[maemo-developers] USB host mode on N900 (was Re: N900 usb host + power charge)

[From: Sarah Newman]

Paul Fertser wrote:
> I'm no USB expert either but given what i already know about it, i
> think more hardware information is needed to be able to give a final
> verdict on the N900 usb host mode functionality. I'm not talking here
> about perfectly correct standards implementation or certification
> issues though i personally would prefer to have a working hostmode
> implementation to having a useless usb logo on the box.

Please take this with a grain of salt as I only spent a couple of hours 
on this.

I don't know at this time what Nokia went through to get their USB 
charging to work, but my expectation is that host mode is not supported 
because the n900 cannot supply power.
We also don't know what errata, if any, are active against the omap3430 
and the PMIC. FYI the beagleboard uses the omap3530 and the tps65950 
which appear to be functionally equivalent.

This http://www.maxim-ic.com/appnotes.cfm/an_pk/1822/ seems like a 
reasonably good introduction to OTG.

That spec as well as the kernel source indicates it is possible for a 
"B" device to act as host.  Note on Figure 5 of that web page above that 
the select for "B" device and peripheral vs. host device can be 
separate.  However it is still necessary for those peripheral/host 
resistors to swap even if the "A/B" resistors do not swap.

Swapping the resistors appears to generally be called SoftConnect.  I 
recall some notes in the kernel to the effect that SoftConnect is not 
supported, but whether the barrier is software or hardware I don't know.

"B" device as host looks to be at least partially implemented in the 
n900 kernel but that doesn't mean it works.  Connecting another device 
with linux and OTG to an n900 should be the easiest way to determine if 
"B" host mode is possible, assuming the other device can also support 
the HNP.

There's also supposed to be a new standard where an OTG can both charge 
and act as host at the same time. I have not looked at it.

In theory you should be able to measure the resistance between the 
different USB lines and that may be easier to do than pure speculation. 
There is also some amount of debugging information for usb available 
with the default kernel.

I think there are true OTG cables which may be an easier route for testing.

Regards, Sarah