I’m just thinking of best method to share the pins for the encoder interface. On the B-G431-ESC1 looks like they are clamping the voltage to 3.3v using diodes and a current limiting resistor on the pins along with 10k pullups. I’m surprised the I2C works with the 1.8k inline and the 10k pullups. I’m thinking of solder jumpers to give access SPI direct access to the pins and possibly the same for the I2C with some 2.2k or 4.7k pullups.
I noticed you have two pins reserved for the opAmps (each). But it looks like the B-G431 board has three pins (for each current sense). Positive and Negative connecting to the shunts and the out-pin for ADC . I may be wrong, just to double check.
You’re right - you can feed the output into a comparator input, and you can use external feedback and filters - but you can also leave this out and use the chip’s internal functions for these things. But you may find the gain values offered there unsuitable, or have other reasons not to do it this way.
Basically the OpAmps and comparators are quite versatile and you can use them in many ways. Matt303 links to one of the App-Notes describing it further above.
@Matt303: might also want to look at the STSpin32G4 chip, used in projects like Field Stack trials (STSpin32G4)
Same processor as the STM32G431 (same speed), up to 40 GPIOs available, ready to drive gate drivers directly (1A sink/source)
I was hoping to use this processor on the lepton 3.0. I think we should collaborate. If you see my previous posts on the topic, I think our needs are similar, although you are looking for considerably higher currents.
See my last posted thread here Beginning of a proposal for Lepton 3.0 - #19 by Anthony_Douglas on the general idea that I was aiming for.
Basically a general purpose motor driver based on the MCU you are talking about.
I think it is very important to realize that this stuff is not actually that easy. The internet is littered with at least a dozen half done projects and only one or two like VESC that are really done because people realize how much work it is after starting and then don’t continue. The Odrive guys realized how much work it was and decided they couldn’t figure out any other way to get paid except to go close source. The same thing happened with bl_heli.
You definitely need ST_link.
In line with how much work this stuff is, the first physical boards should probably be considered a draft. At least 3 iterations can be expected before it’s really solid even if the community collaborates.
In my experience designing IoT boards, if you are purely operating in the digital domain, usually the second revision is ready for production, in rare cases even the first one.
It’s the analog part that is difficult to get right in few tries, especially when high currents and low noise are needed, like in this case. And where PCB layout makes a difference (unlike purely digital, where for the most part the layout is not as critical).
And, yes, as the joke goes, the last 20% of a project takes the last 80% of effort 
Github and Hackaday are a graveyard of half completed projects
Hi @Anthony_Douglas I had a read through your post and I think current sensing is defiantly needed. I use it for sensor less homing etc
The BEMF I currently propose is just a thought for future support but I probably won’t add it at this stage.
I think the best option is we come up with a general pin layout like I have done above and a basic schematic that can be altered easily for people doing other custom boards. This also means one general purpose board can be created for SimpleFOC with all the correct pin definitions.
Have a good look at my pin layout and see if anything is missing for your needs and I will try to adjust it and if we are happy with the selection of pins we can start working on some schematics.
Are you using KiCad?
I don’t think a pin payout for custom boards is enough, as you are seeing there is a long road with a lot of details between that point and a working motor driver. As others have also found. Someone has to traverse that road, and the user can’t do it at the last minute. But this is a start.
Will have a look though. I’m using Easy EDA, either is good but it’s integrated with JLPCB so the actual step of sourcing all parts and getting the board made is more tractable. There are a lot of little tiny parts, soldering things manually is not at all practical.
I think things should generally resemble the B-G431b-esc1 board re pin mapping. It appears that your proposal mostly follows that.
Hm, you don’t really get any spare pins do you? I think the main thing is to break as many pins as practical out and connect the ones that need to be connected to the current sense etc. so even the led pin could be broken out, for instance. To 2.54 mm through hole pads, which can use a dupont, molex or 2.54 mm screw terminal. That way the LED can be used as in/out if you wanted to for instance. Anyone who doesn’t need CAN can use those pins for other stuff. etc.
