oldswamm
100 W
<edit of 2/2, changed drawings><and again on 2/6 and 2/8>
<edited Jan 31 2011> (First post after 'completing' hardware design. Any suggestions would still be easy to implement.)
<edit of 2/8/11> General description.
It's a 3 board stack. The bottom board is the MPU. The next board up is the driver board, and also contains the high voltage to driver voltage (12-15) regulator, which can also supply nearly 5A externally for lights and other circuitry, such as the handlebar unit. The top board is designated the FET board. It contains all the high current parts. FETs, low esr caps, current shunt(s). The largish heatsinks in the sketch should only be needed for sustained high current use (up mountains at full throttle, or pulling trucks?)
Sketch:
Designed so anyone who can successfully build a kit should be able to assemble a kit based on these boards if they read and follow directions (all semi conductors through hole, and SMT are relatively large.)
Hi,
In this design, the FET board is for hi current, period. I assumed from the start that copper wire would have to be soldered to the board to carry the current! (A 6oz trace would have to be 1.5" wide to carry the current I'm hoping for). The driver board will have the high voltage to 12-15v regulator, as well as provision for a very wide variety of drive currents to suit any applicable FET. All along I’ve been trying to design a set of boards that could be built into a WIDE variety of controllers. Versatility is the second objective after ease of assembly.
I’ve decided on the Microchip dsPIC30f4011. Why? It’s the most powerful through hole processor I could find with a minimal motor controller. The 40 pin version because the 28 pin simply doesn’t have enough in/out (they use 8 of the 28 for pwr/gnd). 30 mips. 24 bit program, 16 bit data. One cycle 16bit mem to 40 bit accumulator multiply (using 2 data busses), with provision for a barrel roll, and ‘programmable’ rounding to a 16 bit register. 4 parallel sample and holds with 1mhz ADC. It can monitor phase current at 120khz and phase voltage at 200khz, at the same time. Can do 80khz PWM at 8 bit resolution. It might not be fast enough for SOME motors, but FEW would need a faster processor.
There IS a free c compiler. Look for ‘student edition’. I’m still a student?! Really 60 day ‘shareware’. After 60 days you lose use of 3 modules. I think the linker, simulator, and one other, but not the compiler.
In an effort to maximize current capability, the drain leads are cut flush with the transistor case, and all drain current flows through the ‘tab’/heat spreader. The wide part of the Source lead is installed solid to the copper trace as suggested by lfp, and there is provision for copper wire soldered directly to them and the trace. I’m assuming, at least for high current, that the high side batt lead, and phase leads would be connected to the appropriate heat spreaders rather than the circuit board. Any heat produced at the connections would then pass directly to the heat spreaders rather than the pc board. The gate lead passes right on through to the driver board. I’m told this can’t be done, but not why not…. I see lots of designs that have longer traces than that extra 12mm of gate lead (which in my builds, will have a ferrite bead, thanks, lfp).
Anyway, here’s my latest. Drawn with FreePCB. I fought with killyourselfCAD for 2 days, then downloaded FreePCB, and had a drawing in a few hours.
The copper, top and bottom, for all 3 boards, and the FET board, with and without soldermask cutouts and silk screen. The copper graphics are 5% resized from the output of GerbMagic, and the ones with the SM cutouts and SS are screen captures from the same program.
View attachment 7
There is a ‘buss bar’ of up to 1/16 X 5/8†copper soldered from the first current sensor ‘output’ to the ‘input’ side of the other 2 phase sensors. The (12ga?) wire that wraps around the 8ga negative battery lead, is soldered to both sides of the current sensor, then can be bent up to form a short 1/4†to 5/8" loop, which can then be soldered to the ‘buss bar’, for a potentially higher hardware current limit. There is provision for a wire soldered to both low side source leads of each pair, and to the ‘buss bar’ for phase U, and directly to the current sensor outputs, for V & W. For higher currents, shunts across the Hall sensors for phases V & W could be soldered to the ‘buss bar’ and to the junction of the current sensors and the wires mentioned in the last line. I hope I described it adequately, but it would make for very low board impeadence/heating. If set up for 400A, with the first shunted for 800A, the resistances would be 25uOhm for the first and 50uOhm for the other two.
When I first decided to go with HALL current sensors, I was thinking 1, OR 2, with jumpers in the unused positions. Then after reading Ricky’s controller thread (which I didn’t find till he moved it), I realized there might be an advantage if the first could be left in and used primarily for the hardware over current shutdown in a high amp version. It could be shunted for 600+ amps while leaving the other 2 at ‘only’ 400 for more precise phase current measurement. I should mention that I intend to solder wires (twisted) to the HALL sensors, with connectors that connect to the MPU board, keeping signals off the high current board.
Note provision for heavy wire buildup from the high side sources to the area of the board in contact with the low side heat spreaders in the view with the solder mask cutouts, that carry the phase current.
There are up to seven .62†electrolytic, and provision for as many as 14 MLCC of different sizes (yea, I went a little wild, but for 400A+, they’re good to have).
This is a schematic of the driver/fet stage. Variety of options, hence the doted lines. I provided for a second cheap optoisolated driver to be used as a one part optoisolator, for use with non isolated high side drivers. Would seldom be used, and only by people who aren’t going to worry about the extra $6 or $8…… The bootstrap diode coincides with the DC-DC converter on the board, for ‘either/or’.
The phase voltage/zero crossing detector schematic… Comments? Can be configured with comparators, or with op amps to ADC in. Also a sketch of how the delta/wye sense resistors might be arranged on the MPU board. (FreePCB wouldn’t let me use angles other than 90 degrees, so I couldn’t arrange the resistor in delta/wye like shown)
I can’t find a HI voltage switch mode regulator, let alone through hole and rated for below freezing. The circuit with the 8v zener would provide startup current to the processor, and would effectively shut off after startup. Once the processor is up, it can control preregulation to the LM317 using a spare PWM output and an ADC input. Wouldn’t have to be very accurate, so wouldn’t require any critical processor time. The driver would be the same as others used on the board to avoid confusion. Quiescent current consumption could be reduced with a reduction In available voltage range, by choosing different resistors. Also provision for connection to the serial in/out, so it can be turned on remotely (the PWM would be stopped to shut off). With a better FET, and an LM350, would be good for 5 amp, most of which could be used externally.
All along I’ve been assuming that an ebike controller should fit in the frame, therefore I’ve tried to keep at least one dimension minimal. The FET board is 1.7†X 6.5â€Â, so this controller, so with no finned heat sinks (just the aluminum case) could be around 50mm X 60mm X 170mm. I wanted to keep the length under 1/2 light-nanosecond, but thought it better to make it little longer rather than wider.
My plan is to get a home etched version made (in the next month or two), using a very basic sensored setup, THEN hope someone will want to join me and help with the software for sensorless, FOC, sine, etc. I can’t afford the $50 worth of components right now, let alone the $150 or so to have (real) boards made, but am 'working' on it (might sell an ebike if I could find someone who wants one this time of year). :|
I don’t see any reason why an ebike controller can’t be completely open source. My model is the megasquirt. Where would IT be if the SW was proprietary. In this case, different software modules could be in development by different people. If dozens (or more) people could study the source, at least SOME of the suggestions would surely be of value, so hopefully the various software modules would be more robust, and versatile, as well as being more compatible with each other as a result of the exposure. I agree that the software is more work than the hardware, and while I’m willing to donate my time, I can see where some would want compensation. Does anyone know if the paypal donation links that are associated with some freeware actually provide any income? Perhaps that would be a way to keep the SW open source and still allow you to make some money…..
I figure we could get 5 sets of boards made for around $30 per set, (1oz). If that’s upped to only qty. 15, (so each board is on a different panel), that would drop to about $25, (or $30 with 5 oz copper on the FET board, overkill in most cases (the vias are still 1oz)). If there was sufficient demand, quantity should get it under $10.
