Charging via USB-C

[Luke]

It has become apparent to me that charging via USB-C only works properly (fast) when the PBP is on.
The suspicion is that this has to do with PD negotiation failing when the device is powered off. This results in slow standard USB charging.

The graph below shows the charging pattern with the device on (orange boxes) and off (between the two boxes).
I'll talk to the relevant people and ask about what can be done. In the meantime, if you want your PBP to charge at its optimal using USB-C, make sure that its turned on.

This doesn't affect charging via the barrel jack.
It may be a issue specific to the default Debian MATE or other OSs too - haven't checked. Please report below if you've got experience with other OSs.

[ElektromAn]

The suspicion is that this has to do with PD negotiation failing when the device is powered off. This results in slow standard USB charging.

USB-C PD is handled in the (running) kernel
driver source
drivers/usb/typec/tcpm/fusb302.c

[danielt]

The suspicion is that this has to do with PD negotiation failing when the device is powered off. This results in slow standard USB charging.

USB-C PD is handled in the (running) kernel
driver source
drivers/usb/typec/tcpm/fusb302.c

I've been testing a variety of USB-C chargers combined with an AC power meter (much quicker the observing the gradient),

I'm seeing wildly divergent behaviour:

1. A system that is off will always draw ~2.5W (e.g. 500mA at 5V)
2. Connected to a modern multi-charger type A socket via a USB2.0 cable the PBP draws 10W
3. As #2 but with a USB3.0 cable the PBP draws 13W
4. Connected via a Lenovo 45W laptop charger the PBP draws 15W
5. Connected via an alternative 45W type C to type C charger the PBP draws 2.5W

All but #2 of the above is seen in both 4.4 vendor kernel and mainline (with Tobias' patches). I've only tested #2 using mainline.

[danielt]

5. Connected via an alternative 45W type C to type C charger the PBP draws 2.5W

BTW this is the case I'd really like to fix.

Looking at the fusb302 debug logs the Lenovo charger is doing pull up based "negotiation" whilst the alternative charger uses real PD comms. It negotiates OK but the system never actually draws the additional current. Does anybody have any clue whether software can direct the charger circuitry to draw more current?

[hdk]

Ubuntu Mate Ayufan
Samsung USB-c 15W fast charging adapter
* PBP off = fast charging
* PBP on = not charging at all.

[Luke]

I'll ask @xalius to look into this thread and share some info - he did some research on this yesterday

[zaius]

2. Connected to a modern multi-charger type A socket via a USB2.0 cable the PBP draws 10W
3. As #2 but with a USB3.0 cable the PBP draws 13W

I'm going to take a guess the difference there is happening on the charger side.

[Luke]

2. Connected to a modern multi-charger type A socket via a USB2.0 cable the PBP draws 10W
3. As #2 but with a USB3.0 cable the PBP draws 13W

I'm going to take a guess the difference there is happening on the charger side.

Yes, likely what it negotiates with the PBP

[danielt]

2. Connected to a modern multi-charger type A socket via a USB2.0 cable the PBP draws 10W
3. As #2 but with a USB3.0 cable the PBP draws 13W

I'm going to take a guess the difference there is happening on the charger side.

Yes, likely what it negotiates with the PBP

This is the same charger is both cases and the fusb302 debug logs are identical in both cases.

I'm tentatively reaching the conclusion this is a cable quality/length issue. The slower charging USB2.0 cable is 1.8m, I get faster charging with a 10cm USB2.0 cable and a 0.9m USB3.0 cable. Note that the charger has USB2.0 sockets so the only difference between USB2.0 and USB3.0 cabling should be nice fat conductors that are spec'ed for 900mA.

[ElektromAn]

So
I've bought some USB-C "tester" from the usual sources. Claiming the description
it should be able to track/tap on PD messages ...

AFAICR
USB 2.0 500mA
USB 3.0 ~900mA
But there are some "crazy" USB chargers out there without max. mA limit
Also USB-C is more *dangerous* due max voltage of 20V and 5A with USB-PD
The dealer with big A has dropped all bad USB-C cables

The conductors are crucial, it's the copper aka cable thickness
Some kernel developer, CEC maintainer, has detected bad HDMI cables.