Building the Best Controller

[liveforphysics]

Understood; thank you Eric & Luke both for the clarity.

Generally speaking, slightly off-topic, and with brevity - what would be the effect of high current vs. high-voltage in a coreless motor design, especially in regards to BEMF?

~KF

If the copper fill is identical, then only the input power matters. Doesn't matter if you get your 1000w from 100v at 10amp or 10v at 100amp. The difference is the phase wire size needs to be scaled up to match the current obviously (easy enough), but then the performance is a dead-match. The BEMF in the 10v motor would be growing at 1/10th the rate of the 100v motor, so the speed potential and RPM range overlay each other perfectly.

 

[oldswamm]

I would amend my last post slightly. I would LEAN toward a 6 fet board with no to220 (surely we could find suitable, and cheap enough to247 to satisfy anyone). Limiting it to 6 fets might depend on finding a suitable off board 'pucks', esp to satisfy the 150v, 150A+ people.

As to keeping the operating voltage to 80% of spec, I'm SURE most people who build this controller will do so without ever considering their PAST behavior. :| Are we going to mfg these and sell them with a warranty maybe? And don't get me wrong, I would probably try and stay under say 90% for a daily rider, but this is a controller FOR people who are pushing the edge, so it's likely that many will try one cell more till they find out in real world fashion what the limits really are.

And Luke, thanks for pointing out the 'obvious' ease of changing the FETs on that setup.

Bob

 

[rhitee05]

But, if it's going to be a paralleled fet design... I would like to see the 12fets arranged in an inductively and space logical manner. No stupid single long row across the back edge of the board crap if you're going to do a paralleled FET arrangement. We can make the device much smaller and better performance with a proper layout, and who cares if it takes an extra 20-30mins longer to assemble? It's a hand-built kit controller, that's kinda the idea of using your own man-power to enable better designs that a mass production factory bean-counter would green-light.

100% agree. The layout should be driven by electrical concerns first and thermal concerns second.

In my head, I'm imagining that a two-row design would probably work out well. Similar to how the 24/36-FET boards are layed out, but with all the high-side FETs in one row and the low-side FETs in the second row. There would be wide BATT+ and BATT- traces running along the upper and lower edges of the boards, and there should be plenty of room for wide traces running up/down each phase. Using a 4-layer board you can keep all the high-current traces on the outer layers and use the inner layers for the drive signals, which would allow for lots of unbroken copper area.

I'd probably plan to use heat spreaders to connect the FETs to a heatsink located directly above the board. If you use one heatspreader for the high-side and separate spreaders for each of the low-side phases, then you can move the insulator to the other side of the heat spreader where you've got a larger contact area. I don't think that would complicate assembly, either.

 

[oldswamm]

Eric,
I like the sound of what you see in your head, at least if I'm seeing what you are.

I Like moving the insulators out too, even with a 6 fet, sure. If we use 12 fet, it ties the fets together better as well.

The design could still probably accommodate multiple motors on one end of the stack, either through external fets, or by bending the FETs out, moving the heat spreaders to the sides, and putting the big caps where they could hang off the ends. Sounds a bit cumbersome, but if you use more than 2 motors, there's probably room for a 6" x 6" x 10" controller, and the heatsink arrangement would probably be up to the end user on something that out of the ordinary.

I never found the thread where Jeremy talks about the buffer arrangement. LOTs of other interesting and distracting posts, but.... Could someone give me a hint, would seem relevant?

Bob

 

[rhitee05]

Bob,

Try here: http://endless-sphere.com/forums/viewtopic.php?f=2&t=20913&start=15#p305195

 

[Alan B]

Daydreaming about Board Sandwiches in the Shower

Have been thinking about the physical board modularization. I haven't followed this discussion closely enough to know how close this is to what you are already considering so pardon me for the duplication/overlap there, but this is what I was thinking about that comes together pretty well as a total packaging setup.

