Force measurement with hall sensor

[willib]

I found a way to measure Force with a hall sensor

the sensor can tell me the distance from a magnet

 by outputting a voltage

and

F=kX is the force on a spring

by measuring the force on the spring beforehand , for a given displacement , i believe i can measure Force with a hall sensor.

thus enabling me to measure the force on the stator of my machine and letting me measure ultimatly the torque output of my blades

[jimovonz]

Ummmm... Willib, as you know the hall sensor measures flux density. No doubt you could calibrate a particular sensor to give you distance from a specific magnet along a specific axis for a system in isolation. You could perhaps even use such a setup to  measure the extension of a spring and deduce how much force is being applied (you would obviously have to deal with effect of the proximity of the steel spring would have on the flux from the magnet...) Are you proposing to mount your stator such that it is free to rotate abount its axis against the tension of this spring and use your hall sensor to measure the displacement? The magnetic environment around the alt is not static and would change with speed/load making such a measurement almost impossible. I'm sure there are much easier ways to do it. If your proposing to do away with the spring then I'm in the dark...

[willib]

"Are you proposing to mount your stator such that it is free to rotate abount its axis against the tension of this spring and use your hall sensor to measure the displacement? "

I am going to mount the stator so that it rotates Yes, about 1/2 inch max, or x degrees .

i was thinking of compression on a spring actually, but you have the idea.

[willib]

have ya got a better idea?

[commanda]

Spring balance. Buy them at any good hardware store.

Amanda

[willib]

i got one

thats what i'm gonna calibrate the Force O meter with

all kidding aside i just cant see myself trying to write down the force on the alt as its running

and besides i need instant readings correlated with the wind speed and RPM and current

[commanda]

OK I think I see what you're doing.

I'd forget the hall sensor and use a load cell as a torque arm to stop the rotor rotating. The load cell will give millivolts directly proportional to the force applied.

You will need a stable dc supply (usually 10 volts) and a really good instrumentation amplifier, as load cells output millivolts.

Amanda

[maker of toys]

use a linear potentiometer.  nice robust linear signal, no amps required, no magnetic interaction, very little EMI concerns.

a pendulum and rotary pot/rotary encoder could also work . . . no springs to worry about calibrating/corroding.  

as an aside:  8 bit ADCs with an on-chip serial (send only) interface, suitable for for ucontroller use (not full rs232/rs485) are less than  $3. . . ADC0831 (datasheet: http://www.national.com/pf/DC/ADC0831.html) is one such.  I've used a few, and they react well to field expedients;  a single 7805 (or whatever) can power the ADC and the pot.

Up in the air, you might be able to tap 1 phase with a suitable capacitor and rectifier and send your data down as digital . . .  if you're feeling adventureous, it's not hard to make the digital portion in fiber optic, in case lightning is a concern. (agilent makes a plastic fiber that can be field terminated with a knife, a pair of pliers and a piece of 600 grit sandpaper. . . )

-Dan

[Flux]

You can use a hall sensor, it would be much more linear with two, as Honeywell show in their Lohet data, but you would need to take care of stray fields.

The load sensor from cheap domestic scales would seem easier and have higher output than industrial load cells.

You could strain gage your torque arm, or you could use a linear pot ( conductive plastic preferably) or you could use a linear differential transformer ( expensive but easy to make).

If you can build a torque tube in the systm it gives you even more possibilities with Wideman efect and measuring phase displacement between inductive sensors.

[dinges]

Willib,

I've been following this thread but I have utterly no idea what you & others are talking about.

How about a quick drawing or a picture of what you propose. Is it a dynometer?

[maker of toys]

there's a lot to be said for the differential transformer ideas; I'd think that for home use the linear one would be somewhat easier to make. . . . though as I sit here I can think of two ways to make a rotary DT, and it's only a short step from that to your Wideman idea. . . .  though with the Wiseman, you've increased your signal processing load on the DAQ system.  (at low sample rates that's not really a problem, though)

Too, the biggest problem I see with the load cell route is neatly sidestepped by a repurposed bathroom scale--   by using an existing circuit, you get all the amps and RFI hardening already debugged for you.  I hate when I miss something obvious like that.  <G> bear in mind that some units straingauge more than one support and sum the signals.

