Rickys High Power Flexable motor controller

[liveforphysics]

I'm a bit annoyed today....

Was planning on ordering a few parts from digikey today but as of 4:30am some twat (putting it very politely) tried to use my credit card number on 3 or 4 overseas websites all declined (between $90 and $520 ) but the bank fraud detection immediately locked down my card. It will be 3 - 5 days to I get a new one Grrrrr!!! so basically due to weekends that probably puts the parts I wanted back a week based on typical delivery times from digikey

At least there is other things to work on so not a major issue but always annoying when something interrupts your plan of attack.

Damn thieves.

You are welcome to use my card for buying parts if you want (PM me if you want the number). If I can't trust someone I've never met who is on the other side of the world with my credit card buying info, who can I trust? The enjoyment of excitedly watching this controller develop from the edge of my seat is well worth whatever parts cost you're going to buy.

 

[Ricky_nz]

Damn thieves.

You are welcome to use my card for buying parts if you want (PM me if you want the number). If I can't trust someone I've never met who is on the other side of the world with my credit card buying info, who can I trust? The enjoyment of excitedly watching this controller develop from the edge of my seat is well worth whatever parts cost you're going to buy.

Thanks for the offer Luke,
Not a problem, I calmed down a bit and worked out that if I get the new card by the end of the week which is highly likely I should get the parts a few days later than I had planned but that won't make much difference as I should still get them before the weekend the week after I order and I don't get much time to play with electronics during the week, after a day in front of 3 monitors its not a good thing to come home and solder SMD parts :lol:.

The worst one is If I get delayed till our monday as that pushes every thing back and usually a digikey order on a monday doesn't arrive to the next week here. Sometimes the day difference is a help but in this case not.
Even then I do have other things to work on.

I try to collect up a list of parts as the most economic amount of money to spend at digikey in one go is between $125NZ (Free shipping) and around $350NZ (avoids tax and import processing charges).
I've already though of a couple more parts so I'll just wait till the card arrives.
I've got lots of software to do so It won't hold me up.

Its definitely annoying how some criminal anywhere in the world can be such a pain but it is reassuring that the bank caught it .

Damn now I will have to memorise another card number too many things toi remember.
It was IT's regular forced password update time for me last week at work. Both Windows LAN and SAP(barf).
I think I have the new ones memorised now finally.
On the subject of cooperate IT., why does it take them 3 levels of support operators to solve a mis-configured IE auto logon crap and trash your windows profile and have it only work for 1 day when I solved the problem with 1/2hour of google with profile intact?

 

[Alan B]

Sorry to hear about your problem.

Not sure if you use PayPal, but they can issue a "soft" credit card number via the web interface that can be used for online purchases. I think it is good for just a single purchase, so it cannot be "stolen". I have not used this but it is one way to avoid the problem. Even using PayPal directly avoids the issue since then the store never has your credit card info. So you only have to trust one company with your CC info instead of every one you do business with.

Do you have any idea where you card number "leaked" out? I read the other day that some major bicycle website in UK had lost a bunch of CC numbers.

Great progress on the controller!

 

[Ricky_nz]

Sorry to hear about your problem.

Not sure if you use PayPal, but they can issue a "soft" credit card number via the web interface that can be used for online purchases. I think it is good for just a single purchase, so it cannot be "stolen". I have not used this but it is one way to avoid the problem. Even using PayPal directly avoids the issue since then the store never has your credit card info. So you only have to trust one company with your CC info instead of every one you do business with.

I do use pay pal to buy stuff. I had a quick look arround and couldn't find any mention of the soft credit card feature you describe. Maybe it isn't available out of the US or I looked in the wrong place.
It makes sense though. I wish our banks would offer features like that.

Do you have any idea where you card number "leaked" out? I read the other day that some major bicycle website in UK had lost a bunch of CC numbers.

No I don't know where it leaked out but I had recently brought a few things from chainreactioncycles and a google search turns up that they might have had a leak.