Board #1 - Main CPU, power supplies, most external interfacing with bus interfaces for #2 on BOTH sides (top and bottom, facing FETS to the outside in both cases)
Board #2 - Motor CPU, FET drivers, space for 6 TO247 FETs (optional, not used in all cases)
Board #3 - 12 SUPER FETs (leave FETs off board #2 when using board #3)

Configurations (shown in stack order):

Boards #1 & #2 - small controller for one motor (and sufficient for a successful project)

Boards #2, #1, #2 - a dual motor controller (a small step from single motor)

Boards #1, #2, #3 - a single motor high power controller (board #2 would not have FETs loaded in this case)

Boards #3, #2, #1, #2, #3 - a dual motor high power setup

Notes

Note this retains the option of developing a version of Board #2 with FPGA instead of CPU for vector drive, higher rpms, etc

Note this keeps the FET boards to the outside so the heatsink can piggyback directly on the PC board. The FETs can be mounted so their heatsink surfaces are toward the outside for maximum thermal efficiency

Board #3 could have optional bus bar overlays where needed (it will need them). Board #2 maybe also, or not.

High current leads would come into boards #2 and/or #3. Perhaps these boards are a bit longer than the main CPU and a bus bar/stud could be sandwiched between the boards on the longer end to carry the high current.

Phase current leads would go directly to boards #2 or #3, hall leads to board #2

Note this setup doesn't really do well above 2 motors unless it is designed so that the interconnect between board #1 and #1 can be a cable. In that case it could be a dual stack with one stack having a #1 and the second stack having a mostly empty #1 with a cable over to the first #1. A specific cable connector might be required on the #1 board to facilitate this.

Enough dreaming for now, this message is long enough...

 

[oldswamm]

Thanks Eric.
Been wondering about that since I first saw lfp's 'at the PO' photos.
Looks simple and effective. Might need to tweak the R for different controllers or motors, but Yea.
I would love to put a 100kw or so rated motor in the boat I'm building, but would either need to get a job or rob a bank. Like lfp implied earlier, when the battery bill gets well into 5 figures, you don't need to ask what the FETs cost.
Bob

 

[liveforphysics]

Thanks Eric.
Been wondering about that since I first saw lfp's 'at the PO' photos.
Looks simple and effective. Might need to tweak the R for different controllers or motors, but Yea.
I would love to put a 100kw or so rated motor in the boat I'm building, but would either need to get a job or rob a bank. Like lfp implied earlier, when the battery bill gets well into 5 figures, you don't need to ask what the FETs cost.
Bob

100kw brushless boat???

You're a man that needs this fet badly:
200v, 1600amp, Combined Rth of 0.07C/W.
Make yourself a little 6-fet with those that handles 100kw on a 150v system, and has loads of voltage and current overhead in the FETs. Working the numbers, it would actually be a very cool running 6-fet controller doing 100kw continuously. You could surge to 250-300kw if needed.

http://ixdev.ixys.com/DataSheet/VMO1600-02P.pdf

 

[oldswamm]

:lol:
I've had that datasheet in my 'bike' folder for weeks.
And, I said 100kw 'rated'
The idea of a fast quiet boat!
Of course the downside is that more than 5 or 10 minutes worth of batteries at full throttle would sink the boat, and going the kind of distances we travel by boat (100s of miles), would absolutely require that it be a hybrid (is that word allowed in this forum?).
Bob

 

[Kingfish]

...
Board #1 - Main CPU, power supplies, most external interfacing with bus interfaces for #2 on BOTH sides (top and bottom, facing FETS to the outside in both cases)
Board #2 - Motor CPU, FET drivers, space for 6 TO247 FETs (optional, not used in all cases)
Board #3 - 12 SUPER FETs (or more?) (leave FETs of board #2 when using board #3)
...
Notes
...

Hi Alan_B
Thanks for the visioning; I think it displays what can be accomplished with simple modularity. Please allow me to expand and comment.

Board #1: This is an interesting configuration and slight departure from a separate MPS and MCU. There are two paths we could take here:

• Combine the MPS and MCU together on the same plane: Reduces by one a connection and allows the common interface to become smaller; no harm done.
• Retain the MCU though make the MPS (or reverse) a daughterboard: Keeps the flexibility for present or future use.
• No change from the original plan with all boards shaing a common interface. The placement of the MPS is flexible so long as it does not block the thermal considerations.

Re: Notes
These are excellent observations in-line with my imagination as well.