And I want to reiterate that this is supposed to be a project that anyone who has successfully assembled a kit could assemble. ALL semiconductors are through hole (believe me, that makes it MUCH harder to find components)! All resistors and caps (except electrolytic) are largish surface mount, which almost anyone should be able to handle if they can read and follow instructions.
There’s provision for measuring phase voltage and/or current, phase zero crossing for sensorless operation, a 5 or 12v bit bopped serial interface, and there’s even a CAN interface if you want it.
A couple things that are beyond this project, for me, at least at this point in time, but that I’ve been thinking about a little, (but haven’t honestly researched/studied very thoroughly).
Variable timing. If we program each of 3 16 bit timers to drive the commutation instead of the usual input, and instead use the ‘normal’ input to start the timers wouldn’t that take care of the time critical part? Then the actual time programmed into the timers could be set with a relatively low priority main loop subroutine. Changing motor speed should be easy to deal with. The delay (360e-revolution, less the advance) could be interpolated from a table to save time. Said table could be manually programmed, and/or the result of a ‘learning’ routine.
Sine wave (or other shapes such as the so called trapezoid). Why can’t this also be table driven? Said table adjusted according to feedback from the phase current and/or voltage sensors
Slightly off subject:
I intend to also design a matching set of boards, all for serial com. The bike’s wiring harness will consist of Batt minus (ground?), the serial line, batt+ and/or 12V (no need for HV at the handlebars if 12V is available).
While the controllers will be designed so they can interface directly with the controls (throttle etc.), I personally don’t plan on using them that way (perhaps set up so I could do a ‘hot wire’ if one controller survives a mishap..) I expect to have 2 controllers (2 motors), a combination display/control interface (handlebar mounted), multiple lighting controllers, and interfaces to the BMS and charger, at the minimum, all interconnected via the serial interface. I would also like to design my own boards to control servos, connected to derailleurs (more radicalism, I want a ‘jack shaft’ with a third derailleur, for greater range of speeds, one ‘shift’ button, (the controller knows which way)). I would ALSO like electromagnetic brakes controlled in conjunction with the regen. My point with the last 2, is that I would like the system to be easily expandable, through the serial interface, for yet undreamed of uses. The serial interface also allows far more than the standard '3 way' switches, by far. I intend to have multiple 'user modes', such as youth, road legal, off road, AND multiple '1wd/2wd modes', as well as a standard '3way' control available in each mode.
I might dedicate a unit to traction control, rather than impose it on the handlebar or a controller processor. Would receive throttle info etc, from the HB unit, and based on wheel rpm/wheel acceleration/g force?/tilt?/tip?/steering angle?/??, decide what throttle info to send to each controller. Would be tiny, so could be connected/mounted anywhere. With electronic brake control, would also provide anti skid.
For the display, I intend to use the DE-DP14112 from Sure Electronics, or some similar multicolor LED display. Large (5â€ÂX2.5†display area) (almost too large, but would be readable riding snogo trails in the snow or rain), Bright (for day/night/adverse weather), and will work in the cold (I want everything to be rated for -40, but at least it should work below -20F). I am willing to put up with the low resolution of 32X16 dots, but the HB unit should also be designed to work with a variety of LCD displays, alphanumeric or graphic.
Since the lights on/off/hi, turn signals, 3way, cruise button, throttle, brake, etc. are just inputs to the handlebar processor, they could easily be used to navigate through menus, for field parameter programming, without an extra keyboard (your setup would depend on what controls you actually have, obviously, but could get by with throttle (up), brake (dn), and the display mode button (select).
For the BMS, I think I’m inclined to use a low current shunt during charge type balancer, with a microprocessor monitoring them and controlling charge current. If this setup costs significant charging time (while balancing), then you should cure the problem. IOW, if there is a regular imbalance, it indicates a weak (or leaky) cell, and your battery capacity will be limited to that of the weak cell anyway.
