KWorld UB435-Q Support

KWorld UB435-Q Support

[Robert Cicconetti]

Jarod Wilson wrote:
> With that caveat emptor, here's where the tree that should at least get you 95% of the way
> there with that stick resides:
>
> http://www.kernellabs.com/hg/~mkrufky/lgdt3304-3/

Okay... I built the tip of the archive linked above. It works with my
UB435-Q fairly well, built against 2.6.28-15-generic #52-Ubuntu SMP
x86_64. I've been able to stream QAM256 content for several hours
reliably. Mythfrontend works somewhat... it'll tune the initial
channel, but fails afterward. I suspect it is timing out while waiting
for the RF tracking filter calibration... it adds about 6 seconds to
every tuning operation.

[  812.465930] tda18271: performing RF tracking filter calibration
[  818.572446] tda18271: RF tracking filter calibration complete
[  818.953946] tda18271: performing RF tracking filter calibration
[  825.093211] tda18271: RF tracking filter calibration complete

Initialization:

[  217.319831] em28xx: New device USB 2870 Device @ 480 Mbps (1b80:a340, interfa
ce 0, class 0)
[  217.319842] em28xx #0: Identified as KWorld PlusTV 340U (ATSC) (card=71)
[  217.320062] em28xx #0: chip ID is em2870
[  217.448225] em28xx #0: i2c eeprom 00: 1a eb 67 95 80 1b 40 a3 c0 13
6b 10 6a 22 00 00
[  217.448244] em28xx #0: i2c eeprom 10: 00 00 04 57 00 0d 00 00 00 00
00 00 00 00 00 00
[  217.448260] em28xx #0: i2c eeprom 20: 44 00 00 00 f0 10 01 00 00 00
00 00 5b 1c c0 00
[  217.448275] em28xx #0: i2c eeprom 30: 00 00 20 40 20 80 02 20 01 01
00 00 00 00 00 00
[  217.448290] em28xx #0: i2c eeprom 40: 00 00 00 00 00 00 00 00 00 00
00 00 00 00 00 00
[  217.448304] em28xx #0: i2c eeprom 50: 00 00 00 00 00 00 00 00 00 00
00 00 00 00 00 00
[  217.448319] em28xx #0: i2c eeprom 60: 00 00 00 00 00 00 00 00 00 00
22 03 55 00 53 00
[  217.448333] em28xx #0: i2c eeprom 70: 42 00 20 00 32 00 38 00 37 00
30 00 20 00 44 00
[  217.448348] em28xx #0: i2c eeprom 80: 65 00 76 00 69 00 63 00 65 00
00 00 00 00 00 00
[  217.448362] em28xx #0: i2c eeprom 90: 00 00 00 00 00 00 00 00 00 00
00 00 00 00 00 00
[  217.448376] em28xx #0: i2c eeprom a0: 00 00 00 00 00 00 00 00 00 00
00 00 00 00 00 00
[  217.448390] em28xx #0: i2c eeprom b0: 00 00 00 00 00 00 00 00 00 00
00 00 00 00 00 00
[  217.448405] em28xx #0: i2c eeprom c0: 00 00 00 00 00 00 00 00 00 00
00 00 00 00 00 00
[  217.448419] em28xx #0: i2c eeprom d0: 00 00 00 00 00 00 00 00 00 00
00 00 00 00 00 00
[  217.448433] em28xx #0: i2c eeprom e0: 00 00 00 00 00 00 00 00 00 00
00 00 00 00 00 00
[  217.448447] em28xx #0: i2c eeprom f0: 00 00 00 00 00 00 00 00 00 00
00 00 00 00 00 00
[  217.448465] em28xx #0: EEPROM ID= 0x9567eb1a, EEPROM hash = 0x2888a312
[  217.448469] em28xx #0: EEPROM info:
[  217.448472] em28xx #0:       No audio on board.
[  217.448475] em28xx #0:       500mA max power
[  217.448479] em28xx #0:       Table at 0x04, strings=0x226a, 0x0000, 0x0000
[  217.448488] em28xx #0: v4l2 driver version 0.1.2
[  217.454853] em28xx #0: V4L2 device registered as /dev/video0 and /dev/vbi0
[  217.454889] usbcore: registered new interface driver em28xx
[  217.454895] em28xx driver loaded

Any suggestions? Further data needed?

Thanks!
R C

Re: KWorld UB435-Q Support

[Jarod Wilson]

On Oct 7, 2009, at 9:39 PM, Robert Cicconetti wrote:

