Home Brew Controller, what would it take ?

Ok, my replacement 4840 (4850) Controller has arrived, but I am reluctant to plug it in for fear of Toasting another $250 controller.

Made me think, what would it take to build an indestructible Home-brew Controller?

In looking at Application Notes Like

http://www.onsemi.com/pub/Collateral/MC33033-D.PDF

I only see a handful of components, and nothing that requires special shielding, coil winding, or surface mounting.

If I combined a controller like the above, and a few indestructible MosFets like

http://www.irf.com/product-info/hi-rel/hirelbrushlessdc.html#igbt

or something that was "Fully Protected" such as:

http://www.irf.com/product-info/datasheets/data/ir3310.pdf

A few components, a BIG heat sink, a thermostatically controlled fan, and some breadboarding, and couldn't I build a really really robust controller ?

Forward, Reverse, Dynamic braking, and Reg are all part of some of these controllers... and by moving the MosFets Off-Board, lots of issues are solved (HEAT for one)

Am I being to simplistic?

When I opened the 4840 controller, the parts count was very small, and nothing adjustable or tunable. By moving the POWER section off-board, we can run some REAL wires to/from the Hub/Controller?battery and lower those losses. The power module becomes a "plug-n-play" module that people can drop in what they want to depending on their set-up and budget.

I'm sorry I really don't have any answers but I have wanted to do the same with some of the RC market brushless controllers. I always though it would be cool to buy one built for 45amps and upgrade it to be used with high voltage and high amperage. I think the best thing you can do for your controller is consider the cycle analyst so you can set the current limit below the max of the controller. So if it is a 35 amp controller set it to 30 amps. Also consider upgrading the fets with in the current controller.

the fets in the 4850 revision board are already very good - they're holding up with my near 3kw of power!

rsisson said:

Made me think, what would it take to build an indestructible Home-brew Controller?

Click to expand...

I've thought about it too. The subsystems would be:

- Power switches. For low voltages (<200 volts or so) FETs are generally the best way to go, and there are good FETs available nowadays. I'd add some very high speed diodes to protect them against flyback voltage.

- Power switch driver. Low side drivers are easy. High side N channel drivers are harder. I'd use an off the shelf high side driver with an auxiliary bootstrap power supply to allow operation at very low speeds. The LM5106 gives you a low side and high side driver in the same package, and has a dead time control. (Important to prevent shoot-through.)

- Current sensor. You could do this with a Hall sensor, but probably easier to do with a very low resistance (1 milliohm) resistor in the return line. Then a differential amp turns that into a voltage that can be read by the controller. I'd also add an absolute-max comparator that shuts everything down if current ever exceeds the FET ratings.

Depending on the design I might also add Hall-effect current sensors to each phase. This allows you to better protect the FETs and do sensorless operation.

- Controller. You could go with an off-the-shelf but I think I'd use a programmable controller like a PIC. (The dsPIC30F4011 has a built in three phase motor driver!) Inside the controller I'd put:

-a current mode control, so the throttle controls torque instead of speed

-a "magic sine" synthesizer (above a certain speed) to do sine wave drive (more efficient)

-phase advance for running faster than base speed

-phase inversion for regen braking

And of course it could output any data you wanted to an LCD, and could be programmed via a serial port for things like current limit, max phase advance, low voltage cutoff etc.

FWIW, esteemed member Maxwell built one just fur schitzengiggles:

http://endless-sphere.com/forums/viewtopic.php?t=1001&highlight=

i wish the clyte controllers had connectors like those!!!! that is a work of art!

Seriously looks nice. Almost like something from a business jet electronics system. Do you have some pics of your entire scooter/bike/??? ?

Haven't heard much from Maxwell of late. He might be out enjoying every last drop of good weather on his bike.

Maxwell's controller is a work of art.

