Caveman optical sensor RC controller concept.

[liveforphysics]

Don't look too closely at the drawings, because they aren't to scale or timed or anything. Just crude drawings to comunicate the concept.

Figure out the pole location of your motor, and work out the timing sequence. Once you've worked out what needs to switch when, print out a sticker that wraps around the motor with 6 lines of dashes. Each line of dashes represents when one of the 6 fet banks will switch on. It could be clear coated over, or you could print something like bumpersticker material or whatever it takes to make durable lines. Even having a vinyl sticker layed over it as a mask, and spray painting on some industrial high-temp flat black paint, and pealing off the vinyl or a hundred other ways to get dark/light dashes on a surface.



Next, drill a little scrap of plastic or whatever and mount 6 high speed photo-diodes in a row with the correct spacing to read the lines. Have each photodiode be an input signal to a 6 channel fet driver H-bridge chip that has built-in no pass-through logic.




For throttle control, simply make the common positive to the photo-diodes come from a PWM duty-cycle % that ranges from 0% to 100% depending on throttle position. You could also have a current shunt circuit that can attenuate throttle duty cycle % if you want a current controlled setup, but that would be getting kinda fussy. I like the idea better of just using a massive FET stage and letting current do as it pleases.

Any thoughts? Reasons why this wouldn't work well? Seems like it would provide outstanding control, starting torque, etc. No ability to lose sync, and nothing fussy to fail. Maybe need to wipe off the motor if you spray it with mud or something, but that woudl be easy enough to fix, or install a little spash shield.

[gwhy!]

no reason why this shouldn't work, and it would simplify mounting the sensors as opposed to using halls , But i think you will be better off using just 3 index rows instead of 6 and using white/reflective index marks on a black background. The only potential problem that i can see is the speed at which the opto's will work at they may not be fast enough.

Edit: just re-read the reasoning behind the 6 rows.... nice.

[TylerDurden]

Nice.

Perhaps the diodes could be doubled (vertically) for redundancy? The bands would need to be taller, but not much.

That might be overkill, but any additional fault-tolerance could be a benefit on a massive power setup.

[Thud]

any reason simple led's couldn't provide illumination & move the whole set up inside the can? (assuming you had a custom application for such a device )

[liveforphysics]

Nice.

Perhaps the diodes could be doubled (vertically) for redundancy? The bands would need to be taller, but not much.

That might be overkill, but any additional fault-tolerance could be a benefit on a massive power setup.

It would be easy enough to do double rows, and use "OR" or "AND" logic chips to make sure it has both. Seems like with a little splash shield it would be easy enough to stay reliable. The 6 channel H-bridge FET driver would have logic in it to prevent a possible pass-through event.

[liveforphysics]

no reason why this shouldn't work, and it would simplify mounting the sensors as opposed to using halls , But i think you will be better off using just 3 index rows instead of 6 and using white/reflective index marks on a black background. The only potential problem that i can see is the speed at which the opto's will work at they may not be fast enough.

Edit: just re-read the reasoning behind the 6 rows.... nice.

The 6 rows does make it damn slick. Takes all the brains out of the controller design.

The diodes I was looking at came in 4nS and 9nS types. That's extremely fast. Switching speed would not be a problem.

any reason simple led's couldn't provide illumination & move the whole set up inside the can? (assuming you had a custom application for such a device )

They make little buggers with 3 pins. Common ground, positive leg for the built-in LED, and third leg as the signal out. It seems like it would be easy-peasy to mount them on anything.

[liveforphysics]

Just had another idea. Run 12 rows, 12 sensors, and a pair of 6-channel FET stages, and make it a 6-phase. If done correctly, a 6-phase design can provide the 3-phase torque of wye with the speed of delta.

I think for a trial run, just 6 rows and conventional termination would be a good idea though, just to keep things simple and prove the concept.

[gwhy!]

Just had another idea. Run 12 rows, 12 sensors, and a pair of 6-channel FET stages, and make it a 6-phase. If done correctly, a 6-phase design can provide the 3-phase torque of wye with the speed of delta.

I think for a trial run, just 6 rows and conventional termination would be a good idea though, just to keep things simple and prove the concept.

getting a bit ahead of yourself.. but yes should work. I was just thinking also it would be a very easy way to use a off the shelf sensored controller. Print a sticker out for your desired motor and add a one point adjustable mounting sensor block to your motor mount then jobs a good'un.. as hall sensors seem to frighten quite a few people ( I dont know why )

[swbluto]

Interesting. At first, I assumed the sensors went out to a conventional sensored controller so it'd be another alternative to hall sensors (And could be used as possible redundancy for reliability), but I then I further read that they'd actually control the fet gates directly, so the brains could be eliminated for a bare-bones system.