-Dan

[zapmk]

Hey willib,

Wish I'd seen this post earlier, been setting up my alt dyno to measure the force on the stator, I've always wondered now much force was placed on the stator and mounts,

when shuting down the alt by shorting it out.

I think that you could get a better match to the blades if you knew the torque curve to power output of the alt, or the other way around.

This is something I had in my fishing tackel box that I'am going to gave a try

http://www.americanweigh.com/product_info.php?cPath=23&products_id=485

 -Zapmk

[willib]

I did some repeatability testing

A) is the sensor

B) is a chunk of broken HD neo magnet

C)dial indicator

D)carriage from cd drive

this is a wide shot of the whole setup

next to the salvaged LCD readout is the "power" going to the sensor in the upper left. i left it connected to the sensor for days , by accident , and found the battery in good shape so i checked the current draw from the sensor , its only 4.6mA

and thats the most it draws , it actually goes down the closer the sensor is to a magnetic source. with it right next to the 1/2"dia mag it only draws 2.7mA strange..

those parallels next to the carriage hold the sensor wires

on top of the folders is my wacky wiring system the sensor wires are plugged into a cable just to hold them in place, it needs some work i suppose.

here is the millivolt reading with the dial indicator set at 0

here i moved it in 5 thousanths of an inch

here i moved it back , it didnt always come out exact , but was always within one or two tenths of a millivolt.

With one of my 1/2"x 1/2" dia magnets i could swing the voltage output up to .556V and down to .248V in a span of 1/2 inch.

[willib]

Peter , picture an axial flux machine with a rotatable stator ( a few to 10 degrees anyway)

as the machine puts out power ,the rotors apply torque on the stator it will tend to move in the direction of rotation.

as it moves it butts up against my force meter , and compresses the spring .

the spring will be calibrated so that i know when it has compressed x mm (in) ,it will have a known force on it. F=kX force on a spring.

right now i am working on how to measure x and i think the hall sensor will do the job.

i think i can fit the whole thing inside a plastic tube , but i havnt worked that part out yet

[willib]

hi Dan i looked up the linear potentiometer , it looks real expensive , the ones i was looking at anyway

the chip i'm gonna use has 8 ADCs built in (multiplexed), one adc , 8 single ended inputs)

[willib]

hey Zap thats really cool , i also like their wireless one.

http://www.americanweigh.com/product_info.php?products_id=385&osCsid=7b29c2146823cc24847008f3e78
8fe6e

with that i could make a teleforce meter

[maker of toys]

Agreed. new, scientific or recording quality linear pots are expensive. for a homebrew Dyno, they don't need to be that pricy. Try dumpster diving, which it looks like you're familar with, based on your test rig <g>. . . . or surplus stores/garage sales/your garage, depending.

try one of the following: a home/car stereo unit with a graphic EQ.

Some old boom boxes use linears for volume and balance. . .

Some old computer joysticks also have linear pots.

some 120VAC dimmer units use a linear pot, too. . .

for permanent installation, a differential transformer is the right answer- no moving contacts, easy to construct with materials any of us already have on hand, and easily made weather tight.

[dinges]

Willib,

Is this what you are trying to achieve: ?

http://www.otherpower.com/guemes_notes.html

See the last two pictures & corresponding text. Seems there are much easier ways to build a dynamometer. Unless you're trying to achieve something else.

BTW, I like the fact that you intend to measure such things. I am be very interested in knowing the stator losses of a motorconversion and intend to make measurements like the Dans have done. I'm surprized I haven't see anyone else make such measurements, as they give a direct insight into the efficiency of a genny.

[dinges]

my main gripe being: you're adding an intermediary step (measuring distance) which introduces an extra error.

You could immediately measure force, with a weighing scale (remember, they measure force, not mass). You are instead measuring a distance (x), and converting it into a force by means of the spring equation (F=-k*x). But, you'd need to know the k constant of the spring. Which can be determined, of course. Just adds an extra step, extra errors, etc. Also, you need to calibrate your hall-sensors (distance vs. field-strength/voltage), account for non-linearities of the sensors, etc.

Don't get me wrong, I like the way you use hall-sensors, but in this case, it looks like a solution looking for a problem. BTW, I've just taken apart an old VCR and the hall-sensors have been taken out and carefully put aside.

