Home Brew Controller, what would it take ?

[rsisson]

all,

I JUST finished my first pass at a controller.

It's been 20 years since I did this, so there are bound to be edits...

I optically isolated the MosFets from the Controller
I use a 555 to generate BOOST voltage
I used Optical isolation on BOTH the top and Bottom
I used the New On-systems MC33035 Motor controller
I have a jumper in for separating the Motor return and the current sense as one needs to be BIG while the other is just a signal.

I am not sure about the "closed loop" option, it is Mostly there, but I may need a few more components.

I may put in some more zeners near the FETs for protection, and I need to put in some Electrolitic caps for power filtering, along with more filter caps on the power lines and put values on everything, but I wanted to get the outline done first.

As you can see, the component count is very low... and I think it will be robust....

The layout seperates Power and siganl well, so it should be possible to put the FET's on a sub-boardwith a signal cable to it.

This is only DRAFT 1...so feedback is appreciated.

 

[fechter]

Cool.

The resolution of the image is not quite good enough to make out the details. Are those zeners across the gates?

I have no idea if the optocoupler outputs are fast enough or have enough current to drive the gates. I assume you checked their specs.

The board layout looks nice. Keep in mind the mechanical mounting of the FETs on the heatsink needs to accomodate some thermal expansion. Putting a "L" bend in the leads is the usual approach. Constant expansion and contraction leads to fatigue failures if the leads flex beyond their elastic limit.

 

[maxwell]

Have I missed the obvious? won't both output FETs turn on at the same time?

And I think your turn off times will be very slow (= hot FETs) you need to pull somewhere near an amp to turn off a big FET quickly enough.

Do take a look at the IR drivers it will save a lot of hassle, you have to connect the controller and FETs together somewhere might as well be on the controller board. I use the IR21834 with seperate pins for FET 0V and logic 0V (within 5V anyway) and shoot through prevention. If you really want to use optos I would drive IR21834s with them, some logic optos are quite fast with (funnly enough) logic outputs.

An last of all, use SMT devices as a) they are available, b) less holes to drill c) shorter track lengths (more than a couple of inches on a FET gate and they oscillate with a bad outcome) and d) cheaper.

 

[fechter]

You could put the optocouplers ahead of a IR gate driver chip :wink:

 

[billvon]

all,

I JUST finished my first pass at a controller.

Issues I see:

(Can't really read the schematic so these are based on guesses on what the circuit is.)

FET turn-off will be slow since you only have a pull-down resistor. If you turn off the top FET and it's not completely off before the bottom one turns back on - bang! Blown fuse at least.

A passive pull-down would work for very slow speeds, but for faster motor speeds (especially if you want to PWM) you'll need an active circuit. It can be as simple as a 4050 buffer; they have pretty good drive characteristics. Or it can be a dedicated FET driver. They work better and have more protection.

I didn't see the supplies for the top FETs. Each FET needs its own supply - you can NOT use the same supply for all of them, since it will reference the leg voltage.

 

[rsisson]

I will re-post at a higher resolution if it will let me...

The Upper and lower Fets ARE separately controlled, on the schematic the tool created a BUS where they overlap.

The Chip (u7) has separate outputs for Atop and Abottom.

The Upper Drive Circuit comes right out of the ON-Semiconductors application note for High voltage Interface to N-Channel Power MosFets.

I replicated it for the lower side, so It may need some additional tweaking.

The High side uses a "Boost" voltage which is the Sum of the Motor voltage and the Vcc, or 48+24v...(Vboost)

Not all the supply rails are shown, they are (Vmotor) raw off the batteries, Vcc regulated for the controller and chips, and Vboost to get the gate voltage high enough to overdrive the Fets.

Some of these bits are found in the application note at:
http://www.onsemi.com/pub/Collateral/MC33033-D.PDF

As for surface mounting...not likely... I can't see that well and this is going to be a "HOBBY" board. I can get the board made as a prototype for around $50, and that is fiberglass, 2 layer, painted and labeled.

I can export the schematic in DXF, Netlist for a variety of programs, or Spice Netlist if anyone want to play.


I have been chasing "Dumb" mistakes all day so there WILL be frequest revisions...

