[medved]

Hello

I am learning about servos, and as I searched the forums, the only contex I could find INDEX pulse was in connection with limit switches, that is, to enable more accurate homing using limit switch signal together with first index signal.

But when I first heard about index signal, my first thought was(or I read it somewhere) that it is used for CORRECTING MISTAKES that happen in incremental encoder. I thought that in encoder documentation text sheets there is allways a specification, how big error can be during one full rotation of encoder wheel.


My understanding was, that incremental encoder can "loose steps" because of physical conditions between photodiodes and incoming light, or it can count "too many" steps because of electric spikes(despite protection circuits), and that index pulse is there to enable correction for these mistakes.(micro processor has to check whether sum of up/down counts between two index pulses is zero or exactly the number that corresponds to the number of stripes on the optic wheel, any deviation from this would mean an error).
If I assume there is a worst case mistake of 5 counts every full rotation in 500 line encoder, that means that there accumulates error of one FULL ROTATION every 100 full rotations!! This can be few mm on mill!!
With ABSOULTE encoders, there shuldnt be such problem, even if one position is read wrong, the next position will be read wright and there will be no error to accumulate.


But when I looked closely to information sheets, I could only find specification on CYCLE ERROR which is in electrical angle degrees and it means, as I understand, only variations in width of pulses.
No data on possible miscount every full rotation.


Is my assumption on index use wrong?
Does that mean that optical encoders are PERFECT devices that can not make mistakes that would accumulate?(I doubt about that)
Are there industrial controlers, or hobby brand controlers or homemade contolers that make use of index signal the way I have described it?



Thanks for any answers

[Al_The_Man]

To a great extent, incremental encoders are very reliable accurate source, especially when using differential drivers. There are some older controllers I have seen that has done a cyclical check of the marker against the total per rev. pulses. But the current use is mainly for accurate homing as you mentioned.
The typical explanation of the reading of the graticule head is mostly explained in a simplistic fashion. It is impossible for a LED source and sensor on their own to differentiate between the individual slots in a reading head that has a resolution of more than around 100p/rev. The technique to read very high resolutions is done using a method called the Moire effect.
This is acheived by a the reading head carrying a graticule that is slightly skewed to the main scale which results in the quadrature 'shutter' being much wider than the individual graduations. The rotation of the optical 'shutter' is at right angles to the direction of motion.
Al.

[medved]

There are some older controllers I have seen that has done a cyclical check of the marker against the total per rev. pulses.

Do you remember what brand were those controllers?


I am asking this question for two reasons: once to know how high-end mill systems achieve they accuracy, and secondly because I secretly hope to be able to build my own optical encoders one day. Anticipating "mine" encoders would not be that reliable, I would need to use Index pulse for error detection and correction. But I suppose that controller with encoder error correction capability has to be much more complex and more expensive.

But as a rule, do high-end systems prefer to use absolute encoders(or some hybrid"virtual absolute" encoders which have absoulte trace AND inceremental traces in-between; or optical encoders with analog sine output) or do they fully trust incremental encoders(use no error correction with index pulse)? Or they prefer to use magnetic resolvers?

The typical explanation of the reading of the graticule head is mostly explained in a simplistic fashion.

Well indeed in presentation sheets there are allways two nice drawings explaining working principle, one for interferential principle for super-high resolutions(which I dont dare to try to understand in detail, only in principle) and another for "imaging" principle. This picture looks suspiciously simple: colimated led + wheel + mask-reticle with 4 windows + 4 photocells.
They generously explain differential ("push-pull")circuit(connecting each pair of photodiodes to comparators) but they keep "tricks" like Moire solution for themselves. I am sure there are some more tricks.

If someone knows for more tricks, then please tell about them. On electronics side there must be some filtering solutions, and I think I saw some more complex differential circuits, comparing not just two photodiodes but connecting them with AND logic ladder to others - I suppose reason for that is a "transition point" where two light intensities of two photodiodes meet: one on the way to increase and another on the way to decrease, they must meet somewhere beeing "absolutely" equal for a fraction of a time -and there upredictable things can happen.




Although I think it is possible to read high-resoultion disks even without Moire effect, because diferential circuit is made to sense "statistical"average difference in light intensity between two SHADOWY areas(comprised of MANY stripes) and not to sense transitions of individual stripe... ...but I agree that at higher resolutions there must be a problem: differences in illumination between two windows being too small(both "equally gray") and changes too fast for photodiodes.
If I understand correctly, Moire soultion will create "illusion" of "virtual" stripes, which will be wider(and perpendicular) than actual stripes on optical wheel, hence they will have to "move" faster than actual stripes, but the difference in light intensity between two photodiodes will be greater and hence easyer to detect than it would be without Moire solution.





I dont have any encoders to play with, but has anyone tried to test them: connecting them directly to PC or microcontroler and meassuring CUMULATIVE error(during many hundred revolutions) by comparing actual and ideal counts between two index pulses? This test could tell how reliable particualr encoder really is.