I have been using that card number on the internet for >10 years so It could be one of a large numeber of sites that got hacked. (my pc is fine, no trogens etc).

 

[Ricky_nz]

Finally got the new card . Had a couple of online purchases backed up waiting for it :lol:.

Looks like my bank is finally rolling out the smart card ones as its a smart card. My last card was issued in the middle of last year and it was the old type.
Might have to do with the recent forced upgrade of all eftpos machines in NZ to ones that to support the smart cards in time for the rugby world cup.

By the time I ordered I think it was too late for Digikey so probably get the parts in early April as I never get them the week ordered (Time skew with US and since I missed the window I might as well have ordered on Tuesday next week ).
Oh well its not urgent parts. Just some parts that are kind of optional on the controller but would be nice to fit that I forgot in the last order and a few parts to play with.
I ordered some linear hall sensors to play around with for my brake levers. I've got some nice strong small magnets that I should be able to epoxy onto the underside of the lever if I can find a good place for the sensor with reasonable output voltage range with the magnet mounted there.
As long As I can get a change in voltage within the ADC input range the software can deal with any offsets and the range.

As a side note I heard back from CRC after I sent them an email, they actually rung me, I missed the call but they left a message. Sounds like they found the leak of the credit card numbers and have fixed it. They are bribing those affected with online vouchers.

If anyone else brought from Chain Reaction Cycles this year you might want to check your credit card statements carefully! Fraud detection caught mine but there are stories of others only noticing after they checked their statements.

 

[Alan B]

Glad to hear that CRC has found it, I've been wanting to order something from them.

On the PayPal self-issued credit card numbers for online use, they appear to have suspended that program in 2010. Not sure it was available outside the US. No information about why or if they will restart it. There is software that generates the numbers so perhaps there was a security breach of that.

Interesting about using linear Halls on the brake levers and using variable thresholds. That would be great if they don't drift, or are easy to normalize. I suppose every startup they could be re-determined.

 

[Ricky_nz]

Glad to hear that CRC has found it, I've been wanting to order something from them.

Their description seemed a bit lame "sophisticated real time attack". My guess is this is probably marketing terms for the webserver got hacked and information was siphoned off before the card processing.
I don't have a copy of the email here but I did see a message in a google search that basically came up with its contents.

On the PayPal self-issued credit card numbers for online use, they appear to have suspended that program in 2010. Not sure it was available outside the US. No information about why or if they will restart it. There is software that generates the numbers so perhaps there was a security breach of that.

Perhaps, it would be one of those high value targets for hackers I guess. Pity though as it sounded like a good feature.

Interesting about using linear Halls on the brake levers and using variable thresholds. That would be great if they don't drift, or are easy to normalize. I suppose every startup they could be re-determined.

Yes drift is a concern. I figure the throttles don't drift too much so hopefully it shouldn't be too bad.
A power up recalibration is probably a good way to handle any drift although they suggest a drift in gain with temperature.
I will have to have a play and see how much variation the is, perhaps heating the sensor up to force drift etc.
It may be possible to track drift in real time if the whole sense range shifts. High side limit getting bigger of low side limit getting smaller.

The datasheets seem to suggest 3% error in magnetic strength reading so might not be too bad. They do suggest a shift over the temperature range but I guess our typical temperature range would be -10C to say 40C. a small fraction of the sensors rated -40C to +80C/+125C (depends on version). Ok some people live in snow but they would just have a shifted version so an autocal at start would probbaly work as the temperature is not likley to change too far during a ride.

I guess a lot comes down to making sure that there is a reasonable range of magnetic strength over the leaver movment to keep the signal out of the noise.

I would really like to avoid a user having to move the lever full its full range of motion every time they power on the bike so maybe just check the released position and try and estimate if it has drifted and which way.

 

[rhitee05]

I remember reading somewhere that placing the Hall sensor between a pair of magnets will make the response more linear. I don't remember the original source, but this Allegro app note mentions it:

http://www.allegromicro.com/en/Products/Design/an/an27702.pdf

 

[amberwolf]

That explains why the ScootNGo throttle I had used two magnets at opposite ends of the throttle range.

