HP quadrature decoder?

Hey there.

I have also aquired a couple of HP printers with the very same sensors. One (Q9864) is used for the striped strip running behind the print-head

for position sensing, while the other (Q9845) is used in a similar fasion for a similarly encoded disk to gauge the paper position by the rotation of the rollers. I intend to use them for positioning in a CAD system, and have also been looking for datasheets without success.

I have decided that I will figure them out datasheet or not, as they're

too useful looking to be sat around doing nothing! I'll happily tell you what I learn if your still looking for information when I've done some fiddling about with the oscilloscope. So far I've only had a visual inspection of the components and contemplated their use. I'm fairly sure of the power connections from this first investigation, but can only guess at the function of the other two pins! I'd like to think one pulsed for left movement and the other for right... but that may be too much to hope! Perhaps it's more along the lines of the sensor system in a mouse, where extra circuitry is required to decode direction from the order of the beam interruption.

I have to show my ignorance and admit I have no idea what a 'quadrature' direction decoder is :) But I assume it's effectively a device that provides ready-decoded left and right outputs. Or perhaps it has some bearing on sensitivity??

so far what I have is:

_________ | | 1: -ve | | 2: ? | | 3: +ve ---------- --------- 4: ? |___________________| | | | | 1 2 3 4

Not sure of the voltage, but I'd guess 5v *crosses fingers*

Please let me know if you've got more information than this!

Thanks :)

"PolyVinalDistillate" wrote in news: snipped-for-privacy@e3g2000cwe.googlegroups.com:

You got it. It's not entirely ready as a simple one-bit flag might be for direction, but that's what it does. Not sure about this actual part, but it probably is one. Two sensors, placed so that slots making a 50% duty cycle will place a leading edge on one sensor when the second sensor is still 25% behind, hence quadrature, a phase shift of a quarter cycle. A circuit can test which sensor is trailing, regardless of speed, and get direction from that.

Of course :) I should've guessed that's where the quadrature came from!

*thinks back to his degree... QPSK and the like* Cheers for that :)

It could well be one of those, but the strip passing between the source and the sensor is encoded somewhere in the order of 1/4 to 1/2 mm, so it would need some fairly fancy optics within... The sensor is simply a chunk of silicon around 3mm square under a protective plastic cover. Looking at it now, it seems to me the only way it could possibly be a reliable component is if the source is a laser-diode, and the strip acts as a diffraction-grating, effectively being magnified by the formation of interference fringes that can be readily detected when they fall and move across the sensor. This would also mean there's a possibility that the intervals the sensor can detect would be smaller than those on the strip - but now I'm really just guessing :)

"PolyVinalDistillate" wrote in news: snipped-for-privacy@b28g2000cwb.googlegroups.com:

Without seeing it I can't say for sure, but two things to think about:

1. The sensors don't have to be at 0.25mm apart to read this regular pattern, they could be 1.25, 2.25, so long as that fraction is correct.
2. If you look at a cheap ball-type mouse you'll see slotted wheels that are that finely spaced, it's not expensive tech now.

Usually the sensor surface will have two tiny slots in it. If yours really is a single window, it might be something different. Maybe a two-element detector that sees the shadow passing between two halves, determining direction that way.

I have used the sensors from these mice in the past, and figured out circuitry to decode direction. However with such fine spacing on the encoded strip with the printer sensors, and the 3mm gap between the sensor and the source, I think there would be issues with light diffraction were the light not coherent. By the time light from a noncoherent source reached the sensor through the strip, it is quite believeable that the encoding would be lost and undetectable, though I can't be sure without a proper test of the strip back in the optics labs to gague it's pitch properly.Under magnifying lens and with the calipers I counted 5 lines per mm.

I also studied the sensor surface under the magnifying lens, and it appears to be made up of many sensors on the same axis as the encoded strip. suggesting it may well be more advanced than simply a two-element sensor. Certainly there are not two slots on it's surface.

Cheers for the thoughts, you're pushing me closer to digging out a PSU and firing up the ol' valve-oscilloscope to do some probing of a more electrical nature!

