Arlo's power stage Leaf controller runs and drives page 103

So, you're going to use the aluminum heatsink as a contact... I don't know what the pros think, but I've tried that and had a bad experience doing it, due to aluminum oxide driven poor contact (on the "wire"-heatsink junction). I also don't know if heat-paste can cause unwanted effects on the FETs-heatsink junction. I don't think this kind of contact is reliable enough for the application, and I'm not planning on doing it again in the future.
 
Njay said:
So, you're going to use the aluminum heatsink as a contact... I don't know what the pros think, but I've tried that and had a bad experience doing it, due to aluminum oxide driven poor contact (on the "wire"-heatsink junction). I also don't know if heat-paste can cause unwanted effects on the FETs-heatsink junction. I don't think this kind of contact is reliable enough for the application, and I'm not planning on doing it again in the future.
No paste its graphite paper google pgs graphite. Using the aluminum to conduct out of the back of the fets is a proven its yow my curtis controller does it.
 
i don't know the rule on this either..but here are some thoughts..

i don't think that the mosfet case "drain" pad is really meant to conduct current. i think this because the source pin is very small in comparison, so it doesn't make any sense for the input to be much higher current rated the then output. so why is it conductive at all? probably because that leads to the least amount of thermal barrier.. since it is not electrically isolated probably that means it is direct connect, perfect for cooling! is there any harm with conducting through the drain pad? it could be akward due to the different geometry, more capacitance?

what i have seen is that the drain pad gets electrically isolated (but thermally conductive) by use of a pad. the important part of using a pad (or cream?) is that it can compensate for the none perfect flatness of your heatsink. for a real heatsink, it is suppose to be a certain flatness guarantee and polished (need to avoid hotspots). but how many DIY have a flatness guarantee? looks like some aluminum angle bracket or something is more typical, so maybe something to compensate would be advised.
 
I have a 1204 I upgraded to 500 amps by adding the rest of the mosfets. They simple conduct to the aluminum the mosfet is bolted to!
Highhopes there is a lot of people soldering mosfets to the surface for best cooling and current flow! See here http://www.ferromit.com/controller.html

and that's whats in the dunebuggy [youtube]Mvm6WN7DWY8[/youtube]
 
never said it couldn't be done, just don't think it makes much sense. :wink:

looks like wheely bars are in order! my niece wants a dune buggy.. she's like 12 yrs old or something. been kicking around the idea of using this 5HP induction motor rewound for lower voltage. dunno, too much power? i have no idea.
 
HighHopes said:
never said it couldn't be done, just don't think it makes much sense. :wink:
To me it makes sense because it allows the caps to get closer to were they need to be. If you put the third trace on the board it becomes a pita to get a great cap location.
 
recently there was some discussion about diode reverse recovery adding to the noise & votlage spikes problems.

here is the best solution .. use SiC diode which has no reverse recovery!
new module by CREE is 1200V and 60A rated. digikey sells for around $500.

now maybe this is too expensive .. but there is another use where it may be worth while which is for applications where you want to deliver high power AND high switching frequency, this module will also shine. not much switching losses with very fast switching times. so if you had a very low inductance machine that you wanted a very high switching frequency to control voltage ramps at low speed (startup acceleration) then this is a great module. hint.. for switching frequency > 20kHz, your deadtime will be a killer. it is not enough to compensate for the deadtime because the deadtime will already be introduced (but no harmonic problem cause they got compensated which is good) you must completely avoid the need for deadtime.

part reference: http://www.cree.com/power/products/sic-power-modules/sic-modules/ccs050m12cm2
 
To me it makes sense because it allows the caps to get closer to were they need to be.

good point
 
HighHopes said:
recently there was some discussion about diode reverse recovery adding to the noise & votlage spikes problems.

here is the best solution .. use SiC diode which has no reverse recovery!
new module by CREE is 1200V and 60A rated. digikey sells for around $500.

now maybe this is too expensive .. but there is another use where it may be worth while which is for applications where you want to deliver high power AND high switching frequency, this module will also shine. not much switching losses with very fast switching times. so if you had a very low inductance machine that you wanted a very high switching frequency to control voltage ramps at low speed (startup acceleration) then this is a great module. hint.. for switching frequency > 20kHz, your deadtime will be a killer. it is not enough to compensate for the deadtime because the deadtime will already be introduced (but no harmonic problem cause they got compensated which is good) you must completely avoid the need for deadtime.

part reference: http://www.cree.com/power/products/sic-power-modules/sic-modules/ccs050m12cm2
That's a lot of money for something when someone is trying to make 10+ kw for $500 or less total controller cost.
 
oh yeah, its expensive. just interesting part with interesting feature.

how much does your 2nd generation total inverter cost?
 
The problem with the silicon carbide stuff (beside price) for our applications is that it's high voltage low current.
 
So my plan is to bold 1/8" copper to the - and - on each H bridge.

But about the only way I think that will last is if I solder a flat piece on there first so the copper trace will be thick enough.

My questions are what should I do to the copper to prepare it to be bolted? Tin it? Rub graphite on it? I want to make a good connection that will not oxidize and will last.
 

Attachments

  • IMAG1073 (800 x 450).jpg
    IMAG1073 (800 x 450).jpg
    112 KB · Views: 2,604
  • IMAG1075 (800 x 450).jpg
    IMAG1075 (800 x 450).jpg
    99.5 KB · Views: 2,604
To be honest, I think we're barking up the wrong tree with an output stage like this.... I know, it's the most
logical and all the controllers we've seen are build like this. Lots of FETs in parallel, and FETs for one motor
terminal all close together.

The problem as I see it with such a setup, when all low side FETs are on all the motor currents flow inbetween
the 3 stages through the low supply, and when all high side FETs are on the motor currents flow through the
high supply. The high and low supply wiring through which these currents flow is quite extensive, it includes
all the interconnect power supply wiring between the 3 stages. Every time a switch is made (2 times per
PWM cycle) one wiring inductance is discharged and one is charged. This comes with FET killing spikes

I'm starting to lean towards the other option: build a number (4) of very compact 6 FET output stages and
parallelling thoses towards the motor and the power supply. What this means is that the motor currents are
split over each of the little 6FET stages. Inside the little stages the positive and negative supply are build with
closely spaced bus bar (for low inductance), each little stage has its own 470nF capacitor across the
supply. The result is that the wiring that is charged and discharged is very small (basically the width of 3 FETs)
and has a very low inductance. Spikes will be small and FETs will be happy.

Towards the motor the outputs of the stages are star-connected with equal length wiring, same for the supply.
The currents in these wires are not switched on/off like the currents in the local supply, and have therefore
no or little effect on the spikes.
 
Lebowski said:
I'm starting to lean towards the other option: build a number (4) of very compact 6 FET output stages and
parallelling thoses towards the motor and the power supply. What this means is that the motor currents are
split over each of the little 6FET stages. Inside the little stages the positive and negative supply are build with
closely spaced bus bar (for low inductance), each little stage has its own 470nF capacitor across the
supply. The result is that the wiring that is charged and discharged is very small (basically the width of 3 FETs)
and has a very low inductance. Spikes will be small and FETs will be happy. Towards the motor the outputs of
the stages are star-connected with equal length wiring, same for the supply.
Won't the several 6FET modules have to switch in synch?
 
Njay said:
Lebowski said:
I'm starting to lean towards the other option: build a number (4) of very compact 6 FET output stages and
parallelling thoses towards the motor and the power supply. What this means is that the motor currents are
split over each of the little 6FET stages. Inside the little stages the positive and negative supply are build with
closely spaced bus bar (for low inductance), each little stage has its own 470nF capacitor across the
supply. The result is that the wiring that is charged and discharged is very small (basically the width of 3 FETs)
and has a very low inductance. Spikes will be small and FETs will be happy. Towards the motor the outputs of
the stages are star-connected with equal length wiring, same for the supply.
Won't the several 6FET modules have to switch in synch?
Yes, a 24FET controller would have 1 controller IC, 3 (maybe 6) gate driver IC's and 4 output stages of 6 FETs each
 
Not sure I follow lebowski and I'm not sure what the benifit is.... But I was asking about bolting copper buss bar to the H bridges any idea what to coat the copper with? I have silver....
 
well, i just spent 2 hours "googling" to see what others have done, to see if i could help in some way. but i did not find anything i thought was useful.

i'm wondering if you have the PCB made without a soldermask on that wide copper trace then it would be read to bond with copper bus. a separate copper bar (thickness the same as a mosfet leg, width a bit more than equivalent parallel mosfet legs) could be cleaned, heated, tinned. now you have PCB tinned & copper busbar tinned and just need to join them together.. press firmly & evenly the basbar onto the PCB trace and then heat to melt solder to join? perhaps in a toaster oven?
 
HighHopes said:
well, i just spent 2 hours "googling" to see what others have done, to see if i could help in some way. but i did not find anything i thought was useful.

i'm wondering if you have the PCB made without a soldermask on that wide copper trace then it would be read to bond with copper bus. a separate copper bar (thickness the same as a mosfet leg, width a bit more than equivalent parallel mosfet legs) could be cleaned, heated, tinned. now you have PCB tinned & copper busbar tinned and just need to join them together.. press firmly & evenly the basbar onto the PCB trace and then heat to melt solder to join? perhaps in a toaster oven?
Thanks.

You are a trouper. I think I will solder some strips on end down from the fets to the bolt holes then I will solder 2 copper washers or a small piece of copper and drill it to accept mounting of the big bus bar I just made. It will be double bolted on each + and - connection on each H bridge so It should be ok. This will make it easy to repair for now and if this all goes according to plan I can look at making a riveted or solder connection in the future for a more of a production run.
 
Ok So I asked Dave the question about coating the copper a while back in PM and I will post his answer here I don't think that will be a problem...
bigmoose said:
I think you will be fine with the bare copper. If you solder, you will buff it, flux it and solder. If you use the graphite media, it should be fine. The copper will gradually develop a patina, but I don't think cosmetics are critical in this application. You could always tin plate it. That wouldn't hurt, but I don't think it is critical.

If he gave you liquid tin, it may be similar to the caswell kit, here is a link that may help. They have instructional documents and videos I think.
http://www.caswellplating.com/electroplating-anodizing/tin-plating-kits.html

or electroless

http://www.caswellplating.com/electroplating-anodizing/tin-plating-kits/electroless-tin-plating-kits.html
I have used the electroless at work, it is like magic. Clean immerse, instant tin plating, as they say 0.001 inches per 15 minutes.

All the best,
Dave
Dave by anymeans if this is not to be public let me know or feel free to edit it. But I'm sure its ok.

Dave edit: It's fine Arlo, no problem with you posting it. -bigMoose
 
Using the graphite media, means between copper and PCB and then screw PCB - copper?
I know of a product you put it on a piece of textile and rub copper with it, and it creates a kind of a "tin" coating, but I ear is not very durable, although it is used by hobbists for PCB coating; here it's named "pratex" but I think it's a local product only (if you google it you should find right away).
 
gee.. that stuff is expensive.

one time i saw a guy in the lab solder two busbars together for a quick test. was making busbars for an IGBT 6-pack. it was done pretty basic, just two really huge but hand held soldering irons (one on each copper bus) and then some solder (many individual strands twisted together to get one big strand) and it worked pretty good for lab test purpose, no idea the longevity. he wrapped the copper bus with something to try to keep the bars from acting like big heat spreaders thus concentrating the heat where he wanted. does this help??
 
Anyone know the change in switching times vs temp? I was just looking at the data sheets for 4468 FETs and now I'm interested in 4568 FETs after that ixfk230n20t. I think a lot of my problem has been from dead time changing as thebfets get warmer. I did find the spec showing gate threshold getting lower as temp goes up.!
 
I finished the next build finally today. Its 150v rated fets and 4 in parallel instead of 3. I tested and I have pwm to the gates so I should be up and running in the next few days.
I used graphite paper on the Low side fets and this will work well because the low side are the ones that run hotter. The graphite paper conducts heat 3-5 times what copper does. Its fun to play with you can touch it against a piece of aluminum and it makes the aluminum feel colder then the bare aluminum does on its own.
The hi side is with 1 layer of kapton or what ever you call that stuff. The cap location can't really get any better. So now I need to run tests with the scope under load. I polished the surfaces where the kapton is placed and I bead blasted the rest of the surfaces to help conduct heat away better and once the testing begins I will make a fan mount for this one as well.
I fet matched this one as well. I actually have a lot of confidence in this one. My goal is faster switching (on and off times) and 28-32s lipo with 330-360 phase amps. So I should be able to see ~250-300 battery amps at ~102v under load!
 

Attachments

  • IMG_20131117_194059.jpg
    IMG_20131117_194059.jpg
    65.9 KB · Views: 2,363
  • IMG_20131117_194111.jpg
    IMG_20131117_194111.jpg
    53.1 KB · Views: 2,363
  • IMG_20131117_203128.jpg
    IMG_20131117_203128.jpg
    38.4 KB · Views: 2,363
  • IMG_20131117_203148.jpg
    IMG_20131117_203148.jpg
    53.1 KB · Views: 2,363
  • IMG_20131117_203203.jpg
    IMG_20131117_203203.jpg
    57.3 KB · Views: 2,363
  • IMG_20131117_203212.jpg
    IMG_20131117_203212.jpg
    57.8 KB · Views: 2,363
  • IMG_20131117_203224.jpg
    IMG_20131117_203224.jpg
    70.2 KB · Views: 2,363
  • IMG_20131117_225131.jpg
    IMG_20131117_225131.jpg
    66.8 KB · Views: 2,363
  • IMG_20131117_225204.jpg
    IMG_20131117_225204.jpg
    43.9 KB · Views: 2,363
that looks really good arlo, appreciate sharing the pictures.

i can help you exactly with correlating junction temperature with switching frequency using mathematics. unless you are asking about bench testing, i have never used that mosfet before so can not say. will this math help you ?

you will be hard pressed to get 300A through 4 parallel mosfets. the case lead is limited to 75A, and that is under ideal conditions with DC current. 4 in parallel is 300A... there is no derating here, but there should be. in fact, there MUST be, even if you match the mosfets (just less deratinng if you match).

good luck!!!
 
Back
Top