Re: [PATCH 0/1] xc3028: force reload of DTV7 firmware in VHF band with Zarlink demodulator

[Mauro Carvalho Chehab <mchehab@redhat.com>]

On 07-12-2011 15:25, Gianluca Gennari wrote:
> Il 07/12/2011 17:21, Mauro Carvalho Chehab ha scritto:
>> On 07-12-2011 13:51, Gianluca Gennari wrote:
>>> Il 07/12/2011 16:45, Gianluca Gennari ha scritto:
>>>> Il 07/12/2011 16:05, Mauro Carvalho Chehab ha scritto:
>>>>> On 07-12-2011 12:53, Gianluca Gennari wrote:
>>>>>> Il 07/12/2011 15:20, Mauro Carvalho Chehab ha scritto:
>>>>>>> On 07-12-2011 11:47, Gianluca Gennari wrote:
>>>>>>>> Il 07/12/2011 14:12, Mauro Carvalho Chehab ha scritto:
>>>>>>>>> On 06-12-2011 12:33, Gianluca Gennari wrote:
>>>>>>>>>> Hi All,
>>>>>>>>>>
>>>>>>>>>> I have a Terratec Cinergy Hybrid T USB XS stick (USB 0ccd:0042).
>>>>>>>>>> This device is made of the following components:
>>>>>>>>>> - Empiatech em2880 USB bridge;
>>>>>>>>>> - Zarlink zl10353 demodulator;
>>>>>>>>>> - Xceive XC3028 tuner;
>>>>>>>>>>
>>>>>>>>>> For this device, the ZARLINK456 define is set to true so it is
>>>>>>>>>> using
>>>>>>>>>> the
>>>>>>>>>> firmwares with type D2633 for the XC3028 tuner.
>>>>>>>>>>
>>>>>>>>>> I found out that:
>>>>>>>>>> 1) the DTV7 firmware works fine in VHF band (bw=7MHz);
>>>>>>>>>> 2) the DTV8 firmware works fine in UHF band (bw=8MHz);
>>>>>>>>>> 3) the DTV78 firmware works fine in UHF band (bw=8MHz) but it
>>>>>>>>>> doesn not
>>>>>>>>>> work at all in VHF band (bw=7MHz);
>>>>>>>>>>
>>>>>>>>>> In fact, when the DTV78 firmware is loaded and I try to tune a VHF
>>>>>>>>>> channel, the frequency lock is ciclically acquired for a second
>>>>>>>>>> and
>>>>>>>>>> immediately lost.
>>>>>>>>>> So the proposed patch forces a reload of the DTV7 firmware every
>>>>>>>>>> time a
>>>>>>>>>> 7MHz channel is requested.
>>>>>>>>>> The only drawback is that channel change from VHF to UHF or
>>>>>>>>>> viceversa is
>>>>>>>>>> slightly slower.
>>>>>>>>>> Devices using the D2620 firmwares are unaffected.
>>>>>>>>>
>>>>>>>>> Hi Gianluca,
>>>>>>>>>
>>>>>>>>> The issues with firmware DTV78 x DTV7/DTV8 are old. No matter
>>>>>>>>> what we
>>>>>>>>> do,
>>>>>>>>> we end by having troubles, as the issue is Country-dependent. For
>>>>>>>>> example,
>>>>>>>>> Australia requires a different firmware than Germany, due to the
>>>>>>>>> differences
>>>>>>>>> on the VHF/UHF bands.
>>>>>>>>>
>>>>>>>>> I prefer if you could work into a patch that would add some
>>>>>>>>> modprobe
>>>>>>>>> parameter
>>>>>>>>> to disable the current "autodetection" way, allowing to override
>>>>>>>>> the
>>>>>>>>> firmware
>>>>>>>>> used for VHF and UHF.
>>>>>>>>>
>>>>>>>>> Thanks,
>>>>>>>>> Mauro
>>>>>>>>>
>>>>>>>>
>>>>>>>> Hi Mauro,
>>>>>>>> thanks for the feedback. Unfortunately I do not have any info on
>>>>>>>> which
>>>>>>>> kind of firmware is needed on other parts of the world. All I
>>>>>>>> know is
>>>>>>>> what is happening here in Italy, and what I can understand
>>>>>>>> reading the
>>>>>>>> code. I suppose my findings can be extended to the rest of
>>>>>>>> Europe, and
>>>>>>>> maybe Africa and Middle-East.
>>>>>>>
>>>>>>> Even in Europe, there are some differences.
>>>>>>>
>>>>>>
>>>>>> OK, so the validity of my findings are restricted to Italy.
>>>>>>
>>>>>>>> Can you provide a reference about problems in other continents like
>>>>>>>> Australia?
>>>>>>>
>>>>>>> All I know is from the constant reports at the ML from users. We
>>>>>>> used to
>>>>>>> have a developer in Australia, but he moved away, and it seems that
>>>>>>> he lost
>>>>>>> interest on DVB development, as we were unable to contact him ever
>>>>>>> since.
>>>>>>>>
>>>>>>>> Do you think a simple module parameters that allows to
>>>>>>>> enable/disable
>>>>>>>> the usage of the DTV78 firmware would do the trick?
>>>>>>>
>>>>>>> Perhaps one or two module parameters to allow forcing a certain
>>>>>>> firmware
>>>>>>> for
>>>>>>> VHF and UHF.
>>>>>>
>>>>>> Seems reasonable.
>>>>>>
>>>>>>>> Eventually, do you agree that the default solution should be to
>>>>>>>> DISABLE
>>>>>>>> DTV78 firmware, since this seems to be the more robust solution, and
>>>>>>>> let
>>>>>>>> the user enable it through the kernel parameter if it is working
>>>>>>>> in his
>>>>>>>> country? Or do you prefer the other way around, so by default  DTV78
>>>>>>>> firmware is enabled, and users with problems can disable it
>>>>>>>> through the
>>>>>>>> kernel module parameter?
>>>>>>>
>>>>>>> AFAIK, DTV78 should be used in Spain and in Germany. Changing the
>>>>>>> current
>>>>>>> default doesn't look a good idea, as it will cause regressions, if
>>>>>>> the new
>>>>>>> way is not backward-compatible.
>>>>>>
>>>>>> With the proposed patch DTV78 will be used in UHF band, while DTV7 in
>>>>>> VHF band. Will this make any difference in Spain or Germany?
>>>>>
>>>>> Not sure. I don't live there ;)
>>>>>
>>>>>> What about a kernel parameter to specify the country?
>>>>>> Something like:
>>>>>>
>>>>>> country={0-4}
>>>>>>
>>>>>> DEFAULT=0,ITALY=1,GERMANY=2,SPAIN=3,AUSTRALIA=4
>>>>>>
>>>>>> