RE PM/email vs posting here: Agreed. But if for some reason you prefer PM or Email I'm game too.
No not altium. Eagle.
Why not DRV8302? Well, I calculated the 12V current for VESC and found we were already close to or over the allowed current for the DRV8302. I asked TI what could be done about about this, suggesting an external 12V source to prevent the DRV getting hot from the linear regulator. They indicated that this was not an option, and that using fets as big as the ones on the VESC was not something the DRV was designed for. Eventually I need to go above 50A, bigger fets are not possible so I'll have to put them in parallel too! That is going to rule out the DRV completely.
As it stands VESC already uses the DRV outside the allowed safe-operating-area. Because Benjamin and I think the 30mA limit for the gate drives is due to the power dissipation of the linear regulator, we agree that it's probably OK if you don't go up to 60V but stay with 50.4V max.....
In my experience, if you try a big project without any intermediate steps, you will fail. "I'm going to build myself a castle, I'm going to live like a king". If you don't have any building experience that will fail. In electronics projects I have to be a bit careful, even though I have experience. When I take on something "too big" it always ends up getting stuck in a not-working state. So I'm slowly moving up the power-scale towards my target. At 12 or 24V I can run off my lab powersupply and prevent frying the mosfets. During those tests, the lab powersupply would click into currentlimit when reprogramming the STM. Turns out the fet drive signals float to "high" during programming. Oops. A pulldown solved that. Starting immediately with a 12S 5A LIPO would have cost me a whole bunch of FETs before I'd have figured out what was wrong. So building up slowly means a more reliable road to the full-power version later on. (And I know myself I get irritated when I blow a bunch of hardware and don't know why, and then the project gets scrapped).
So my current testing stands at 35V with the ebay DCDC converters. I don't share your concerns with "noise" on those regulators. The 12V is just meant to drive the gates. Anything between say 10V and 18V is fine. 12V +/- 1V ripple? Great!
The 5V is used... nowhere. Well, almost. In my design there is an MCP9700 on the power board that measures the temperature there. The critical stuff runs off the 3.3V regulator, which is a nice stable linear one. I'm planning to measure the current draw on the 12V line. To verify the calculations that we've done. Similarly, I want to know the current on the 5V line. I expect to replace the 5V DCDC converter with a 7805. The space for a DCDC converter is not worth the power savings.
If someone wants a 55V version tomorrow, I'd have to speed up those measurements and run a test with my own DCDC stepdown. As I'm currently using a "60V" stepdown chip, if it turns out a significant number of people are interested in >60V versions, then I'd have to think about that. I have several options:
* find a chip that works up to 100V. Possibly use a cheaper version for < 60V.
* Allow external stepdown for > 60V operation.
* a few more.
Re: split boards: Yes, I'm seriously considering splitting the power part in the final version. One of the advantages is that you end up with a large surface with the power-part, that is good for cooling. IRF has told me: Cool the back of the PCB, not the top of the FETs. (Benjamin are you listening?) Experimental verification required!
Re: 072... I have other boards that are able to use STM32FxxxR chips. My notes list 23 chips that I think are compatible, I've physically tried 6 of them. The list is outdated. e.g. I haven't looked at any STM32L devices for compatibility. The 072 is a nice "beginners model". It has USB, plenty of flash and RAM and is not as peripheral-starved as the earlier "value line" chips. (e.g. it has a DAC that previously was only available on more expensive chips).
My CPU board is '072 (and '405 and probably 21+ other) CPU-compatible. When bringing up the CPU board, there was something stupid. The USB connector was bad or something like that. So then we built another '405 board and that didn't work either. So then we built two of those boards with the '072. Those worked, showing that the board-layout was OK. But running VESC on them is going to be a challenge as they only have one ADC peripheral. And no FPU. And runs 2.5 times slower. However it CAN be done. ST's FOC library is able to do it.
Anyway, Benjamin's decision to simply go with the more powerful CPU means that you have the ease-of-programming not worrying about the flash filling up, etc. That's a good engineering decision. When you're going to make thousands that decision needs to be rethought (not necessarily revised!). The "expensive" '405 also has FP. In this application FP is not useful at all! Any variables can be scaled to work with integers. So instead of storing 23.4V as (float)23.4, you store 23400 mV. (or 1000x more in microvolts). There is never going to be a huge range of numbers required. And if the difference between 51.1 and 51.2 matters to you, so does the difference between 12.2 and 12.3. A fixed-point format works just fine. But it requires a bit of "effort" on the part of the programmer. And as our lead-programmer is doing it for the fun of it we must respect his decision to use floats and simplify the programming a bit. Totally worth it!