My new 18 FET TO-247 layout riding video page 10

[peters]

Thanks for your comments, I was just showing my board as another example for horizontally placed MOSFETs. It is quite rare as I see, in most controllers (as the usual Chinese ones) they are vertical.
My design, as always, is a trade-off among power, efficiency, dimensions, compactness, connector placement, robustness, etc. With my board I did not want to go beyond 120V, 50A battery and 150A phase current, but I wanted to make it small, and it is fine for my application. I did a lot of measurements and tests, there is really some noise due to parasitic effects, but acceptable. It is robust in the range of my specs and I've been using it for 1.5 yrs in my bikes without issue.
Of course my power bus is laminated, just not rectangular-shaped, as I tried to describe. My gate drivers were placed as proposed in an International Rectifier app note, in order to reduce the effect of the stray inductances in the driver loop. The Miller cap problem is compensated by slowing down the turn-on of the FETs and a small value turn-off resistor, it is acceptable on my board. Stronger clamping would help to improve efficiency, of course.

I'm an EE, so not fully dumb for the topic, and I know what I am doing (usually) In this profession there is always something new to learn, noone can say he knows everything, so good to beware of the EE guys who are too self-confident.
My first power design was a DC motor controller, then I burnt a lot of FETs and some gate drivers, that is obvious, but none on this 2nd board so far. As I'm specialized in microelectronics, before these projects 100amps was close to infinite for me.

 

[Lebowski]

hey, I'm in microelectronics too, what do you do ? I'm an analog / mixed signal IC design engineer...

 

[peters]

I worked on ASIC layout for a short time, then in FPGA and digital design for years, but now I am on a long vacation

 

[zombiess]

Thanks for your comments, I was just showing my board as another example for horizontally placed MOSFETs. It is quite rare as I see, in most controllers (as the usual Chinese ones) they are vertical.

Out all all the layouts I've designed trying to meet the strict layout rules for power stage I think I like this configuration the best because it can scale well. I could add a 4th or 5th device without a lot of work if this design works well. It allows for a large aluminum heat spreader bar that then can connect to a finned aluminum case. I'm surprised by the weight of everything, heavier than I thought. I don't think my thermal setup is good for continuous high power unless it has forced air cooling. The thermal mass of everything should make it decent for some high power bursts which is what I am after in this size. Makes for a very compact layout. Your controller must be pretty compact since you laid the MOSFETs flat as well. What kind of case did it end up in? I'm trying to find a larger heat sink type case for another project.

I've got 2 different Chinese controllers here that have the MOSFETs laid flat, but 1 of them is a large RC controller, only ones I've seen do it.

 

[crxguy52]

I don't think my thermal setup is good for continuous high power unless it has forced air cooling. The thermal mass of everything should make it decent for some high power bursts which is what I am after in this size. Makes for a very compact layout. Your controller must be pretty compact since you laid the MOSFETs flat as well. What kind of case did it end up in? I'm trying to find a larger heat sink type case for another project.

For what it's worth, I looked for a while for a case that had decent cooling. I found that all the cool kids make their own extrusions, since they're making thousands of controllers. Not so well suited to us DIYers. The solution I came up with on my controller is a decent compromise, but the next controller is going to use a curtis style layout, since it allows any sized heatsink to be attached:

The first one is the one I'm actually basing my controller on. That bottom plate attaches to a giant heatsink and a plastic case seals against the heatsink.

 

[liveforphysics]

It is important not to confuse the things you find inside a Kelly controller with good engineering design. Zombiess power stage is already radically better than that Kelly, not to mention the Lebowski brain driving it is ahead of ALL production controllers.

I don't think my thermal setup is good for continuous high power unless it has forced air cooling. The thermal mass of everything should make it decent for some high power bursts which is what I am after in this size. Makes for a very compact layout. Your controller must be pretty compact since you laid the MOSFETs flat as well. What kind of case did it end up in? I'm trying to find a larger heat sink type case for another project.

For what it's worth, I looked for a while for a case that had decent cooling. I found that all the cool kids make their own extrusions, since they're making thousands of controllers. Not so well suited to us DIYers. The solution I came up with on my controller is a decent compromise, but the next controller is going to use a curtis style layout, since it allows any sized heatsink to be attached:

The first one is the one I'm actually basing my controller on. That bottom plate attaches to a giant heatsink and a plastic case seals against the heatsink.