> Jarod Wilson wrote:
>> With that caveat emptor, here's where the tree that should at least  
>> get you 95% of the way
>> there with that stick resides:
>>
>> http://www.kernellabs.com/hg/~mkrufky/lgdt3304-3/
>
> Okay... I built the tip of the archive linked above. It works with my
> UB435-Q fairly well, built against 2.6.28-15-generic #52-Ubuntu SMP
> x86_64. I've been able to stream QAM256 content for several hours
> reliably. Mythfrontend works somewhat... it'll tune the initial
> channel, but fails afterward. I suspect it is timing out while waiting
> for the RF tracking filter calibration... it adds about 6 seconds to
> every tuning operation.
>
> [  812.465930] tda18271: performing RF tracking filter calibration
> [  818.572446] tda18271: RF tracking filter calibration complete
> [  818.953946] tda18271: performing RF tracking filter calibration
> [  825.093211] tda18271: RF tracking filter calibration complete
>
> Initialization:
>
> [  217.319831] em28xx: New device USB 2870 Device @ 480 Mbps  
> (1b80:a340, interfa
> ce 0, class 0)
> [  217.319842] em28xx #0: Identified as KWorld PlusTV 340U (ATSC)  
> (card=71)
> [  217.320062] em28xx #0: chip ID is em2870
> [  217.448225] em28xx #0: i2c eeprom 00: 1a eb 67 95 80 1b 40 a3 c0 13
> 6b 10 6a 22 00 00
> [  217.448244] em28xx #0: i2c eeprom 10: 00 00 04 57 00 0d 00 00 00 00
> 00 00 00 00 00 00
> [  217.448260] em28xx #0: i2c eeprom 20: 44 00 00 00 f0 10 01 00 00 00
> 00 00 5b 1c c0 00
> [  217.448275] em28xx #0: i2c eeprom 30: 00 00 20 40 20 80 02 20 01 01
> 00 00 00 00 00 00
> [  217.448290] em28xx #0: i2c eeprom 40: 00 00 00 00 00 00 00 00 00 00
> 00 00 00 00 00 00
> [  217.448304] em28xx #0: i2c eeprom 50: 00 00 00 00 00 00 00 00 00 00
> 00 00 00 00 00 00
> [  217.448319] em28xx #0: i2c eeprom 60: 00 00 00 00 00 00 00 00 00 00
> 22 03 55 00 53 00
> [  217.448333] em28xx #0: i2c eeprom 70: 42 00 20 00 32 00 38 00 37 00
> 30 00 20 00 44 00
> [  217.448348] em28xx #0: i2c eeprom 80: 65 00 76 00 69 00 63 00 65 00
> 00 00 00 00 00 00
> [  217.448362] em28xx #0: i2c eeprom 90: 00 00 00 00 00 00 00 00 00 00
> 00 00 00 00 00 00
> [  217.448376] em28xx #0: i2c eeprom a0: 00 00 00 00 00 00 00 00 00 00
> 00 00 00 00 00 00
> [  217.448390] em28xx #0: i2c eeprom b0: 00 00 00 00 00 00 00 00 00 00
> 00 00 00 00 00 00
> [  217.448405] em28xx #0: i2c eeprom c0: 00 00 00 00 00 00 00 00 00 00
> 00 00 00 00 00 00
> [  217.448419] em28xx #0: i2c eeprom d0: 00 00 00 00 00 00 00 00 00 00
> 00 00 00 00 00 00
> [  217.448433] em28xx #0: i2c eeprom e0: 00 00 00 00 00 00 00 00 00 00
> 00 00 00 00 00 00
> [  217.448447] em28xx #0: i2c eeprom f0: 00 00 00 00 00 00 00 00 00 00
> 00 00 00 00 00 00
> [  217.448465] em28xx #0: EEPROM ID= 0x9567eb1a, EEPROM hash =  
> 0x2888a312
> [  217.448469] em28xx #0: EEPROM info:
> [  217.448472] em28xx #0:       No audio on board.
> [  217.448475] em28xx #0:       500mA max power
> [  217.448479] em28xx #0:       Table at 0x04, strings=0x226a,  
> 0x0000, 0x0000
> [  217.448488] em28xx #0: v4l2 driver version 0.1.2
> [  217.454853] em28xx #0: V4L2 device registered as /dev/video0 and / 
> dev/vbi0
> [  217.454889] usbcore: registered new interface driver em28xx
> [  217.454895] em28xx driver loaded
>
> Any suggestions? Further data needed?

Nothing off the top of my head, no. But I've got a UB435-Q of my own  
now, sitting on my desk waiting for me to poke at it... Not sure when  
I'll have time to actually poke at it though. :\

-- 
Jarod Wilson
jarod@wilsonet.com

Re: KWorld UB435-Q Support

[Robert Cicconetti]

On Wed, Oct 7, 2009 at 10:08 PM, Jarod Wilson <jarod@wilsonet.com> wrote:
> On Oct 7, 2009, at 9:39 PM, Robert Cicconetti wrote:
>> Okay... I built the tip of the archive linked above. It works with my
>> UB435-Q fairly well, built against 2.6.28-15-generic #52-Ubuntu SMP
>> x86_64. I've been able to stream QAM256 content for several hours
>> reliably. Mythfrontend works somewhat... it'll tune the initial
>> channel, but fails afterward. I suspect it is timing out while waiting
>> for the RF tracking filter calibration... it adds about 6 seconds to
>> every tuning operation.
>>
>> [  812.465930] tda18271: performing RF tracking filter calibration
>> [  818.572446] tda18271: RF tracking filter calibration complete
>> [  818.953946] tda18271: performing RF tracking filter calibration
>> [  825.093211] tda18271: RF tracking filter calibration complete
>>
>> Any suggestions? Further data needed?
>
> Nothing off the top of my head, no. But I've got a UB435-Q of my own now,
> sitting on my desk waiting for me to poke at it... Not sure when I'll have
> time to actually poke at it though. :\

A little further poking yields that RF_CAL_OK in EP1 is 0, which is
why it keeps recalibrating.

