Stingray Robot Kit (#28980) and Wheel Encoders

[GordonMcComb]

Quadrature assumes two signals 90 degrees out of phase, and these are more like two 50% duty cycle clocking signals with a very slight delay between them. You could still derive direction information by noting which clock transition arrives first.

With these gear motors, it's highly unlikely you'll not know what direction they are turning unless they are being forcibly back-driven. In these instances, quadrature adds value because 1) it increase the effective resolution of the disc by either 2X or 4X (depending on how you count the transitions) and 2) it allows you to better discern between valid and invalid transitions. If the disc is sitting right at a transition the output of the optical vane could stutter and send long streams of pulses, even though the bot is not moving. You know the transitions are not valid if the opposite channel is also not changing.

The disc with 36 "fins" will also not quite produce equal mark/space signals, as the distance between the fins is larger than the width of the fins. This is probably okay as you can compensate in software (or the software just doesn't care). It *could* be a problem in hardware quad decoders, like the ones US Digital sells. Their chips are pretty tolerant of timing delays, and they include internal pulse stretchers. These discs are probably okay, but if the mark/space ratio is ever too asymmetrical, the chips no longer perceive the A/B channels as quadrature, and the output is erroneous. Since these discs are being done on a laser there's always the opportunity to fine-tune their geometry, should it be the case that hardware decoders can't handle them. But again, the spacing of the 36-fin wheel is probably close enough to do the job.

-- Gordon

[Matt Gilliland]

Gordon's exactly right. I'll add to what he said, to clarify even more....

The OLD Encoder had two separate sensor modules that were thru-hole soldered after the boards had gone through the Surface Mount. We were able to space the distance between the two sensor "beams" to be exactly the width of each fin (see the pic above). The space between two fins (the "valley") was the same distance as well. This yielded a perfect (as perfect can be) 90 degree out of phase signal pair.

Now, look at the Stingray. Although the "wings" are (should be) the same width as the "valleys", it doesn't matter, because the Single Surface Mount Sensor that we're using has the "two eyes" that are really close together. What this means is that we'd have to have the Stingray disc with fins that are VERY, VERY, TINY (probably rectangular slots), and LOTS of them. Now, since this is attached to the drive motor (very fast), that would be a huge amount of pulses for very little actual distance traversed. As it stands, at full motor voltage the second wing trips approximately 50uS after the first wing does.

Also, because the fins would have to be symmetrically spaced, and very close to each other, the sizes of the "valleys" would be super-tiny, and the "fins" would be really thin - easily bent or broken.

But yes, you'll still get direction information.

Now to the Madeusa/Eddie platforms. The encoders are on the drive-wheel output shaft. Since the quadrature output is properly produced when we have the valleys and fins spaced the same as the distance between the two sensors, and in this case our two sensors are VERY close together, the fins and valleys would be VERY tiny as well - not mechanically reliable. (Especially because these are not enclosed devices).

Now, that being said, and like Gordon suggested, knowing your speed and the necessary "correction factor", you could derive a very close value for the quadrature data in code.

Also, when you factor in the backlash within the drive-motors, we'd have way more granular data than would be necessary due to the 1/2" of "slop" in the drive system itself.

Either way, you'll still see "staggered pulses" that tell you which direction, and of course the rate that you're bot is traveling - make the calculations and you'll certainly be close enough to not care - simply because of the backlash (which in this case cannot be mitigated).

Yeah, more than I wanted to know too.
-MattG

[W9GFO]

I've been working on a quadrature encoder to use with my PropBOE-Bot. It uses a slotted disc and modified QTI sensors. What I have noticed is that the holes need to be slightly wider in order for the "on" time to be equal with the "off" time. Because of the way the reflective sensor works it "sees" the holes for a shorter period of time.

[Matt Gilliland]

W9- Kind of like painting with a spray can - "over-spray"!
-MattG

[Lev]

I am confused about Matt's post #18. The 12VDC replacement motor (lower rpm) does not have an extended motor shaft for the encoder. Will the motor spec be changed?

http://www.parallax.com/Store/Robots...8/Default.aspx

[Chris Savage]

I am confused about Matt's post #18. The 12VDC replacement motor (lower rpm) does not have an extended motor shaft for the encoder. Will the motor spec be changed?

http://www.parallax.com/Store/Robots...8/Default.aspx

That motor is being replaced with one that does have the motor shaft extended out, we just haven't received the correct samples yet.

[GordonMcComb]

I've been working on a quadrature encoder to use with my PropBOE-Bot. It uses a slotted disc and modified QTI sensors. What I have noticed is that the holes need to be slightly wider in order for the "on" time to be equal with the "off" time. Because of the way the reflective sensor works it "sees" the holes for a shorter period of time.