I suppose I could give a brief description of the ebike I’m planning for next summer. Frame welded mostly from heavy (cheap) bicycle frames. Homemade swingarm (I’m thinking of using Heim joints for the pivot points). Good dbl air shock (I would also like an overload arrangement for cargo or a passenger). 2WD designed for ‘all conditions'. The front motor will tentatively be a 9C laced to a motorcycle front wheel, probably 21", and will be mounted in a light, air over, dampened motorcycle front end. The rear motor would be an x5 or 9c, laced to a fairly wide 17" MC wheel (I'm looking at the new Crystalyte motors). I'm thinking of using a second front crank with the cranks cut off/removed, as a jack shaft. 18 evenly spaced gears. Gives me the ability to pedal from under 5mph (think about a 100# bike with dead batteries), all the way to 38mph+, at a cadence of 60rpm, and almost exactly the same cadence change for each shift
I want to be able to ride snogo trails, summer and winter. . Anti spin will be a high priority. I want to be able to load it in a boat and take it anywhere (boats, airplanes, and snogos are the only way in or out of town). Could hang both hind quarters of a moose across the top tube, and walk along with the left hand on the seat and the right on the throttle, Ho Chi Mien trail style (If you've ever had to carry a moose a mile, you would understand). I've also decided a sound system is necessity. When riding a MC or snogo, or even walking, bear will hear you coming, and will hide (and hide her cubs), but I fear I would be in danger of riding up on them on an ebike.
Well, I probably covered about half of what I’ve thought about posting here over the last 2 months, enough for now.
Bob
From the original post:
To begin with, let me make it clear that I'm not claiming expertise in controller design. Virtually everything I know about BLDC motors and controllers I've learned in the last couple months on ES.
Also, I sometimes make statements when I should be asking questions, so consider apparent 'statements of fact' to be opinions or better yet, questions.
Anything I publish here will be true open source, and anyone who contributes to this thread should consider their contributions to be in the public domain. What I mean by 'open source', is that you, or anyone else can use this design, or any portion of it in any way. They can build it, change it, even sell it, about the only thing you can't do is copyright or patent it, unless of course you can show prior usage, and/or....
As this evolves I will come back and edit this post, so this SHOULD contain current info.
All that said, with help, I think I/we can design an inexpensive controller that anyone who can assemble an electronics kit could build. I think we can buy all the parts for a basic 100A+ controller for less than $50, not counting the boards, which only 'cost' pennies if you make them.
In keeping with the DIY theme, the ICs and transistors should all be through hole. I have access to an SMT rework station, and SMT would make the driver board so much easier, but I really think there’s a ‘market’ for a simple DIY controller project.
I do use some ‘larger’ surface mount resistors, capacitors, and maybe diodes, but nothing any skilled electronics hobbyist couldn’t handle safely (safe to the parts that is).
I see this project as much more do it yourself than BBC or Alan's 'commuter controller', as well as wanting it to be potentially ready to PUSH THE LIMITS of a '3 legged' fet design. It could be expanded to 12 fets, but I think trying to find a single fet that will do the job is really the way to go.
(edit 1/31/11) Can't find the FETs I want for a 300A+ 6 fet, so expanded to a 12 fet design.)
If we used DS1822 type temp sensors, as an option, we could use as many as we want, and it would only cost us 1 cpu pin. Epoxy them to the fets, the heat spreaders, the heat sinks, the pc board, the caps, etc., especially for the prototype stage, (with some minimal logging of course). Know just what was getting hot just before the smoke came out.
Yet another thought. This could be used to control up to 3 dc motors with no hardware mods couldn’t it? Maybe replace one fet with a diode . Or 2 motors with one reversible. Not a comment on this one either?
Thanks for looking, and for reading through my ramblings!
Let me know if you think this is worth pursuing. And by all means point out my mistakes and misconceptions. (Did you notice I managed not to use the word cheap?)