I've commented out the part of the code that recalibrates if RF_CAL_OK
is 0; EP1 always seems to be c6... and now mythfrontend is happy. :)

This is not a long term solution, but as ugly hacks go it was pretty
straight forward. :)

-Bob

Re: KWorld UB435-Q Support

[Jarod Wilson]

On 10/09/2009 09:27 PM, Robert Cicconetti wrote:
> On Wed, Oct 7, 2009 at 10:08 PM, Jarod Wilson<jarod@wilsonet.com>  wrote:
>> On Oct 7, 2009, at 9:39 PM, Robert Cicconetti wrote:
>>> Okay... I built the tip of the archive linked above. It works with my
>>> UB435-Q fairly well, built against 2.6.28-15-generic #52-Ubuntu SMP
>>> x86_64. I've been able to stream QAM256 content for several hours
>>> reliably. Mythfrontend works somewhat... it'll tune the initial
>>> channel, but fails afterward. I suspect it is timing out while waiting
>>> for the RF tracking filter calibration... it adds about 6 seconds to
>>> every tuning operation.
>>>
>>> [  812.465930] tda18271: performing RF tracking filter calibration
>>> [  818.572446] tda18271: RF tracking filter calibration complete
>>> [  818.953946] tda18271: performing RF tracking filter calibration
>>> [  825.093211] tda18271: RF tracking filter calibration complete
>>>
>>> Any suggestions? Further data needed?
>>
>> Nothing off the top of my head, no. But I've got a UB435-Q of my own now,
>> sitting on my desk waiting for me to poke at it... Not sure when I'll have
>> time to actually poke at it though. :\
>
> A little further poking yields that RF_CAL_OK in EP1 is 0, which is
> why it keeps recalibrating.
>
> I've commented out the part of the code that recalibrates if RF_CAL_OK
> is 0; EP1 always seems to be c6... and now mythfrontend is happy. :)
>
> This is not a long term solution, but as ugly hacks go it was pretty
> straight forward. :)

Finally got around to poking at this again. Forward-ported the patches 
to the current v4l-dvb tip, and gave 'em a spin with my own UB435-Q, as 
well as a 340U that Doug gave me when he was in town a bit ago. Both are 
working just fine with my QAM feed here at the house, albeit with the 
same lengthy delay when changing channels you (Robert) mentioned. At a 
glance, I was hoping simply setting rf_cal_on_startup for the 
card-specific tda18271_config would remove the delay, but neither a 0 or 
a 1 seems to particularly help with tuning delays. Hoping maybe Mike has 
an idea on this part...

In related news, I actually managed to get my original 340U with the C1 
tuner to work briefly as well, and with the same code, no tuning delays. 
Seems either the PCB is cracked or the usb connector is just that bad, 
and it only works when positioned just so...

-- 
Jarod Wilson
jarod@wilsonet.com

Re: KWorld UB435-Q Support

[Jarod Wilson]

On 11/15/2009 11:46 PM, Jarod Wilson wrote:
> On 10/09/2009 09:27 PM, Robert Cicconetti wrote:
>> On Wed, Oct 7, 2009 at 10:08 PM, Jarod Wilson<jarod@wilsonet.com> wrote:
>>> On Oct 7, 2009, at 9:39 PM, Robert Cicconetti wrote:
>>>> Okay... I built the tip of the archive linked above. It works with my
>>>> UB435-Q fairly well, built against 2.6.28-15-generic #52-Ubuntu SMP
>>>> x86_64. I've been able to stream QAM256 content for several hours
>>>> reliably. Mythfrontend works somewhat... it'll tune the initial
>>>> channel, but fails afterward. I suspect it is timing out while waiting
>>>> for the RF tracking filter calibration... it adds about 6 seconds to
>>>> every tuning operation.
>>>>
>>>> [ 812.465930] tda18271: performing RF tracking filter calibration
>>>> [ 818.572446] tda18271: RF tracking filter calibration complete
>>>> [ 818.953946] tda18271: performing RF tracking filter calibration
>>>> [ 825.093211] tda18271: RF tracking filter calibration complete
>>>>
>>>> Any suggestions? Further data needed?
>>>
>>> Nothing off the top of my head, no. But I've got a UB435-Q of my own
>>> now,
>>> sitting on my desk waiting for me to poke at it... Not sure when I'll
>>> have
>>> time to actually poke at it though. :\
>>
>> A little further poking yields that RF_CAL_OK in EP1 is 0, which is
>> why it keeps recalibrating.
>>
>> I've commented out the part of the code that recalibrates if RF_CAL_OK
>> is 0; EP1 always seems to be c6... and now mythfrontend is happy. :)
>>
>> This is not a long term solution, but as ugly hacks go it was pretty
>> straight forward. :)
>
> Finally got around to poking at this again. Forward-ported the patches
> to the current v4l-dvb tip

Meant to include this:

http://wilsonet.com/jarod/junk/kworld-a340-20091115/

> , and gave 'em a spin with my own UB435-Q, as
> well as a 340U that Doug gave me when he was in town a bit ago. Both are
> working just fine with my QAM feed here at the house, albeit with the
> same lengthy delay when changing channels you (Robert) mentioned. At a
> glance, I was hoping simply setting rf_cal_on_startup for the
> card-specific tda18271_config would remove the delay, but neither a 0 or
> a 1 seems to particularly help with tuning delays. Hoping maybe Mike has
> an idea on this part...
>
> In related news, I actually managed to get my original 340U with the C1
> tuner to work briefly as well, and with the same code, no tuning delays.
> Seems either the PCB is cracked or the usb connector is just that bad,
> and it only works when positioned just so...

I'll give all three sticks I've got here a quick spin with an OTA signal 
tomorrow too. But I think I'm not seeing any significant reason to not 
move forward with trying to get this code finally merged.

