatment after the assemley to reduce the stress