[willib]

good ! keep them handy because they are really cool devices, very tiny arent they ,not much larger than a microdot

last night i checked the flux from my magnets ( the 1" x .875"dia N40 ones)all were pretty much the same , in the range of .634 to .639 V , but one was .644 V

it is interesting if you turn the sensor on edge to the magnetic field the voltage can be set to zero. i suppose this is because the flux difference from one side to the other is 0?

i agree it is not going to be easy , but it sure is fun

it so happens that i do have a linear pot , a bunch actually, after Maker of toys mentioned the graphic EQ , i found one a few months ago ( garbage picking )lol

but it still will need a spring so i'm going to stick with the hall sensor .

"Also, you need to calibrate your hall-sensors (distance vs. field-strength/voltage), account for non-linearities of the sensors, etc."

i agree the hall sensor output is not linear over a large range , because the magnetic field is not linear wrt the distance from the magnet.

but that can be taken into account when calibrated ie,  y force = x voltage

ie the applied force will be directly translated into a voltage when calibrated.

[Ungrounded Lightning Rod]

Want to measure the torque on the alternator?  Turn the problem inside out:

 - Mount the alternator on a pair of bearings, allowing the stator to rotate on the same axis as the shaft.

 - Resist the rotation with a spring scale between a lever on the alternator (at a known distance from the axis) and a suitable anchor.  The scale should be pivoted at both the anchor and the attachment to the lever, and should be at right angles to the line from the center of the axis to the lever attachment point.

 - Make coils of your power wiring, like the coil-cords on a telephone handset, so they don't apply significant force VARYING WITH POSITION to the alternator stator.

Torque on the shaft (and thus the stator) is force times distance, where distance is from the axis to the joint between the lever and the scale, and force is the scale reading under load minus the scale reading at rest.

But why are you bothering to measure this?  If this is a radial-flux machine with negligible parisitics (decent bearings, no cores in the coils, no significant reverse-current leakage in the diodes, and reasonably thin wire - i.e. not pipe or buss-bar - so the coil eddy currents are negligible), torque is DIRECTLY proportional to the current in your alternator windings.  Using an average-reading ammeter (a shunted analog meter) in your rectified output lead will give you a reading proportional to torque.  Calibrate it at one speed and it will be good for all speeds.

[Flux]

I assume you mean axial as most radial machines have iron.

I strongly suspect that with a star connected airgap alternator the torque will be proportional to dc load current near enough for normal use. This would simplify things quite a bit. I wouldn't expect this to hold for delta connection with circulating currents.

For an ac load into resistors the relationship should be pretty exact, I think I can set up and check the dc case fairly easily to see if the rectifiers do anything queer.

Flux

 

[Flux]

Within normal experimental error it seems as though dc load current is near enough to torque for an air gap machine. There is an offset at zero no doubt due to the diode drop and the torque is higher than indicated by amps into a battery compared to a short circuit, but results will be within the errors expected from a torque measuring circuit using cheap options.

The graphs show that delta is slightly worse than star but still within reason.

The low curve in each case is short circuit current. The higher one is into a simulated 12v battery ( constant 13V)

Don't expect any of this to hold for iron cored machines but for axials you have a simple method of measuring mechanical power.

Flux

[willib]

Flux did you use Power / RPM  to get your torque data

this sure does simplify the problem

[Ungrounded Lightning Rod]

I assume you mean axial as most radial machines have iron.

Yep.  (Once in every week, Braino in every brain!)

[Flux]

Torque was measured direct with a spring balance just as you planned to do.

This is a direct comparison between true torque and dc stator current.

Flux

[willib]

you probably cant post a picture of the setup but i would really like to see it

how did you make your stator moveable , cut slots in it near the perimeter?

i was thinking how i would do it , all the mount points would be changed to slots , all it has to do is pivot (around the center )a little bit .

[Ungrounded Lightning Rod]

There is an offset at zero no doubt due to the diode drop and the torque is higher than indicated by amps into a battery compared to a short circuit, but results will be within the errors expected from a torque measuring circuit using cheap options.

