1792 réponses à ce sujet

[garfield2142]

I knew the 3b+ would consume less, but didn't expect this much of a difference...

PowerSupplyTest4.mp4   9,01 Mo

[firened]

Huhh wow

[firened]

Does it run over the same cables and dupont power distribution board?

[garfield2142]

I didn't need the distribution board for this one, but the power supply cable is the same

[garfield2142]

But you're right, for a fair comparison I should've ran it through the distribution board

[firened]

Even with that in mind, i thought the pi4 only uses 20-25% more.
I wonder if there's something else unaccounted for, still.

[garfield2142]

I should just do a naked raspberry pi 3 & 4 with the power supply @5 v and only a camera.

[garfield2142]

Hmm, you were right, there were way too many variables to make a conclusion. With the pi4 I used a powerpack pro which is discontinued by now. When I run the pi with only the lab bench power supply, the difference is not that significant

[Mike118]

@Garfield2142

[firened]

@Mike118 could cutting the i2c line on the charge chip be a good test?

If the i2c collision is the problem, it could be moved to the second i2c bus the Pi has on the next revision for future implementation if necessary

[Mike118]

I just made a software patch and tested it succesfully. Now the robot will not crash anymore when charging.
I guess cutting the I2c line on the charger might work too, but in order to charge at full capacity I think we might need to send few I2C commands to the charger ...

[firened]

On 2 robots already, after switching to the 4 pin magnetic cable instead of the 2 pin, the geekworm UPS v3 stopped charging.

Here's a hypothesis as to why, i estimate a 30% chance of it being correct, but really hope it's not the case:

- geekworm UPS v3 has 2x ICs: (1) a PD negotiation IC and (2) the BQ25... Charge and boost IC.

- the magnetic 4 pin plug, under specific circumstances, when unplugging, could perhaps accidentally connect 12 V to one of the data D+ D- lines, which may kill the PD negotiation chip. From this point forward it 95% of the time stays at 5V.

- due to unknown reasons (perhaps the 5V sags too much), the BQ25.. doesn't want to use more than about 1A @ 5V. Which effectively disables charging.

- by using a different PD negotiation IC (external PD decoy module), we can have 9 V again.

- due to unknown reasons, by manually setting the charge current limit in the BQ25.. i2c registers, it only then starts charging. Restoring proper functionality.


I'll test it more to see if this fixes it

[firened]

What the I2C registers look like

[firened]

[U+2191]the above only worked when force feeding 9V. It still didn't charge on 5V.

There are multiple things that can disable charging on the bq25895 charge IC, all have to be set correctly:
- voltage drop between USB no-load and under-load cannot be more than 0.5V by default. With a magnetic plug and running robot, this is normally the case. <- this drove me up the walls, i didn't understand why it doesn't want to charge on 5V. Status: if VDPM_STAT is 1. Fix: Increase VINDPM_OS to 1V.
- total input current limit. Status: if IDPM_STAT is 1.
- on i2c read a watchdog is started automatically, runs out after 40 s and resets most registers. Fun!
- ADC's don't update by default.


Here are my notes so far, sorry it's not clean yet:

UPS geekworm v3 (BQ25895)
on 9V supply
```
i2cset -y 1 0x6a 0x07 0x8d #disable i2c watchdog
i2cset -y 1 0x6a 0x03 0xba #reset watchdog fail flag
i2cset -y 1 0x6a 0x02 0x6d #enable 1s adc conversion
i2cset -y 1 0x6a 0x00 0x58 #set charge current
```

on 5V supply
if REG13 0x80 `VDPM_STAT` is 1
increase `VINDPM_OS`: `i2cset -y 1 0x6a 0x01 0x0a`
if 0x40 `IDPM_STAT` is 1
increase `IDPM_LIM`: how? it's read-only. -> appears to be controlled by same `IINLIM` 0x00

[firened]

Some common caveats and troubleshooting is listed here:
https://gist.github....d2a2f2a00b18c7f

[garfield2142]

I just found this UPS that I never heard before. What do you think? https://www.pisugar....aspberry-pi-ups

[firened]

Nice.
Smaller battery.
Fuel gauge?

[garfield2142]

It's very similar to the wittypi I tried last year. It also has an RTC and its own webUI. Addresses are 0x57/0x68 https://docs.pisugar...3/pisugar-3-i2c

[garfield2142]

I'll order the 2 and 3 sometime soon. The first version does not support battery reading from software, only the 2 and 3 version.

[wowhik1]

The capacity is quite realistic, but it doesn't make much sense.
Inside this module there is a single 3.7V / 4.2V cell, and the voltage is boosted to 5V via a step-up (boost) converter.
It may only be useful for powering a Raspberry Pi, and nothing more. When connecting servos or motors, the voltage will sag.
(For robots or other equipment, a voltage of around 8.4V is preferable, especially for motors, etc.)

So it depends on the purpose