Re: No HDMI signal from Nitrogen6x board with daisy

[Eric Nelson <eric.nelson@boundarydevices.com>]

Hi Peter,

On 05/28/2014 09:09 AM, Peter Bergin wrote:
> Hi Eric,
>
> On 05/28/2014 04:03 PM, Eric Nelson wrote:
>> Thanks Peter,
>>
>> On 05/27/2014 11:51 PM, Peter Bergin wrote:
>>> Hi Eric,
>>>
>>> On 05/27/2014 11:11 PM, Eric Nelson wrote:
>>>> Hi Peter,
>>>>
>>>> On 05/25/2014 11:43 PM, Peter Bergin wrote:
>>>>> Hi Eric,
>>>>>
>>>>>    <snip>
>>>>>
>>>>> In order to understand more what happens during the process I have
>>>>> enabled dynamic debug in the kernel and turned on messages from mxc_hdmi
>>>>> and mxc_ipu. I have attached my /var/log/messages if someone more
>>>>> familiar with the drivers can have a look.
>>>>>
>>>> Thanks for the detailed information.
>>>>
>>>>> IPU_INT_STAT_5 is classed as Error Interrupts in the Reference Manual.
>>>>>
>>>>> May 26 06:02:30 nitrogen6x user.warn kernel: imx-ipuv3 2400000.ipu: IPU
>>>>> Warning - IPU_INT_STAT_5 = 0x00800000
>>>>>
>>>>> This seems to be a IMDAC_NFB4EOF_ERR_x. NFB4EOF - New-frame before
>>>>> end-of-frame.
>>>>>
>>>>> How can I get more understanding what's wrong in the IPU? Any ideas?
>>>>>
>>>> Nothing concrete. Perhaps the Freescaler's have some idea.
>>>>
>>>> The only time I've seen the NFB4EOF error has been when I've
>>>> had a bug (in a camera driver). I believe it means that there's
>>>> an initialization problem.
>>>>
>>>>
>>>> I've been trying to compare against what I see here,
>>>> and the two primary suspects are:
>>>>
>>>> 	- some quirk of your EDID
>>>> 	- the fact that you're running an older processor revision
>>>> 	(TO 1.0).
>>>>
>>>> I haven't yet found a TO 1.0 board to test against.


After trying everything else (hard-coding your EDID, and using a DVI
cable), I managed to find a board with Silicon Revision 1.0, and
it's failing more or less as you described:

	- 640x480 works
	- 1920x1080p fails

I also see those "timeout waiting for flip irq" messages whenever
switching to a resolution that doesn't function.

I'm not sure why, but it's probably time to grep through those pesky
tests for tapeout:
	imx_get_soc_revision() == IMX_CHIP_REVISION_1_0

TO 1.0 **did** work at 1080P in the older kernels.

Regards,


Eric