Need MOSFETs that have much lower miller charge to reduce switching transition distortion and filters on the phase voltage dividers.
I tried navigating the source code, it was very difficult for me to follow the process flow. I’m just bad at working with third-party software (there are other libraries in other fields I couldn’t figure out how to use).
The phase voltage isn’t even used in the FOC algorithm except when un-driven, and for measuring resistance you can set duty cycle thus voltage is deterministic so how does phase voltage filters help?
The estimated and actual voltages don’t match up, I have absolutely no clue why, but that’s what happens.
I was pretty annoyed when I discovered this myself.
The filters also allow for a VERY robust BLDC integration algorithm that doesn’t need tuning.
I use a smaller resistor of size 2.2k and filter cap of 15-20nF. The bigger resistor can scale up indefinitely because the time constant is the product of the equivalent parallel resistance times the capacitance.
This allows me to sample at 300kHz asynchronous to PWM and integrate the BEMF.
Averaging these samples with corresponding current samples during static PWM provides an accurate resistance measurement, down to 3% accuracy.
Wow I’d really like to read about the benefits of this new tool.
@Gamer43 I feel you are not happy with this improved new tool, I wonder why? It’s offered to the world for free and with a lot of effort behind the scenes, most of the people around here feel really grateful and fortunate to have such smart people moving forward with esk8 electronics.
My question to the developers is, will BFI be better than hall sensors? Or would it be mainly a work around for sensorless motors?
It’s just going to depend on the sensor and the motor. In general I think a sensor will still be nice when it works, but it is cool that now there’s more options on the market of usable motors if a clean takeoff is possible without them.
I also think the algorithm is brand new and probably will see a bit more growth and improvement as we get more people testing it. I think the number one thing that’s missing right now is a routine to select parameters automatically, we’ll be working on that it shouldn’t be terribly difficult ( famous last words).
This attitude isn’t ideal. We could use less people competing and more people collaborating to make awesome stuff. Especially when the problems are challenging and interesting. Vedder has taken the time to make a nice video explaining exactly how this new method is done which is more than you can say about most developers and I believe it deserves more respect than what you’re showing.
Yeah, Metr is usually quick to update just give it a week or two. It is definitely nice you don’t always have to update anymore though. Will be especially nice once some of the quirks with that are ironed out.
It’s on a shaky foundation. Here are examples: the firmware doesn’t take into account that PWM is preloaded, so the observer uses yet another set of erroneous values. There are no sanity checks on the speed estimators. Sensored/open-loop to sensorless is a hard switchover, there is no synchronization procedure. When sensorless is running, it doesn’t compare against sensored as a sanity check.
The code is full of so many if-else (and lacking in comments) I have no clue how to go about addressing these issues in the source code.
This is one of the reasons why I decided to implement my own FOC algorithm (I wanted to do BLDC a really long time ago, before I got into this forum, and I accomplished it).