Bob
<edited Jan 31 2011> (First post after 'completing' hardware design. Any suggestions would still be easy to implement.)
<edit of 2/8/11> General description.
It's a 3 board stack. The bottom board is the MPU. The next board up is the driver board, and also contains the high voltage to driver voltage (12-15) regulator, which can also supply nearly 5A externally for lights and other circuitry, such as the handlebar unit. The top board is designated the FET board. It contains all the high current parts. FETs, low esr caps, current shunt(s). The largish heatsinks in the sketch should only be needed for sustained high current use (up mountains at full throttle, or pulling trucks?)
Sketch:Designed so anyone who can successfully build a kit should be able to assemble a kit based on these boards if they read and follow directions (all semi conductors through hole, and SMT are relatively large.)
Hi,
In this design, the FET board is for hi current, period. I assumed from the start that copper wire would have to be soldered to the board to carry the current! (A 6oz trace would have to be 1.5" wide to carry the current I'm hoping for). The driver board will have the high voltage to 12-15v regulator, as well as provision for a very wide variety of drive currents to suit any applicable FET. All along I’ve been trying to design a set of boards that could be built into a WIDE variety of controllers. Versatility is the second objective after ease of assembly.
I’ve decided on the Microchip dsPIC30f4011. Why? It’s the most powerful through hole processor I could find with a minimal motor controller. The 40 pin version because the 28 pin simply doesn’t have enough in/out (they use 8 of the 28 for pwr/gnd). 30 mips. 24 bit program, 16 bit data. One cycle 16bit mem to 40 bit accumulator multiply (using 2 data busses), with provision for a barrel roll, and ‘programmable’ rounding to a 16 bit register. 4 parallel sample and holds with 1mhz ADC. It can monitor phase current at 120khz and phase voltage at 200khz, at the same time. Can do 80khz PWM at 8 bit resolution. It might not be fast enough for SOME motors, but FEW would need a faster processor.
There IS a free c compiler. Look for ‘student edition’. I’m still a student?! Really 60 day ‘shareware’. After 60 days you lose use of 3 modules. I think the linker, simulator, and one other, but not the compiler.
In an effort to maximize current capability, the drain leads are cut flush with the transistor case, and all drain current flows through the ‘tab’/heat spreader. The wide part of the Source lead is installed solid to the copper trace as suggested by lfp, and there is provision for copper wire soldered directly to them and the trace. I’m assuming, at least for high current, that the high side batt lead, and phase leads would be connected to the appropriate heat spreaders rather than the circuit board. Any heat produced at the connections would then pass directly to the heat spreaders rather than the pc board. The gate lead passes right on through to the driver board. I’m told this can’t be done, but not why not…. I see lots of designs that have longer traces than that extra 12mm of gate lead (which in my builds, will have a ferrite bead, thanks, lfp).
Anyway, here’s my latest. Drawn with FreePCB. I fought with killyourselfCAD for 2 days, then downloaded FreePCB, and had a drawing in a few hours.
The copper, top and bottom, for all 3 boards, and the FET board, with and without soldermask cutouts and silk screen. The copper graphics are 5% resized from the output of GerbMagic, and the ones with the SM cutouts and SS are screen captures from the same program.
View attachment 7
There is a ‘buss bar’ of up to 1/16 X 5/8†copper soldered from the first current sensor ‘output’ to the ‘input’ side of the other 2 phase sensors. The (12ga?) wire that wraps around the 8ga negative battery lead, is soldered to both sides of the current sensor, then can be bent up to form a short 1/4†to 5/8" loop, which can then be soldered to the ‘buss bar’, for a potentially higher hardware current limit. There is provision for a wire soldered to both low side source leads of each pair, and to the ‘buss bar’ for phase U, and directly to the current sensor outputs, for V & W. For higher currents, shunts across the Hall sensors for phases V & W could be soldered to the ‘buss bar’ and to the junction of the current sensors and the wires mentioned in the last line. I hope I described it adequately, but it would make for very low board impeadence/heating. If set up for 400A, with the first shunted for 800A, the resistances would be 25uOhm for the first and 50uOhm for the other two.
