Torque Transducer

[Mariss Freimanis]

Hi,

I'm redesigning my dynamometer. I had been using a servomotor as a torque measuring load but I want to replace it with a rotating torque transducer and a frictional brake load. This is to reduce the load moment of inertia.

Torque transducers are expensive; $3,000 and up for a magnetostrictive 5Nm to 10Nm device and they are easily damaged by torque overload, typically 150% of rated value.

It occurred to me to look at making my own. Hooke's Law says there is a linear relationship between applied torque and resulting torsional angle of any elastic material. My idea is to take a long thin steel rod, apply torque to it and measure the resulting torsional angle. As long as it stays within the elastic limits, the angle will be proportional to torque. This angle would be measured by differencing the counts from encoders mounted at both ends of the torsion rod.

The drawback to using a spring is it stores mechanical energy. Rate damping must be used to prevent undesirable ringing (oscillation) in the spring. I have elected to use a viscous oil and vanes to provide 2nd-order damping. See the attached pdf.

The torsion rod (piano wire?) would have to be of sufficient diameter and length to require +/-500 to +/-1,000 in-oz (3.5Nm to 7Nm) torque to give a +/-45 degree torsion angle and still remain within elastic limits. A 2,048 line encoder at both ends would give +/-1,000 counts over that angle range.

The design of the rotor and stator vanes is such that they would mesh and prevent any rotation much beyond +/-45 degrees, protecting the torsion rod from over-torque. The torsion rod (red) is thin so it would be susceptible to 'whip' at high RPMs. That is why it is sheathed by the rotor (green) and stator (blue) assemblies. The torsion rod would be fixed (set screw?) at the ends but free to torsionally rotate within the respective sheaths. They would also add stiffess to the assembly, taking bending stresses off of the torsion rod.

I'm thinking this thing should have only an oz-in or so of bearing friction yet be able to measure 1,000 in-oz torques. Vane length and oil viscosity would have to be adjusted to give a critically damped mechanical response.

What do you think? All wet or just half-baked?

Mariss

[Bubba]

Mariss,
My immediate knee jerk reaction is the problem of sealing the oil and the effect any seals would have on the output reading? No seal would mean oil in all the wrong places and a tight seal would affect the torque reaction and therefore the reading.

What kind of rpm range are you looking at? How about an Eddy Brake type of machine?

[georgef8]

If you can tip it up on its left end, it won't leak.

Will the damping have to be retuned for different motor inertias or can you overdamp it to tolerate various resonant frequencies?

George

[Mariss Freimanis]

Bubba,

Yeah, no great effort has gone into the seal problem. An 'O' ring might take care of it but it would have to introduce very little friction or otherwise it would add a hysteresis error to the readings. On the other hand, the oil would go a long way to lubricate and limit 'O' ring rotational friction.

RPM? 3,000RPM is a reasonable goal. My guess is the aspect ratio (diameter Vs. length) will be greatly influenced by the shear modulus of the torsion rod. Once calculated, this thing might be bigger in diameter, shorter in length. That will increase the critical RPM.

Eddy current brakes have a large moment of inertia. I'm trending towards proportional frictional brakes. Looking at hobby-type RC model nitro-car disc brakes. Those cars can get huge.

georgef8,

Damping is entirely a function of the spring constant of the torque transducer, not the test motor. Being that, it is a part of the transducer design, it has no effect on the test motor. You simply want the transducer to return to its zero-torque position without overshoot or ringing once the applied torque is suddenly removed.

Mariss

[georgef8]

Mariss,

Look at it using electrical equivalents, which makes it easier for me to understand. Say its a series LRC resonant circuit:

damping ratio = 1/2Q = 1 for critical damping i.e. just barely no ringing.

Q = w0*L/R = 1/SQRT(L*C) *L/R = 1/R * SQRT(L/C)
where w0 is the resonant frequency in radians per second, so the Q depends on all three elements.

Now pick L to represent the torsional spring and C the inertia - doesn't really matter which way round.

If you want to change inertia while keeping the Q constant, you need to change R.

If you purposely overdamp, you can avoid ringing even when the test motor has a different inertia within reason.

