low inductance output stage contruction
- Thread starter Lebowski
- Start date
Futterama
1 kW
Take a look at the pictures again, those where it is all finished, there is a piece of mica under the FET's :wink:
Futterama
1 kW
Yes, I used the one Lebowski linked to and I also tried another one, same issue. Damn.bearing said:
Lebowski
10 MW
I do not use the heatsink as a conductor, I put little mica plates under the FETs...
About the inductance calculator. What is important is the trend when you change a dimension, not so much the actual inductance.
From playing around with the calculator (or looking at the equation for the inductance) you can extract simple design rules like
as short as possible, wider is better and put as close together as possible.
About the inductance calculator. What is important is the trend when you change a dimension, not so much the actual inductance.
From playing around with the calculator (or looking at the equation for the inductance) you can extract simple design rules like
as short as possible, wider is better and put as close together as possible.
HighHopes
10 kW
- Joined
- Mar 28, 2013
- Messages
- 929
when the screw goes through the heatsink to mount mosfet.. does it touch ever so slightly the inner surface where it passes through the tab and (assuming tab is drain connected) thus become a conductive path to the heatsink via the screw? i ask because i did what you did once and did not realize i had to also isolate the screw until after i had trouble int he lab.
liveforphysics
100 TW
For the hot-rodding-type folks (like myself) who see this as another puny 6-fet and aren't impressed, this is a path forward for powering motors that isn't intrinsically expansion limited like conventional typologies.
If you want to make a mosfet based DIY 500hp BLDC controller, groupings of these modules could potentially do it. Our traditional topology doesn't really work well with respect to current sharing, and because of that it's scaling-up to higher power levels has a bit of a soft-ceiling on it.
A topology like this doesn't inherently have a ceiling as I'm seeing it. That's pretty exciting for hot rodding.
If you want to make a mosfet based DIY 500hp BLDC controller, groupings of these modules could potentially do it. Our traditional topology doesn't really work well with respect to current sharing, and because of that it's scaling-up to higher power levels has a bit of a soft-ceiling on it.
A topology like this doesn't inherently have a ceiling as I'm seeing it. That's pretty exciting for hot rodding.
Yup this is what I was talking about with the big stator I have. Maybe splitting it into 12 or 15 different 3 phase motors wound on one stator. Have each one on a fuse... If its one solid stator it would be ok to have dummy sections with a master section where only one section does the current measurement and coil position detection while the rest are a perfect copy of that.
liveforphysics
100 TW
You could match top fuel power densities with 27 of this module properly cooled (LN2? Dry ice?) forming a 9 lebowski stand alone units and a motor with some multiple of 9 teeth and each set kept separate so the motor ends up with 27 phase leads in 9 sets of electrically separated motors to power with the 9 stand alone, phase current controling and BEMF sensorless 6fets. Maybe throttle and enable signal to each unit is done over fiber. Intrensicly balanced current sharing possible with out the need for symetry in layout/packaging that makes the alternative approach too impractical to manufacturer. That type of array seems likely to become a high power EVs drive geometry to enable intrinsically safer lower voltage and higher current EV drives. It's going to be more efficient, cheaper, higher reliability with comparatively safe fails while-running situations (can limp on remaining non-failed functioning stages).
http://www.digikey.com/product-detail/en/MIXA450PF1200TSF/MIXA450PF1200TSF-ND/4321809
I might build a controller with these. $450 for 3 is not bad, $75/half bridge that does >400A at 1kV and has package isolation and die temp monitoring and massive flyback diodes seems like a bargain. I'm personally into non-leathal voltages for anything that is daily transportation stuff, but for purpose-built electric drag racing with an actually good HV safety rule book for aiding safe design, bring on the voltage! You could possibly have a Sevcon size 4 sized BLDC controller that does over 500kW rather than 50kW.
http://www.digikey.com/product-detail/en/MIXA450PF1200TSF/MIXA450PF1200TSF-ND/4321809
I might build a controller with these. $450 for 3 is not bad, $75/half bridge that does >400A at 1kV and has package isolation and die temp monitoring and massive flyback diodes seems like a bargain. I'm personally into non-leathal voltages for anything that is daily transportation stuff, but for purpose-built electric drag racing with an actually good HV safety rule book for aiding safe design, bring on the voltage! You could possibly have a Sevcon size 4 sized BLDC controller that does over 500kW rather than 50kW.
liveforphysics
100 TW
Its hard to get a lower inductance layout than direct dice-bond wire-dice.
http://ixapps.ixys.com/Viewer.aspx?p=http%3a%2f%2fixapps.ixys.com%2fDataSheet%2fMTI145WX100GC.pdf
Your topology exploits this chip very well. It would be ideal for pretty much every ebike from 250w to >5kW ebikes just depending on if you wanted a match-box sized few ounce passively controller for roadbikes, and a giant fan cooled CPU heatsink or something for people trying to get >10kW or whatever, but everyone could use the same tiny Lebowski control PCB, and you just vary the caps and heatsinking and maybe leave spaces on the PCB to solder in the small current measurement chips and the biggest allegro or whatever.