 

[fechter]

I can at least read it now.

Two things I can see:
you should have a resistor in series with the gate to prevent oscillations.
The gates need to be actively driven by transistors both high and low. A pull down resistor is not going to be fast enough.

 

[rsisson]

The Gate drives were the next dumb mistake...

The zener and resistor were supposed to go to the drain of the upper leg...

 

[rsisson]

Ok... I will re-draw it with the IRS218's driving the MosFets rather than the Opto-isolators.

Question... should I use BOTH the Isolators AND the Drivers, or are the drivers robust enough to protect the controller ?

Also...I found TO-220 SOCKETS... would they be worth using, or is the current load to high?

Lastly, using the driver chips, what does it take to Parrallel MosFets? Does the driver have enough umph, or should I use a chip per MosFet?

 

[fechter]

The gate driver chips are usually rated for up to 600v. I'm not sure what would be gained by feeding them with isolators (I'm not saying it's a bad idea).

I have seen the driver chips blow up after a FET failure, but the commutator feeding them was intact. I suppose some kind of protection between them might be advisable.

You can do the math on the gate drivers and FETs to figure out if you can pair them up. Based on experience with the IR2101's, I'd guess you can parallel at least two FETs off of each driver.

Sockets? I don't think they would be happy at over 10 amps or so. If the rest of the design is good, the FETs should never blow (in theory).

 

[lawsonuw]

Have any of you guys looked at the OSMC? This is a proven motor controller than can handle 300Amps at ~48V. The cool thing is that it's Open Source so the schematics are freely available. The power stage from the OSMC would be a really good reference for an e-bike controller's power stage.

http://www.robotpower.com/downloads/
http://tech.groups.yahoo.com/group/osmc/
http://www.robotpower.com/downloads/osmc3-22sch-clean.pdf


Oh yea, regenerative braking is really easy to do. With a FET driven PWM speed controller Synchronous Rectification enables the two way power flow needed for regeneration. So for power from the battery to the motor the controller looks like a Buck power converter. Reverse the power flow, motor to battery now, and the same circuit looks like a Boost power converter. (assuming synchronous rectification is used) This works with DC motors and should work just as well with a BLDC motor.

Sincerly,
Marty

 

[rsisson]

Ok, working from the OUTPUT backwards...

How about this for 1-phase?

The IRS2183 uses INVERTING input for the lowside, so I used the Opto-isolators to Invert the LO side only (Pull up vs pull down)

I show 4-Mosfets, but you could use 2 or 4. The only difference would be that the C2 cap would need to be larger for 4 Mosfets as it would be feeding twice the boost current for the high side...


Let me know what you think....

 

[fechter]

That looks good.

Only thing I would check on is the switching speed of the opto's. Some of them are a bit slow. Fast ones are available. They would guarantee brain survival in the event of output stage failure.

 

[rsisson]

The robo controller looks interesting... Looks like a 2-phase controller with no buffering between the Controller and the MosFets...

My idea is similar..use an off the shelf controller, without the need for any software... and make it modular...need more current, add more MosFets.

If it is all done here, I would leave it here, Ie public, and it would be a HOBBY controller meant to be more robust than usual for its size, and CHEAP and easy to build...

If I can figure out how to do it, you could make the 1=phase modular, keep adding them until you have the current you need... the CHips will go to 600V...

 

[rsisson]

Most of the ones I was looking at were in the 2-4us range....

Those were analog... the digital output ones were faster.... but they wern't as robust or flexible in voltages

 

[maxwell]

risisson

I would turn the opto transistors the other way up, assuming Vcc is a positive voltage. Or look for logic optos, no need for slow pull up(down)s. To protect the brain part of a motor controller you can simply put resistors in series with the inputs to the drivers, I never have and despite a few blow ups never killed the micro.

I would still recomend the use of the IR21834 driver for it's driver and logic seperated comon lines and the shoot thorough protection, it is so easy to write the wrong bit of code and BANG. The com/vss difference can be up to +/-5V with full operation and more without blowing it up. Handy hint... some of the 14 pins are not connected these can be used to simplify tracking on a PCB.