When I first decided to go with HALL current sensors, I was thinking 1, OR 2, with jumpers in the unused positions. Then after reading Ricky’s controller thread (which I didn’t find till he moved it), I realized there might be an advantage if the first could be left in and used primarily for the hardware over current shutdown in a high amp version. It could be shunted for 600+ amps while leaving the other 2 at ‘only’ 400 for more precise phase current measurement. I should mention that I intend to solder wires (twisted) to the HALL sensors, with connectors that connect to the MPU board, keeping signals off the high current board.
Note provision for heavy wire buildup from the high side sources to the area of the board in contact with the low side heat spreaders in the view with the solder mask cutouts, that carry the phase current.
There are up to seven .62†electrolytic, and provision for as many as 14 MLCC of different sizes (yea, I went a little wild, but for 400A+, they’re good to have).
This is a schematic of the driver/fet stage. Variety of options, hence the doted lines. I provided for a second cheap optoisolated driver to be used as a one part optoisolator, for use with non isolated high side drivers. Would seldom be used, and only by people who aren’t going to worry about the extra $6 or $8…… The bootstrap diode coincides with the DC-DC converter on the board, for ‘either/or’.
The phase voltage/zero crossing detector schematic… Comments? Can be configured with comparators, or with op amps to ADC in. Also a sketch of how the delta/wye sense resistors might be arranged on the MPU board. (FreePCB wouldn’t let me use angles other than 90 degrees, so I couldn’t arrange the resistor in delta/wye like shown)
I can’t find a HI voltage switch mode regulator, let alone through hole and rated for below freezing. The circuit with the 8v zener would provide startup current to the processor, and would effectively shut off after startup. Once the processor is up, it can control preregulation to the LM317 using a spare PWM output and an ADC input. Wouldn’t have to be very accurate, so wouldn’t require any critical processor time. The driver would be the same as others used on the board to avoid confusion. Quiescent current consumption could be reduced with a reduction In available voltage range, by choosing different resistors. Also provision for connection to the serial in/out, so it can be turned on remotely (the PWM would be stopped to shut off). With a better FET, and an LM350, would be good for 5 amp, most of which could be used externally.
All along I’ve been assuming that an ebike controller should fit in the frame, therefore I’ve tried to keep at least one dimension minimal. The FET board is 1.7†X 6.5â€Â, so this controller, so with no finned heat sinks (just the aluminum case) could be around 50mm X 60mm X 170mm. I wanted to keep the length under 1/2 light-nanosecond, but thought it better to make it little longer rather than wider.
My plan is to get a home etched version made (in the next month or two), using a very basic sensored setup, THEN hope someone will want to join me and help with the software for sensorless, FOC, sine, etc. I can’t afford the $50 worth of components right now, let alone the $150 or so to have (real) boards made, but am 'working' on it (might sell an ebike if I could find someone who wants one this time of year). :|
I don’t see any reason why an ebike controller can’t be completely open source. My model is the megasquirt. Where would IT be if the SW was proprietary. In this case, different software modules could be in development by different people. If dozens (or more) people could study the source, at least SOME of the suggestions would surely be of value, so hopefully the various software modules would be more robust, and versatile, as well as being more compatible with each other as a result of the exposure. I agree that the software is more work than the hardware, and while I’m willing to donate my time, I can see where some would want compensation. Does anyone know if the paypal donation links that are associated with some freeware actually provide any income? Perhaps that would be a way to keep the SW open source and still allow you to make some money…..
I figure we could get 5 sets of boards made for around $30 per set, (1oz). If that’s upped to only qty. 15, (so each board is on a different panel), that would drop to about $25, (or $30 with 5 oz copper on the FET board, overkill in most cases (the vias are still 1oz)). If there was sufficient demand, quantity should get it under $10.