You'd still have some logical interaction in the controller for interacting with the throttle, but it sounds like this might be a nice way to go about it. What are the specific advantages of this over a hall sensored setup?

[gwhy!]

Interesting. At first, I assumed the sensors went out to a conventional sensored controller so it'd be another alternative to hall sensors (And could be used as possible redundancy for reliability), but I then I further read that they'd actually control the fet gates directly, so the brains could be eliminated for a bare-bones system.

You'd still have some logical interaction in the controller for interacting with the throttle, but it sounds like this might be a nice way to go about it. What are the specific advantages of this over a hall sensored setup?

The main advantage is mounting the sensors. With halls they will have to cover at least 120 degrees around the motor can and need to be movable in unison where as using the sticker/opto solution the sensors need only cover the width of 1 sensor ( 1 mounting post ) so it make the hardware so much more manageable.

[vanilla ice]

Another thing, iirc usually optical sensors have better resolution than hall sensors.. you may even see some small efficiency and/or performance gains.

[Jonathan in Hiram]

Another advantage is that you can play with the timing simply by peeling off one sticker and putting on another.

You would want to put a permanent zero timing mark on the motor can, a scratch or dimple or something like that.

[gwhy!]

Another advantage is that you can play with the timing simply by peeling off one sticker and putting on another.

You would want to put a permanent zero timing mark on the motor can, a scratch or dimple or something like that.

you will need to make the sensor post adjustable by a couple of degrees for fine tuning. but yes replace the sticker for playing around with pulse widths.

[GGoodrum]

I love the simplicity. I wonder how hard it would be to apply this to an inrunner design, like the AstroFlight 32xx. Could the stripes be arcs on a disc? If so, everything might fit inside the can.

-- Gary

[fechter]

I've seen some motors with optical sensors. Usually they use an interrupter disc, but I don't see why a reflective sensor wouldn't work. Just need to shade it so it doesn't screw up when it's in the sun. I think you could do it with just 3 sensors and use a standard sensored controller. I bet you could run the optical sensors directly to the hall wires (with some resistors). If your pattern was black against a shiny metal background, I don't think you'd need reflective paint. Timing could be adjusted by moving the senor mount.

A similar approach would be to use a disc rather than a drum for the pattern. This may allow for mounting the encoder on the motor shaft externally or on the end face of the rotor.

With the 6 sensor approach and the right FETs, you could essentially turn a brushless motor into the equivalent of a brushed motor by using the sensors and FETs for commutation. This is like the eCycle silicon commutated motor. http://www.ecycle.com/motorgenerator.html With this setup, the commutation is done separately from the PWM and you just use a (cheap) brushed motor controller. Of course it would make sense to just gate the high side FETs with the PWM signal to avoid the losses in a separate controller. In this kind of a setup, I think you'd want to add some kind of logic latch that would prevent shoot through if a sensor screwed up.

[spinningmagnets]

I am a newb to electronics, and my old brain is slow, but...

I believe gas prices will spike up again, and every time that happens, hub-kits get quickly sold out and back-ordered. I believe non-hubs will eventually provide an E-bike solution that is better and cheaper than the most common hub-kits. Towards that end, I think the next year here on ES will result in open-source plans for a DIY-motor that is scaleable in size and power.

Though I am no expert by any means, I think the axial-flux motor with optical sensors will end up the best DIY by far. Removing hall sensors from the inside of a hub might mean adding external trigger-magnets, which could then accumulate debris. Optical sensors have no problem being external.

I have been assured by an RC enthusiast that his very affordable DIY optical sensors ran a motor at 6,000-rpm with no switch timing issues.

In humid weather, moisture is almost impossible to keep out of the inside of a hall-sensored motor, and on occasion I recall someone posting that his halls shorted out. Though I "think" DIY halls could be potted, having external sensors completely eliminate any brushless sensor heat issues.

External photo-sensors should be easy to make completely rain-storm ridable. I have heard of an LED shining light onto a white/black decal, which reflected back onto a photo-sensor. Also a disc and caliper arrangement where the light shined through holes in the disc. Apparently both types worked well.