-- 
Jarod Wilson
jarod@wilsonet.com

Re: KWorld UB435-Q Support

[Jarod Wilson]

On Nov 15, 2009, at 11:55 PM, Jarod Wilson wrote:

> On 11/15/2009 11:46 PM, Jarod Wilson wrote:
>> On 10/09/2009 09:27 PM, Robert Cicconetti wrote:
>>> On Wed, Oct 7, 2009 at 10:08 PM, Jarod Wilson<jarod@wilsonet.com> wrote:
>>>> On Oct 7, 2009, at 9:39 PM, Robert Cicconetti wrote:
>>>>> Okay... I built the tip of the archive linked above. It works with my
>>>>> UB435-Q fairly well, built against 2.6.28-15-generic #52-Ubuntu SMP
>>>>> x86_64. I've been able to stream QAM256 content for several hours
>>>>> reliably. Mythfrontend works somewhat... it'll tune the initial
>>>>> channel, but fails afterward. I suspect it is timing out while waiting
>>>>> for the RF tracking filter calibration... it adds about 6 seconds to
>>>>> every tuning operation.
>>>>> 
>>>>> [ 812.465930] tda18271: performing RF tracking filter calibration
>>>>> [ 818.572446] tda18271: RF tracking filter calibration complete
>>>>> [ 818.953946] tda18271: performing RF tracking filter calibration
>>>>> [ 825.093211] tda18271: RF tracking filter calibration complete
>>>>> 
>>>>> Any suggestions? Further data needed?
>>>> 
>>>> Nothing off the top of my head, no. But I've got a UB435-Q of my own
>>>> now,
>>>> sitting on my desk waiting for me to poke at it... Not sure when I'll
>>>> have
>>>> time to actually poke at it though. :\
>>> 
>>> A little further poking yields that RF_CAL_OK in EP1 is 0, which is
>>> why it keeps recalibrating.
>>> 
>>> I've commented out the part of the code that recalibrates if RF_CAL_OK
>>> is 0; EP1 always seems to be c6... and now mythfrontend is happy. :)
>>> 
>>> This is not a long term solution, but as ugly hacks go it was pretty
>>> straight forward. :)
>> 
>> Finally got around to poking at this again. Forward-ported the patches
>> to the current v4l-dvb tip
> 
> Meant to include this:
> 
> http://wilsonet.com/jarod/junk/kworld-a340-20091115/
> 
>> , and gave 'em a spin with my own UB435-Q, as
>> well as a 340U that Doug gave me when he was in town a bit ago. Both are
>> working just fine with my QAM feed here at the house, albeit with the
>> same lengthy delay when changing channels you (Robert) mentioned. At a
>> glance, I was hoping simply setting rf_cal_on_startup for the
>> card-specific tda18271_config would remove the delay, but neither a 0 or
>> a 1 seems to particularly help with tuning delays. Hoping maybe Mike has
>> an idea on this part...
>> 
>> In related news, I actually managed to get my original 340U with the C1
>> tuner to work briefly as well, and with the same code, no tuning delays.
>> Seems either the PCB is cracked or the usb connector is just that bad,
>> and it only works when positioned just so...
> 
> I'll give all three sticks I've got here a quick spin with an OTA signal tomorrow too. But I think I'm not seeing any significant reason to not move forward with trying to get this code finally merged.

All three check out with an OTA signal as well. I'll try to poke at the tuning delay thing a bit more tonight, but I'm inclined to send off formal patches Real Soon Now.

-- 
Jarod Wilson
jarod@wilsonet.com

Re: KWorld UB435-Q Support

[Michael Krufky]

Jarod Wilson wrote:
> On Nov 15, 2009, at 11:55 PM, Jarod Wilson wrote:
> 
>> On 11/15/2009 11:46 PM, Jarod Wilson wrote:
>>> On 10/09/2009 09:27 PM, Robert Cicconetti wrote:
>>>> On Wed, Oct 7, 2009 at 10:08 PM, Jarod Wilson<jarod@wilsonet.com> wrote:
>>>>> On Oct 7, 2009, at 9:39 PM, Robert Cicconetti wrote:
>>>>>> Okay... I built the tip of the archive linked above. It works with my
>>>>>> UB435-Q fairly well, built against 2.6.28-15-generic #52-Ubuntu SMP
>>>>>> x86_64. I've been able to stream QAM256 content for several hours
>>>>>> reliably. Mythfrontend works somewhat... it'll tune the initial
>>>>>> channel, but fails afterward. I suspect it is timing out while waiting
>>>>>> for the RF tracking filter calibration... it adds about 6 seconds to
>>>>>> every tuning operation.
>>>>>>
>>>>>> [ 812.465930] tda18271: performing RF tracking filter calibration
>>>>>> [ 818.572446] tda18271: RF tracking filter calibration complete
>>>>>> [ 818.953946] tda18271: performing RF tracking filter calibration
>>>>>> [ 825.093211] tda18271: RF tracking filter calibration complete
>>>>>>
>>>>>> Any suggestions? Further data needed?
>>>>> Nothing off the top of my head, no. But I've got a UB435-Q of my own
>>>>> now,
>>>>> sitting on my desk waiting for me to poke at it... Not sure when I'll
>>>>> have
>>>>> time to actually poke at it though. :\
>>>> A little further poking yields that RF_CAL_OK in EP1 is 0, which is
>>>> why it keeps recalibrating.
>>>>
>>>> I've commented out the part of the code that recalibrates if RF_CAL_OK
>>>> is 0; EP1 always seems to be c6... and now mythfrontend is happy. :)
>>>>
>>>> This is not a long term solution, but as ugly hacks go it was pretty
>>>> straight forward. :)
>>> Finally got around to poking at this again. Forward-ported the patches
>>> to the current v4l-dvb tip
>> Meant to include this:
>>
>> http://wilsonet.com/jarod/junk/kworld-a340-20091115/
>>
>>> , and gave 'em a spin with my own UB435-Q, as
>>> well as a 340U that Doug gave me when he was in town a bit ago. Both are
>>> working just fine with my QAM feed here at the house, albeit with the
>>> same lengthy delay when changing channels you (Robert) mentioned. At a
>>> glance, I was hoping simply setting rf_cal_on_startup for the
>>> card-specific tda18271_config would remove the delay, but neither a 0 or
>>> a 1 seems to particularly help with tuning delays. Hoping maybe Mike has
>>> an idea on this part...
>>>
>>> In related news, I actually managed to get my original 340U with the C1
>>> tuner to work briefly as well, and with the same code, no tuning delays.
>>> Seems either the PCB is cracked or the usb connector is just that bad,
>>> and it only works when positioned just so...
>> I'll give all three sticks I've got here a quick spin with an OTA signal tomorrow too. But I think I'm not seeing any significant reason to not move forward with trying to get this code finally merged.
> 
> All three check out with an OTA signal as well. I'll try to poke at the tuning delay thing a bit more tonight, but I'm inclined to send off formal patches Real Soon Now.
> 


