PDA

View Full Version : Solved Mysterious rotation sensor disk



Ron Czapala
09-04-2011, 03:50 AM
I had an old Canon iP4000 inkjet printer that quit working.
Thinking that I might be able to scavange some stepper motors and other parts, I took it apart.

I was surprised to see that there were no steppers - just various regular DC permanent magnet motors and IR/photoresistor sensors to precisely move the paper and inkjet carriage.

There are transparent disks glued to the gears and sensors positioned to monitor the track on the disks.

The strange part is that the track appears to be consistently the same darkness instead of alternate light and dark segments.

So how can this setup work???

http://forums.parallax.com/attachment.php?attachmentid=84807


84807

Beau Schwabe
09-04-2011, 04:00 AM
Is the grey area on the disk polarized? ... look at a crystal wine glass through it ... does it have a rainbow or prism effect... have any of the newer 3d glasses handy? not the red/blue, the ones that are polarized. Try looking through them and rotating your disc.


EDIT: a Google search for that part number in the Photo indicates a 'FILM, TIMING SLIT DISK" ... there may be fine marks in the grey area that you can not see. It produces a very fine quadrature signal to the receiver.

Ron Czapala
09-04-2011, 04:20 AM
Is the grey area on the disk polarized? ... look at a crystal wine glass through it ... does it have a rainbow or prism effect... have any of the newer 3d glasses handy? not the red/blue, the ones that are polarized. Try looking through them and rotating your disc.


EDIT: a Google search for that part number in the Photo indicates a 'FILM, TIMING SLIT DISK" ... there may be fine marks in the grey area that you can not see.

I looked at using a bright six Sony IR video camcorder light and could not detect any variation.
I thought maybe it had a dye that reacted to infrared light...

I also looked a crystal wine glass thru it - couldn't see anything there.

Curious...

RobotWorkshop
09-04-2011, 04:25 AM
Most of the HP inkjet printers have similar encoder disks as well. For the carriage that goes back and forth with the ink cartridges there is usually a flat strip with the same markings that work as a quadrature encoder for that axis. Back in July 2008 I used one of these in a SERVO article on encoder matching. It is a nice example of a high resolution quadrature encoder. As I recall you only need to supply +v, ground, and then you can read each channel right off the sensor. If you run the sensor at 5V and connect it to a Propeller then I'd probably add a 2.7K or something similar in between the sensor and prop pins.

Robert

Beau Schwabe
09-04-2011, 04:27 AM
Check my update EDIT... When I google it, it appears to be a timing disk. I can find tons of places that are willing to sell it for as low as $1, but I can't find a data sheet on it.

If you have a flat bed scanner, place it on the scanner and set it for the finest resolution... I bet then you can see the marks... maybe.

Lawson
09-04-2011, 04:34 AM
Any ordinary magnifying glass and a Digi-cam should help resolving the lines on that disk. The magnifying glass will let the camera focus closer than it otherwise would be able to. Otherwise, take two of the disks and set the outer gray strips roughly on top of each other. A course pattern of light and dark areas should form and move around as one disk is shifted relative to the other.

Lawson

Ron Czapala
09-04-2011, 04:37 AM
Most of the HP inkjet printers have similar encoder disks as well. For the carriage that goes back and forth with the ink cartridges there is usually a flat strip with the same markings that work as a quadrature encoder for that axis. Back in July 2008 I used one of these in a SERVO article on encoder matching. It is a nice example of a high resolution quadrature encoder. As I recall you only need to supply +v, ground, and then you can read each channel right off the sensor. If you run the sensor at 5V and connect it to a Propeller then I'd probably add a 2.7K or something similar in between the sensor and prop pins.

Robert

I did not find a strip type encoder. The sensors only have three wires - ground, v+ for the emitter and the wire to the photo sensor. If it was a quadrature setup it would need two emitter-sensor pairs.

It must really be high resolution! I need a microscope...

RobotWorkshop
09-04-2011, 04:48 AM
Hello Ron,

That disk looks just like the ones HP uses. Maybe Cannon is using something different. Do you have any pictures of the sensor that has the slot for this encoder disk? Maybe they were only using one channel. Very odd.