I can't afford a proper RC build yet, and I won't have my own shop again for about a year, but...I CAN build a DIY axial-flux with optical sensors inside an apartment, and I think a lot of other people can too.

Whether adding sensors to an RC motor or a DIY axial-flux, I think optical sensors will end up being the hot ticket.

[Kburn77]

I really like the idea of a interupter/encoder style disc, here's some really cool optical sensors that would be easy to implement on an encoder disc. http://www.jameco.com/webapp/wcs/stores ... eyCode=PDF
it uses a 5v supply but maybe the controller already uses 5v for hall sensors?

[Jonathan in Hiram]

The trick to getting phototransistors to respond quickly is making the load impedance as low as possible. One way to do that is using an op amp as a current to voltage converter, with a current to voltage converter, the load impedance is essentially zero.

An LM3900 or an LM324 quad pack op amp will work, the 3900 has a bit better specs but requires a bit more external circuitry, they both will work on a single supply. It's been at least a decade since I fooled with this stuff, I'm sure there are better choices out there today.

[vanilla ice]

A splash guard thingy could help with sunlight too. I know the sun screwed with our optical rear wheel slip sensors at work.

[John in CR]

LFP,
It sounds great. There's something about the hall sensored commutation that just doesn't seem right at take off, so a more positive control over commutation may prove very beneficial. Also, going with a brushed type commutation without the brushes may prove to be a real winner too, because firing all the coils all the time instead of only 1/3 or 2/3rds has to be a more effective use of the copper. Cheap labor down here too, hint, hint.
John

[fechter]

Optical sensors will be just as prone to moisture damage as a hall sensor. Either one should be fully submergible if properly sealed.

On the other hand, I can think of a few motors that got hot enough to fry the hall sensors, which are typically located right in the windings. By using optical sensors outside and away from the heat generating windings it should be immune to that failure.

I'm not sure if they would need a buffer amp. That might help. You might be able to do something real easy with a logic inverter chip. This would also speed up the swtiching.

Here's one that Mouser has for $0.76 ea. http://www.vishay.com/docs/83760/tcrt5000.pdf
That's cheaper than a hall sensor. I bet you could interface those directly with a "standard" brushless controller by just adding a few resistors.

A few other thoughts:
The disk or drum with the pattern needs to be fairly large in diameter to maintain precise timing with a reflective sensor. You might be able to use a single pattern and stagger the senors to get the right output phasing. This is the equivalent of how hall sensors would be placed. This would require a greater spacing between the sensors. With a single pattern, the dark and light sections would correspond exactly to the north and south magnet poles.

Keeping the sun out is essential, and not such an easy thing to do in some cases.

The power consumption of the IR emitters is quite a bit higher than your typical hall sensor, so the 5v supply may be strained. The forward voltage of the emitters is low enough that you could possibly put 3 in series and still run them off 5v, which would cut the current draw by a factor of 3. I think you'd want to run them somewhere between 10-20ma.

I think it make sense to try this with a regular brushless controller. The cave man commutation scheme would be great if you had to make the whole controller from scratch, but it' so much easier to just get one already built (I know, it takes half the fun out of it).

[liveforphysics]

I've seen some motors with optical sensors. Usually they use an interrupter disc, but I don't see why a reflective sensor wouldn't work. Just need to shade it so it doesn't screw up when it's in the sun. A bit of felt draped over the sensor holder and glued to be snug up against the motor should solve ambient light problems I think you could do it with just 3 sensors and use a standard sensored controller.I'm not all that impressed with TO-220 fet packages used in all other controllers, not too impressed with poor use of space in board layout and weak traces. I bet you could run the optical sensors directly to the hall wires (with some resistors). If your pattern was black against a shiny metal background, I don't think you'd need reflective paint. Timing could be adjusted by moving the senor mount. Yep, that's what I currently do with my external hall sensor retro-fit RC outrunner.

A similar approach would be to use a disc rather than a drum for the pattern. This may allow for mounting the encoder on the motor shaft externally or on the end face of the rotor. More needless complexity IMO.

With the 6 sensor approach and the right FETs, you could essentially turn a brushless motor into the equivalent of a brushed motor by using the sensors and FETs for commutation. This is like the eCycle silicon commutated motor. http://www.ecycle.com/motorgenerator.html With this setup, the commutation is done separately from the PWM and you just use a (cheap) brushed motor controller. Of course it would make sense to just gate the high side FETs with the PWM signal to avoid the losses in a separate controller. In this kind of a setup, I think you'd want to add some kind of logic latch that would prevent shoot through if a sensor screwed up. Yep, hence the part where I suggested driving a 6 channel H-bridge fet driver with built-in no-passthrough protection.