Jarod,

I havent had time to review this entire email thread yet... but:

 >>>>> [ 812.465930] tda18271: performing RF tracking filter calibration
 >>>>> [ 818.572446] tda18271: RF tracking filter calibration complete
 >>>>> [ 818.953946] tda18271: performing RF tracking filter calibration
 >>>>> [ 825.093211] tda18271: RF tracking filter calibration complete


If you see this happen more than once consecutively, and there is only 1 
silicon tuner present, then it means something very bad is happening, 
and there is a chance of burning out a part.  I still wouldnt not 
recommend any mainline merge until you can prevent this behavior -- I 
suspect that a GPIO reset is being toggled where it shouldnt be, which 
should be harmless ... but until we fix it, we cant be sure what damage 
might get done...

The RF tracking filter calibration is a procedure that should only 
happen once while the tuner is powered on -- it should *only* be 
repeated if the tuner indicated that calibration is necessary, and that 
would only happen after a hardware reset.

This still looks fishy to me...

-Mike

Re: KWorld UB435-Q Support

[Robert Cicconetti]

On Tue, Nov 17, 2009 at 12:55 AM, Michael Krufky <mkrufky@linuxtv.org> wrote:
>>>>>> [ 812.465930] tda18271: performing RF tracking filter calibration
>>>>>> [ 818.572446] tda18271: RF tracking filter calibration complete
>>>>>> [ 818.953946] tda18271: performing RF tracking filter calibration
>>>>>> [ 825.093211] tda18271: RF tracking filter calibration complete
>
>
> If you see this happen more than once consecutively, and there is only 1
> silicon tuner present, then it means something very bad is happening, and
> there is a chance of burning out a part.  I still wouldnt not recommend any
> mainline merge until you can prevent this behavior -- I suspect that a GPIO
> reset is being toggled where it shouldnt be, which should be harmless ...
> but until we fix it, we cant be sure what damage might get done...
>
> The RF tracking filter calibration is a procedure that should only happen
> once while the tuner is powered on -- it should *only* be repeated if the
> tuner indicated that calibration is necessary, and that would only happen
> after a hardware reset.
>
> This still looks fishy to me...

It happened at every tuning operation, and made mythfrontend unhappy
(unable to tune after the first channel). I disabled the check for
RF_CAL_OK which triggered the recalibration, and mythfrontend worked.
The stick has been plugged in for a few months, so presumably would've
caught on fire by now if it was going to. It would be nice if the
tuning delay went away, though.. it still takes ~6 seconds to switch
frequencies.

I have not yet compiled and tested the lastest patches from Jarod.

R C

Re: KWorld UB435-Q Support

[Jarod Wilson]

On Nov 17, 2009, at 1:03 AM, Robert Cicconetti wrote:

> On Tue, Nov 17, 2009 at 12:55 AM, Michael Krufky <mkrufky@linuxtv.org> wrote:
>>>>>>> [ 812.465930] tda18271: performing RF tracking filter calibration
>>>>>>> [ 818.572446] tda18271: RF tracking filter calibration complete
>>>>>>> [ 818.953946] tda18271: performing RF tracking filter calibration
>>>>>>> [ 825.093211] tda18271: RF tracking filter calibration complete
>> 
>> 
>> If you see this happen more than once consecutively, and there is only 1
>> silicon tuner present, then it means something very bad is happening, and
>> there is a chance of burning out a part.  I still wouldnt not recommend any
>> mainline merge until you can prevent this behavior -- I suspect that a GPIO
>> reset is being toggled where it shouldnt be, which should be harmless ...
>> but until we fix it, we cant be sure what damage might get done...
>> 
>> The RF tracking filter calibration is a procedure that should only happen
>> once while the tuner is powered on -- it should *only* be repeated if the
>> tuner indicated that calibration is necessary, and that would only happen
>> after a hardware reset.
>> 
>> This still looks fishy to me...

Agreed. I did manage to dig into this some more last night, something is definitely still awry. Here's a dmesg dump with some extra debug spew added in key spots:

...
em28xx driver loaded
tda18271 4-0060: creating new instance
TDA18271HD/C2 detected @ 4-0060
tda18271: R_EP1 is 0xce
cal is not initialized (cal_initialized=false)...
tda18271: performing RF tracking filter calibration
tda18271: RF tracking filter calibration complete (0xde)
DVB: registering new adapter (em28xx #0)
DVB: registering adapter 0 frontend 0 (LG Electronics LGDT3304 VSB/QAM Frontend)...
em28xx #0: Successfully loaded em28xx-dvb
Em28xx: Initialized (Em28xx dvb Extension) extension

1st tuning attempt

tda18271: R_EP1 is 0x00
cal is not initialized (cal_initialized=true)...
tda18271: performing RF tracking filter calibration
tda18271: RF tracking filter calibration complete (0x00)
tda18271: R_EP1 is 0x00
cal is not initialized (cal_initialized=true)...
tda18271: performing RF tracking filter calibration
tda18271: RF tracking filter calibration complete (0x00)

2nd tuning attempt

tda18271: R_EP1 is 0x00
cal is not initialized (cal_initialized=true)...
tda18271: performing RF tracking filter calibration
tda18271: RF tracking filter calibration complete (0x00)
tda18271: R_EP1 is 0x00
cal is not initialized (cal_initialized=true)...
tda18271: performing RF tracking filter calibration
tda18271: RF tracking filter calibration complete (0x00)

I'll try tweaking the GPIO reset mask and whatnot, definitely does seem like something's getting reset that shouldn't, because you can clearly see that cal *was* initialized, then R_EP1 got zeroed out.