Also put (big) shottkey diodes across the FETs they reduce the PWM losses a lot, my controllers have never even got warm let alone hot.

All this comes from experience with 3 controllers so far with a fourth in the pipeline, just about got it right now.

A home designed and built controller is NOT the cheap option but can be the best.

To give an idea of what can be done my MKIV has...
100V
140A
Serial link to the throttle/display
Serial link to the battery
400mA LED driver
Temperature sensors

The display can show...
Battery volts
Battery current
Motor/battery/heatsink temperatures
Ah remaining (or used)
Speed
Distance
Mode (I have a 250W 15mph limit for the UK that is switchable, this is motor output and therfore takes efficiency into account)
Individual cell (parallel sets anyway) voltages

The battery pack also has a disconnect relay so if I go down hill too fast (for either charge curent or cell volts) it is disconnected from the controller.

And anything else that can be calculated from...
Battery current
Battery volts
Controller rail volts
Motor RPM
Time
Throttle position
The three temperatures

 

[rsisson]

Thanks for the inputs...

Part of this exercise is to exercise my old tired brain... It is getting COLD here now...34deg F this morning, a bit nippy for a spin, so a good time for tinkering.

I will look into the logic isolators, and the IR21834's unused pins sound interesting.

Another thought I had was putting the MosFets on small "Sub-cards" with Power edge connectors. Each track on the connector can take 30A...very modular, and good for tinkering...not really practical, but fun to think about...

 

[fechter]

A home designed and built controller is NOT the cheap option but can be the best.

Absolutely.

We can hope that someday a design like Maxwell's would go into mass production so the price would be much lower. There are very few decent controllers on the market in this voltage / power range.

 

[lazarus2405]

We can hope that someday a design like Maxwell's would go into mass production so the price would be much lower. There are very few decent controllers on the market in this voltage / power range.

Perhaps when you EEs are done, you could contact one of the players in the industry about just such production. Perhaps it's wishful thinking, but one would think that a competitive company would have a hard time turning down a product for which the target audience has already done all the R&D. Think of it as outsourcing the engineering to the customer. Hell, hobbyists don't mind, especially if in the end they can get a product specifically tailored to their needs.

 

[rsisson]

First, International Rectifier JUST released a new 3-Phase High voltage Gate driver IC. IRS23364D

It has 3 charge pumps, its own current sensing stuff, and more.

It removes several components from the mix... Yea.

Question....

Should the Charge pump capacitor be roughly the same capacitance as the MosFet Gate capacitance, larger or much larger ?

For the IRF4110 the Gate capacitance is 9620 pf, or ~10nf. Is that the size capacitor I should use for the Bootstrap charge pump capacitor?


next, someone recommended LARGE schottky diodes across the MosFets...Uhm, how large is LARGE... they come pretty big, and they are not expensive....

Bob

PS... when I am done with this I WILL be posting the schematic, drawings, traces and so forth here, along with information so that people could order them. I get get 2-boards (bare) for $30... so they might be really cheap in moderate quantities...

 

[maxwell]

I had a look at the IR three phase drivers, the main problem is trying to track them on the PCB, at least one pair (quad) of FETs ends up a long way away. Also, the drive is a bit low for multiple FETs, if you blow it up all three go (I have got home at least twice on two phases before), more legs to de-re-solder when changing, more expensive and harder to get.

They are OK for high voltage three phase applications where the current is lower but not realy suited to our requirements.

Shottkeys, I use 100V 2X20A types paralleled (one device per FET (or however many you parallel)), they come in a TO220 case, never killed one of those!

 

[rsisson]

OK, I allowed enough space for either a really big diode, or another to220.

Hum, about the "distance" from the controller to the MosFet...

Other than slowing down the PWM freq, how do the really big controllers remote mount their MosFets? the frequency of the PWM is in the 10Khz range, adjustable from 2-80Khz via your choice of RC parts... At those frequencies, their shouldn't be a problems.

The IR chip has separate Gnds for the signal and Power sides, so some caps will get the spikes, along with your Schottky's. As for driving multiple MosFet's, don't have a solution for that...