Robert

kwinn
09-04-2011, 04:58 AM
That is a high resolution encoder disk and a mask with the same density of lines is placed on front of the detector. When the disk lines are aligned with the mask lines the IR gets through the clear sections to the detector. When the disk lines are aligned with clear mask areas little or no IR gets through.

Since the same circuitry that reads the pulses from the encoder also controls the motor it does not need a quadrature signal to tell it what direction the motor is going.

Ron Czapala
09-04-2011, 05:01 AM
Check my update EDIT... When I google it, it appears to be a timing disk. I can find tons of places that are willing to sell it for as low as $1, but I can't find a data sheet on it.

If you have a flat bed scanner, place it on the scanner and set it for the finest resolution... I bet then you can see the marks... maybe.

I was thinking that exact thing before I saw your reply.

I set my scanner software to scan Positive Film and removed the back panel so the light would shine thru from the scanner cover.

I had a hard time getting the scanner software to recognize the "film" but finally scanned it at 1600 dpi.

You can see the lines now! Wow!!

http://forums.parallax.com/attachment.php?attachmentid=84808

84808

Beau Schwabe
09-04-2011, 06:48 AM
"When the disk lines are aligned with the mask lines the IR gets through the clear sections to the detector. When the disk lines are aligned with clear mask areas little or no IR gets through." - Exactly !! ... you can 'see' a similar effect when you have two perforated circuit boards and hold them up to the light to look at the 'holes' slightly twisting one board over the other.

"Since the same circuitry that reads the pulses from the encoder also controls the motor it does not need a quadrature signal to tell it what direction the motor is going." - True, but the same sensor can have two detectors built into one unit and have a single 'mask' with two sections one section purposefully aligned 90 deg out of Phase. The result is a high precision quadrature output ... available when it's needed. depends on the application.

Loopy Byteloose
09-04-2011, 07:09 AM
Several years ago, I managed to damage my HP inkjet printer by knocking it off a desk and decided to salvage parts. It did not have a disk, but had a strip of similar 'micro-lines' that positioned the carriage.

It seems that a standard IR LED and sensor can handle such high resolution with the right firmware.

Gadgetman
09-04-2011, 11:37 AM
You should take a look at HPs 'Large Format' printers, such as the Designjet series of inkjet plotters. The sensor strip is a thin steel strip mounted behind the carriage. There's actually warnings in the manual about touching the strip as it's SHARP!

Heater.
09-04-2011, 12:42 PM
I do believe that you guys are talking about Moiré pattern. As described here:http://en.wikipedia.org/wiki/Moir%C3%A9_pattern

kwinn
09-04-2011, 05:45 PM
@Heater, it may be that they are using the encoder disk and detector mask to produce Moiré patterns but it is more likely that the lines on the disk or strip are parallel rather than offset 5 degrees. The active area of the sensor is larger than the line spacing of the disk so if the mask were not there it would pick up the average intensity of several lines and clear areas.

Over the years I have taken apart many optical encoders and for those where the detector could be removed without disturbing the mask the lines of the mask were always aligned to block or pass the IR, not to generate a Moiré pattern.

Beau Schwabe
09-04-2011, 05:50 PM
Heater,

"I do believe that you guys are talking about Moiré pattern." - Yes, also termed a 'picket fence effect' which would fall under the definition of two overlapping patterns that are the same pattern, however one is a slightly different size. Fore-shortening allows you to sometimes observe this phenomenon when driving down a country road. Regularly planted crops such as corn can also create similar interference patterns to the observer.

kwinn,

The 5 Deg is just used as an example... even at 0 Deg, the block or pass is still considered a Moiré pattern. It's actually not to far of a stretch to how Radio works when you heterodyne the desired signal you want to receive with the local oscillator to create a 'beat frequency'

GordonMcComb
09-04-2011, 06:07 PM
Double-grated encoders don't have to use moire, though many do. Some use something called the Lau effect, and there are others (Talbot is an interesting alternative; I used a variation some 20 years ago to demonstrate a distance measuring technique that's been used in optical disc players more than a decade before that -- nothing is new, you know).