Same PCB board and code and tuning software makes you a plasma breathing hot-rod ebike controller with better than sinus control and no sensor BS needed for precise torque control, or a matchbox sized 250w 36v controller setup that would survive forever with a bare minimum of caps because of a high safe voltage ceiling and low currents and essentially no need to even heatsink it at the currents in a 250w system. That would be like the God of all EV PCB's. A single one already is the best sensorless ebike motor control system in the world that I've ever even heard of at any price. With one PCB you could offer that level of control in very easy, simple and DIY friendly to assemble (bolt one already isolated package to something that serves as a heatsink, solder on already professionally reflowed lebowski control boards using modern super fine pitch SMD components, with just a number of big (and small) plated through holes. If you're making a tiny controller for your road bike, maybe stuff some tiny polymer cap footprints for use with up ~30v-ish stuff (which is huge IMHO, including 90% of RC and robotics and multi-rotors etc).
http://www.panasonic.com/industrial/includes/pdf/POSCAP_NPI_Sheet.pdf
Along with the usual arrays of various different sizes of caps for wide freq coverage transient absorbing, and TVS diode footprints etc. All the stuff typical ebike controllers don't do to save $20 in premium quality components or design, and as a result they fail when pushed.
You could offer both better performance (smoother, sensorless, perfect torque control, easily adjusted/tuned) than anything else I've ever seen, and cover a very wide range of ebike power needs all with the same tiny board we buy from you already populated to the level where it's just missing the heatsink and non-SMT caps and current sensors. Then how you finish the build becomes the big factor in its size and power.
Gang 4 of the boards together built up for power and maybe have a 50hp drive. Maybe 8-20 of them power your car and enable awesome high power performance and efficiency, but at non-lethal voltages. Or use one with a tiny heatsink positioned in the propwash of multi-rotor on a board populated with kick ass solid state caps.
Maybe leave the expensive parts unpopulated in the isolated daisy-chainable coms bus so it only adds that cost to people who are going to gang up groups of them, because I think most applications won't need it, that single chip is tough enough for really any non-racing ebike needs if you cool it, and its an easy to cool package.
http://ixapps.ixys.com/Viewer.aspx?p=http%3a%2f%2fixapps.ixys.com%2fDataSheet%2fMTI145WX100GC.pdf
Your topology exploits this chip very well. It would be ideal for pretty much every ebike from 250w to >5kW ebikes just depending on if you wanted a match-box sized few ounce passively controller for roadbikes, and a giant fan cooled CPU heatsink or something for people trying to get >10kW or whatever, but everyone could use the same tiny Lebowski control PCB, and you just vary the caps and heatsinking and maybe leave spaces on the PCB to solder in the small current measurement chips and the biggest allegro or whatever.
Same PCB board and code and tuning software makes you a plasma breathing hot-rod ebike controller with better than sinus control and no sensor BS needed for precise torque control, or a matchbox sized 250w 36v controller setup that would survive forever with a bare minimum of caps because of a high safe voltage ceiling and low currents and essentially no need to even heatsink it at the currents in a 250w system. That would be like the God of all EV PCB's. A single one already is the best sensorless ebike motor control system in the world that I've ever even heard of at any price. With one PCB you could offer that level of control in very easy, simple and DIY friendly to assemble (bolt one already isolated package to something that serves as a heatsink, solder on already professionally reflowed lebowski control boards using modern super fine pitch SMD components, with just a number of big (and small) plated through holes. If you're making a tiny controller for your road bike, maybe stuff some tiny polymer cap footprints for use with up ~30v-ish stuff (which is huge IMHO, including 90% of RC and robotics and multi-rotors etc).
http://www.panasonic.com/industrial/includes/pdf/POSCAP_NPI_Sheet.pdf
Along with the usual arrays of various different sizes of caps for wide freq coverage transient absorbing, and TVS diode footprints etc. All the stuff typical ebike controllers don't do to save $20 in premium quality components or design, and as a result they fail when pushed.