 

[Ricky_nz]

http://www.allegromicro.com/en/Products/Design/an/an27702.pdf

Thanks for the link, Some good information in there.

I'm not sure I can arrange 2 magnets with minimal changes to the brake leaver as I don't want to mount the hall sensor on the leaver. Maybe 2 magnets on either side of the leaver moving by the sensor I guess but might be hard to hide. I want to try and keep the hall sensor and magnets reasonable tidy and not too noticable if possible. I can't modify the brake leaver much as its part of my avid elixer R hydraulic brakes.

 

[Alan B]

Linearity may not be important here anyway, and if so you can always do that in software.

 

[Ricky_nz]

Linearity may not be important here anyway, and if so you can always do that in software.

yes, software can fix a lot of hardware problems, just don't tell the hardware engineers :lol:.


Also for those interested here is a link to the outcome of the chaine reaction cycles fraud issue
http://www.theregister.co.uk/2011/03/25/chain_reaction_fraud_follow_up/

 

[rhitee05]

I actually don't think linearity is all that important either way. Your brain is pretty good at compensating for such things, so long as the response isn't too twitchy. A more useful aspect of using two magnets is probably that it seems to let you use a wider range of mechanical motion, and you get a larger signal over that range. You can compensate for both of those, too, but it would probably make the action smoother and easier to control.

Depending on how your bike is setup, perhaps you could put the magnet(s) and sensor somewhere on the cable? Put the sensor right at the end of the cable sheath and somehow attach a magnet to the cable itself, for example.

 

[Ricky_nz]

As to linearity and magnets I'll have a play when I get the sensors to see what works.
Youre right the users brain should adapt. Still want to try and achive somthing that is reasonably smooth and predictable/natural to use especially when the disc brake begins to be activated in parallel with the regen.

Depending on how your bike is setup, perhaps you could put the magnet(s) and sensor somewhere on the cable? Put the sensor right at the end of the cable sheath and somehow attach a magnet to the cable itself, for example.

Since I want to do it on a bike with hydraulic brakes I think it has to go at the leaver unless I get hold of some dot 4 compatible pressure sensors and manage to plumb them in. Probably possible but not cheep and would have difficulty detecting very light leaver application.

Other news:
I have started assembling a second power board as a spare. This time I soldered down the SMD 1uF caps first as adding them after the FETS on the first board was really a pain due to access and the amount of heat that was needed. A few on the first board ended up not flat against the board but electrically well connected.

I'm not sure which FETs I'll fit yet,
I have both IRFP4368 (75V) and IRFP4468 (100V). The first board has IRFP4368 (75V) devices.
I guess I could build the second with the 100V devices although a 12FET 100V is probably way overkill for the 80-100 180Kv motor :lol:. If pushed to maximum RPM at that level I'm sure the core losses would be rather high. Maybe it would be entertaining to cook my GM 1KW hub although probably not the safest as the bike its on has v brakes and a notchey headset bearing.
On the other hand a second 75V version would be good in case I explode the first one pushing the current up. Remember I have kept it down to approx 100A RMS/phase at this point of testing but it definitely has more up its sleeve .

A 100V version would have slightly higher losses in the FETS due to the slightly higher on resistance.
IRFP4368 (75V) typ Ron = 1.46mR
IRFP4468 (100V) typ Ron = 2mR

I should probably order a few more gate drive chips so I can build up a third board or repair one of the first two if something goes wrong.

I really need to go and buy so aluminium and try and build up the reduction for my new bike so I can get the 80-100 on a bike.

