Encoders for Coupling Investigation 1

[RichPWade]

So I need to setup a system for measuring a Wankel engines shaft velocity on either side of an elastic coupling in order to investigate issues with them tearing. I want to setup an encoder on both sides of the coupling to measure shaft accelerations on either side in order to characterize the force applied to the coupling. I've looked at a few options but the location is challenging. Essentially I'm trying to find an encoder with at least a 720 pulse per rev resolution (Though greater than 1800 would be preferred), a relatively large setup tolerance, and preferably a differential output. I've looked at solutions from RLS (Renishaw in the states) with their magnetic incremental encoders which would work but they have extremely tight air gap tolerances (0.1 to 1 mm). I'm also leaning more toward optical encoders as the location I need to mount them is very close to the ignition coils and so are a very electrically noisy environment.

Anybody have any leads they can pass my way on Encoders that might work for me?

Thanks,
Richard Wade

 

[GregLocock]

I can't believe you need that many teeth for a durability issue, but OK. For driveline work I used ABS tonewheels and the associated sensors, but I don't know what technology they are. Mostly on engines I've used optical probes.

You are aware you'll need to FM demodulate the signals? The easy way is to use a TV analyser instrument but it is an easy bit of code to write.




Cheers

Greg Locock


New here? Try reading these, they might help FAQ731-376

http://eng-tips.com/market.cfm?

 

[SomptingGuy]

It's been a while, but we always used variable reluctance probes (cheap, robust, simple, passive) targeted at existing gears in the driveline. It just requires finding suitable flanges and housings to drill and tap. It's not too difficult to write the FM demodulation code to process their signals to resolve gear motion at higher than tooth-passing frequency (not limited to zero-crossing detection algorithms).

Steve

 

[btrueblood]

I built my own encoders, once upon a time, using IR optical gates (torn out of simpler encoders) and laser-printed sheets of acetate. And thought I was so smart when I wrote my own code to process the signals. Then I joined this forum and learnt from Steve and Greg that it's ho-hum standard. Do realize that the gate (transistor) switching speed limits either your rpm or pulses/rev; faster is better but can cost.

 

[FeX32]

In our labs and test cells/ dynos we usually use similar to what Steve mentioned.
Hall effect sensors measuring the passing of each tooth of an existing gear. The is then processed into a digital input and processed in real time to determine the time between each rising edge of the digital signal. This can then be used to measure speed, accel, or angular displacement. We do use 1020 encoders as well, but most of the time 40-100 teeth are enough since the encoders use up much more space.
The hall effect sensors are positioned with custom mounts that are modular and usual can be setup very quickly no drilling or tapping or even touching the existing machinery.

 

[FeX32]

Steve, I never tried processing speed resolution at higher that tooth passing frequency.
I might look into this.

Also, we did at one time have a small issue with drift. We were measuring timing error between 2 shafts using the same real-time clock and using tooth passing freq, but the noise in the signal caused some drift in the angular displacement after integration.

 

[SomptingGuy]

@FeX32:

An old colleague of mine was obsessed with demodulation. The basis is that the probe and the tooth width (& gap) are similar in size, so the signal produced is a pretty clean carrier wave. At each sample, a gear position can be estimated, rather than at each tooth. He used the hilbert() function in Matlab to obtain the analytic signal. The unwraped phase of this is proportional to the gear's angular position: +2pi per tooth. Gets a bit fussy with missing or broken teeth, but gear teeth are usually very precise, regular and clean.

(It also can get a bit complicated if the gear has appreciable movement radially on its bearing, but those extra DOFs can be resolved with additional probes around the circumference.)

Steve

 

[FeX32]

Interesting. I might try this if I get some spare time. In some applications it is useful to have more accuracy.
Particularly, when computing timing errors between 2 shafts or gears, I have found in the past that 2 gears with differing numbers of teeth can give drifts when because of inherent accuracy differences.
Say one gear has 20 and the other has 40 teeth. The one that updates its speed and displacement twice as much has more accuracy. But if you are then using this to say measure backlash between the gears you have to measure timing error (2*N1/N2). Due to the differences in accuracy, a drift in this timing error signal over time can be observed.
I always wanted to find a better way to get rid of this easily.