USB Device Mode

[nghia]

Hello R-Pi team!

Sorry to ask again about USB device mode on R-Pi.

Is the R-PI HW (version A or B) capable of supplying a USB device port?
If so, is there a plan to have the SW support this USB device port?

If not, is there a plan to add a new version (version C?) of R-Pi with support for both USB device port and USB host port?

Thank you
Rgds
Nghia
PS: Are you still on schedule for early Nov for the initial batch?

[abishur]

The answer to the end of the post is that the schedule was for November in general, a couple people have assumed it means early November The scheduled release date is for late November / early December.

The r-pi will only act as a host device. It's a full fledged PC, so I'm not expecting it to act as a USB device

[ukscone]

Quote from abishur on September 29, 2011, 04:20
The r-pi will only act as a host device. It's a full fledged PC, so I'm not expecting it to act as a USB device

I asked in the hardware questions on the front page and the answer was.

the model b no way no how as the chip that supplies the ethernet and two usb ports gets in the way. on the model a it might be possible using something to the method used on pxa2xx based phones (pam treo) and other devices might work but it hasn't been tried so we'll have to wait until we have r-pi's in hand and give it a go and see what happens.

[Lob0426]

The LAN9512 chipset stated it could be set to device or host mode. Wonder why it gets in the way? Makes sense it might be easier to try to setup an A model as a device. It will be interesting to see what happens when they are available.

[ukscone]

probably because it blocks direct access to the SoC supplied single usb. on all the devices where i've seen the usb host/otg hack work (hardwarewise) it's because the usb slot being manipulated is part of the SoC with only a resistor or two between it and the usb slot.

[Lob0426]

In reality if the A model had the 256 MB of memory it would be the best of the two to use for just about everything anyway. An external hub directly to the single USB host port would work just fine. Hardwire the hub to the RasPi and you would not even notice the difference between the two models except for the meory size.

[ukscone]

yes it's the model A i am more interested in even with only 128MB ram. Of course I do want a model B or two but the Model A is more interesting to me and if it had 256MB then i'd be one it quicker than if it was a bag of twiglets left unattended and unguarded

[gsh]

Quote from Lob0426 on October 6, 2011, 03:43
The LAN9512 chipset stated it could be set to device or host mode. Wonder why it gets in the way? Makes sense it might be easier to try to setup an A model as a device. It will be interesting to see what happens when they are available.

From my reading of the spec there is no OTG support in the 9512.

[Vindicator]

The B model is the more functional out of the box, but with the 2 usb and Ethernet in reality sharing the same resources there is going to be collisions and that will probably result in lower performance, the up side is you can plug in the Ethernet and a keyboard and a mouse, also video, and you are ready to go.
On the A model side you only have one usb, by using the powered hub for all the connections I suggest the throughput of the A might actually be better in the end. As the hub will probably control some of the collisions in the usb bus.
The fun part will be seeing if this theory pans out.

[MarkSmith]

Quote from abishur on September 29, 2011, 04:20
The answer to the end of the post is that the schedule was for November in general, a couple people have assumed it means early November The scheduled release date is for late November / early December.



Will that leave enough time to get a batch lot off to father Christmas?

A lot of children (small and grown up) are hoping to find one under the tree.

[nullstring]

ukscone, I would like to request jumpers for the connection between the USB port and the USB hub for Model B.
It would allow us to hook up to the USB port directly and disconnect the hub for circumstances like this.

It might be too late for this, but it sounds like a good idea to me.

[abishur]

ha ha that might be an issue. For starters ukscone, like myself, is not a part of the r-pi team, we just volunteer our time to help out on the forums as such we don't have access to an alpha board or pin out information. Secondly, I'm fairly certain there's not jumper connection between the usb port and usb hub. See the usb port is integrated into the ARM chip itself (think cell phone). The usb hub is directly connected to the correct pins on the ARM with all data lines buried somewhere in one of the layers of the PCB.

[Gert van Loo]

As the PCB is the same between A and B there are resistor jumpers which connect the BCM2835 USB port directly to the USB connector (Version A). However the BCM2835 USB port is always connected to the HUB so you have to either unsolder the HUB it or put a knife to the tracks. (As the PCB is not yet ready I don't know if you could get at those track...) You could try to to lift the 9512 pins.

[kme]

I'm getting confused. Could someone please clarify the relation between LAN9512 and BCM2835. Isn't the LAN9512 a USB hub (with a LAN adapter)? And if one throw a LAN9514 (four USB ports) into the mix?

[Burngate]

Is the PCB the same between A & B? Earlier I understood they were different, to the extent of 4 / 6 layers. If they're the same it would make Christmas more fun (fires up the surface-mount rework station) Do we yet know if the connectors are still all through hole?

[kme]

I'm with Burngate. I have the impression that the A and B are different PCBs, but my impression was they are physically different sizes. The 4 vs. 6 layer discussion was another one, as they had to give up the cheaper 4 layer PCB and swallow the more expensive 6 layer.

Uh, we are something like six weeks from announced deadline. We could really use some basic facts. They should be nailed down by now.