 

[Ricky_nz]

Did a bit more work on the 2nd power board this evening.
This time I have decided to try a double thickness of RG58CU Braid to build up the board.
The first board used a single layer.
There is enough clearance for this with the same PCB stand-offs and extra copper is good.
It looks thicker and I am happier with it. I still need to do the other bus and add some braid across to the terminal.
I used solder paste under each piece of braid this time to get things started and as I warmed the board up I filled in with normal solder. Seems pretty solid and nothing is too high so it won't interfere with the FET placement.
It goes pretty easy once the board warms up a bit, my 70W iron was fine. As the board heats the large copper areas hold a lot of heat. I probably should have dug out an old electric frypan I have in the shed with stuffed teflon to preheat the board. Might try that one next time but even without preheating it went pretty fast.


On the bus done there is 3 doubled up runs of braid so 6 runs of braid from one end of the board to the other in parallel with 4 ounce copper and 3 ounce copper except for the small gap around the mounting hole.

Doing this board with twice the copper used to beef it up will be an interesting comparison because the board is warmed by the FETS as well as from the copper so hard to tell where the heat comes from most.

Also looks like my digikey package arrived in Auckland this morning and appears to have made it through customs so all things going well I'll have it tomorrow . If so that would be record time from order processed email from digikey to delivery for me.

Now time to wash any lead residue off my fingers and get something to eat before bed.
I don't muck around with that "lead free crap", real solder for me.

 

[mdd0127]

I can't wait until you get these ready for sale! I love your ideas so far.


I did just have a thought when looking at your beefed up traces I thought I'd throw out there though. Not that there's anything wrong at all with building them up that way. My controller is done the same way. I just thought that your electronics and architecture deserved a little prettier trace solution.

What about getting copper plates machined in the same shape as the board traces with a bunch of little holes drilled in them for the components and down the middle to allow the solder to flow through and stick? You could heat the whole piece up off the board and tin it, flow a puddle along the trace, set the plate on, and heat until it all flows together. If the board allowed for it, it could even bolt on and avoid the soldering altogether.

Probably been done before and more effort than it's worth. It would be neat though. Just a thought.

Now I need to find a cad drawing of a 12 fet infineon board and get some copper plate coming......for now at least :wink:

 

[Ricky_nz]

I did just have a thought when looking at your beefed up traces I thought I'd throw out there though. Not that there's anything wrong at all with building them up that way. My controller is done the same way. I just thought that your electronics and architecture deserved a little prettier trace solution.

Yep it would be nice to use copper.

What about getting copper plates machined in the same shape as the board traces with a bunch of little holes drilled in them for the components and down the middle to allow the solder to flow through and stick? You could heat the whole piece up off the board and tin it, flow a puddle along the trace, set the plate on, and heat until it all flows together. If the board allowed for it, it could even bolt on and avoid the soldering altogether.

Probably been done before and more effort than it's worth. It would be neat though. Just a thought.

Copper would allow the bolt termintions to the board to be better too .

I had thought that instead of copper added to the bottom of the board a simple copper strip with legs could be soldered to the top side either side of the DC bus film caps but it would need to be bent to get to to the DC input terminals. (Might upload a sketch later)

I think the soldering of the copper to the bottom of the board board may require some form of jig to ensure correct placment although if solder paste placed between the copper and the PCB and then a couple of tacks of normal solder on the edge to hold it in place may allow the whole board + copper to be heated to the melting point of the solder paste allowing the copper to be completely soldered. The paste melts at a lower temperature.

One risk with copper soldered onto one side of the board is it can cause the board to bow. I noticed a small bow on the board I just soldered all that braid to, not much but definitly want to do that before the SMD components are fitted.

How well does copper machine?

Now I need to find a cad drawing of a 12 fet infineon board and get some copper plate coming......for now at least :wink:

If you can't find a cad drawing you can always try scanning the bottom of the PCB (my old hp scanner hapially focus several mm above the glass. You could clean up the scanned image and then get a vector outline of it.
It will be interesting to see how you get on machining the copper for the infineon board.

I'm already thinking about another power board design that would just need simple copper strips with minimal or no machining but it would be a single phase board so there would be 3 required (stacked). Haven't thought about all the details yet like how to get the DC bus to all three boards nicely. I've got more software to do before to start that one though :lol:.