It should certainly be possible to build a 'bulletproof' controller.
All the things billvon has listed are possible. Just look at the RC brushless controllers! They have a microcontroller of some sort, a voltage regulator, and a power stage. Most of the functions are handled by the microcontroller. The power stage is relatively easy. That's more mechanical.

Rather than buid the thing entirely from scratch, I perfer to take a cheap one and modify it into what I want. Some of those microcontrollers have demo boards you can get that would save you the hassle of etching a board. The microcontroller approach is the wave of the future. The price of the hardware is getting pretty reasonable.

A current sensor for each phase wire (or at least two, since you can calculate the third), could be used to limit the output in the case of a fault.

In our dreams, manufacturers could provide separate "brains" and power stages with *standardized*, compatible interfaces. Plug and play...

Most of the New uP based controller have some form of current sensor in the return leg and do a pulse-by-pulse current limit.

I have seen one design that used current reporting Power MosFets...

There are also "protected" MosFets that internally limit current, and others that are just plain MASSIVE that they would be hard to hurt.

What all of the "Safe" MosFets give up for the protection is conductivity, in that they have much higher "On" resistance (Rds). The Best in the market now is in the 4-6mOhm range, some of the others are in the 120 range, a HUGE difference. That means the POWER dissipated INSIDE the MosFet is going to go up by that amount as well...

Since Power - I^2 x R
The power dissipated INSIDE the MosFet would be:
40 Amps^ 2 x .004 = 6.4 Watts.
40 Amps^ 2 x .120 = 192 Watts OUCH ! Now we NEED a FAN....

This is why people want to run at Higher voltages... to get more of the power INTO the motor, Power and Current are linear (Power = Voltage x Amps) but it is also the SQUARE of Current x resistance. By DOUBLING the voltage, you QUARTER the necessary current to deliver the same power.

This is also why people want to run larger than stock wires from the Battery through to the Motor...

12Ga stranded wire is .00532 Ohm/meter
8 Ga stranded wire is .00210 Ohm/meter
6 Ga stranded wire is .00132 Ohm/meter

The power lost in the WIRES drops by a factor of 4, as is the voltage lost to the wire, yada yada yada

I like the massive FET approach. If they're big enough, something else will blow.

In looking at some of the circuits, on of the problems is going from the controller chip's Vcc (5v-18v) the the Power MsoFets voltage (48v+) and the driver circuits, boost voltaqges and that rot...

Why can't to do away with all that rot and put in an Optical Isolation chip on BOTH sides, high and low. We then can drive the c##p out of it directly from the rails.

Not to mention that we can now Isolate the POWER section from the signal section completely, maybe even on a seperate board connected by short runs of Coax or twisted pair...on two 3 pair connectors...

Hi folks, yes been a bit busy.

Thanks for the complements on the controller, somewhere here there is all the insides too (search under author). Currently working on a) rewinding a 40X motor to give less copper loss ans b) an even more unbreakable controller. The new controller is needed as the work commute could give 100+volts of BEMF (a couple of big downhills, on the way home luckily), more FETs!!

rsisson said:

Why can't to do away with all that rot and put in an Optical Isolation chip on BOTH sides, high and low. We then can drive the c##p out of it directly from the rails.

Not to mention that we can now Isolate the POWER section from the signal section completely, maybe even on a seperate board connected by short runs of Coax or twisted pair...on two 3 pair connectors...

Click to expand...

That would work and it's a common approach on some of the larger car sized controllers. The IR gate driver chips work well, though. They can handle up to 600v on the high side.

On the high side, you still need to drive the gates 10-12v above the battery voltage, so you need a bootstrap circuit or a charge pump.

There may be some advantage to having the two sides isolated. There would be less chance of false triggering due to voltage transients.

Why couldn't we just use one of the bigger RC controllers? What about them is different?

vanilla ice said:

Why couldn't we just use one of the bigger RC controllers? What about them is different?