> It happened at every tuning operation, and made mythfrontend unhappy
> (unable to tune after the first channel). I disabled the check for
> RF_CAL_OK which triggered the recalibration, and mythfrontend worked.

Yeah, tuning is much quicker here if I skip that check as well, but its definitely not the proper fix.

> The stick has been plugged in for a few months, so presumably would've
> caught on fire by now if it was going to. It would be nice if the
> tuning delay went away, though.. it still takes ~6 seconds to switch
> frequencies.

Wait, it still takes that long with the check gone? I didn't poke for very long with the check disabled, mostly focusing on trying to figure out why things are going haywire.

> I have not yet compiled and tested the lastest patches from Jarod.

Really shouldn't be any difference from what you've got, they're just rebased to the latest v4l-dvb tree.

-- 
Jarod Wilson
jarod@wilsonet.com

Re: KWorld UB435-Q Support

[Devin Heitmueller]

On Tue, Nov 17, 2009 at 9:15 AM, Jarod Wilson <jarod@wilsonet.com> wrote:
> On Nov 17, 2009, at 1:03 AM, Robert Cicconetti wrote:
>
>> On Tue, Nov 17, 2009 at 12:55 AM, Michael Krufky <mkrufky@linuxtv.org> wrote:
>>>>>>>> [ 812.465930] tda18271: performing RF tracking filter calibration
>>>>>>>> [ 818.572446] tda18271: RF tracking filter calibration complete
>>>>>>>> [ 818.953946] tda18271: performing RF tracking filter calibration
>>>>>>>> [ 825.093211] tda18271: RF tracking filter calibration complete
>>>
>>>
>>> If you see this happen more than once consecutively, and there is only 1
>>> silicon tuner present, then it means something very bad is happening, and
>>> there is a chance of burning out a part.  I still wouldnt not recommend any
>>> mainline merge until you can prevent this behavior -- I suspect that a GPIO
>>> reset is being toggled where it shouldnt be, which should be harmless ...
>>> but until we fix it, we cant be sure what damage might get done...
>>>
>>> The RF tracking filter calibration is a procedure that should only happen
>>> once while the tuner is powered on -- it should *only* be repeated if the
>>> tuner indicated that calibration is necessary, and that would only happen
>>> after a hardware reset.
>>>
>>> This still looks fishy to me...
>
> Agreed. I did manage to dig into this some more last night, something is definitely still awry. Here's a dmesg dump with some extra debug spew added in key spots:
>
> ...
> em28xx driver loaded
> tda18271 4-0060: creating new instance
> TDA18271HD/C2 detected @ 4-0060
> tda18271: R_EP1 is 0xce
> cal is not initialized (cal_initialized=false)...
> tda18271: performing RF tracking filter calibration
> tda18271: RF tracking filter calibration complete (0xde)
> DVB: registering new adapter (em28xx #0)
> DVB: registering adapter 0 frontend 0 (LG Electronics LGDT3304 VSB/QAM Frontend)...
> em28xx #0: Successfully loaded em28xx-dvb
> Em28xx: Initialized (Em28xx dvb Extension) extension
>
> 1st tuning attempt
>
> tda18271: R_EP1 is 0x00
> cal is not initialized (cal_initialized=true)...
> tda18271: performing RF tracking filter calibration
> tda18271: RF tracking filter calibration complete (0x00)
> tda18271: R_EP1 is 0x00
> cal is not initialized (cal_initialized=true)...
> tda18271: performing RF tracking filter calibration
> tda18271: RF tracking filter calibration complete (0x00)
>
> 2nd tuning attempt
>
> tda18271: R_EP1 is 0x00
> cal is not initialized (cal_initialized=true)...
> tda18271: performing RF tracking filter calibration
> tda18271: RF tracking filter calibration complete (0x00)
> tda18271: R_EP1 is 0x00
> cal is not initialized (cal_initialized=true)...
> tda18271: performing RF tracking filter calibration
> tda18271: RF tracking filter calibration complete (0x00)
>
> I'll try tweaking the GPIO reset mask and whatnot, definitely does seem like something's getting reset that shouldn't, because you can clearly see that cal *was* initialized, then R_EP1 got zeroed out.
>
>> It happened at every tuning operation, and made mythfrontend unhappy
>> (unable to tune after the first channel). I disabled the check for
>> RF_CAL_OK which triggered the recalibration, and mythfrontend worked.
>
> Yeah, tuning is much quicker here if I skip that check as well, but its definitely not the proper fix.
>
>> The stick has been plugged in for a few months, so presumably would've
>> caught on fire by now if it was going to. It would be nice if the
>> tuning delay went away, though.. it still takes ~6 seconds to switch
>> frequencies.
>
> Wait, it still takes that long with the check gone? I didn't poke for very long with the check disabled, mostly focusing on trying to figure out why things are going haywire.
>
>> I have not yet compiled and tested the lastest patches from Jarod.
>
> Really shouldn't be any difference from what you've got, they're just rebased to the latest v4l-dvb tree.