I'm guessing the discs are fairly thick (1.5mm is thick for diffraction effects), and maybe the optical alignment isn't quite perfect. That and if the LED source isn't terribly bright, I can well see that the light going spends more time below 50%, resulting in longer off times (assuming a 50% threshold). Making the slots/holes larger could work, but it seems to me you can't undo a bigger hole, should the design need changing. What are you using to detect low/high values? I've found comparators to offer better flexibility than just relying on the threshold values of input pins. More parts, though.

Matt is using fairly thin acetal. I could never cut that thin on my CNC, but it's child's play on a laser. And that's exactly what's going on -- Matt is like a child playing with that laser machine!

-- Gordon

[Matt Gilliland]

Matt is like a child playing ...

I resemble that remark.
O' course that comment's coming from an old guy that still plays with robots...ya' gotta consider the source.

-MattG

[W9GFO]

I'm guessing the discs are fairly thick (1.5mm is thick for diffraction effects), and maybe the optical alignment isn't quite perfect.

The discs (wheels) are 5mm crystal clear acrylic and the slots are about 1.7mm, with about 1.5mm between. It is the on time that is longer, hence larger holes to increase the off time. The alignment will likely never be perfect since the sensor is diffused and sits four slots widths away. That would only affect the phasing anyway, not the on/off time balance.

[Whit]

This is exciting news! Great work guys...

[Matt Gilliland]

Here's the latest pic for the new encoder.

As shown, the same encoder is used on each Stingray wheel, it's just reversed. The mounting brackets and hardware will be available as a separate kit.

The Encoders will be sold separately, without any mounting hardware. Then there will be a "Mounting Kit" for the Madeusa & Eddie robots, and another mounting kit for the Stingray (as depicted here). Who knows what other robot mounting kits we can come up with?

This is the prototype set in the pic. A *lucky*(?) customer will be getting this prototype in a couple of days to "test drive" with no documentation, no directions, and no code samples, and...and...let's see, what else can't I give them, to make it a real challenge?



-MattG

[Whit]

This is the prototype set in the pic. A *lucky*(?) customer will be getting this prototype in a couple of days to "test drive" with no documentation, no directions, and no code samples, and...and...let's see, what else can't I give them, to make it a real challenge?
-MattG

Sounds like a job for Phil or erco - but then again there is no wood...

[Roger Lee]

Matt,

Nice photograph. That picture explains it all.

That one (or one very similar ) should go on the product page.

[Ttailspin]

but then again there is no wood...

There will be...


-Tommy

[Stempile]

Here's the latest pic for the new encoder.

As shown, the same encoder is used on each Stingray wheel, it's just reversed. The mounting brackets and hardware will be available as a separate kit.

-MattG

On the StrayRay mounting kit as pictured you have a little block of plastic for each sensor. What was the consideration for that over just a couple stand-offs?

ms

[Matt Gilliland]

but then again there is no wood

So you can rest-assured that ol' cellulose boy is OUT of the running...

Actually, I'm not sure that the "secret customer" is even on these Forums. The recipient is going to be encouraged to give us some updates here, as he "plays with it".

-MattG

There will be...

et tu Tomme? Then short-circuit MattG!

[Matt Gilliland]

On the StrayRay mounting kit as pictured you have a little block of plastic for each sensor. What was the consideration for that over just a couple stand-offs?

Great question Stempile.

My desire for any "retrofit" to a Legacy Product (like the Stingray), is to come up with ways in which there is no need for any modifications to the original Product.

Normally, standoffs would certainly be easier, but the hole locations on the Encoder Board weren't quite right for alignment to any of the existing holes on the Stingray chassis.

The Encoder's first order of priority is to be a "drop-in" replacement for the original Position Controller that is currently on the Madeusa and Eddie platforms. I wanted to leverage the design effort to have the same Encoder footprint work in the Stingray too.

So, the block was a necessary evil because it accommodates the offset of the Encoder mounting holes with the "stock holes" already present on the Stingray -

Viola! No drilling of new mounting holes required

-MattG

PS: Yes, I know I spelled a musical instrument

[Stempile]

Oh, yeah... DUH. Chris said that at the start of this thread. Forgot. Thanks.

[RobotWorkshop]

Hello Matt,

The encoder disks look pretty thick. How much clearance is there between the encoder and the sensor? If there isn't much then if there is any end play in the motor shaft that can become an issue. I've run into some in the past where a sensor looked ok but when the motor ran back and forth the end of the shaft would move in and out a but and cause the encoder wheel to touch the sensor. I always check that when installing them to ensure there is proper clearance.

Robert

[Chris Savage]

Oh, yeah... DUH. Chris said that at the start of this thread. Forgot. Thanks.

That's okay...as Matt G. can verify, much of what I say gets forgotten.