Configuration for more that 2WD I think is beyond the scope of V1, although it is fun to develop a plan:

• What if you could have a two-channel #2 Board? Now the configuration could support up to 2WD a side, and with high-power by pairing #3 Boards. The physicality of the FETs and thermal taps is retained.

It’s nice to dream. Yours allows for dual-path and off-board FET support.
Good one, KF

 

[rebelpilot]

Modularity, I like that....

I would like to add this: think outside the box and don’t put everything in one box.

I apologize if the ideas I present here have already been discussed, but I glaze over badly when trying to catch-up on a long thread. (And yet here I am making it much longer and harder to read. I was learning google sketchup last night but I’m still too slow and too impatient.)

I’ve been pondering my next build from a systems design standpoint, and I want to do everything I can for a clean looking, low maintenance install. It is not convenient to connect every wire on a bike to one box. I don’t want to use another Chinese controller. (I saw the inside of mine, it isn’t pretty)

Personally, I would like a handle bar mounted LCD to display: speed, maxspeed, avgspeed, actual time, trip time, odometer, trip odometer, amps, maxamps, volts, minvolts, watt-hrs, Wh/km, Wh/mi, battcycles, avgWh/mi, etc, and a keypad to allow me to set the controller parameters and the throttle response, cruise control etc. and to control my lights. I don’t want to use a separate programmer for the control setup, and I don’t want any other devices mounted on my handle bars.

A display mounted on the handle bars would provide a convenient junction box to connect the brake levers, throttle, headlamp, blinkie, front turn signals, and turn signal switch. The wires would be short and only the 12V, ground and comms wires would run off the handle bars to the rear.
(Kingfish: having just completed your light setup you can maybe appreciate where I am coming from, a major drawback with ebikes are the large bundles of wire flexing every time the handlebars are turned). (By the way, were my suggestions on the converter thread helpful?)
The display module would have:
-a backlit LCD – I don’t know if graphics would be possible, or if that pushes the processor to a whole other level, but it would be nice. Actually it could make a huge difference in bling factor, and readability.
-a simple, small linear regulator for the processor 5 volt supply.
-protected-driver 12 volt outputs for lamps, or current limited LED supplies, either way there are lots of chips and modules available for either, both could be an option on the pcb.
https://ec.irf.com/v6/en/US/adirect/ir?cmd=eneNavigation&No=0&N=0+4294835973&numRecs=99
http://www.dealextreme.com/details.dx/sku.13557
(I apologize for placing a deal extreme link in your thread Kingfish)
-a keypad that has actual mechanical switches, or at least a membrane with good tactile feedback. (I wear gloves whenever I ride)
-The display module must be absolutely waterproof. (the last two items are almost mutually exclusive aren’t they?)
-I would also like to shoe horn in a 433 Mhz transmitter and antenna with a tilt sensor into my display.

The small three wire cable from the display would connect to another box mounted to my seat post.
This box would be designed as a light bar with the rear blinkie and turn signals on the back surface and the heatsink on the front, the only connections to this box would be the aforementioned comms/power cable to the display; plus wires for the halls, phase wires, and the battery wires, all exiting the bottom. (I see you’ve been considering turn signals too Kingfish)
Inside the box would be:
- the DC-DC converter, 12 volts out should be all that’s necessary,
- a small linear regulator for the processors,
- current limited supplies for the LEDs.
- a processor for comms and the battery monitor and running the rear signal lights and talking to the motor control processor.
- motor processor and FET driver board,
- FET board (I’m liking LFP’s surface mount suggestion, we all build our bikes in the kitchen, why not bake our FET boards there too?),
- a 75 watt 12 volt output for heated clothing. (not many people would need this).

I’m thinking an aluminum extrusion would be good for this box, and a design that allowed removing the light panel to provide access to everything without disturbing any wires or heat sink connections would be a priority.
Since the light bar won’t be all that wide it would be nice if the turn signals were moving arrows <<< (*) >>>

I don’t expect that anyone will share the exact same vision as me for the system design. One benefit of an open source project is that others can modify the design to suit their needs. My intention is to help stimulate the discussion and encourage a flexible modular design.
Of course there are many variations possible, like moving the motor control board to yet another box, or dropping the turn signals, maybe use part of the bike as a heatsink. Where should the main power switch and key switch be mounted, front or rear? Could the turn signal switch reside on the side of the display or would restricting the module mounting location to one side of the handle bars be a problem? Maybe the display graphics/text could rotate so you could flip it to mount on either side of the bars, or even turn it 90 degrees?