"PolyVinalDistillate" wrote in news: snipped-for-privacy@h48g2000cwc.googlegroups.com:

Any motivation to use an oscilloscope is good motivation. :)

That sensor is interesting me a lot now, if it has several sensors built in. It's possible they depend on equal spacing to make the same detection as each other, in a way to reduce noise by taking a mean average or something. Also, I don't know if the light has to be coherent, it might be enough to be very narrow bandwidth. Some LED's are almost monochromatic, I had a few from Hewlett Packard that were like that. Coherence can bring troubles of its own, mainly speckle, which dramatically increases noise.

I've had the sensor hooked up to 5v. Get a pretty, dull red glow from the source, and pins 2 and 4 are indeed quadrature outputs almost exactly 90 degrees out of phase (judging from the poor trace on my valve oscilloscope). Direction encoded like so:

Strip Direction:1->2->3->4. 2 is 90 degrees before 4

The outputs appear to be TTL compatible (I have hooked them up to the pushbutton circuit to a little micro for pulse counting through a

74LS148), and the device hasn't burned out on 5v :) I've also rigged up the vernier caliper with the strip running through the device. Due to my setup, I only have 1cm translation in which to measure pulses, but after moving back and fourth 4 times through 4 cm, I've clocked up 237 pulses. This works out to around 0.17 mm per pulse, and a 50/50 duty judging from the oscilloscope readout. The quadrature output seems to push the accuracy to under 0.1mm... I like these parts, they have potential :)

I'd overlooked speckle... though I'm not sure of it's impact were there a narrow enough beam of light emmitted... Cretainly it appears to be general illumination in the near-infrared that is emitting from the source, and not a beam, which would fit in with it being relatively monochromatic as opposed to coherent. Were it coherent, I'd have expected the sensor to be yet more sensitive :)

Looks like I've got enough information to use the sensor myself now - hope it's useful to others cannibalising HP inkjets!

And just to ensure my pinout is useful, here it is again (even though it'll be all out of alignment when it's posted):

Top View

______ | -> | Phase1 leads Phase2 90deg |

"PolyVinalDistillate" wrote in news: snipped-for-privacy@b28g2000cwb.googlegroups.com:

Nice, basic part, a good module to have. It rules out the more complex business of handling analog differences, as you say it's already close to TTL level switching.

Coherence would have been a real problem I think, not a benefit. The narrower the beam, the worse it gets. If you have a focussable laser module, try focussing to a pinpoint at two feet, then drawing the spot across a white wall, or even a fine white ceramic to show how even the finest surfaces make huge noise. The jumps in apparent brightness will be fierce, far worse than if you spread the light in a 3mm spot and try the same thing.

I recognise it now, I've seen them in various boards I've salvaged in the past. Some have the sensors on one side, as in this case, others have a slot between two projections to work by transmission instead of relection, and I think most have the same 4 wire connecting plan.

I've no such laser module at my disposal to experiment with, but I imagine you're correct. It would make sense :)

The only versions of these sensors I've seen before the inkjets were the same style as those in ball-mice, with two clearly visible slits and one light emitter. These ones are far snazzier than any of the others in my collection! Also, these are the first I've had that are so friendly they even save you the trouble of a couple of transistors and schmidtt triggers to convert to TTL :)

I have thrown together a little decoder for the sensor to give me left/right pulses at minimum spacing possible. This has given me an accuracy of 600 devisions-per-inch (approx. 1/24 mm). The kind of number I'd have expected if I thought about it for any length of time :)

I think that's all there is to say about this little device now, except mebbe 'I want more of them' :) Thanks for all the motivation to stop me being a lazy arse, and actually bother to figure it out! I'm awfully pleased - it's just the thing I wanted for a CAD Milling machine. Now I just have to build the milling machine... heh

"PolyVinalDistillate" wrote in news: snipped-for-privacy@m7g2000cwm.googlegroups.com:

It keeps me going too. Got to draw on anything I can to motivate me, on some things... I'm trying to re-aqquaint myself with some trigonometry to rotate an array of points. Interesting distractions are as helpful as trying to focus, and definitely more fun.

Nice Information,

I have found a datasheet of a similar device from the same manofacturer (Agilent). Search for Agilent HEDS-9710. Those that PolYVinalDistillate wrote is true. This is a link of the datasheet: