Building differential electronic levels?

[fobyellow]

i think the reason that they all make short pendulum is the natural frequency, the total "noise" of the electronic level consist up of 2 parts, one is the LVDT readnoise, the other one is forced-oscillation.

the lvdt part could do better than 20bits resolution nowadays, the circuit noise could almost be ignored, the remaine noise is DC+AC magnite noise from the enviroment, the eddy current noise of the LVDT core, they are hard to remove.

the forced-oscillation is because of the vibration of earth or building, it mainly exist in 0.15-15Hz in frequency domain, so you have to design the natural frequency of the pendulum as far away as possible from 0.15-15Hz. the seismic observation equipment always set the natural frequency to more than 100Hz, and they give a definite transfer function of the device, you can fix the errors, and get the "true ground".

the LVDT used in federal-mahr and taylor-hobson are aera-vary type, the advantage is its linearity while the disadvantage is low sensitivity, the air-gap-vary type sensor is more sensitive, but non-liner, and its rarely used.

somehow the non-linear capacitance sensor is widely used, the Wyler-AG use a 20mm diameter archimedes-spring as the pendulum, the natural frequency could be more than hundreds herz, but i think its very difficult to DIY

 

[ballen]

This repeat problem has troubled me for months, i tried different way to adjust the suspending point, but have no result.

I've been following your posts with interest. Did I understand correctly that the main issue is hysteresis in the suspension spring? This means that the zero point (low energy state) of the spring shifts after it has been moved or flexed. I'm not sure that heat treating to remove stress will help this. As you point out, the thin wires used in the Talyvel are less likely to have this problem than the much thicker/wider strips that you are using.

 

[Mark Rand]

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.

To quote my instrumentation lecturer when I tried to use that reasoning at university. "You can't repeal Hooke's Law, Boy!.

Bearing stiction is a different thing though...

 

[fobyellow]

I've been following your posts with interest. Did I understand correctly that the main issue is hysteresis in the suspension spring? This means that the zero point (low energy state) of the spring shifts after it has been moved or flexed. I'm not sure that heat treating to remove stress will help this. As you point out, the thin wires used in the Talyvel are less likely to have this problem than the much thicker/wider strips that you are using.

sorry for my poor english

i think the “hystersis” is the final result of the suspension uncertainty, there are some possible reasons

1, plastic deformation, the LVDT-core stroke is +-0.5mm in my design, so its not likly

2, creep, but it wont happen in short period test

3, the LVDT uncertainty, i tried to limit the LVDT stroke to -0.5mm~0, and the test result of going 0 is very good, about +-0.15um/m 10 times, thats about 0.01um in absolute distance of LVDT

4, the suspention position uncertainty, in federal's design, imagine magnify the "zero point" 1000 times, the sharp vertical edge of the base and platen are no more straight or flat, so if the "zero point" shifts at different times, the result would magnify about 10 times in my design(its decided by the pendulum length).

for this issue, i dont have very good solution to verify, i did tried several times to grind the suspension posotion, using sand-paper or surface-grinding-machine, i even tried to polish them using Si-C or diamond, but i didnt get a definite related result. still trying.

5,the deformation of the spring
there are 2 different situations
5.1, the suspension base and platen is not flat, it would make a small "band tape"(i dont know if this expression is correct, see the following pic)

this could be a disaster
5.1 the suspension base and platen is flat enough, but when you try to tighten the screw, the platen and the base may slide, what would the reed-spring in the middle of the 2 parts do?? this would be serious if there are 2 or more bolts
federal designed a pair of groove, it could improve the "zero point" position-certainty and the "slide", i havent tried it yet

6,damping fluid, this is not of my major, i'm not sure if there are more mystry
i did made a prototype using talyvel oil-drop-damping, just a change the long thin wire to the short reed spring, and it leads to parasitic-oscillation. the repeat is likely 5um/m in my test, i thought the damping fluid all over the container could have a better result , and here is where i am



like i posted, the repeat is erratic in my test. if the stroke is +-0.5mm(thats almost +-5000um/m is my design), the uncertainty could be 0~100um/m, sometimes it shifts 0 in 2-time test, sometimes it goes to hell. and i fould tapping the shell is also sensitive the the repeat.

i tried to limit the variety to much smaller zone, like +-150um/m, the uncertainty is much smaller.
the following result is tested by pressing and lifting the granite-plate, the level stay-still on the plate

the data goes to +150um/m while pressing the plate, thats my weight. the data goes to only -50um/m when i try to lift the plate, its toooo heavy.

magnify the chart to nearly zero, you can see the repeat drops to 3um/m while limiting the stroke in my test.

so far, i dont have more ideas except the heat-treatment, hope you can have more ideas

 

[fobyellow]

long term test

i did a 116 hours test, the level stays still on the granite surface. and i'm prety sure i tightened all the screws.



there are some sudden-changes, it cannt be caused by outer factor. so i'm thinking about the silicone-oil, it may react with the polyester (enameled wire used in LVDT coil) or epoxy resin(used to bond the framework and steel-case of the LVDT coil). or some kind of glue used to bond the ferrite and steel of the LVDT iron-core.

 

[ballen]

I don't know how to fix your problem. But I thought you might find it interesting to see a hysteresis and stick-slip problem in a bubble level. I am monitoring the position using a capacitive sensor, and comparing it to a Talyvel 4.

This was sitting on a platform overnight, which was slowly shifting as the building cooled and warmed.

 

[fobyellow]

i did followed the NASA paper to test the bubble level, they could achieve very high resolution. but it share the same disadvantage, the long reading period. and in my test, it would took mitutes before it stables. i think thats because of the viscosity of the liquid, and the flaw of the inner glass surface. i read a paper describing the hydrophobic coating, its a king of steel like platinum or silver??

i have an idea to improve this. you can try it.



replace the air to another kind of liquit. usually they would have different dielectric constant, the edge aera of the 2 liquid may distort due to the surface tension, but the middle aera might be linear. my daughter have a similar toy we baught from Tokyo-Tower, and i observed the settleing time after movement, it seem shorter than a bubble. if you are going to apply for patent, please add my name

 

[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

This was very useful, thank you for making this!

One small error (also pointed out in Zerspanungsbude): the values for the three potentiometers given in the schematic are incorrect. See photo below. The correct values are:

(1) Balance (screwdriver): 250 Ohms
(2) Zero adjust (knob): 500 Ohms
(3) Gain (screwdriver): 20k Ohms

The values given in the schematic for (1) and (2) are not identical to above, but close enough that it would not matter. The value given for (3) on the schematic is too small by a factor of 100.

On the original these are wirewound potentiometers, which are more temperature stable. I'm using normal carbon composition ones, should be good enough for normal shop use over a few hours at fairly constant temperature.

Cheers,
Bruce

 

[Mk1_Oz]

Is Nate still on the forum? Did he ever finish his level?