 

[oldswamm]

KF, till your last post, I've been thinking that I personally would never use more than 2 channels and have just been 'lobbying' for versatility, but It just occurred to me how nice a 'real' electric 4 wheel drive ATV would be! There's literally hundreds of them (gas powered) running around this town. Maybe for version 1.2, a dual 6 fet, for 3/4 motor or economy/low power 2 motor?
Bob
Rebelpilot, your pretty close on to most of what has been said already. I INTEND to use a serial cable for all communication to/from the handlebars, and probably also to a separate lighting controller (incl. turn signals), and will be lobbying for same when the time comes, but as has been pointed out, all that pretty much all comes after ironing out the FET stage. As to heated clothing, you and I might be in the minority there.

 

[rhitee05]

Rebelpilot,

I think a separate "head unit" is a good idea. It's really a necessity if you want to do any of the more advanced features, like a useful cruise control, the ability to change settings on-the-fly, or CA-like functionality.

We had previously discussed having a serial bus available for add-on, and I think that was included into the official spec, for exactly this sort of reason. Adding a control unit to the design might be biting off a little too much for the first round, but we'd want to make sure all the hooks are there so it could be added later.

From a design standpoint, it does make a lot of sense to have two or more smaller boxes with a common bus between them. This is exactly how the electronics in a car are set up and it's a good model to emulate.

 

[oldswamm]

Eric,
Were getting a little ahead of the project here, but it seams to me that the head unit, and a lighting controller would be very easy to design, program, and integrate. I feel confident even I could handle that part, (hopefully with a little help).
As to a control keyboard, we could use what's there. For example, move the 3 way from 2 to 3 to 1 then back to 2 while holding the cruise button, and of course parked, to enter program mode, then use the most intuitive controls to navigate through menus.
I hadn't actually thought about it, and don't think it's been mentioned till rebelpilot brought it up, but a method of field reprogramming would seem to be a natural.

 

[Kingfish]

KF, till your last post, I've been thinking that I personally would never use more than 2 channels and have just been 'lobbying' for versatility, but It just occurred to me how nice a 'real' electric 4 wheel drive ATV would be! There's literally hundreds of them (gas powered) running around this town. Maybe for version 1.2, a dual 6 fet, for 3/4 motor or economy/low power 2 motor?
Bob
Rebelpilot, your pretty close on to most of what has been said already. I INTEND to use a serial cable for all communication to/from the handlebars, and probably also to a separate lighting controller (incl. turn signals), and will be lobbying for same when the time comes, but as has been pointed out, all that pretty much all comes after ironing out the FET stage. As to heated clothing, you and I might be in the minority there.

hehe 8)
I used to work for a little tiny itty bitty smallish toy company, and one of the first jobs I had was to re-engineer a three-channel animation board into two-channels; we literally scraped off the third channel and shot new plastic where one servo drove two actuating parts: Teddy Ruxpin’s mouth and nose became linked rather than move independently; little kids never noticed as we raised the profit margin by a third.

As soon as I read about Alan_B’s board-stacking – the two and three-channel animation popped right back into my tiny pointed head. Makes sense doesn’t it?

Clothes heating:
This is why we have an independent MPS – so you can customize your Aux Circuits as required! Personally, I’d like plasma balls of lightning to fly from my gloved hands for obliterating Z O M B I E S that cross my path – but to each his own

Zombies beware, KF

 

[rebelpilot]

Kingfish, I had to look up what you meant by "independent MPS".

