Dont forget dimensioning of allowed gate cap in accordance with driving capability of said drivers 
Funny how symmetry plays such a great role in electronics. From silicon chips and even PCB design
5 Likes
Why add parts to the design if the parts cost more than replacing blown units?
Yes, particularly if you don’t replace them anyway, or charge a ridiculous price for replacement insurance.
1 Like
IT’S A FOCKING BEAST!!!
And it’s very much alive. Tests are great so far - Low noise levels, and overall very convincing.
12 MOSFETS, each rated at 100V 207A. Local PNP turn-off circuit at each pair of fets for fast turn off.
MCU board solder-only sandwiched together with main board on back side (easier than I thought.
Finally, a footprint comparison. Only thing I had lying around was an old burned-out CFOC2.
BUT, there are a few pads I need to move further apart to avoid erroneous boards from jlcpcb. And then a few cosmetic things.
Now, I need to test that the IMU, CAN, SRV-port (for remote control), NRF-24 addon works, etc.
Release-party coming up.
5 Likes
The board upon arrival from jlcpcb. Sorry, you will need a saw… Needs to be like this to avoid getting extra charged by the pcb manufacturer. I will try to minimize the paths to be sawed in coming orders, to see how far I can push it. Don’t know if I can make a deal with them, because it is only one board (in modules).
Att., 12 holes for solid mounting of each mosfet to aluminum back plate.
6 Likes
Excellent! Glad to see this moving forward.
I know that everyone quotes the room temp junction pulse current rating, it’s such a wonderfully high number. But that 207A rating for your FETs is a theoretical number for a microseconds long pulse (at most) that leaves the junction at room temp. It can’t be used for anything practical.
1 Like
Sweet! Dumb question though… Did you do a “solder-only” attachment of the processor board to avoid noise, prevent it coming loose, or just for the heck of it? How safe do you think it would be to mount it on pin headers and sockets - if also mechanically secured?
Also, I’m assuming JLCPCB did all the surface mount stuff, including the processor, right?
1 Like
- Yeah, the chinese did all the surface mount stuff, including processor.
- I’m sure you can attach the processor board with pins in a socket, without having to worry about noise. Many of the signals are digital, and those that are analog are amplified to span the whole range from 0-3.3V. However, I might claim something here that is not correct, because I’m not an electric engineer. (so please comment, anybody). My reference for the claim is the MP2 design, where they also use pins and sockets.
- Well, it was just easier to attach it without using pins. But, if I were to detach it again, I’m not sure how I would do that. About vibrations… I think your solution will do.
1 Like
Me neither! But you seem to know what you’re doing 
I know the board is not going to run at 2x207A = 414 Amps! That’s an insane number. At 75V the losses to heat would be enough to heat a house at winter time, and obviously everything would desolder.
Electromagnetic induced noise and oscillations are what eventually will kill the board as power turns in eastern directions. A short circuit when a mosfet is self triggered at some point.
1 Like
Good comments here! Thank you for pointing out that I do in fact not have a seperate gate resistor for each FET. This is a mistake from my side, and I do need to fix that. SDFJAHKT!!
Luckily, I only need to reorder the “main board” part, so I guess it will be 1/3 of the price.
3 Likes
This picture shows the design of a single phase in the schematics of the MP2 design. Mr. Battery_Mooch, I would like to hear your comments on this… ?
I don’t know what I’m doing, but I know that I’m doing something… (btw, Is this quote already taken?)
HUGE MISTAKE !!!
We’re starting to get into consulting territory here. 
But I’m happy to give some quick thoughts…
- This might seem silly to some but just seeing a readable schematic, with notes, is a great sign. So often I’ve looked at ones that have made understanding the circuit almost impossible. That kind of short-sightedness can really hurt the designer at some point later when they’ve forgotten all of the little details but changes need to be made.
- I haven’t looked in detail at that chip’s specs/requirements but quick glance seems okay. I’m sure you’ve done this but review all connections ten times.
- No Rboot? I don’t know if your design needs it, just inquiring.
- You’ll have to test whether your 15Ω gate resistors give you the turn-on/off times you’re looking for, something I’m sure you were going to do.
- Be aware of which JEDEC layouts the FET datasheet uses for their theta-jc and theta-ja specs. You’ll need to compare the area/weight of your copper planes against the JEDEC ones that the FET manufacturer uses. Oops…wait…TO-220 FETs…nvm.
- Take into account that the FETs are close enough to raise the “local ambient” temp for each FET.
- Add onto that the potential for a high ambient temp inside the enclosure…think hot Summer day in the sun and add in a hot battery pack too.
- Calculating FET heating is tough since it’s a switching application but the lack of moving air and small space needs to be taken into account. You probably won’t have convective cooling due to that and radiative cooling onto surfaces so close won’t help.
- IMO, the pcb layout of the power stage is more important than the component and component value choices as this is where the problems usually happen. Reference designs and online tools have helped us make great component choices but often leave us in the dark regarding truly effective layouts. OnSemi does have some recommendations and that’s great to see. External limits (cost, max allowable pcb size, parts availability, etc.) often play a big part in the layout though and often force us to make compromises. This is the part that frustrates me most as a power electronics engineer.
- It’s great to see the attention spent on gate charge, turn on/off times, and snubber component values! Do you have a scope good enough to evaluate the effectiveness of the initial design choices and any changes made?
7 Likes
IMO this will lead to disaster most of the time.
You can take out most of the inter-connecting board material using a line of spaced out slots and then adding “mouse bites” using unplated holes in the remaining material segments in order to make snapping the boards apart much easier.
You only need to leave enough material to hold the boards together during manufacture/assembly. Make sure any copper planes are kept out of this space between boards.
Take into account the forces used to snap the boards apart and how that might bend the boards and stress the component solder joints. Consider where fingers might be placed when doing this. You might need to have a graphic showing where to place fingers and how to safely snap them apart without stressing the board.
Sorry for doing the drawing in MoochCAD…
6 Likes
I would think a Dremel cut-off wheel would safely separate them - no?
1 Like
It could work but not safely IMO…one slip could damage the board or a component. It also creates huge amounts of static which can easily fry components. Not everyone has a Dremel too.
3 Likes
I looked on youtube for that old SNL fake commercial where they perform a circumcision in the back of a Lincoln Town Car, but couldn’t find it. 
2 Likes
I really appreciate your help here… Thanks!
But, this one I tried already, and were to be charged an extra 50 dollars for adding mouse bites.
A dremel cut off wheel is impossible to control, use a small manual saw like this:
1 Like
Use small manual saw
2 Likes