Whatever the actual sensing these double-graters rely on diffraction and fringe effects to measure optical displacement. Because of the fineness of the optics, the encoder discs are virtually useless without their corresponding sensor. Good luck getting any information on the sensor.

-- Gordon

Heater.
09-04-2011, 07:34 PM
Ron,
Have you tried shining a laser pointer through that grid? Just wondering if it's fine enough to get a diffraction pattern out of.

Ron Czapala
09-04-2011, 09:38 PM
I stuck the disk onto a hard drive spindle mounted in a block of wood.
Then I made some small aluminum brackets and mounted the sensor next to it.
I identified the IR emitter leads ans have it working but I haven't hooked up the sensor yet.

The sensor has transparent windows on the top and bottom halves so you can see thru it.
There is a 100 ohm resistor in series with the IR emitter and capacitor
(which looks like a resistor with colored rings like a resistor) across the power and ground.

The sensor has four connections GND, V+ and two others I assume to be the sensor output(s).

More later...

http://forums.parallax.com/attachment.php?attachmentid=84814


84814

84813

RobotWorkshop
09-04-2011, 09:55 PM
That sensor looks just like the ones I've pulled out the HP printers. There are four wires going there. I suspect that is a regular quadrature encoder.

Ron Czapala
09-04-2011, 10:01 PM
That sensor looks just like the ones I've pulled out the HP printers. There are four wires going there. I suspect that is a regular quadrature encoder.

It might be - I was looking at some other sensors when I stated that there were only three wires - there were actually five optical sensors in the printer.

GordonMcComb
09-04-2011, 11:23 PM
With the (apparent) high resolution of the disc I'm thinking it's probably not a standard slit optical photocoupler. Diffractive sensors are capable of single-channel or quadrature outputs, just like regular photocouplers, but instead of detecting light-dark transitions they work by analyzing the differences caused by fringing.

Look up polaroid diffractive encoder and you'll find numerous patents from the late 80s onward regarding this stuff.

The industrial side of Canon makes/made a sensor of this type, but using a laser for its collimated beam. They are capable of over 75,000 transitions per revolution. Maybe there's a small picowatt laser in that thing, though more likely it's just an LED. In the high-end Canon encoders the output is a bi-phase sine wave -- a bit more work to interface than standard digital quadrature pulses.

This may be neither here nor there, but as yours was a Canon printer it makes sense they're using Canon sensors.

-- Gordon

Ron Czapala
09-05-2011, 12:19 AM
With the (apparent) high resolution of the disc I'm thinking it's probably not a standard slit optical photocoupler. Diffractive sensors are capable of single-channel or quadrature outputs, just like regular photocouplers, but instead of detecting light-dark transitions they work by analyzing the differences caused by fringing.

Look up polaroid diffractive encoder and you'll find numerous patents from the late 80s onward regarding this stuff.

The industrial side of Canon makes/made a sensor of this type, but using a laser for its collimated beam. They are capable of over 75,000 transitions per revolution. Maybe there's a small picowatt laser in that thing, though more likely it's just an LED. In the high-end Canon encoders the output is a bi-phase sine wave -- a bit more work to interface than standard digital quadrature pulses.

This may be neither here nor there, but as yours was a Canon printer it makes sense they're using Canon sensors.

-- Gordon

Gordon,
I think you may be right.

I usually use my digital camera to view IR emitters since the camera sensor will pick up the IR frequencies, but I could actually see a red glow in the sensors with the naked eye - probably an LED.

I don't think I'm going to pursue it now. I was mainly interested in disk's characteristics.

I had no idea that such a high resolution was possible with these devices.

The other two disks are lower resolution - the lines are clearly visible.
In fact, they are marked 150 lpi - 150 lines-per-inch I assume...

I think I was trying to use the wrong sensor with the high resolution disk. Two of the modules are identical so they probably match the low resolution disks.

Thanks for everyone's input!
- Ron

Ron Czapala
09-05-2011, 01:34 AM
Here is an oscilloscope image of the two output channels.
I turned the disk slowly by hand and grabbed the image.

