Absolute Position Encoder

[scrat]

Forgive me if this sounds like a silly question.

I thought Sin/Cos resolvers/synchros where absolute position encoders. So what is all this talk about reading absolute position during start up only? I know I must be missing something here.

Thanks

A sin-cos higher "frequency" trace (i.e. with many periods per turn) is added. This allows to read the position at higher resolution within one period of this trace, while the absolute position in the turn comes from the main (1 period/turn) traces. In practice, the ADC resolution gets multiplied by the number of periods of the secondary trace. This is, at least, how they explained it to me.

[Leo Bodnar]

They are absolute with respect to one turn.  If the resolver is used in a manner that it makes more than one turn, then some method to count turns is needed, software or hardware.

There are multiturn absolute encoders that can typically uniquely resolve 4096 turns.  They just have a gear train with few additional sensors in addition to main encoder disk.

[Leo Bodnar]

Here is some useful information on Hyperface encoders:
http://www.sick-automation.ru/images/File/pdf/Hyperface_e.pdf

EnDat encoders:
http://www.auto-met.com/heidenhain/Pin_outs/EnDat%2021_22.pdf

[RayJones]

I get the impression with the hyperface units that they fit onto the motor shaft and can provide both the motor turns info, along with an absolute, geared down encoder.

Pretty nifty device, BUT depending upon the final drive system may be susceptible to gear backlash and worn gearing for the absolute readout.

An absolute encoding system on the very last driven element, fitted with anti-backlash gears itself if required, is sure fire.

[am2pgs]

Thanks everybody for the clarification regarding absolute position. I now undrestand the issue.

[Leo Bodnar]

I get the impression with the hyperface units that they fit onto the motor shaft and can provide both the motor turns info, along with an absolute, geared down encoder.
Pretty nifty device, BUT depending upon the final drive system may be susceptible to gear backlash and worn gearing for the absolute readout.
An absolute encoding system on the very last driven element, fitted with anti-backlash gears itself if required, is sure fire.

Not sure if you are talking about the whole machinery or the internal design of the encoder.

Multiturn encoders have both optical high density pickup mounted directly on the shaft and gear train for full turn counts only.  Single turn encoders are exactly the same but without extra gear train.  These gears only have to count full turns so their backlash or wearing out is non-issue.

I have only had experience with AC servo motors that have absolute encoders fitted in by the manufacturer.  It's not impossible to retrofit the encoder to existing motor but with such a high precision it is difficult to create perfect mounting.

 

[FreeThinker]

I get the impression with the hyperface units that they fit onto the motor shaft and can provide both the motor turns info, along with an absolute, geared down encoder.
Pretty nifty device, BUT depending upon the final drive system may be susceptible to gear backlash and worn gearing for the absolute readout.
An absolute encoding system on the very last driven element, fitted with anti-backlash gears itself if required, is sure fire.

Not sure if you are talking about the whole machinery or the internal design of the encoder.

Multiturn encoders have both optical high density pickup mounted directly on the shaft and gear train for full turn counts only.  Single turn encoders are exactly the same but without extra gear train.  These gears only have to count full turns so their backlash or wearing out is non-issue.

I have only had experience with AC servo motors that have absolute encoders fitted in by the manufacturer.  It's not impossible to retrofit the encoder to existing motor but with such a high precision it is difficult to create perfect mounting.

 

I don't understand that statement, surely if the resolver is changing from (say) 360⁰ to 0⁰ and the turns counter does not change at the same time then it will be 1 complete cycle out until it does change? This would need to be accounted for in it's error correction system somehow.

[Leo Bodnar]

I don't understand that statement, surely if the resolver is changing from (say) 360⁰ to 0⁰ and the turns counter does not change at the same time then it will be 1 complete cycle out until it does change? This would need to be accounted for in it's error correction system somehow.

Two systems have an overlap to account for this.  Technically, geared multiturn part counts down to half turns.
Have a look at Fig.5 and Fig.6 of http://www.sick-automation.ru/images/File/pdf/Hyperface_e.pdf
It's a similar situation with aligning sin/cos period and a full period counter.
E.g.
angle=359⁰, gears = 0.5 or 1.0, result: 359⁰
angle=0⁰, gears = 0.5 or 1.0, result: 360⁰
etc.