Main Power Supply:

* Main Battery Power, Unregulated, 36V to >100V, with voltage test pads
Option to have high-charge and discharge buffer (SuperCap or High-C batteries)
* 12V Branching Regulated PS
o Driver PS, Regulated, Isolated, 12V for FETS
o Controller/Brain PS, regulated (5V or less)
o On/Off circuitry (cut power to Brain or Driver supply), with slow-ramp/spark suppression
o Auxiliary PS, regulated for lights and accessories (may be a mix of 3.3, 4.1, 9V), with optional capacitance when main power is lost. Example: Headlights still work for another hour if the main batteries loose power and you have to pedal home.
o Scalable power output to meet or exceed need.
* Separate Charger & BMS connections

Are there stil plans to include a discharge buffer? I can't imagine how it could be practical. Wouldn't it be gigantic and costly? Can't people just parallel some lipo with their LifePo, outside the box?
I agree 12 volts is needed - for everything.
Why have a 5 volt regulator on a different board than the processors?
A mechanical on/off switch would have to be massive to support any appreciable DC current, and if a huge relay or solid state device were to be used there is a definite safety problem if the contacts weld or the solid state device shorts, so why bother? I think the only practical way to control the main battery current in an emergency is to pull the battery plug (which should already have sufficiently large contacts to safely break the load).
Again, I don't think it is practical to run several wires around the bike with several different voltages, and the controller enclosure is not likely to be located in a convenient spot to make multiple connections.
Scalable power - why? 12 volts at few amps should cover all bases.
Separate charger and BMS connections? I don't know if I understand what this means. Why run all these wires from the battery pack, or require the use of a custom battery pack with comms, A/D etc? Why try to support different chemistries etc. A BMS system does not belong in the controller.
The only part of the "MPS" that I would keep is the 12 volt switcher.

 

[Kingfish]

Rebel, you need to read this thread from the beginning. Take your time, read thoroughly, and imagine how some of the features can be wisely accomplished without the presumptions you have made.

Bottom-line: You can elect to stuff the board the way you wish - custom-fit the way YOU want it. 8)

Enjoy, KF

 

[rebelpilot]

(By the way, were my suggestions on the converter thread helpful?)

http://endless-sphere.com/forums/viewtopic.php?f=3&t=16623&start=45
Kingfish, Sorry if I annoyed you. Was I at least able to help with the ohm's law problem?

 

[Kingfish]

(By the way, were my suggestions on the converter thread helpful?)

http://endless-sphere.com/forums/viewtopic.php?f=3&t=16623&start=45
Kingfish, Sorry if I annoyed you. Was I at least able to help with the ohm's law problem?

OT: Absolutely! I cut the resistor out. I didn't want to do it... but you were right and it was necessary for proper expansion. I'm working on the next project: A new battery bag now - trying to do the layout correctly in CAD: Lots of odd angles - making it unlike any sheetmetal pattern I have done before (except this will be a fabric bag though the methodology still applies). I bought a walking-foot industrial sewing machine to stitch it together. A bag for P0, and then one for P1. After that ~ I shall conquer the world with lil' triangle lipo battery bags. One small step for LiPo, one giant leap for mankind. I made that up just now...

All the best to you and yours, KF

 

[rhitee05]

I just think KF was trying to point out that there are valid design reasons why we made certain choices. Which is not to say that comments aren't welcome.

For example, the separate MCU and MPS boards. One reason was that this could make it possible to take an off-the-shelf board, say a PIC or Atmel reference board, and use it as the MCU. We can concentrate all the application-specific circuitry on the MPS and place fewer requirements on the MCU. That's not to say it wouldn't be possible to design a single board that contains both functions. From an engineering standpoint, there are advantages to having two small boards instead of one larger board.

I get the impression you may have the wrong idea about how the various boards fit together. We're not talking about "several wires all around the bike" here - the 3 boards would be stacked together inside the controller and share a common interface via pin-and-socket connectors. It makes good design sense to cluster all the DC voltage supplies together, for example. It's not really a disadvantage to send multiple voltages over the common interface to the MCU, and in any case it was a design decision to make all the supplies available through the interface to facilitate expandability. There might, for example, be analog circuitry on the MPS board which requires the 5V supply.

I think there was a design assumption early on that it's BYO circuit protection devices for the battery. You're right that it's not practical to size a mechanical switch or even a fuse for all potential currents (nor desirable). Keeping the fuse physically and electrically close to the battery is better, anyway. I think the "power switch" would really be a control for the 12V and various secondary supplies, so an on/off switch for the controller rather than a battery cut-off.

I can't speak for KF, but I think there are still a few notional features scattered throughout the spec descriptions that probably won't make it through the design process.

 

[Kingfish]

I just think KF was trying to point out that there are valid design reasons why we made certain choices. Which is not to say that comments aren't welcome.