And I want to reiterate that this is supposed to be a project that anyone who has successfully assembled a kit could assemble. ALL semiconductors are through hole (believe me, that makes it MUCH harder to find components)! All resistors and caps (except electrolytic) are largish surface mount, which almost anyone should be able to handle if they can read and follow instructions.
There’s provision for measuring phase voltage and/or current, phase zero crossing for sensorless operation, a 5 or 12v bit bopped serial interface, and there’s even a CAN interface if you want it.
A couple things that are beyond this project, for me, at least at this point in time, but that I’ve been thinking about a little, (but haven’t honestly researched/studied very thoroughly).
Variable timing. If we program each of 3 16 bit timers to drive the commutation instead of the usual input, and instead use the ‘normal’ input to start the timers wouldn’t that take care of the time critical part? Then the actual time programmed into the timers could be set with a relatively low priority main loop subroutine. Changing motor speed should be easy to deal with. The delay (360e-revolution, less the advance) could be interpolated from a table to save time. Said table could be manually programmed, and/or the result of a ‘learning’ routine.
Sine wave (or other shapes such as the so called trapezoid). Why can’t this also be table driven? Said table adjusted according to feedback from the phase current and/or voltage sensors
Slightly off subject:
I intend to also design a matching set of boards, all for serial com. The bike’s wiring harness will consist of Batt minus (ground?), the serial line, batt+ and/or 12V (no need for HV at the handlebars if 12V is available).
While the controllers will be designed so they can interface directly with the controls (throttle etc.), I personally don’t plan on using them that way (perhaps set up so I could do a ‘hot wire’ if one controller survives a mishap..) I expect to have 2 controllers (2 motors), a combination display/control interface (handlebar mounted), multiple lighting controllers, and interfaces to the BMS and charger, at the minimum, all interconnected via the serial interface. I would also like to design my own boards to control servos, connected to derailleurs (more radicalism, I want a ‘jack shaft’ with a third derailleur, for greater range of speeds, one ‘shift’ button, (the controller knows which way)). I would ALSO like electromagnetic brakes controlled in conjunction with the regen. My point with the last 2, is that I would like the system to be easily expandable, through the serial interface, for yet undreamed of uses. The serial interface also allows far more than the standard '3 way' switches, by far. I intend to have multiple 'user modes', such as youth, road legal, off road, AND multiple '1wd/2wd modes', as well as a standard '3way' control available in each mode.
I might dedicate a unit to traction control, rather than impose it on the handlebar or a controller processor. Would receive throttle info etc, from the HB unit, and based on wheel rpm/wheel acceleration/g force?/tilt?/tip?/steering angle?/??, decide what throttle info to send to each controller. Would be tiny, so could be connected/mounted anywhere. With electronic brake control, would also provide anti skid.
For the display, I intend to use the DE-DP14112 from Sure Electronics, or some similar multicolor LED display. Large (5â€ÂX2.5†display area) (almost too large, but would be readable riding snogo trails in the snow or rain), Bright (for day/night/adverse weather), and will work in the cold (I want everything to be rated for -40, but at least it should work below -20F). I am willing to put up with the low resolution of 32X16 dots, but the HB unit should also be designed to work with a variety of LCD displays, alphanumeric or graphic.
Since the lights on/off/hi, turn signals, 3way, cruise button, throttle, brake, etc. are just inputs to the handlebar processor, they could easily be used to navigate through menus, for field parameter programming, without an extra keyboard (your setup would depend on what controls you actually have, obviously, but could get by with throttle (up), brake (dn), and the display mode button (select).
For the BMS, I think I’m inclined to use a low current shunt during charge type balancer, with a microprocessor monitoring them and controlling charge current. If this setup costs significant charging time (while balancing), then you should cure the problem. IOW, if there is a regular imbalance, it indicates a weak (or leaky) cell, and your battery capacity will be limited to that of the weak cell anyway.