It does look like a quadrature encoder

http://forums.parallax.com/attachment.php?attachmentid=84822

84822

GordonMcComb
09-05-2011, 02:04 AM
Looks like the right sensor for the disc. At the resolution of the disc, if it were a standard slit sensor the output would be erratic at best, because many stripes would be visible through the slot. It's also clear you're seeing conditioned pulses, not signals straight off the detectors.

If you can see a red glow it could be a near-IR laser -- they're not that expensive these days -- or it could be an LED with a very narrow output wavelength. The more coherent the light output the better. My money would be on a laser. They're actually cheaper in the long run.

Though I can't imagine many robotic applications would benefit from this resolution you've got a pretty nice encoder there!

-- Gordon

Ron Czapala
09-05-2011, 02:33 AM
Looks like the right sensor for the disc. At the resolution of the disc, if it were a standard slit sensor the output would be erratic at best, because many stripes would be visible through the slot. It's also clear you're seeing conditioned pulses, not signals straight off the detectors.

If you can see a red glow it could be a near-IR laser -- they're not that expensive these days -- or it could be an LED with a very narrow output wavelength. The more coherent the light output the better. My money would be on a laser. They're actually cheaper in the long run.

Though I can't imagine many robotic applications would benefit from this resolution you've got a pretty nice encoder there!

-- Gordon

I used a BS2 to count pulses on one channel over a 2 second interval and tried to do 1 revolution.

It looks like approx 2500 pulses in 360 degrees!

$WMc%
09-05-2011, 03:12 AM
Nice find Ron
'
I have the need for a high resolution encoder.
'
Now I know where to find one.(I have a couple old printers laying around I kept for parts)
'
Thanks for the thread and info.

GordonMcComb
09-05-2011, 04:01 AM
It looks like approx 2500 pulses in 360 degrees!

With quadrature that'll resolve to 10,000 cpr. Great for your inkjet printing bot!

-- Gordon

kwinn
09-05-2011, 11:14 PM
With quadrature that'll resolve to 10,000 cpr. Great for your inkjet printing bot!

-- Gordon

That may be a bit of overkill for a robot wheel sensor but sounds great for a precision cnc system.

Duane Degn
09-05-2011, 11:31 PM
I was thinking it would be great to use with a laser and camera combo. You might be able to get more than just one distance point by moving the laser and keeping track of were the dot shows up on the camera. Or you could precisely move the camera laser combo.

I'm going to start looking through some of my old printers for these. Thanks for posting the information.

Duane

Mark_T
09-06-2011, 12:19 AM
@Heater, it may be that they are using the encoder disk and detector mask to produce Moiré patterns but it is more likely that the lines on the disk or strip are parallel rather than offset 5 degrees. The active area of the sensor is larger than the line spacing of the disk so if the mask were not there it would pick up the average intensity of several lines and clear areas.

Over the years I have taken apart many optical encoders and for those where the detector could be removed without disturbing the mask the lines of the mask were always aligned to block or pass the IR, not to generate a Moiré pattern.

The obvious way to use this technique is have the sensor grating at a slightly different spacing from the encoder disk - then the Moiré pattern can be quadrature encoded by placing the sensors 1/4 wavelength apart at the difference (spatial) frequency.

If the sensor's grating were at the same spacing you would lose direction information and miscount on direction changes.

kwinn
09-06-2011, 03:04 AM
The obvious way to use this technique is have the sensor grating at a slightly different spacing from the encoder disk - then the Moiré pattern can be quadrature encoded by placing the sensors 1/4 wavelength apart at the difference (spatial) frequency.

If the sensor's grating were at the same spacing you would lose direction information and miscount on direction changes.

True, but this also requires high precision in placing the sensors. A lot of equipment does not require that level of precision. For the lower resolution (100/200 lines/inch) encoders I have dismantled the encoder disks or strips and the detector masks look like the attached diagram. Can't say with 100% certainty that the sensor grating spacing was exactly the same as the disk or strip but it certainly looked like it was.

I have made an optical encoder using a laser printer and transparencies with 64 lines per revolution and equal spacing for the disk and sensor grating. It worked well for several years.