DM6467T USB Issues

I see the same behavior on my board as well. As you also pointed there are lot of errors with regards to udev - device manager which could be cause of this issue. Attached boot up log on my board.

Regards,

Prateek

USB is not supported on DM6467T at this point in time.  There is a silicon bug that prevents a reliable operation of USB.  A new DM6467T revision with this bug fixed is expected shortly (I think end of April but not sure).

regards

swami

Thanks Swami for clarification.

USB will work on current DM6467T EVMs up to 60C case temp. Control the case temp by having a fan and/or heat sink to cool the part.

Regards, Srirami.

There is a black anodized aluminum finned heatsink on the DM6467T SoC already. 

We added a 3" box fan blowing through ducting to force air past the heatsink.

We tried connecting a simple USB serial mouse (known to work with DM6467) and also tried connecting the USB mouse to a powered USB hub, in case the problem is related to sourcing power to USB devices.

The SoC heatsink is cool to the touch even after running the DVSDK 3.10 H.264 demo video for a long time.

However, the USB is still non-functional despite the additional attention to cooling.

Eric

Eric,

From your description, it looks like the case temperature is not the problem.

The USB issue with DM6467T parts is that the USB enumeration fails at high temp (> 60C). Provide the log file.

Swamy,

From the log file, does it look like an USB enumeration fail? If not, its a different issue.

Regards, Srirami.

Hi Srirami,

Do you want to see the boot-time logfile (dmesg)?

Eric

Eric,

What is the intended mode of operation Host ?? If so there is a configuration issue with the kernel.  The log shows that the kernel is configured for Device mode.  Pl. correct it.

regards

swami

Hi Swami,

The intended mode is Host. We need the ability to hook up a mouse, keyboard, and possibly flash drives.

At boot the kernel does start USB in the device mode, is there a way to change this to Host operation without rebuilding the kernel?

Thanks,

Jeff

Hi Jeff,

My understanding is you need to change it at the kernel build stage. Further, you will need USB OTG mode support to move between host and driver at run time that is not available on DM6467 EVM.

Swami - please validate my understanding.

Thanks,

Prateek

You would have to choose the mode of operation when you are compiling the kernel (under USB menuconfig).

regards

swami

It worked. Thanks Swami.

Under make menuconfig, I have to make two changes

a) change the mode from device to host mode under Device Drivers-> USB and

b) Disable the USB gadget support. 

Prateek

We haven't solved our original problem quite yet. 

The USB mouse gets powered up now, but we have not verified that it actually works. 

And, somehow we broke our networking altogether when we tried to recompile without IPv6 (we want to use just IPv4 right now).

Eric

OK, after a little playing around with it, we believe that the USB mouse is working.  However, it is no longer at:

/dev/input/mouse0 (or, /dev/input/mice).

Instead, the first USB mouse is now created with device entry /dev/input/event0.

So, this means that code which runs on old and new distributions will have to check for the different device nodes.  This brings up a couple of interesting questions:

Why did the device node name change from mouse0 to event0?

Why is there no combined node in case of mutliple mice?  (if mouse0 and mouse1 both exist, then 'mice' includes mouse inputs from both, in the old distribution).  Now, if there is more than one mouse, they are enumerated as event0, event1, etc.  But there is no combined input node (like 'events').  Also, I checked the latest Ubuntu 9.10 behavior, and it too uses the 'event0' naming, but it also includes 'mouse0' and 'mice', preserving the legacy behavior.  How can we do that with Arago/DVSDK 3.10?

Eric

TMS320DM6467TZUT1 devices produced after May 1, 2010 have full USB support.  Our technical documentation has been updated to reflect this and to help customers identify when their product was produced.

For the original situation, it turns out that the default kernel was not configured to act as a USB host - only a USB device.  After recompiling the kernel the USB devices are accessible.

Thanks for the information.  How do I know when the chip on my evaluation board was produced?  What technical doucment? 

Look at Figure 1 (On Page 6) in DM6467T Silicon Errata (http://www.ti.com/litv/pdf/sprz307a) for more detais on how to decode package markings to get the manufacturing date.

Note that, old devices (that are dated prior to May 1, 2010) also support USB if you keep case temp less than 60C (using some fan), which will be good for the developmenet purpose.

Regards, Srirami.

Thank you for your quick response.  My chip has 07 in place of the YM code.  Was this produced in July 2010 or 2000?  (I expected an A for 10.)

There is a typo in the Silicon Errata. In "YM", "Y" is in Decimal (shows last digit of the year) and "M" is in hex (A=Oct, B=Nov, C=Dec).

Since we show only last of the year, "07" shows both 2000 and 2010. Since TI was not building DM6467T in 2000, assume it as 2010.

The typo will be corrected in the next revision of the Silicon Errata.

Regards, Srirami.