I suppose I could give a brief description of the ebike I’m planning for next summer. Frame welded mostly from heavy (cheap) bicycle frames. Homemade swingarm (I’m thinking of using Heim joints for the pivot points). Good dbl air shock (I would also like an overload arrangement for cargo or a passenger). 2WD designed for ‘all conditions'. The front motor will tentatively be a 9C laced to a motorcycle front wheel, probably 21", and will be mounted in a light, air over, dampened motorcycle front end. The rear motor would be an x5 or 9c, laced to a fairly wide 17" MC wheel (I'm looking at the new Crystalyte motors). I'm thinking of using a second front crank with the cranks cut off/removed, as a jack shaft. 18 evenly spaced gears. Gives me the ability to pedal from under 5mph (think about a 100# bike with dead batteries), all the way to 38mph+, at a cadence of 60rpm, and almost exactly the same cadence change for each shift
I want to be able to ride snogo trails, summer and winter. . Anti spin will be a high priority. I want to be able to load it in a boat and take it anywhere (boats, airplanes, and snogos are the only way in or out of town). Could hang both hind quarters of a moose across the top tube, and walk along with the left hand on the seat and the right on the throttle, Ho Chi Mien trail style (If you've ever had to carry a moose a mile, you would understand). I've also decided a sound system is necessity. When riding a MC or snogo, or even walking, bear will hear you coming, and will hide (and hide her cubs), but I fear I would be in danger of riding up on them on an ebike.
Well, I probably covered about half of what I’ve thought about posting here over the last 2 months, enough for now.
Bob
From the original post:
To begin with, let me make it clear that I'm not claiming expertise in controller design. Virtually everything I know about BLDC motors and controllers I've learned in the last couple months on ES.
Also, I sometimes make statements when I should be asking questions, so consider apparent 'statements of fact' to be opinions or better yet, questions.
Anything I publish here will be true open source, and anyone who contributes to this thread should consider their contributions to be in the public domain. What I mean by 'open source', is that you, or anyone else can use this design, or any portion of it in any way. They can build it, change it, even sell it, about the only thing you can't do is copyright or patent it, unless of course you can show prior usage, and/or....
As this evolves I will come back and edit this post, so this SHOULD contain current info.
All that said, with help, I think I/we can design an inexpensive controller that anyone who can assemble an electronics kit could build. I think we can buy all the parts for a basic 100A+ controller for less than $50, not counting the boards, which only 'cost' pennies if you make them.
In keeping with the DIY theme, the ICs and transistors should all be through hole. I have access to an SMT rework station, and SMT would make the driver board so much easier, but I really think there’s a ‘market’ for a simple DIY controller project.
I do use some ‘larger’ surface mount resistors, capacitors, and maybe diodes, but nothing any skilled electronics hobbyist couldn’t handle safely (safe to the parts that is).
I see this project as much more do it yourself than BBC or Alan's 'commuter controller', as well as wanting it to be potentially ready to PUSH THE LIMITS of a '3 legged' fet design. It could be expanded to 12 fets, but I think trying to find a single fet that will do the job is really the way to go.
(edit 1/31/11) Can't find the FETs I want for a 300A+ 6 fet, so expanded to a 12 fet design.)
If we used DS1822 type temp sensors, as an option, we could use as many as we want, and it would only cost us 1 cpu pin. Epoxy them to the fets, the heat spreaders, the heat sinks, the pc board, the caps, etc., especially for the prototype stage, (with some minimal logging of course). Know just what was getting hot just before the smoke came out.
Yet another thought. This could be used to control up to 3 dc motors with no hardware mods couldn’t it? Maybe replace one fet with a diode . Or 2 motors with one reversible. Not a comment on this one either?
Thanks for looking, and for reading through my ramblings!
Let me know if you think this is worth pursuing. And by all means point out my mistakes and misconceptions. (Did you notice I managed not to use the word cheap?)
Bob
Would copper work if large enough that it didn't heat significantly (<.2mOhm)? Twisted pairs and careful op amp selection.... Wouldn't be to hard to temperature compensate with a temp sensor on or near them either.