Ah.  Parasitic loads (air friction, eddy currents, etc.) are linear or near it and start as soon as the genny is turning.  But current through diodes starts after you've got 1.2 volts to overcome diode drop.  So: a kink at startup.  An offset that corresponds to the torque from all non-output-current sources at cutin speed.

If you project the slopes back to the Y axis the intersection gives you that parisitic cutin torque.  Notice that it's very small for star (which has only mechanical friction and the tiny eddy current heating from currents circulating within the wire diameter), much larger for delta (which also has the circulating harmonic current).  This give delta a pronounced kink, star almost none.

Delta has a lower slope than star because with star all the load current goes through two coils, producing two units of drag, while with delta about 2/3 of the load current only goes through one coil, producing only one unit of drag for that fraction.  Resulting slope for delta is about (2 * 1/3 + 1 * 2/3) / (2 * 1) = 2/3 that of star, which is about what the graph shows.

I expect delta to show a little deviation from linearity along the curve, falling off a tad at the right, because heating of the coils reduces the circulating current's parasitic load to something below what its linear-with-speed projection would predict.  But the harmonic current in this machine is low enough, and the temperature-related resistance change small enough, that it doesn't show up.

Since parasitic loads are proportional to speed and speed is higher for a given current when charging batteries than when shorted you get a higher torque for a given output current for chargng than for short.  But since parasitic loads are a very small fraction the deviation is small, nearly in the noise.  And it's higher for delta.

[Flux]

The set up is simple. The front rotor of the alternator is mounted to a countershaft driven by a dc motor. You can do exactly the same thing by mounting it in a lathe. Without restraint the stator and stator mounts would just rotate with the rotors but the stator is held from rotating by a spring balance.

Torque is just spring balance reading times radius of attachment.

I can probably get a picture if you still cant follow it. This is purely a test rig, not an operational wind turbine.

Flux

[zapmk]

Willib,

I found a Torrington roller bearing to fit the axle, drilled a hole in the stator bracket for a lite press fit, used a very light weight oil, it rotates very easy.

The scale that Iam going to use I beleave uses a load cell, no noiceable movement at all.

 -Zapmk

[Flux]

This parasitic torque is interesting, I know there is an issue with delta and some machines could be significantly worse so let's forget delta for now.

This alternator I did the test on has tapered roller bearings and has always showed lower efficiency on load tests than I would have expected so this issue has not been a complete surprise.

Up to the point where diodes conduct the bearing torque is too low to read on the spring balance. As soon as diodes conduct the torque becomes real and easily measurable even when the electrical power delivered is negligible. I can't think of any reasonable electrical reason ( unlike the delta case where there is good reason)

I have found that this torque is influenced by the setting of the bearings so It may be mechanical or a combination of electrical and mechanical. I suspect there is some transverse vibration of the stator that may affect the bearing loading.

It is small enough not to show in normal operation and also with prop thrust the bearings are loaded differently.

This factor seems to be significantly less with machines with ball bearings unless by chance the machines with ball bearings are electrically different in some way.

It just means that for using dc current for torque you need to have a calibration at two points to allow for the offset.

Flux

[willib]

Your torque measurements are Kg-m ?

i dissagree with the statement "The graphs show that delta is slightly worse than star..."

because at 4 Amps star used 7 to 7.5 ? kg-m of torque , while delta only took 4.5 to 5 kg-m of torque to output the same amount of current?

there may have been other factors like RPM , to make you say delta was worse than star , but from here it looks like delta was more efficient than star , in that for the same current delta put less torque on the stator , and ultimatly would use less torque from the blades ?

[Flux]

This was a test to determine the relationship between dc current and torque.

It has nothing to do with power or speed. When I say that delta was worse than star I was just referring to the linearity of the relationship and the fact that with delta the difference between short circuit current and current into a battery were greater.

Obviously the star winding will produce more amps into a short or into a battery at lower speed and will take more torque to produce the same power.

I gave you no speed readings so you have no way of knowing about anything other than torque, You can form no opinion on power or efficiency.

The delta speeds were higher in each case, the short circuit figures were at a crawl and the 12v figures were around 150 rpm for star and 300 rpm for delta, changing with current to some extent.

Did I confuse you when I said this gives an easy way to measure power, by that I meant that you could derive torque from dc current and use the speed from frequency to have the 2 components to measure power.

Flux