pbio/drv/battery/battery_ev3.c: Implement EV3 battery driver #375
Conversation
This reads the values from the ADC and scales them according to the ideal scaling factors.
Without this, the hub can falsely detect low battery when booting.
|
Done. Also added a workaround for detecting low battery on bootup (it's added in the ADC driver, unlike the REVISIT in the STM32 implementation). |
| // Battery current is read across a 0.05 ohm equivalent shunt resistor | ||
| // which is then connected to an op-amp configured with a gain of 16 | ||
| // (non-inverting). This yields 1 A = 0.8 V. | ||
| *value = ((uint32_t)(value_raw) * 5000 * 10) / (1023 * 8); |
There was a problem hiding this comment.
It would be nice to move scaling to mV to the ADC driver on all platforms. Can be a different PR though. I think David might be working on ADC at the moment.
|
I tried this with 3 EV3s so far and the error is kind of all over the place. 
|
| // Battery voltage is read through a voltage divider consisting of two | ||
| // 100 K resistors, which halves the voltage. The ADC returns 10 LSBs, | ||
| // where full scale is 5 V. | ||
| *value = ((uint32_t)(value_raw) * 2 * 5000) / 1023; |
There was a problem hiding this comment.
In the Linux kernel, for the battery voltage, I also took into account the VCE of Q19B that I somehow figured to be 50 mV (don't remember how I came up with that number). And I also added a bit more to account for the shunt resistor which depends on the current.
|
I also think we need to first slow down the SPI on the ADC a bit more to get more accurate readings. Below are some data points I took (by enabling debug prints in the DCM). As we can see, the {W,C2T,T2C}DELAYs are already helping quite a bit, but the values are still skewed a bit compared to running at a slower SCLK rate. It looks like 1MHz is the magic value to let us actually read a saturated 5V on the ADC. (Note: the values are what I set |
This reads the values from the ADC and scales them according to the ideal scaling factors.
Voltage was tested and reads in the correct order of magnitude, slightly lower (~100 mV) when compared against a cheap DMM without traceable calibration.
Current appears reasonable (116 mA) and increases when I squeeze a running motor (up to 500 mA).
Fixes pybricks/support#2311