Hi there,
First post - so if my question is completely stupid, tell it politely ;)
Most of the failures seem to happen on the power side. that means blown FETs or gate drivers (whichever that might be).
The control signals to and from the main MCU is more or less the same no matter how the power side is built, so my question would be - why isn't anyone building the power side separately and leave the MCU on a different board? I knw that will most likely increase the size a bit but IF you blow the power stage you can just pop another one in (and don't repeat whatever you were doing) and you will keep the settings from before.
I'm thinking about building a power stage with IRF7749 FETs and the DRV8323. In another post it was suggested that the 8323 might not perform as good as the 8301 with external current sensors but it might simplify the design process a bit.
And if it don't work out - then I would only need to make a new power stage.
Another idea could be to have an EEPROM on the board so the main MCU could identify safe values for that specific stage.
Is this the worst idea ever?
It is a fine idea, the problem occurs ensuring good signal integrity between the boards.
You have some very sensitive signals that need to not be run a huge distance.
Also the expensive part of most controllers is the power stage, the mcu is very cheap in comparison. So it doesn't make economical sense to split into two boards like that, as it will also increase complexity of the design and require connections between the boards.
I use 2 boards on the A200S due to a lack of space, they are connected by a samtec connector which is quite expensive. That board only has power supplies and protection circuitry though.
For a two board design you need a bunch of signals for a DRV83xx design at a minimum you need a 15 pin connection. 6x PWM, 3x Phase current, 3x phase voltage, 3x SPI lines.
So essentially it comes down to cost, size and design time. If you have the time to make separate boards go for it.
Thanks for the comments!
Yes I'm aware that some of the signals are very sensitive and that the economic gain is pretty much non-existing.
After creating the thread I saw a similar request in another so I can't be the only one seeking this feature.
Without having looked at the MCU an option could be to introduce an external ADC that could share the SPI bus (if timing allows it) or run on a separate bus. However that would make the interface less than standard and the whole construction more expensive in the end.
Another reason for doing a two-PCB job would be to change the substrate for the power-stage to a thick-film design with a ceramic base so cooling from both sides could be more efficient.
i have built my ESC with two PCB's
Power PCB -> Just MOSFETS (and one Shunt)
Logic PCB -> Everything else (also Gate Drivers)
3 Connections per Phase (without Current Sense -> without FOC)
2 Connections for Power
2 Connections for Current Sense
200A with 10S with 80mm x 40mm
The big advantage: You could order your Power Board with aluminum Core or thicker copper
This is what my design has/had.
A 3oz 2mm aluclad PCB with MOSFETs and high power components. The Aluclad allows for optimum cooling.
A 1oz 1.6mm FR4 PCB with logic for lowest cost.
While the connections are sensitive, they're not impossible to design around and by splitting the boards you can make far more cost effective PCBs.
Thanks guys,
Makes me think that the idea is not that far fetched after all. Better get to it then!