In other words, the first stop is making things programmable and flexible.
We definitely want SPI , I2C, and ST-link, UART broken out, the op amps connected to pads where shunt resistors can go (ideally with the option to omit the shunt resistors, e.g. a through hole near each pad that the resistor would solder to could be used to help put a wire in place instead of the shunt, or maybe we can just say use the pads an a piece of wire).
Might be good to have an area with pull up resistors for i2c, whose traces can be cut if needed, as Runger notes I think he mentioned that was a good way to to i2c pull up resistors. Most people are not using more than one device on the I2C line and an extra set of resistors is I think ok most of the time, so if the terminal device is an as5600 board of the usual type for instance it should still work, they contain 10k ohm pull ups (the experiments with the lepton indicated the lack of pullups was a thing, however that may have been incorrect, it may have been the noise issue on the lepton which the extra pull ups masked, and in any case that one board which I thought maybe didn’t have pull ups was a non-standard as5600 board that is not commonly available).
Well, it’s a good time to be designing G431 based boards. LCSC has them very cheap: Search by "STM32G431C"
If you like the QFN48 format of the STM32G431CBU6 you can get them for less than $1.50
This depends on your design. My personal pain-point is currently around 0603 size. But for most components you can choose the size, and can choose not to make everything 0402 and smaller. Using a bit bigger components greatly increases solderability.
Of course once you have a design you’re happy with and want to produce many, the turnkey assembly makes a lot of sense. But @robca already mentioned the difficulty of designing the analog part - if you can self-assemble a few prototypes a single cheap ($5?) JLC order for PCBs will let you try out dozens of variations of your analog design using different component values/brands/types…
If I may put in a preference for connectors:
Stemma/QT (JST-SH1.0 4 pin) for I2C
6-pin JST-SH1.0 for SPI (like here: https://github.com/runger1101001/Magnetic_Encoders/blob/master/Encoder_SPI/pics/SPI_pinout.png)
4-pin JST-SH1.0 for UART
ARM-14 header for SWD + Serial
Consider if you need the UART connection. I would instead put it on the SWD (where it is then automatically connected to. the PC via the ST-Link). Or if you want to use 2 extra pins on the MCU consider USB, which can also transport serial streams and is more versatile (e.g. also able to provide power, has shielded connection, additional USB profiles and more).
Use a “solder jumper”. I like to use them “normally closed” so the user only has to cut the trace if he doesn’t want the option (in this case the pull-ups). It’s a zero cost solution and most PCBs have extra space on the back to put the solder jumpers.
I personally would not do it this way. If you want this, then why not use the SimpleFOC shield on the Nucleo64 G431 board? All the pins will already be broken out, plus the Nucleo has a built-in ST-Link.
This approach suggests the board can be used for other things, but in fact it will be 100% busy with motor control.
My opinion is the driver board should focus on the motor control. The interfaces it offers should be used as command inputs for the motor driver, or as interfaces to the required sensor, or as outputs for telemetry. What functions it has, it should offer on-board, and it should not try to also be a general purpose MCU board at the same time. The driver board then anyway interfaces with a controlling MCU or PC, and this other system can worry about the other stuff.
nucleo plus shield is way too expensive, though. It’s like $60 usd for the shield plus a nucleo board. it also has poor current capability and is lacking on many other counts compared to what could be provided for $15 usd. IDK how available it is. It was a great start, I don’t want to knock it, but I think it’s high time for a second generation.
Previous estimates for a well designed board with much better features, 6 pin instead of 3 pin (which I gather allows some features such as regenerative braking? Also perhaps finer control over dead time etc.)
Secondly, in any case I think it would be important to at least settle on an exactly nucleo board, as it has become clear there are a lot of features which have to be sorted out on an mcu by mcu basis.
For a one of project I agree the shield and nucleo board makes sense, but I think it warrants a sucessor. I also think it would be advisable to share this idea of using a nucleo board, not an uno, more prominently in the documentation as I think due to the shield design and so forth a lot of people are thinking that an uno can do at least something but it’s really not up to much.