[kfong]

Luke,

Cool idea, optos should have no problem in this application. There are already opto sensors that have transmitter and receiver pairs so it will reflect on the surface easily. If you know what the hall effect outputs are, it might be just an easy substitution. This optocoupler uses plain paper as the contrast media. Would be easy to create a pattern on the motor and secure it with clear tape or paint. You would need to know if the motor algorithm uses edge or pulse width detection. Keep the sensor on the underside or on top so it gets less sunlight, so as long as there is good contrast it should work out fine. It already has a daylight filter.

http://search.digikey.com/scripts/DkSea ... -1239-2-ND

With the large motors it would be easy to create a circuit board with a strip of these. Let me know if you plan to pursue this. I can help by making a simple 2 layer circuit board with my milling machine. Just attach it length wise to your motor and print out a pattern.

Kin

[kenkad]

Kfong,
Thank you for the Osram SFH 9240 reference. I was planning to use the RPR-220, which has a slower response and no Schmidt trigger. I have changed my layout to use the SFH 9240. Again thanks. I have enclosed a JPG of how I will be mounting my motor controller to the motor shell and sensing the reflective spots on the rotor disk. You can see the reflective spots in the circular slots in the JPG. I also plan to be able to adjust the timing by rotating the motor (+/- 10 degrees) controller on the motor shell. The motor controller (as shown) is setup to allow up to six 3phase drives. This is a long term research project for me. I have not started to cut any material yet. I still have some small details to work out. I have provided this JPG simply to give others ideas they can implement in their designs. Keep up the good discussion.
kenkad

[deVries]

November 29, 2009...

I've seen some motors with optical sensors. Usually they use an interrupter disc, but I don't see why a reflective sensor wouldn't work. Just need to shade it so it doesn't screw up when it's in the sun. A bit of felt draped over the sensor holder and glued to be snug up against the motor should solve ambient light problems I think you could do it with just 3 sensors and use a standard sensored controller.I'm not all that impressed with TO-220 fet packages used in all other controllers, not too impressed with poor use of space in board layout and weak traces. I bet you could run the optical sensors directly to the hall wires (with some resistors). If your pattern was black against a shiny metal background, I don't think you'd need reflective paint. Timing could be adjusted by moving the senor mount. Yep, that's what I currently do with my external hall sensor retro-fit RC outrunner.

A similar approach would be to use a disc rather than a drum for the pattern. This may allow for mounting the encoder on the motor shaft externally or on the end face of the rotor. More needless complexity IMO.

With the 6 sensor approach and the right FETs, you could essentially turn a brushless motor into the equivalent of a brushed motor by using the sensors and FETs for commutation. This is like the eCycle silicon commutated motor. http://www.ecycle.com/motorgenerator.html With this setup, the commutation is done separately from the PWM and you just use a (cheap) brushed motor controller. Of course it would make sense to just gate the high side FETs with the PWM signal to avoid the losses in a separate controller. In this kind of a setup, I think you'd want to add some kind of logic latch that would prevent shoot through if a sensor screwed up. Yep, hence the part where I suggested driving a 6 channel H-bridge fet driver with built-in no-passthrough protection.

Luke,

Cool idea, optos should have no problem in this application. There are already opto sensors that have transmitter and receiver pairs so it will reflect on the surface easily. If you know what the hall effect outputs are, it might be just an easy substitution. This optocoupler uses plain paper as the contrast media. Would be easy to create a pattern on the motor and secure it with clear tape or paint. You would need to know if the motor algorithm uses edge or pulse width detection. Keep the sensor on the underside or on top so it gets less sunlight, so as long as there is good contrast it should work out fine. It already has a daylight filter.

http://search.digikey.com/scripts/DkSea ... -1239-2-ND

With the large motors it would be easy to create a circuit board with a strip of these. Let me know if you plan to pursue this. I can help by making a simple 2 layer circuit board with my milling machine. Just attach it length wise to your motor and print out a pattern.

Kin

Bump.

Has anyone pursued this optical sensor-controller to get a working model yet?

Or, does someone have an ETA when this might happen for RC motors?

TIA