Hey Jarod,

I haven't seen your exact GPIO config but I noticed something
recently:  the em28xx driver runs the dvb_gpio sequence whenever
starting streaming, not just whenever opening the DVB frontend.  This
means that if your dvb_gpio definition strobes the tda18271 reset (as
opposed to just taking it out of reset), then the chip will get reset
whenever the streaming is started (a real problem if multiple tuning
attempts are performed without closing the frontend first).

Mauro seems to think this is intended behavior, although I cannot see
how this could possibly be correct, especially since the .init()
callback is not called in that case.  I setup a tree to remove the
call, but never got far enough into the testing to confirm whether it
broke any improperly configured boards depending on the incorrect
behavior.

As a test, you might want to check your dvb_gpio config and see if you
are pulling anything low and then high, and just remove the line that
sets the pin low and see if the recalibration still occurs.

Devin

-- 
Devin J. Heitmueller - Kernel Labs
http://www.kernellabs.com

Re: KWorld UB435-Q Support

[Robert Cicconetti]

On Tue, Nov 17, 2009 at 9:15 AM, Jarod Wilson <jarod@wilsonet.com> wrote:
>> It happened at every tuning operation, and made mythfrontend unhappy
>> (unable to tune after the first channel). I disabled the check for
>> RF_CAL_OK which triggered the recalibration, and mythfrontend worked.
>
> Yeah, tuning is much quicker here if I skip that check as well, but its definitely not the proper fix.
>
>> The stick has been plugged in for a few months, so presumably would've
>> caught on fire by now if it was going to. It would be nice if the
>> tuning delay went away, though.. it still takes ~6 seconds to switch
>> frequencies.
>
> Wait, it still takes that long with the check gone? I didn't poke for very long with the check disabled, mostly focusing on trying to figure out why things are going haywire.

Okay.. couple of unscientific tests later show I was wrong above:
First tuning, ~5-6 seconds to lock.
Later tunings, ~3 seconds to lock.

This is with my hack to remove the recalibrations.

-Bob

Re: KWorld UB435-Q Support

[Jarod Wilson]

On Nov 17, 2009, at 9:37 AM, Devin Heitmueller wrote:

> On Tue, Nov 17, 2009 at 9:15 AM, Jarod Wilson <jarod@wilsonet.com> wrote:
>> On Nov 17, 2009, at 1:03 AM, Robert Cicconetti wrote:
>> 
>>> On Tue, Nov 17, 2009 at 12:55 AM, Michael Krufky <mkrufky@linuxtv.org> wrote:
>>>>>>>>> [ 812.465930] tda18271: performing RF tracking filter calibration
>>>>>>>>> [ 818.572446] tda18271: RF tracking filter calibration complete
>>>>>>>>> [ 818.953946] tda18271: performing RF tracking filter calibration
>>>>>>>>> [ 825.093211] tda18271: RF tracking filter calibration complete
>>>> 
>>>> 
>>>> If you see this happen more than once consecutively, and there is only 1
>>>> silicon tuner present, then it means something very bad is happening, and
>>>> there is a chance of burning out a part.  I still wouldnt not recommend any
>>>> mainline merge until you can prevent this behavior -- I suspect that a GPIO
>>>> reset is being toggled where it shouldnt be, which should be harmless ...
>>>> but until we fix it, we cant be sure what damage might get done...
>>>> 
>>>> The RF tracking filter calibration is a procedure that should only happen
>>>> once while the tuner is powered on -- it should *only* be repeated if the
>>>> tuner indicated that calibration is necessary, and that would only happen
>>>> after a hardware reset.
>>>> 
>>>> This still looks fishy to me...
>> 
>> Agreed. I did manage to dig into this some more last night, something is definitely still awry. 
...
> Hey Jarod,
> 
> I haven't seen your exact GPIO config but I noticed something
> recently:  the em28xx driver runs the dvb_gpio sequence whenever
> starting streaming, not just whenever opening the DVB frontend.  This
> means that if your dvb_gpio definition strobes the tda18271 reset (as
> opposed to just taking it out of reset), then the chip will get reset
> whenever the streaming is started (a real problem if multiple tuning
> attempts are performed without closing the frontend first).
> 
> Mauro seems to think this is intended behavior, although I cannot see
> how this could possibly be correct, especially since the .init()
> callback is not called in that case.  I setup a tree to remove the
> call, but never got far enough into the testing to confirm whether it
> broke any improperly configured boards depending on the incorrect
> behavior.

This tree, I presume.

http://kernellabs.com/hg/~dheitmueller/em28xx-modeswitch/

I just tacked on the very last patch there onto my local tree to test with one of these sticks. Behavior is the same though, and the tda18271 reg dumps look equally bad -- they're all reported 0x00.

> As a test, you might want to check your dvb_gpio config and see if you
> are pulling anything low and then high, and just remove the line that
> sets the pin low and see if the recalibration still occurs.

I'm pretty sure you explained how to do this to me once before on irc, but its been a while, and that knowledge has since leaked out of my brain... Currently, I have:

/*
 * KWorld PlusTV 340U and UB435-Q (ATSC) GPIOs map:
 * EM_GPIO_0 - currently unknown
 * EM_GPIO_1 - LED disable/enable (1 = off, 0 = on)
 * EM_GPIO_2 - currently unknown
 * EM_GPIO_3 - currently unknown
 * EM_GPIO_4 - TDA18271HD tuner (1 = active, 0 = in reset)
 * EM_GPIO_5 - LGDT3304 ATSC/QAM demod (1 = active, 0 = in reset)
 * EM_GPIO_6 - currently unknown
 * EM_GPIO_7 - currently unknown
 */
static struct em28xx_reg_seq kworld_a340_digital[] = {
        /* only diff from default gpio is to keep 1 clear to turn on LED */
        {EM28XX_R08_GPIO,       0x6d,   ~EM_GPIO_4,     10},
        { -1,                   -1,             -1,     -1},
};

I've tried various combinations in here today, all without any significant change in behavior. But I suspect I'm missing something I should be trying. Ah well. Bed time. More poking tomorrow...