EDIT:
My parts arrived . Digikey processed them early Tuesday our time and I have them by 4pm Friday. Not bad since it takes a day for packages to go from Auckland to Napier and be delivered. Thats the fastest delivery I have had from digikey. It only stopped in aussie for about an hour rather than the worst I have had of nearly 2 days stuck there. Its got the parts I need to finally finish assembling the first power board so I can bolt it down to its heatsink for hopefully the last time and some parts I needed to finish assembling the second board. Its also got some linear hall sensors etc.

 

[Ricky_nz]

Oops. damn probe slipped while probing the aux power supply late last night .
I was running the whole thing off a 30V 2.5A bench supply at the time.

After much concern at the rising cloud of smoke I discovered that a tantalum crowbar device worked well and managed to protect the rest of the controller and power electronics.
The device in question was a 20V 10uF SMD tantalum capacitor. Sure put out a lot of smoke :lol:.

I disconnected the power supply section from the rest, replaced the capacitor and powered up off a bench supply and its all good, no damage apart from the power supply .

Now I just need to figure out what I fried in the PSU. There is not much too it. 2 FETs, a diode, the controller chip and a bunch of passives. The controller chip is currently outputting 4.3V where it should be 7V for its VCC so its either the chip or the diode connected t that pin. I think its the LP5116 thats toast though . I'm sure its going to suck removing that from the PCB as its got one of those thermal pads under it soldered to the board (LM5116 20 pin TSSOP EP).

When I'm certain its the fault I'll cut its legs off and heat from the back of the board through the vias and hopefully be able to pluck the chip off with a pair of tweezers.

I really need to choose a clamp device for the 15V rail but I haven't found one that meets my requirements and is available in single quantities yet.

I do have some fuses to put inline with the converter now that came in my digikey order.

On the plus side I have assembled a bit more of the second power board today. I will order a heatsink for it on monday.

 

[amberwolf]

The device in question was a 20V 10uF SMD tantalum capacitor. Sure put out a lot of smoke :lol:.

I'm pretty sure that tantalum caps violate some law of physics in how much mass is converted to smoke vs how much they started with. :lol:

 

[salty9]

Conservation of mass is not violated. You have to add oxygen and maybe nitrogen but the reaction is probably not energetic enough so that argon can be neglected

 

[rhitee05]

I really need to choose a clamp device for the 15V rail but I haven't found one that meets my requirements and is available in single quantities yet.

A TVS diode is probably a good fit for your requirements. They're fast, can absorb moderate energy spikes, and are pretty inexpensive. You could add an MOV as well for an extra degree of protection, they are slower but can absorb more energy (also pretty cheap).

The problem with finding a clamping device would be how much room you have between the nominal and max voltages. If you're trying to clamp the 15V rail and you need to protect devices that can tolerate say 18V, that will be tough. For a 15V rail, you'd probably want to select a TVS with 17V or 18V clamping voltage and 20V or so max clamping voltage. That wouldn't provide much protection to those 18V devices.

If you tell us your specific requirements, we might be able to help you find a solution.

 

[Ricky_nz]

A TVS diode is probably a good fit for your requirements. They're fast, can absorb moderate energy spikes, and are pretty inexpensive. You could add an MOV as well for an extra degree of protection, they are slower but can absorb more energy (also pretty cheap).

Yep, I had been looking for a crowbar type device but there is much less choice of them available just because I wanted to limit the power dissipated in the device but since there is a fuse in series the TVS clamp diodes probably will handle it for the time required and there does seem to be a much larger range of them available in small quantities.

The problem with finding a clamping device would be how much room you have between the nominal and max voltages. If you're trying to clamp the 15V rail and you need to protect devices that can tolerate say 18V, that will be tough. For a 15V rail, you'd probably want to select a TVS with 17V or 18V clamping voltage and 20V or so max clamping voltage. That wouldn't provide much protection to those 18V devices.

If you tell us your specific requirements, we might be able to help you find a solution.