[abishur]

Actually, if I'm remembering correctly it's the alpha board that have 6 layers but the model A/B board will both be 4 layers. I believe the PCB layout is the same with the LAN9512 chip and ethernet port removed on the model A and one USB port instead of two. That said, without the LAN9512 chip the Model A PCB must be slightly different as the USB hooks up straight to the ARM....

[nullstring]

Quote from Gert van Loo on October 13, 2011, 13:02
As the PCB is the same between A and B there are resistor jumpers which connect the BCM2835 USB port directly to the USB connector (Version A). However the BCM2835 USB port is always connected to the HUB so you have to either unsolder the HUB it or put a knife to the tracks. (As the PCB is not yet ready I don't know if you could get at those track...) You could try to to lift the 9512 pins.

Thats the thing. It would be nice if we didn't have to unsolder/knife anything

If the BCM2835 USB port and the 9512 were connected by jumpers, we could easily disconnect the 9512 by removing the jumpers and this would also give us solder points for the BCM2835 USB port.

I forsee the ability to add a switch (on my side) to enable/disable the hub.
This way I could have USB client when I need it and an ethernet port when I don't.

I understand it's probably too late for this, but please take it under consideration.
It's simple change as far as the schematic and it could have many benefits assuming we get this USB mode working.

[Burngate]

If the boards are the same, and they're doing what I would think of doing - replace the 9512 with a 0R link or two - then I can still use the rework station. And if the tracks are buried only one layer down from either side, then I've got a nifty trick with the dentist's drill, 28 guage wire and some araldite ...

[Lob0426]

@abishur they ditched going to a four layer boards last I heard. They will be 6 layers.

The BCM2835 is a cell phone chip. The LAN9512 uses the single USB that comes out of that chip and breaks it into an Ethernet and 2 USB ports. The LAN9512 is a USB hub chip. The board is not designed to take another chip such as the LAN9514 (1 Ethernet and 4 USB ports) as a direct replacement. If you look it over enough you may very well be able to solder a LAN9514 onto it after removing the LAN9512. You may be able to carry the other two ports out as dongles. I don't know, is the pin count and locations the same between the LAN95xx chips? I have not look at the data sheet on the LAN9514 to compare it.

I would suspect that disconnectiong the Data lines through a switch then routing then direct would allow direct USB access to the BCM2835 chip. It all depends on if the lines are buried underneath the BCM2835! Some delicate work no matter how you look at it.

That USB hub with Ethernet is a good bet for an A model. If the A model could be had with the 256MB I would use it with one of those type of hubs stripped down to essentials. I have been looking at several similar models.

[kme]

The board is not designed to take another chip such as the LAN9514 (1 Ethernet and 4 USB ports) as a direct replacement.
Can you explain why not? The BCM2835 has one USB-lane to the LAN 9512 USB hub. Why would the LAN9514 USB hub be wildly different - seen from the BCM2835's point of view? The PCB would obviously need a redesign to handle the extra traces, but that's -isn't it?

[abishur]

Isn't that a fairly big thing in and of itself? I mean that's an extra 8 lanes (4 per USB port) that would need to be traced onto the PCB. Plus that would also mean that the two chips have a different amount of pins and therefore probably a different pin placement as well. All that adds up to being rather unlikely for a hobbyist to be able to swap these chips out themselves. Or as lobo was saying for you to be able to desolder one chip, and solder in the new chip as a direct replacement and immediately have everything up and running.

Also, thanks for the tip lobo, I must have missed/forgotten about the increased layer thing!

[Lob0426]

Can a LAN9514 be done as a hardware hack? maybe. But it will not be for the faint of heart.

http://www.smsc.com/media/Down.....s/9512.pdf
ftp://smsc.com/pub/Data_Sheets/9514.pdf

just looking at the pinouts it might be possible. They have the same number of pins but there are more of the pins used on the LAN9514. The locations of the pins that have comparable names are in the same places. There are 6 pins that are different. That would be 4 data, 2 for each USB port, plus 1 power pin for each port so the chip can control the USB devices eg. turn them off during eject and power management. 4 pins in middle of bottom and 2 lower right pins. all the other pins are labeled exactly the same. So you may be able to solder into these 6 pins to get the other 2 USB ports up. The grounds are common I believe. But you will still have to remove the LAN9512 to give it a try and you may damage the board trying.

[kme]

Well, I didn't suggest the 9512/9514 swap to be done by the end user on a final B board, but rather have it default on a future C (or whatever) board and was mere speculating if the board size itself would allow for such a complication.

In the end it will be much neater and cheaper than the extra external USB hub most users will be forced to add to the B board.

[Gert van Loo]

A and B version boards are the same, just less components on the A-board (and two 0-ohm resistors to connect the USB port).
As to the number of layers. We hoped to get the design done in 4 layers to reduce cost. But it turned out to be very, very difficult. Also there was a serious risk of it not working reliably and we would have to do a re-design. So in the worst-case scenario we would also have missed the end-of-November deadline, probably by two months. Thus the board will be 6-layers and, if nothing goes wrong, you have something to play with during Christmas.