You could offer both better performance (smoother, sensorless, perfect torque control, easily adjusted/tuned) than anything else I've ever seen, and cover a very wide range of ebike power needs all with the same tiny board we buy from you already populated to the level where it's just missing the heatsink and non-SMT caps and current sensors. Then how you finish the build becomes the big factor in its size and power.
Gang 4 of the boards together built up for power and maybe have a 50hp drive. Maybe 8-20 of them power your car and enable awesome high power performance and efficiency, but at non-lethal voltages. Or use one with a tiny heatsink positioned in the propwash of multi-rotor on a board populated with kick ass solid state caps.
Maybe leave the expensive parts unpopulated in the isolated daisy-chainable coms bus so it only adds that cost to people who are going to gang up groups of them, because I think most applications won't need it, that single chip is tough enough for really any non-racing ebike needs if you cool it, and its an easy to cool package.
My rule of thumb is 40-50% of the data sheet coolaid. I mean if a mosfet is rated for 200v and 230 amps you will be very lucky to get 23,000 watts with a brushless controller using 6 of them heck even the simple math worked out to 44% for the sevcon power switches!
liveforphysics
100 TW
Arlo1 said:
That's why for the drag racing application, you start with them all screwed to the right thickness of Al block that is chilled with dry ice or LN2 or whatever. That is to get a few seconds of data-sheet BS performance numbers. Think about it, if you are finished with the quarter mile in 4seconds, and you have a window to safely swing upwards150degC, you don't even need that thick of aluminum to eat that whole heat load just in the specific heat capacity of the heatsink with no need to shed any heat to air or water or whatever at all.
For the ebike power chip, I agree you would need to sink it very well to draw >100A continuous phase current, but its got outstanding package Rth and a layout that contributes to it being very easy to sink well, its even isolated on the sinking tab already for you.
Futterama
1 kW
So this is kinda the idea I have for a controller layout. The "busbar" conductor is a round copper plate with cuts splitting it up in 6 pieces. The battery wire is attached in the middle. The teal lines are cuts in the copper, seperating the individual 6-FET stages, these are then connected together in the middle where the battery wire attaches. Only 3 FET's are shown, the other 3 are on another plate on the opposite side, like a Prince chocolate biscuit, the biscuits is the copper plates and the chocolate is the space between them. One plate is +VBAT and the other is -VBAT. On the circumference there is the phases, these are then connected in 3 star connections, one for each phase: A, B and C.
So the high current from the battery is split up in 6 parts when entering the cut up copper plates and the plates can be placed real close together, something like 0.2-0.3mm apart. So the current is 1/6 for each 6-FET pair and the busbars (round copper plates) are placed close together. This should give low energy spike, right?
So the high current from the battery is split up in 6 parts when entering the cut up copper plates and the plates can be placed real close together, something like 0.2-0.3mm apart. So the current is 1/6 for each 6-FET pair and the busbars (round copper plates) are placed close together. This should give low energy spike, right?
Attachments
liveforphysics
100 TW
I think that is brilliant and balanced inductance, balanced resistance both measured and apparent.
Have you though about cap placement and types? Cooling?
Have you though about cap placement and types? Cooling?
Futterama
1 kW
I hope I can fit some SMD style caps in at the yellow dots, all way around. I just have to make a clearance for them in the phase copper blocks. See the picture. Since the 2 plates are close together, the SMD's would fit nicely.
But what about electrolytic caps you may ask. They should also be very close to the copper plates but they would not be right next to both of them. They could all go on one side, and then they would have one leg attached close to the plate that is closest, and they would have the other leg go to the other plate (pink dots), and this would create a distance from the second plate to the cap housing, but no more than the copper plate thickness + copper plate spacing and I think that would end up at something in between 4-8mm. I might be a bit worried about this since I saw something about when the electrolytic caps have long legs, like when you lay them down on the side on a PCB, the legs could melt from the high currents.
Regarding cooling, the SMD FET's are about 1mm tall, and I would mill a recess for them to fit in, so the rest of the middle copper plates and also the phase blocks, would have a flush surface to fit a heatsink on. The 2 copper plates would be seperated by a layer of thermal pad, and I would also need a thermal pad between the heatsink and the copper. I'm aware that the most heat is generated in the drain of the FET so the copper plate closest to the heatsink should be +VBAT.
This brings me to the things I haven't figured out yet. I think the phase blocks should be attached to the middle plates with electric isolation in between, like a thin thermal pad, to keep them from applying mechanical stress on the FET's. Also the plates should be held together, all to avoid the stress on the FET's.
I can't really use steel screws as they are conductive. If I should use conductive screws/bolts, I would have to isolate them from the phase blocks using some plastic spacers, like the ones used on e.g. the flange on TO-220 FET's. These plastic spacers should be heat resistant, since I have to screw them in place before soldering and I think the soldering would heat the whole arrangement to soldering temperature.
Another option is to use heat resistant composite screws/bolts but I have to find somewhere to purchase those, I haven't used them before.
To get the best surface for the heatsink, I'm thinking about bolting the copper parts together and put the whole thing in the mill before soldering. Then I can mill the copper surfaces flat and hopefully they will stay that way after soldering.
All this work is because I want a small controller and I would like to use a CPU heatsink. The CPU heatsink I got has a 66 x 66mm flat surface for attaching the power stage so I have to/would like to limit the size to fit this.
But what about electrolytic caps you may ask. They should also be very close to the copper plates but they would not be right next to both of them. They could all go on one side, and then they would have one leg attached close to the plate that is closest, and they would have the other leg go to the other plate (pink dots), and this would create a distance from the second plate to the cap housing, but no more than the copper plate thickness + copper plate spacing and I think that would end up at something in between 4-8mm. I might be a bit worried about this since I saw something about when the electrolytic caps have long legs, like when you lay them down on the side on a PCB, the legs could melt from the high currents.
Regarding cooling, the SMD FET's are about 1mm tall, and I would mill a recess for them to fit in, so the rest of the middle copper plates and also the phase blocks, would have a flush surface to fit a heatsink on. The 2 copper plates would be seperated by a layer of thermal pad, and I would also need a thermal pad between the heatsink and the copper. I'm aware that the most heat is generated in the drain of the FET so the copper plate closest to the heatsink should be +VBAT.
This brings me to the things I haven't figured out yet. I think the phase blocks should be attached to the middle plates with electric isolation in between, like a thin thermal pad, to keep them from applying mechanical stress on the FET's. Also the plates should be held together, all to avoid the stress on the FET's.
I can't really use steel screws as they are conductive. If I should use conductive screws/bolts, I would have to isolate them from the phase blocks using some plastic spacers, like the ones used on e.g. the flange on TO-220 FET's. These plastic spacers should be heat resistant, since I have to screw them in place before soldering and I think the soldering would heat the whole arrangement to soldering temperature.
Another option is to use heat resistant composite screws/bolts but I have to find somewhere to purchase those, I haven't used them before.
To get the best surface for the heatsink, I'm thinking about bolting the copper parts together and put the whole thing in the mill before soldering. Then I can mill the copper surfaces flat and hopefully they will stay that way after soldering.
All this work is because I want a small controller and I would like to use a CPU heatsink. The CPU heatsink I got has a 66 x 66mm flat surface for attaching the power stage so I have to/would like to limit the size to fit this.
Futterama
1 kW
I may have to scrap this whole idea of soldering the FET's directly to the copper busbar and then use double sided PCB with aluminium core and some flexible thermal padding, like the RC controllers usually do, but I'm still working with the solid copper idea.
Oh, by the way, I will mill a clearance for the gate pins and use twisted wires for the gate drive, like Lebowski do it.
Oh, by the way, I will mill a clearance for the gate pins and use twisted wires for the gate drive, like Lebowski do it.
h0tr0d
1 kW
- Joined
- Apr 28, 2012
- Messages
- 460
liveforphysics
100 TW
I love the cap position. Looks very nice! I hope it works awesome for you guys!!!
h0tr0d
1 kW
- Joined
- Apr 28, 2012
- Messages
- 460
Thanks Luke.
And I hope see that 3 IXYS IGBT's inverter made by you some day!
I'm really a fan of laminated buss bars.
http://www.stormcopper.com/products/Relevance.html
IMO, laminated buss bars are way more important than super close cap location.
Example:
Lebo has 250nH/m with his geometry
Using the same 0.6mm copper but 30mm wide and a 0.01mm separation (two layers of kapton tape) gives us 25.6nH/m.
That's 10x lower, so the caps can be 10x further away for the same spiking result.
Now let's imagine we have a 4 layer buss bar (2 positive, 2 negative), 30mm wide but 0.3mm copper sheet (the same copper cross section), same 0.01mm separations.