George

[TOTALLYRC]

I don't know what you are on using for the dampening oil, but take a look at rc car shock oil. The silicone based stuff is available in a wide range of viscosities and is very temperature stable. We use it in rc cars so the dampening rate won't change as the oil temp increases thru use.

We use the same type of idea only stationary for measuring the motor torque on rubber band airplane motors as they are being wound because it is more accurate and easier than counting turns. Of course we don't use encoders and such but the basic principal is the same.

Mike.

[georgef8]

Here's a manufacturer of eddy current dampers: http://www.cda-intercorp.com/products.cfm?cid=5

They probably cost the earth - there is a picture of a sattelite on the cover of their catalog.

If the torsion wire winds up 45 deg for 3Nm thats 3.8Nm/rad. A typical Nema34 step motor with 4.6Nm holding torque has 140E-6 kgm^2 inertia for a resonance frequency of 165 rad/s or 26Hz.

As far as I can tell critical damping happens for a damper value of 1.25E3 Nms/rad. You have to use one of the dampers with a gearbox to get there which makes it fairly unattractive.

[skippy]

Maybe this is a dumb suggestion but the first thing that occurs to me is my wife's excercise bike. The whole bike cost me $200 and it's got 12vdc controlled and variable resistance against the pedalling force (braking without any friction) to make it get harder and easier as required or by preset programs in the handlebar mounted minicomputer. Buy one, ditch the bike, do your own electronics and you have a dyno! Well, maybe not if I'm the one doing the electronics......

[David Campbell]

Mariss,

I crunched a few numbers and it appears workable but right on the limit as far as steel yield stress and critical speeds (the answers I came up with was 1/4" diameter, 27" long rod, 2950 RPM critical speed - back of the envelope stuff)

With a high tensile steel rod you could reduce the diameter a bit smaller (double the yield strength and the rod diameter drops to 3/16" and the length reduces to about 3" from memory, critical speed is somewhere around 20K RPM). However go much above 1000 oz-in either way and the shaft will have a permenant twist (exceed the yield point).

It has been a while since I have done torsional problems so I could of dropped a factor somewhere...

Have you considered mounting a pair of strain gauges (+/-45º to the axis, wheatstone bridge, etc), a small SMD op-amp and three slip rings (power + signal)? This would remove the spring effect and eliminate the need dampening. Since the signal coming off the slip ring is analog you could use a CRO to display any resonance in the system.

Given that modern op-amps use practically zero current, you could reduce the slip ring count to two by having the op-amp drive something as a current sink (transistor or FET?). Therefore you could read the torque by measuring the supply current (current may be more accuracte across slip rings than a voltage signal).

If you really want to get funky (given your experience with electronics you would only find it a mild challenge) is to supply power / return signal by inductive means (no slip rings).

Just throwing a couple of ideas out there...

David Campbell

PS: Here in Aus strain gauges are available for about USD$6 each, you would need two for a half bridge configuration (the other half of the bridge would be a pair of 120 ohm resistors).

[FPV_GTp]

How did your redesigned dynamometer test rig workout ? Would love to see some pictures of these motor dynamometer test rigs

Any of you electronics wizards interested in designing a adaptable cnc dynamometer controller for a IC engine or chassis dynamometer wether it be a water brake , eddy current or a inertia dynamometer configuration.

cheers

[AustinT]

A few years ago I made a an auto dyno using a semi trailer axle with super single tires mounted in place of the dual wheels. The axle was secured to a footing so that the top of the tires were at ground level. An approach was poured so the car being tested could back onto the tires sticking out of the ground. The car was then chained so that it could not move forward. In the chains I used load cells to determine how hard the car was pulling. I also logged the rpms so that I could determine Hp using the torque and rpm. A typical dyno run went something like this.
Back the car onto the dyno tires and block the front wheels so it could not roll back. Fasten the chain containing the load cell to the axle of the car being tested. Get into the car and put it into 4th gear( 1-1 ratio ) . Start the computer logging the rpms of the motor and the force on the loadcell. Apply the breaks slowly to the semi axle so that the car would pull harder on the loadcell. Test at various rpms to get a good HP curve.
Hope that makes sense, Here is a thread on the camaro forums

http://www.thirdgen.org/techboard/di...ne-trying.html