Those BEMF inputs don’t have ADC IN.
Are you planning to support stepper motors (I see 4 PWM timers)?
Maybe you could use TIM1_BKIN as a Enable/Disable Motor (PC13/PB10/PB12, etc)?
Hi @dzid26 I probably won’t use BEMF at this stage, was just for future proofing. The 4th Timer is for another half bridge for a braking resistor. I don’t see why it can’t be used to drive a stepper or 2x DC motors if needed. Why use TIM1_BKIN as enable/disable?
Hi @runger I will take note of preferred connectors and use them. I have just had a massive project come up for work so this has delayed me on this project, so hopefully get something on paper in the next few weeks
TIM1_BKIN can be programmed to “physically” 
stop the timers (even if the CPU halts) which is a neat safety feature (emergency stop).
I know this project is a long way from completion anyway and I would surely have to modify things significantly anyway, and swapping one connector for another isn’t too hard on the pcb level. However if I may respectfully disagree with Rungers notes regarding connectors, I am sure it is sound from an engineering standpoint, but as a user with little $, connectors of even slightly sophisticated nature, and especially any significant variety, is a real bother. The lepton 2.0 for instance, the connectors used the picoblade, iirc, did cause significant turbulence. And the thing is I believe these types of complications multiply by each other, rather than being merely additive. So things get out of hand really fast.
It may seem like just another drop in the bucket to use a more advanced and slightly less common connector, but imo it is one thing that can be trimmed back. IMO 2.54 mm through holes are the winner in this case. However there is clearly a lot else to do here and it would make no sense to get bogged down in connector selection and frankly I don’t even know what most of the connectors Runger mentioned are.
But if you recall we had problems with the Picoblade, they pre-made connectors are ridiculously overpriced, like $10 each. For all 4 connections on the lepton it would have been $40 to buy them from digikey. By mere coincidence there are little kits on amazon that you can use to assemble a ton of connectors yourself from pre crimped wires, but it was touch and go and hard to discover that option. That was a close shave.
I agree, 2.54mm is the most convenient. I can still crimp 2mm, but below it’s a gamble if the crimp holds. Premade cables typically are too long, too short or need to be converted to other connectors, they hardly ever fit my needs.
Both the Picoblade and JST SH type connectors are very commonly used, with many alternative vendors that provide fitting plugs and sockets.
For this reason there are also vendors on AliExpress that provide pre-crimped cables of various lengths and types, for very cheap eg <$1 in quantities of 10 or more.
Of course blank headers are cheaper still, but that means soldering for the user, and a big additional source of error.
I agree that self-crimping these sizes isn’t easy, but premade cables are available. Having a connector rather than just a DIY header makes the board much more finished and easy to use IMHO.
Of course, useablity always has a certain price, so it’s a trade-off.
I thought picoblade was copyrighted or something? I probably didn’t check aliexpress because I needed them fast. I do recall complaining that digikey can’t just stock stuff at reasonable prices. The idea of a one stop electronics store seems to have escaped them. If people can offer something on aliexpress for $1, you should be able to buy ten of them from digikey for $20.
For the 1.25mm pico blade connectors there is an equivalent JST part, but also other vendors. I don’t know how licensing for those things works, or if the patents have expired, or the connectors are in some non-essential way “different” and hence compatible but not the same.
Anyway, I don’t think it matters. The important thing for me is that there are people selling cables for them
Hi @Matt303 ,
Thank you for sharing this OPAmp setup.
I know you were planning to work on this board. Just checking to see if it this setup worked out for you (for the higher volt and current requirements)?
I’m trying to do something similar (based off of the STM32G431CBU6 version, similar to the B-G431-ESC1. But not sure on what resistors to choose for the OPAmp network.
Can you please share how you were able to find the value of the internal feedback resistors R5(2.7k) and R4(1k), as the gain is primarily dependent on these.