A thought was to stack multiple boards with through connectors ....
Want more current, add another board...Board fries, take it out and replace it... You connect the High power traces with metal threaded standoffs.. The fan out of the CPU is about 10, so it SHOULD drive about 10 sub-boards... The logic connection between boards is low-speed and could be standard IDC stuff. Each sub-board would be well within its "Safe" range at all times, but the Assembly could deal with MASIVE amount of power

10 x 50 Amps/board is a LOT.

All that would be on each board would be the 3-phase driver, its caps and stuff, 6 MosFets, and the Current control for THAT board. An IDC for signal in, and one for out. A Standoff Pad/Connector/mount for the power, and Each phase out and a duplicate to carry to the next board. The Fault line is open collector, so you could "or" them so if any board had a problem, they would ALL shut down...

The 4110 should be good to about 80V...80v x 500A... OUCH...

Substituting a high Voltage Mosfet could raise the voltages to crazy levels... There is a 120V 80A 18mohm Mosfet option.. As EACH mosfet is now carrying less, a larger Rds is not as big a problem...

You would connect the leads to the top/bottom board with ring connectors to the standoffs... You would have a modular power "Brick" A small fan and you could probably run a CAR on it...

Now I am sure there are reasons it WON'T work.... but its an idea...

 

[rsisson]

Here is a ROUGH idea of what Envisioned as a STACKABLE board

The IDC's would go from board to board, with the last one going to the controller board. Even if you needed to put some pull/up/down on the board and a buffer on the main CPU board, it still would make a powerful assembly.

The #8 holes would connect each board to its neighbor with a Metal threaded standoffs, staggering each board.

The easiest method would be to have the standoffs a TO-220 tall.

Each of these boards could take 80V and 50A. (trace limitation, fet can go to 180A..Ha !) The standoffs are heavier than a #6 so they wouldn't be a limitation.

Obviously, the board could be made smaller still, shrinking the distance from the driver to the MosFets, and rearranging stuff. But just an idea...

 

[fitek]

Wow I wish I had waited for this thread about six months ago. I am not an EE, I am a CSE, and I have been piecing my own homebrew controller together for about half a year now. I have to admit I am pretty ignorant on driving the mosfets. Looks like I just have one more issue to overcome though (if you read on).

My board is based on an Atmel ATMEGA88. It takes as input direction, throttle (via a pot), and the hall sensor outputs. It outputs control signals for 3 h bridges and has protection in software against shoot through.

The software works, but I'm still working out the connections to the MOSFETs. I tried driving them directly. That didn't work. So I bought IR2101s, but I am still having a problem because the "logic" inputs on the IR2101 take 10v-20v as input. The ATMEGA88 runs at 5V. I'm not sure how to boost the 5v to get another 5 to 15v.

It does make my motor spin. Just the output from the power board is really wimpy.

I've tried to keep the part count low since I'm terrible at soldering, and the cost too, because I'm cheap. So far I'm at about 40 bucks I think. The OSMC is pretty expensive.

This weekend I will have to read through this discussion more thoroughly.

 

[rsisson]

I am WAY out of date, thats why I am asking for help here... but I can tell you what I know...

The Bottom MosFet is not neccessarily the problem, as you have the full output of the ATM to ground through the gate...

The problem is the TOP MosFet, as it is more than 5V above ground, so your logic output doesn't drive it strong enough. Typically you take the HI output and take it into a BOOST circuit that takes the Gate signal Above the Top rail voltage enough to make sure it is solidly turned on.

Take a look at Figure 24 on page 13 of the following aplication note.
http://www.onsemi.com/pub/Collateral/MC33035-D.PDF
And then at Figure 25 on the next page. You can use a regular transistor there and get the same effect...

(The entire note is worth reading)

There are several other ways to deal with "Vboost" but almost all of them require at least a CAP, a Diode and another transistor or a device to charge the Cap. When the Upper MosFet is Off, the Cap is charged to 5v. When it is turned ON, the voltage that is applied is the Motor voltage PLUS the 5v on the Cap

Some others can be found by googling Mosfet Bridge and your ATM part # like...
http://homepages.which.net/~paul.hills/SpeedControl/SpeedControllersBody.html