View Single Post
joerg_rw's Avatar
Posts: 2,222 | Thanked: 12,651 times | Joined on Mar 2010 @ SOL 3
#643
Originally Posted by sarahn View Post
Having not been able to see the schematic before posting that, I thought perhaps it showed the twl4030 as the phy. That's not the case.

So the twl4030 (cribbing from the tps65950 datasheet) can drive vbus using an integrated charge pump. In order to drive vbus, cp.in needs to be connected to vbat. It is. r6 (cp.gnd) is grounded which is also good. However, I don't see any indication of the CP flying capacitor being present. According to wikipedia, the source of all knowledge, a charge pump uses a capacitor to operate. AFAIK that would not be integrated with the twl4030, so the charge pump can't operate. I also don't see vbus going to any other power supply.
.
I thought it'd be clear by my previous post that I've checked these details. It's not just wild guessing only when I suggest to try and patch kernel driver to set the correct bits in GAIA registers. AIUI there are all needed components present for the chargepump in twl4030 gaia. (see C4213,14,15). The charge pump CAN operate. So VBUS going to GAIA R8 should be sufficient for +5V supply on USB-host.

Originally Posted by sarahn View Post
So the question is, why not have the charge pump? The 4 things I can think of is a) it added too much noise b) it doesn't actually work c) it can't supply enough power to meet spec d) it's stupid and tries to charge itself off its own charge pump.

Or perhaps there is no integrated charge pump on the chip they used. As I understand charge pumps are expensive and they may have decided the demand didn't justify the additional cost.
No, the real question is why Nokia decided to have a dedicated PHY chip for USB instead of using the builtin in TWL4030 GAIA. There might have been silicon erata (bug in chip).

Originally Posted by sarahn View Post
Also, even though the ID pin is connected to the twl4030, I suspect that normally this interrupt needs to be signaled through the ulpi interface (I think that's what normally happens) which the twl4030 can't do.
If something creates such a need, then it's the kernel driver. I was under the impression we intended to fix it to work with the given hardware. Also nota bene that detecting ID pin is more of an OTG feature and absolutely optional for the manually activated hostmode I suggested (exactly like it's usually done on N810 as well, as it seems nobody is able to get such a magic cable with grounded ID pin. And even if you could find such a cable, it had a micro-A plug while the receptacle of N900 is a micro-B - N810 is micro-AB)

jOERG
 

The Following User Says Thank You to joerg_rw For This Useful Post: