Rotary optical quaderature encoders like the agilent HEDS series lack fine accuracy because of eccentricity in the mounting, the motor bearings, and the dimensions of the interacting parts. A back of the envelope calculation reveals the effect of eccentricity.


For 10 thou of eccentricity, with a radius of .5 inches:


ErrorAngle = Eccentricity*(360/2*pi*r)= 1.145 degrees of error


For 1 thou, (which would be very hard to achieve) there would be 0.1145 degrees of error


Which may be acceptable in some situations. Plus, these encoders have very good repeatability. But what if accuracy is desired? The error could be mapped, however this requires special equipment. I have also heard of methods that use two encoder read heads instead of one.


I am looking for some sort of alternative to this that will provide better accuracy when mounted to a motor. What sorts of encoder products exist that tackle this problem?

 

Cheap encoder tends to produce cheap results, there are three basic methods of motor position encoding, simple quadrature counts, the higher the resolution the higher the accuracy and smaller input increment.
The resolution can also be increased by using all four edges of the quadrature pulses to achieve x4 the basic resolution.
Then the quadrature encoder that uses the source sine/cosine signal and decodes using the arc tangent function for high resolution.
Then the magnetic resolver, that usually uses the same arc tangent function.
It usually pays to use the finest resolution the controller or processor is capable of, even though it may be a smaller resolution than required.
Industrial motion systems are now using 100k pulses/rev encoders.
Max.

 

Just use a better encoder. We have wafer positioning systems that use encoders for up to 360k counts using internal interpolation and external quadrature for large disks.

http://www.microesys.com/specifications/packaged-encoders

 

Those mention quadrature so presumably they use the quadrature sine wave/arctangent detection method?
Max.

 

Those mention quadrature so presumably they use the quadrature sine wave/arctangent detection method?

You sure you don't mean cosine?

 

Those mention quadrature so presumably they use the quadrature sine wave/arctangent detection method?
Max.

This encoder is used on a few systems. Internally it might be using that to provide the digital output.
http://www.microesys.com/sites/default/files/Packaged-Encoders-M35-Data-Sheet.pdf

 

You sure you don't mean cosine?

Should be sine/cosine.
Max.

 

I am not sure if you guys understood my post fully. I was talking about eccentricity and how a very small amount of it, either in the bearings of the motor, or in the dimensional interaction between the code wheel and the reader, would make the readings very inaccurate.

Encoder manufacturers acknowledge this. Take this quote from US digital for instance:
"the accuracy of an encoder is primarily defined by the precision at which the code information is placed on the disc and how concentric the pattern rotates with respect to the encoders sensing element"

"Accuracy and resolution of an encoder are not connected in any way"
http://www.usdigital.com/company/encoder-white-paper

Of course, you guys might be talking about modular encoders which might not have these problems? There may be other considerations with these types of encoders, particularly in how they are coupled

 

I am not sure if you guys understood my post fully. I was talking about eccentricity and how a very small amount of it, either in the bearings of the motor, or in the dimensional interaction between the code wheel and the reader, would make the readings very inaccurate.

The model that I linked to 'M35' has "Dual read optics for disc eccentricity error cancellation. So yes, it's well understood and handled with the right encoder.

 

Modern single chip encoder/interpolators allow alignment to compensate for error sources. There are several areas which can cause an error - see page 10 of this paper on signal conditioning and interpolation: http://www.ichaus.de/upload/pdf/WP7en_High-Precision_Interpolation_140124.pdf as well as page 8 here: http://www.ichaus.de/upload/pdf/WP6en_Magnetic_vs_optical_17092012.pdf.

 

In addition to the encoder we normally have a resolver in tandem to generate the sine and cosine position signals for exact speed and velocity control at high rates then switch to the encoder control loop for exact positioning when stopped.
http://www.analog.com/library/analogDialogue/archives/48-03/resolver.html

 

I think I might look into those chip encoders that IC-man mentioned.

I guess the point I would like to make in this thread, is that the resolution of an encoder has nothing to do with its accuracy. Accuracy specs aren't even given on the most expensive encoders available, like the M35 that nsaspook mentioned. They describe it as being high accuracy, and they mention the feature of two read heads for correction, but they give no specifications for this accuracy. I believe this is because it can vary with how the encoder is set up.
It can not be assumed that an encoder is accurate to the count.

 

...
For 1 thou, (which would be very hard to achieve) there would be 0.1145 degrees of error
...

I think you are exaggerating the problem. Decent servomotors like ones I have here have shaft runout FAR less than a thou, maybe 10 times less, and trust me I have a dial gauge.

And even the older glass encoders on them are optically printed. Remember that optic printing is used to make everythign these days right down to the finest silicon with billions of transistors on it.

In other words; they know how to line up optics these days!

And good encoder discs are friction pressed onto the servomotor shaft so there is little to no runout from mounting it, not like the old days with a grub screw etc.

I would be astounded if any decent servomotor and encoder set had "0.1145 degrees of error". (We are talking radial error here?)

 

Yes, radial runout of 0.010" is a very large amount, like someone mounted the encoder wheel by wrapping the motor shaft with duct tape, and then mashed the wheel over it.

I would expect far less error in most applications.

 

we have a lot of servo systems here, from lots of different manufacturers. using resolvers, encoders, and magnetic. some of the best are from fanuc and siemens. some of them do as mentioned before, with a n encoder or resolver on the motor that takes over at high speeds and also supplies operational feedback for the digital drives. any shaft runout or gear slop is fixed quickly here, you cant have machines hold a .0001 tolerance with backlash or runout.

 

Hi,

I am making some quadrature encoders for a tracker. I made some film disks 9 dark 9 light, and disk mounts (with grub screws) and attached them to the gearbox end of a motor 100:1. This gives a resolution of 0.1 degree. (If perfect)

Even though I spent time centring the disks I am finding some eccentricity with the disks. Also on an oscilloscope, with the interrupters on opposite sides, it appears that the two signals are nowhere near opposite.

It appears that the oscilloscope is using the signal from one interrupter, to show both interrupters, and errors of my old equipment added to my errors, show all sorts of wobbles and errors.

Having said that, if I set it up the best I can, I assume it could be 'say' better than 0.3 degree out, and I only need 1 degree of accuracy. Should I just accept the errors and move on?

See: http://forum.allaboutcircuits.com/showpost.php?p=750615&postcount=421

Camerart.

 

I don't know how HP managed but they have an encoder on a spinning mirror on my 1200 dpi laser printer. The shaft diameter is 3mm so, via you can calculate the allowable runout but remember, the spinning mirror must be aligned much better than the 1/1200" because the distance travelled by the reflected beam.

There may or may not be an encoder on that mirror motor but the alignment and shaft concentricity issues are the same whether a mirror an encoder.

All that on a 15-year-old printer with 100,000 page count and running fine.

 

I found a broken encoder fixture for a M35 (hub shaft broken from excessive belt pressure) for a large disk direct drive motor positioning system. A 1:1 Kevlar timing belt drive from the motor shaft to the encoder drive shaft/bearing with a flex coupling to isolate small offsets to the encoder shaft. The encoder here is a 8192X10 model with 4X edge counting. +-5 counts is the normal positioning error.

 

The encoder here is a 8192X10 model with 4X edge counting. +-5 counts is the normal positioning error.

I assume this is due to mechanical error external to the encoder and/or PID resolution, not actual reading error of the encoder?
Max.

 

Belt and flex coupling.