For example, the separate MCU and MPS boards. One reason was that this could make it possible to take an off-the-shelf board, say a PIC or Atmel reference board, and use it as the MCU. We can concentrate all the application-specific circuitry on the MPS and place fewer requirements on the MCU. That's not to say it wouldn't be possible to design a single board that contains both functions. From an engineering standpoint, there are advantages to having two small boards instead of one larger board.

I get the impression you may have the wrong idea about how the various boards fit together. We're not talking about "several wires all around the bike" here - the 3 boards would be stacked together inside the controller and share a common interface via pin-and-socket connectors. It makes good design sense to cluster all the DC voltage supplies together, for example. It's not really a disadvantage to send multiple voltages over the common interface to the MCU, and in any case it was a design decision to make all the supplies available through the interface to facilitate expandability. There might, for example, be analog circuitry on the MPS board which requires the 5V supply.

I think there was a design assumption early on that it's BYO circuit protection devices for the battery. You're right that it's not practical to size a mechanical switch or even a fuse for all potential currents (nor desirable). Keeping the fuse physically and electrically close to the battery is better, anyway. I think the "power switch" would really be a control for the 12V and various secondary supplies, so an on/off switch for the controller rather than a battery cut-off.

I can't speak for KF, but I think there are still a few notional features scattered throughout the spec descriptions that probably won't make it through the design process.

Agreed, especially the last line: Design Specs are like Wish Lists ~ you stuff all that you can think of into them and cross your fingers that you can achieve 50%. I believe we have been pragmatic with most of the objectives; this is after all design-by-community

Best, KF

 

[Kingfish]

Briefly, let’s summarize:

• We’re good on the MPS.
• We have a modicum of understanding with the MCU leaning towards a Main Processor supported by Motor Drivers.
• Three possible FET board choices with tantalizing opportunities regardless of which path is taken:
  • SMD FETs – massive amps, ≤ 75V
  • TO-247 (or like) 6-FET ≤ 150V ≤ 150A
  • TO-247 (or like) 12-FET/Dual-Row > 150A

Note: It seems possible to support off-board FETs regardless of choice.

If I had to prioritize:
For Version 1, let’s do the 6-FET first, and right behind it in parallel have the 12-FET ready in the wings, then follow up with the SMD FET.

Can we agree on that and keep the ball rolling?

Best, KF

 

[Alan B]

Some candidate AVRs for motor control. Based on a quick search of their app notes here:
http://www.atmel.com/products/AVR/mc/?family_id=607

AT90PWM3 AT90PWM3B AT90PWM316
ATtiny261/461/861
ATmega48/88/168
ATmega32M1 <<== this one is probably the best

Development board ATAVRMC100 (pricey at 250) includes software for sensor and sensorless BLDC motor control. Appears to include a small motor. board contains AT90PWM3B cpu, drivers, etc.

 

[rebelpilot]

Alan, I have been considering the AVR DRagon and the ATAVRMC301 Motor Control Board for ATtiny861.
http://store.atmel.com/PartDetail.aspx?q=p:10500053
http://store.atmel.com/PartDetail.aspx?q=p:10500150
$49 and $69 respectively. Would this be a useful development platform for a beginner?
Does it have useful debug capabilities?

Thanks

 

[Alan B]

Alan, I have been considering the AVR DRagon and the ATAVRMC301 Motor Control Board for ATtiny861.
http://store.atmel.com/PartDetail.aspx?q=p:10500053
http://store.atmel.com/PartDetail.aspx?q=p:10500150
$49 and $69 respectively. Would this be a useful development platform for a beginner?
Does it have useful debug capabilities?

Thanks

I have a dragon board. Very powerful little development board.

Great find on the motor control devel board. That one is much better priced than the one I found above. It doesn't have the power gear onboard, the ATAVRMC300 board has that, but that might be better done a different way anyhow, and even if bought it is cheaper than the other board I mentioned before.

One tip on the Dragon board. Don't touch it while powered on. There is a voltage inverter that can be damaged by just touching it when powered. Unless they improved it. But it is a great board.

The debug capability is significant. I have not used that feature a lot, but it is there when needed.

How does the motor control hardware on this AVR cpu compare to the PIC chips?