I can set the converter output between 12 and 15V. Its used for gate drive and small accessories such as LED's etc. I'm currently running it at 12.5V which seems adequate. I have an absolute max of 18V for the linear regulator on the control board. (This is the lowest voltage rating of anything on the rail).
The output of the converter is max 3A and the input supply range is from 20V - 100V so I'll probably fuse that at 2.5A.
There are 2 failure modes I want to protect against:
a) MOSFET failure (short) full input voltage to output
b) open regulation loop.
I imagine the open voltage loop could cause the fuse to blow quickly due to increased power requirments into a TVS clamp diode.

I found these Digikey SMCJ13CABCT-ND http://search.digikey.com/scripts/DkSearch/dksus.dll?Detail&name=SMCJ13CABCT-ND Bourns SMCJ13CA
this morning. These are nice in that their package size allows my to replace one unnecessary capacitor on the rail with it avoiding a PCB change.
I don't think they can keep the voltage below 18V under large currents though.
I have accidentally taken out 5V regulator due to 20V in from bench supply and the 5V rail was ok (Regulator turned into a resistor).
Maybe I can find a better regulator with the same pinout as the ZLDO1117G50TA though.
The other devices susceptible to over voltage are IRS21864 gate drives, 20V rated with internal clamps.

 

[rhitee05]

You could try something like the On Semi 1SMA11AT3. It's technically an 11 V device, but you can operate it right on the edge in exchange for a lower clamping voltage. The nominal spec is 1 mA @ breakdown voltage of 12.85 V, so the DC losses should still be pretty small. The spec range is 12.2-13.5 V, though, so it's possible you could end up on the wrong end of that spread and higher loss. The clamping voltage is 18.2 V at 22 A, so it should be able to blow your fuse before anything else.

A slightly better solution would be if you could adjust your rail voltage down slightly to 12 V instead of 12.5 V and find a 5 V supply that would take at least 20 V like the gate drivers. Reducing the rail voltage should keep you clear of the TVS breakdown range, and finding a more robust regulator would allow the TVS to clamp almost 2 V before anything reaches its absolute max spec. Obviously you'd want to make sure your fuse of choice is sized so it blows before the TVS will pop. This TVS is rated 400W for 1 ms, you can get the larger packages rated for 1500W if it seems necessary.

You could also help out the TVS by placing a relatively low value resistor in series between the 12V supply and the TVS. The idea would be to choose a value so that the drop is small for nominal current draw, but large enough to drop a couple of volts or so when the TVS starts to clamp.

 

[Ricky_nz]

A slightly better solution would be if you could adjust your rail voltage down slightly to 12 V instead of 12.5 V and find a 5 V supply that would take at least 20 V like the gate drivers. Reducing the rail voltage should keep you clear of the TVS breakdown range, and finding a more robust regulator would allow the TVS to clamp almost 2 V before anything reaches its absolute max spec. Obviously you'd want to make sure your fuse of choice is sized so it blows before the TVS will pop. This TVS is rated 400W for 1 ms, you can get the larger packages rated for 1500W if it seems necessary.

Thanks,
Just found a different 5V regulator that is rated for 20V the NCV1117ST50T3G from on semi. Looks pin compatible so a simple swap. Actually the same basic part number but from a different supplier with higher maximum input voltage :lol:. I can definitely dial the voltage down a tad. I have had no problems with 12.5V gate drive so down to 12 shouldn't cause any issues Don't want to go lower as I want things like motor bike LED indicators and stop lights to work off the supply as well.

I'll add some TVS diodes to my list of parts to order along with the better voltage regulators. That basically puts every component on that rail at 20V rated or higher. I think some of the caps are 20V etc.

I did plan on getting round to the protection for the converter but nothing accelerates it like a close call :lol:.
I ran out of time today to remove the chip today. I was working on one of my bikes though.

 

[rhitee05]

I think that should work out for you just fine. Like I said earlier, you could add an MOV in parallel with the TVS to absorb more energy, but it doesn't seem like that should be necessary for you here. TVS diodes are very handy little parts.