 

[HighHopes]

I've got 2 different Chinese controllers here that have the MOSFETs laid flat, but 1 of them is a large RC controller, only ones I've seen do it.

tesla motors lies their 14 parallel IGBTs flat too.. so there must be something to it.

 

[zombiess]

tesla motors lies their 14 parallel IGBTs flat too.. so there must be something to it.

I like my MOSFETs like I like my women... yup, something to it alright.

 

[peters]

Your controller must be pretty compact since you laid the MOSFETs flat as well. What kind of case did it end up in? I'm trying to find a larger heat sink type case for another project.

Unfortunetaly the large electrolytic caps are still high, so not as flat as I would like, about 50mm altogether.
On my 2nd bike (in my signature) it was in a hammond 1590WR1FL box, but it was not ribbed. I selected this one because it was sealed. But that worked at a lower power, only 1400W. Air cooling from the wind was fine, the max. internal temperature was about 70 degrees celsius.
In my 3rd bike the controller is not boxed, but mounted on a large aluminium L-profile together with the BMS, and this "power module" is mounted inside the frame under the top plate, contacting with its whole surface, that is about 400x80mm. The steel frame is not a good heat conductor, but the surface is large, so the frame is the box in fact. Until now the max. temperature raise in the controller above the ambient was 40 deg. (celsius).
I choosed this way because I prefer to have the controller inside the frame, but there may be a problem in the summer hot weather, especially if it is in direct sunlight for some time, I will see...
But I need to dissipate only a few percent of 4000watts (maybe 200watts, I did not measure the efficiency), and it never lasts for very long time. For several hundreds of amps this method would not be enough for sure. I'm thinking to put a fan inside the frame, which would circulate the internal air and spread the heat from the MOSFETs around everywhere, then no need to open the frame for ventillation, so it would remain waterproof.

 

[Alan B]

Is there any consensus on a good driver to use with these FETs?

 

[HighHopes]

that's actually a really good question Alan. practically any gate driver will do if you are below 1kW, maybe even below 5kW under good conditions. but as soon as you hit >10kW, the game changes significantly and you really start to notice the difference between a good gate driver and all the rest.

in terms of most robust, there is no question that IXYS supplies the world's best. they are also very fast. unfortunately they acheive robust by eliminating a lot of functionality which you must have in most high performance drives.. so you have to add that functionality in as an external analog design. makes the gate driver a bit bigger and a bit more expensive (like $30 more), but it will be bullet proof (i almost mean that literally).

TD350 from STmicro, you wouldn't expect this gate driver chip to be as good as it is.. but it is. and it is well proven in industry for 10+ years. the device is now on the "mature status" list too. this device does not switch fast, so 20kHz is about as high as you can go in reasonable way though i think you could push it to 25khz or maybe even 30kHz under good conditions. this is the chip i suggested for zombiess in his first gate drive design that worked.

Infineon is supplying the gate driver chip i am using (for the first time) but it looks like a good one, part number 2ED020I12FA.

Alegro has been trying to get into this market for long long time and i think just now they might have something worth talking about; part number ACPL-333J. i suggested it for Arlo cause it met all his needs. we'll have to wait until he is done his design before we know it is any good or not. but i have some faith in Alegro, and they have been seriously trying to make successful gate drive IC for long time so hopefully they finally learned how to do it.

the above list should be enough to support any high performance motor drive above 10kW and less than 100kW. so no need to experiment with unknowns.

having said all that.. having a good gate drive IC is not enough for a working gate driver.. but it is for sure a step in the right direction.

 

[Futterama]

Highhopes, is it possible to use any of those drivers without isolated supplies? Is it possible to supply them using bootstrapping?

And by the way, IXYS website is crap. Most parts are obsolete, links from datasheets is not working, IXYS parts in the datasheet cannot be found using the part search ect.

 

[zombiess]

The td350e driver I use can be boot strapped. I choose not to so I can maintain good isolation. It would save me 5 isolated supplies which would translate into $10-20. I am building one off so saving $20 vs having the best design and higher chances of success matter more to me.