-- 
Jarod Wilson
jarod@wilsonet.com

Re: KWorld UB435-Q Support

[Jarod Wilson]

On Tue, Nov 17, 2009 at 11:59 PM, Jarod Wilson <jarod@wilsonet.com> wrote:
> On Nov 17, 2009, at 9:37 AM, Devin Heitmueller wrote:
>>>>> If you see this happen more than once consecutively, and there is only 1
>>>>> silicon tuner present, then it means something very bad is happening, and
>>>>> there is a chance of burning out a part.  I still wouldnt not recommend any
>>>>> mainline merge until you can prevent this behavior -- I suspect that a GPIO
>>>>> reset is being toggled where it shouldnt be, which should be harmless ...
>>>>> but until we fix it, we cant be sure what damage might get done...
>>>>>
>>>>> The RF tracking filter calibration is a procedure that should only happen
>>>>> once while the tuner is powered on -- it should *only* be repeated if the
>>>>> tuner indicated that calibration is necessary, and that would only happen
>>>>> after a hardware reset.
>>>>>
>>>>> This still looks fishy to me...
>>>
>>> Agreed. I did manage to dig into this some more last night, something is definitely still awry.
> ...
>> Hey Jarod,
>>
>> I haven't seen your exact GPIO config but I noticed something
>> recently:  the em28xx driver runs the dvb_gpio sequence whenever
>> starting streaming, not just whenever opening the DVB frontend.  This
>> means that if your dvb_gpio definition strobes the tda18271 reset (as
>> opposed to just taking it out of reset), then the chip will get reset
>> whenever the streaming is started (a real problem if multiple tuning
>> attempts are performed without closing the frontend first).
>>
>> Mauro seems to think this is intended behavior, although I cannot see
>> how this could possibly be correct, especially since the .init()
>> callback is not called in that case.  I setup a tree to remove the
>> call, but never got far enough into the testing to confirm whether it
>> broke any improperly configured boards depending on the incorrect
>> behavior.
>
> This tree, I presume.
>
> http://kernellabs.com/hg/~dheitmueller/em28xx-modeswitch/
>
> I just tacked on the very last patch there onto my local tree to test with one of these sticks. Behavior is the same though, and the tda18271 reg dumps look equally bad -- they're all reported 0x00.
>
>> As a test, you might want to check your dvb_gpio config and see if you
>> are pulling anything low and then high, and just remove the line that
>> sets the pin low and see if the recalibration still occurs.
>
> I'm pretty sure you explained how to do this to me once before on irc, but its been a while, and that knowledge has since leaked out of my brain... Currently, I have:
>
> /*
>  * KWorld PlusTV 340U and UB435-Q (ATSC) GPIOs map:
>  * EM_GPIO_0 - currently unknown
>  * EM_GPIO_1 - LED disable/enable (1 = off, 0 = on)
>  * EM_GPIO_2 - currently unknown
>  * EM_GPIO_3 - currently unknown
>  * EM_GPIO_4 - TDA18271HD tuner (1 = active, 0 = in reset)
>  * EM_GPIO_5 - LGDT3304 ATSC/QAM demod (1 = active, 0 = in reset)
>  * EM_GPIO_6 - currently unknown
>  * EM_GPIO_7 - currently unknown
>  */
> static struct em28xx_reg_seq kworld_a340_digital[] = {
>        /* only diff from default gpio is to keep 1 clear to turn on LED */
>        {EM28XX_R08_GPIO,       0x6d,   ~EM_GPIO_4,     10},
>        { -1,                   -1,             -1,     -1},
> };
>
> I've tried various combinations in here today, all without any significant change in behavior. But I suspect I'm missing something I should be trying. Ah well. Bed time. More poking tomorrow...

Or a few months later. About two weeks ago, I finally poked at these
sticks some more, after getting a bit of info from another user, and
we've finally got an actual fix for this problem -- .deny_i2c_rptr = 1
just needed to be set in the lgdt3305_config struct, as the device's
tuner isn't actually behind an i2c gate. With that change, the stick
behaves quite well w/o any alterations to the tda18271 code. Patches
are here:

http://wilsonet.com/jarod/junk/kworld-a340-20100218/

They're in Mike's hands now, since they rely so heavily on the lgdt3305 driver.

-- 
Jarod Wilson
jarod@wilsonet.com

Re: KWorld UB435-Q Support

[Amy Overmyer]

On Tue, Nov 17, 2009 at 11:59 PM, Jarod Wilson <jarod@wilsonet.com> wrote:

snip

>>Or a few months later. About two weeks ago, I finally poked at these
>>sticks some more, after getting a bit of info from another user, and
>>we've finally got an actual fix for this problem -- .deny_i2c_rptr = 1
>>just needed to be set in the lgdt3305_config struct, as the device's
>>tuner isn't actually behind an i2c gate. With that change, the stick
>>behaves quite well w/o any alterations to the tda18271 code. Patches
>>are here:
>>
>>http://wilsonet.com/jarod/junk/kworld-a340-20100218/
>>
>>They're in Mike's hands now, since they rely so heavily on the lgdt3305 driver.

I compiled up this change on my setup and the Kworld is working wonderfully for me now. I was able to do a full ATSC us center freq scan pretty quickly and then added it to mythtv and its working well there, too. It's doing OTA capture.

Thanks, Jarod.