[Al_The_Man]

As far as I am aware, most of the major CNC manufacturers have been using incremental encoders for many,many years, and if there was an inherent problem of eratic output, I am sure they would have adopted another method. It seems that the robotics industry has a tendency to use absolute encoders.
Alot of the major manufacturers now use high res (100,000p/rev) in a type of serial transfer method.
The problem with resolvers has been that although more reliable and less noise-prone, they are not capable of high resolution and at one time were geared up from the motor shaft for this reason.
I have not seen any report from current manuf. to indicate they do cyclical checks on encoder info.
The one I remember that did was an early Allen_Bradley system, that corrected the count if it did not agree at the marker pulse time.
You can visibly observe the moire shutter effect if you get hold of an old encoder that can be scrapped and expose the reading head area, by turning the shaft under a bright light, the 'shutter' rotation can be seen by eye.
As far as I know you only need 2 cells for quadrature detection.
Al.

[unterhaus]

Most electronics for reading encoders now includes digital filtering to avoid the noise. Look at the data sheet for the LSxxxx encoder counter at U.S. Digital. I don't think there really is a lot of trouble reading encoders nowadays, unless you're using a very simpleminded scheme.

A lot of motor drives don't even use the index pulse. I think those that do often just have it to pass through to a control that wants it.

[One of Many]

Hello

I am learning about servos, and as I searched the forums, the only contex I could find INDEX pulse was in connection with limit switches, that is, to enable more accurate homing using limit switch signal together with first index signal.

This would be correct in my experience, at least for industrial grade machines. During the homing function the servo is set to travel at a pace and direction to find the homing switch, then slow to continue on to see the marker pulse. The control then loads the position as an absolute envelope to the machines parameters. Most other moves are done incremental to an origin with a watchdog error handling if a postion is commanded outside the soft limits.

There are all types of error correction schemes. Although I have not heard of the marker being used like that. I have heard Anot and Bnot pulses used as a comparison to validate the A and B channels. If many in a run of 3 pulses are missed, a flag will be raised. That would indicate that the secondary channels are not just inversions of the A and B pulses, but independent sensors themselves.

A more common servo motion error handling scheme is dubbed Following Error. From what I gather, this has to do with commanded position verses actual position based on the velocity feedback of the tac in relation to tracking the frequency of the encoder pulses? It is monitored so as to maintain the error within a certain amount during a move to flag over run or delays. This may be a simple explanation, but it is along these lines.

DC

[Geof]

...During the homing function the servo is set to travel at a pace and direction to find the homing switch, then slow to continue on to see the marker pulse...

Just to be a bit picky you have it homing in the wrong direction. The machine moves positive until the home switch opens then moves negative very slowly until the home switch closes and looks for the first index pulse after the switch closes for the home position. If it doesn't find an index pulse within a certain number of encoder counts following the home switch closing it alarms out. If the home switch malfunctions the axis just moves very slowly to either hard limit and alarms out with a servo overload.

[One of Many]

Just to be a bit picky you have it homing in the wrong direction. The machine moves positive until the home switch opens then moves negative very slowly until the home switch closes and looks for the first index pulse after the switch closes for the home position. If it doesn't find an index pulse within a certain number of encoder counts following the home switch closing it alarms out. If the home switch malfunctions the axis just moves very slowly to either hard limit and alarms out with a servo overload.

Since we are being picky...... . So, I am not technically wrong, I just forgot a disclaimer .

I suspect that function is machine dependant. Our BP VMC only travel in a positive direction after the home switch is found, but does slow down and stop on the marker. These will fault out against a hard stop if no marker is found.

Our Milltronics do a reverese once the switch is made, then reverses again in a positive direction at a slower pace to find the marker pulse. These too will fault out against a hard stop if no marker is found.

I have yet to see either of these machines use the over travel limit switches if the pulse is missing. I think the home and OT switches are too close together to make this feasible.

Another system I worked on actually used the counts between the switch and the marker pulse to compensate an offset for the machine limits. I think this was done because the actual max travel of the axis was past the home switch in normal use. The marker pulse was setup to happen 1/2" prior to the hard stop. This system would error out if the marker pulse was not found within 1 revolution of the encoder after the home switch.

One gantry router we have reverses about 3-4 times between prox switches to marker pulses to square the bridge since it has 2 servos and 2 homing prox's on the x axis rack and pinions. That is a goofy dance, let me tell ya!

I am sure there are more homing sequence strategies, but to be fair I suppose a few should be mentioned to show there are no real standards in all applications.

DC

[Al_The_Man]

I am sure there are more homing sequence strategies, but to be fair I suppose a few should be mentioned to show there are no real standards in all applications.
DC

Your right, Nearly all the systems I have come across and implemented allow the system implementer to set up the homing sequence in several cenarios, move negative, move positive, come back off the limit, go past the limit, etc,etc.
Also the ability to grid-shift once the marker is found, effectively moving the marker pulse is another feature.
Al.