 

[Futterama]

zombiess, can you post a link to your supplies?

 

[Alan B]

Thanks for the suggestions HighHopes.

I looked at these and a few other drivers. So far the one that appeals to me most is the 333J.

The Avago ACPL-333J that was suggested above looks to be excellent. It has built in optical isolation, miller clamping, desaturation detection with isolated signal back to the micro, and is readily available from DigiKey. But it is only one side (high or low side) and is a bit expensive. Six required.

The Infineon didn't appear to be available so I didn't look at it in detail.

The TD350 also looks good (requires external optocouplers and/or transformer), has desat detection and miller clamp, and the price was about half. Readily available from Mouser.

So for minimum parts count the 333j is better. It costs more but by the time the external parts are added the TD350 may not be cheaper.

I'm considering building a 6 or possibly 12 TO247 FET power stage for learning and trying some motor control programming, so these drivers may be on the high end, but I like the hardware protection as software development may result in some chip-destroying moments without good hardware protection.

 

[zombiess]

Alan,

Not sure if you realize it or not, but I have already completed and posted ALL the work including schematics, PCB layouts, etc to get a TD350 up and running with a layout that has been tested. You can see 6 of them pictured in this thread.

It would be much faster for you to tweak my debugged setup to fit your needs, it's pretty compact. You won't need the boost stage, technically neither do I but I needed to build with it to verify.

If you want to use my layout I need to repost some of the files as I have made some changes from issues I found from my first run.

I'm working on a new Lebowski controller layout and I finished up a 2 board stacked layout design last night that is 1.5 x 3.0 x 0.75", so nice and compact. I was thinking about making a low power (burst to 200A phase max) 6-12 FET controller for my wifes bike (depends on enclosure available). This new layout was done because I need to fit it in my 18 FET controller and I have very little space to work with.

 

[Alan B]

I should look more closely at what you have done.

I'm thinking about driving it with a Teensy 3.1 CPU board, a 72 MHZ ARM, and have an open and well supported platform to do some software experimentation.

 

[zombiess]

Why the Teensy platform?

 

[HighHopes]

can these drivers be run with boot strap? yes i think they all can, but hold on a moment. usually bootstrap design is for cost sensitive, low performance. if you want low cost but high performance you will run in to the same problem that has plagued this forumn for years which is, inverter fails or must operate at lower power, or must operate at lower performance.

so you have to have it clear in your mind what your application needs vs your design priorities and then find the optimal solution. if you want 300A high performance and then cheap out with a bootstrap, don't be surprised when it doesn't work. but if you want 300A and you put 400A worth of silicon with a killer thermal heatsink and then switch super slow... maybe bootstrap can work. so you see it can (probably) be done with bootstrap and high power but the design has to be thought out and be balanced to make sense.

 

[zombiess]

so you have to have it clear in your mind what your application needs vs your design priorities and then find the optimal solution. if you want 300A high performance and then cheap out with a bootstrap, don't be surprised when it doesn't work. but if you want 300A and you put 400A worth of silicon with a killer thermal heatsink and then switch super slow... maybe bootstrap can work. so you see it can (probably) be done with bootstrap and high power but the design has to be thought out and be balanced to make sense.

The short answer is, don't do it because it's not worth it if you are after high performance, and if you are designing something on your own, you are:
1. into learning
2. nothing off the shelf meets your requirements

 

[Alan B]

Why the Teensy platform?

Good question.

I want to do an ARM based embedded project as a learning experience. I've done many others, but nothing yet on an ARM. As an electronics and software engineer this is an interesting and fun project for me. I'd like to have a controller I had more control over.

I'm looking for an ARM board that has open software and is easy to work with and will be available for awhile. That lets out all the cheap OEM boards that are gone in a flash. Paul, the maker of the Teensy series of boards is involved in a bunch of projects and provides a lot of good support libraries for his boards, and the code is open. He continues to make them over a long period of time.

I have a fancy TI InstaSpin motor control setup, and I'm not happy with the way they tangle up the software licensing. The hardware is good, but they discontinue it faster than you can learn it, and the software licensing is complex and potentially expensive.