That would give a 13nH/m inductance BUT, since we have 4 layers (3 separators in between), it behaves like 3 parallel inductors, giving a 4.3nH/m buss bar inductance (!!!).
Less then 2% of the original inductance, like having a cap 50mm away with inductance of 4.3nH or a cap 1mm away with 250nH.
Lebo, buy some of this and report back, please! :wink:
I used this calculator: http://www.eeweb.com/toolbox/broadside-trace-inductance/
Note that trace separation H is the sum of the copper sheet + separator(s) thickness's.
And I hope see that 3 IXYS IGBT's inverter made by you some day!
I'm really a fan of laminated buss bars.
http://www.stormcopper.com/products/Relevance.html
IMO, laminated buss bars are way more important than super close cap location.
Example:
Lebo has 250nH/m with his geometry
Using the same 0.6mm copper but 30mm wide and a 0.01mm separation (two layers of kapton tape) gives us 25.6nH/m.
That's 10x lower, so the caps can be 10x further away for the same spiking result.
Now let's imagine we have a 4 layer buss bar (2 positive, 2 negative), 30mm wide but 0.3mm copper sheet (the same copper cross section), same 0.01mm separations.
That would give a 13nH/m inductance BUT, since we have 4 layers (3 separators in between), it behaves like 3 parallel inductors, giving a 4.3nH/m buss bar inductance (!!!).
Less then 2% of the original inductance, like having a cap 50mm away with inductance of 4.3nH or a cap 1mm away with 250nH.
Lebo, buy some of this and report back, please! :wink:
I used this calculator: http://www.eeweb.com/toolbox/broadside-trace-inductance/
Note that trace separation H is the sum of the copper sheet + separator(s) thickness's.
Lebowski
10 MW
I'm too poor to spend 20 euro on a 5meter roll of tape
I'll stick with ducktape or (if I can find a scrap piece) some very thin acryllic sheet material.
Keep in mind that the width of 3 TO220's is much less than that of the big IGBT's you're using, so my inductor length is much less.
I see you've bent the bottom of the busbar, so in effect you're having two back-to-back L pieces. Is that allowed in a bus bar ?
I'll stick with ducktape or (if I can find a scrap piece) some very thin acryllic sheet material.
Keep in mind that the width of 3 TO220's is much less than that of the big IGBT's you're using, so my inductor length is much less.
I see you've bent the bottom of the busbar, so in effect you're having two back-to-back L pieces. Is that allowed in a bus bar ?
Futterama
1 kW
My circular design will have quite some inductance, around 800nH/m and that is mainly because I need to use something like 8mm think busbars so there is enough material to fasten the phase blocks to. I'm thinking maybe I need some other means of holding the copper blocks/busbars together. Maybe a steel or aluminium ring around the phase blocks to bolt them onto. If I do that, I could make the busbars as thin as 1mm and I could get something like 130nH/m.
My nH/m value may be high, but the distance from the caps to the FETs is low, I hope this could help a bit.
My nH/m value may be high, but the distance from the caps to the FETs is low, I hope this could help a bit.
h0tr0d
1 kW
- Joined
- Apr 28, 2012
- Messages
- 460
If you are honestly speaking :? , PM me your PayPal or address, you have to have that tape.Lebowski said:I'm too poor to spend 20 euro on a 5meter roll of tape
The only thing I ask in return is that you build the thing and report back to us.
Ducktape doesn't resist 400ºC like kapton does and acrylic only goes to 160ºC, so soldering the caps/mosfets afterwards not recommended...Lebowski said:
But if you know how to go around those problems, go for it! :wink:
Just trying to help to build a better thing, so you can have more available power with the same Mosfets and caps...Lebowski said:
The setup shown in the photos is less than ideal BUT it works for the power we wanted, cheap and a monkey could assemble it.Lebowski said:
Having the L's back-to-back isn't perfect, the separator could be a LOT thinner but it's easier and less time consuming...
With a 8 layer buss bar and better component placement we could go up 50VDC, only an extra 25kW...
speedmd
10 MW
Look it up under polyimide K instead of the name brand stuff. Much better cost results.
http://www.tapesolutionsinc.com/951..._content=pla&gclid=CK2bsbbMprwCFckWMgod5kIAKQ
http://www.tapesolutionsinc.com/951..._content=pla&gclid=CK2bsbbMprwCFckWMgod5kIAKQ