[FreeThinker]

Ok I see it now, but how would the system know it's position after a power loss? Would it have to do a home? I was assuming that it didn't need to and that was throwing me off I think.

[Leo Bodnar]

Ok I see it now, but how would the system know it's position after a power loss? Would it have to do a home? I was assuming that it didn't need to and that was throwing me off I think.

Absolute encoder has immediate knowledge of its position. 
Incremental encoder has two signal tracks and two pickup sensors (3rd track is for index mark):


4096 positions absolute encoder would have 12 tracks and 12 sensors providing 12-bit position value:


It's a bit simplified but in essence is how it works.

[FreeThinker]

I've worked on KUKA robots and if they were powered down mid cycle they needed to (quite literally) unwind back to a known point in it's 3d world. Accuracy was maintained because the effective  length of the encoder was greater than maximum movement (ie it never overflowed) but still needed to re reference on power up. When originally commissioned it was necessary to mount calibration modules on each axis to define this point mechanically. So while when powering  up it thought it knew where it was (from it's encoders) it never knew exactly until it re zeroed all of it's axis, do these not have this problem?

[RayJones]

I get the impression with the hyperface units that they fit onto the motor shaft and can provide both the motor turns info, along with an absolute, geared down encoder.
Pretty nifty device, BUT depending upon the final drive system may be susceptible to gear backlash and worn gearing for the absolute readout.
An absolute encoding system on the very last driven element, fitted with anti-backlash gears itself if required, is sure fire.

Not sure if you are talking about the whole machinery or the internal design of the encoder.

I was talking about the *system*.

You can have ultra precise counting of motor shaft revolutions, but due to tolerances, wear and backlash in the drive train from the motor to your final output, it may actually take a few degrees of shaft rotation before the output actually begins to move upon reversals.
This is why relying on a geared down shaft encoder at the motor itself may not match reality - this includes using motor shaft counting schemes.

However when it comes to the internal gearing of such multi turn encoders, I have no doubt that they would use spring loaded anti backlash gears as minimal load is transferred through these "data take off" gears.

For reference, I'm thinking about our radar systems that have continuous 360 degree azimuth rotation and spend most of their lives rotating in one direction, but are required to accurately position at times.
This positioning mode would more than likely require a motor drive reversal, thus exposing the danger of relying upon motor counts alone due to gear train tolerances.
These systems have a single turn absolute positioning encoder with spring loaded anti backlash gears on the final drive gear that is fixed to the axis you observe.
The motor is always connected via gearboxes.
We never home, but rely upon accurate encoder position information from the final axis.

The actual encoder does not need to be precisely aligned. Offsets to the reported position are determined during a calibration procedure in software.

Now due to the continuous azimuth rotation, slip rings are employed to gather information from the elevation axis.
The issue we had was the elevation would creep out of position due to corrupted motor counts which were derived from the resolver feedback of the brushless motors passing over these slip rings - large antennae if you like.
Various filters and error checks were employed in the servo drives, but the problem remained.

After a software change, the system now uses the absolute encoder on the final elevation axis and it does not creep out of alignment over time.
The encoder was always there, but was only used to read back the position, and was only ever used by the servo drive upon startup.

This is a real world example.
I'm sure in a perfect environment the concept may work fine, BUT not in our real world.

[Leo Bodnar]

However when it comes to the internal gearing of such multi turn encoders, I have no doubt that they would use spring loaded anti backlash gears as minimal load is transferred through these "data take off" gears.

They are very basic gears with no special treatment.  If you shake the encoder you can hear them rattle.  They only need to count with half-a-main-shaft-turn precision as I have described above.  The main high precision optical disc is rigidly attached to the encoder shaft (and subsequently a motor or load shaft.)

Nice description of the radar troubleshooting!  Slip rings still rule 

[RayJones]

Slip rings still rule 

Indeed they do, and for the price per unit you could have a fine existence making them

The modern rings are very fine, think of 1mm spring wires running in narrow grooves.
The original ones are very chunky, think chunks of graphite 10 x 8mm on fat metallic rings.

But yeah, fixed cables are simply non workable. For sure you could get smarter these days with wireless/optical methods, but you still need hefty currents for motors, and simple volts to run any smart units beyond the slip rings.

Tesla coils would induce too much RFI and spoil the -112dBm+ receiver sensitivity