The Teensy 3.1 appears to have the needed motor control hardware support, but I'm still researching the details there to make sure it doesn't have some fatal flaw. It has a Cortex M4 at 72 mhz, supported by both open Arduino and standalone GCC environments with a fair number of I/O pins and pricing under $20. I've used other Teensy processors and experienced short learning curves and rapid development cycles. The TI kit is faster to get started on with all their motor control GUIs and libraries, but you don't learn much about motor control, just about using TI libraries. I want to learn more about the lower level details. I want to get away from closed source tools and library codes and try some things myself, and be able to share that with other interested parties.

If the Teensy doesn't pan out, I will choose something else. This is merely a familiar starting point that appears to meet requirements, but is subject to change as I verify that.

I think there is a place for a good trapezoidal controller. We have less than ideal trapezoidal controllers and we now have a bunch of improved sinewave controllers, but we don't have any great trapezoidal BLDC controllers, and they should be simpler to make than the sinewave models. The Teensy could do sinewave control (as far as I know), but that's not my interest now.

So, at the moment, I'm motivated to try something along those lines. I'm thinking 6-FET with TO247's, really simple. Not really quite like what you are building - not pushing the power levels at all. Whatever six well driven TO247's can do. Something with good hardware protection (to support safe software development), but no paralleling of FETs to worry about. Keep it Simple. Something that can be programmed easily and tested daily on a real commuting ebike, so it must be packaged for on-bike use. (I need to look at that aluminum box you are using, it looks excellent for this, maybe a smaller version - good choice.)

That's my thoughts at the moment. Subject to change, of course.

 

[HighHopes]

but we don't have any great trapezoidal BLDC controllers, and they should be simpler to make than the sinewave models.

that could make an interesting project. a year ago i wouldn't even discuss it, but now i have a better appreciation for cost sensitivity.

if i were doing a trapezoidal BLDC controller, i would design it to be slow switching, bootstrap, self commutating, NO DIGITAL CONTROLLER. on the cheap, but robust (for what it is). anyway.. we're hijacking zombiess' thread .. again

 

[Alan B]

I hope Jeremy doesn't mind.

In fact, this (a non CPU DIY controller) has been done on ES by another Jeremy (Haris?).

He used a controller chip and made a super simple controller. Perfect for a simple roll your own. He had trouble with the bootsrap starting up, I don't recall how he solved that. The thread is around here somewhere.

I'm aiming for something with much more refined throttle control and a few programmable features. But if my motor has trapezoidal back EMF, sinewaves aren't really the best waveform, and from a learning perspective, probably not the best place to start either. Maybe work up to that eventually. That's what Lebowski did, but he did it in PIC assembler and he's not sharing sourcecode (which is fine). I'd rather work with something in C on an ARM.

Anyway, thanks to everyone for these excellent threads and all the expert advise, I don't normally work with this type of electronic design, so I'm soaking up all I can. Since I like to write embedded software, design electronics and PC boards, and make things open and available this seems like a good project. A modest but bullet proof output stage is important for software development, as well as for reliable commuting. I've just had my 24 FET commuter ebike controller fail again, and I'm not satisfied with the throttle control either. My average power is under 1kw but peaks hit 5-6kw at 75V, so it is pretty modest to design something for. Now I'm running a 12 FET at 3kw peak and it really bogs down on the 15% grades.

I could build Jeremy's design but it seems to be significant overkill. It would be fun to have all that power, but it probably would get me in trouble.

 

[HighHopes]

what is your concern with instaspin? i did not notice any licensing issues when i read about it, seemed to me that the software is freely available. some of their code is proprietary and you can't see it, but i do not recall reading anything about licensing. can you explain?

 

[Alan B]

Their software tools and libraries are severely licensed, and quite expensive. When you buy one of their kits you get some access to the software, but it is limited. The one I have doesn't include the newer features, for example, there's no free or inexpensive software upgrade, so I would have to buy another $300 kit. Their debugging tools are expensive as well. The kit I purchased a year ago is already discontinued and it will soon not be possible to buy more of the ControlCard CPU cards that it contains. So if I put effort into that it will not be shareable for very long. It is probably fine for a business to develop a quick product, but not a great fit for DIY if you want to build more than one. It is much less open than it appeared to be.

Not to say there is anything wrong with it, but it just doesn't fit well with where I'd like to invest my time and energy.