Click to expand...

they are ment to be cooled by large amounts of airflow, and generaly don't have a heatsink. generaly the voltage is relitivly low compared to Ebikes, and the componants are built so close together that high voltages would jump the traces on the boards. They also use a signal driven throttle, so a special throttle would need to be built.

on the up side, they are sensorless, and programable for things not thought of in Ebikes, like advance, LVC, andsoft start options.

Most of the RC controllers are limited to something like 36v. At lower FET voltages, you can get some really low on resistances, so heating should not be too bad. The larger ones have heatsinks and cooling fins.

It's not too hard to interface to the servo input with a servo tester.

Some of them have an odd startup sequence and usually require messing with the throttle to calibrate them every time you turn it on.

The maximum voltage limitation is the biggest drawback. I thought about grafting the brain off one to the power stage of a Crystalyte controller to make a high power sensorless box.

So for a 36v bike an RC controller is an option. But are they affordable?

Fecher,

Most of the IR chips come with a bootstrap built in, you just have to make sure the fets switch to 0V on a regular basis to charge it. In my controller the maximum PWM is limited to 1/1024 so there is at least a wee bit of charge every cycle. I have only killed one by putting 40+ volts on the 12V line (the other two survived).

Right. That's why you do the PWM on the high side. At low speeds, the PWM charges the bootstrap. At full throttle, the commutation does it.

fechter said:

Most of the RC controllers are limited to something like 36v. At lower FET voltages, you can get some really low on resistances, so heating should not be too bad. The larger ones have heatsinks and cooling fins.

It's not too hard to interface to the servo input with a servo tester.

Some of them have an odd startup sequence and usually require messing with the throttle to calibrate them every time you turn it on.

The maximum voltage limitation is the biggest drawback. I thought about grafting the brain off one to the power stage of a Crystalyte controller to make a high power sensorless box.

Click to expand...

Castle Creations has a 50V/125A version that I have used with 16s a123 setups. These come with pretty hefty heatsinks and are designed to be air-cooled. They are also working on a new series, for massive 50% RC planes (more like UAVs... ), that are good for 100V. One version will handle 200A, and the "big boy" will do 400A.

Knotronik also makes a model called a PowerJazz that will work up to 63V and 120A continuous and 200A for up to 15 second bursts.

-- Gary

vanilla ice said:

So for a 36v bike an RC controller is an option. But are they affordable?

Click to expand...

Ggoodrum Is more up to date than I am. The biggest I knew of was the Phoenix 125. it could handle 125 amps continous, but a max of 6s lipos, 25.2V. Still, thats 3150 watts continous. More than 4 horsepower, for as long as the batteries would hold out.

Afordable is relitive. they do more, but they're more expensive. But then there is a lot to be said for a controller with an internal resistance of 0.0006 ohms

rsisson said:

Why can't to do away with all that rot and put in an Optical Isolation chip on BOTH sides, high and low. We then can drive the c##p out of it directly from the rails.

Click to expand...

You can, but you need to get power to the FET drivers somehow. Since the FETs have significant amounts of gate capacitance, you need a decent amount of power to drive them - on the order of a watt for a big power FET that's switching above audio frequencies.

You can do this with a custom power supply that has a single forward converter primary and six secondaries, all going to feed one of the FET switches. (You could actually get away with four.) It's not trivial though, and there are other issues (like speed of the optical isolator, shoot through protection etc.)

I found a Shareware Schematic & PCB generation tool and am in the process of trying to build the skeleton of a modular home-brew controller.

As soon as I get all the obvious mistakes out I will post my plans for review. After a few cycles, I will have some prototype boards made to play with.

I am designing it with Opto-Isolation on both HIGH and Low along with all of the other things I have learned from this board.

For those interested, DIPTRACE offers a commercial level Schematic & PCB program FREE for designs under 250 leads with an EXCELENET library and I have built the custom devices for the controller and other devices.

Look for some plans & drawings in a few days.

Rsisson,

Here is the design for my controller, you can 'read' it by downloading Seetrax Ranger 2XL from seetrax.com, you don;t need a licence to view.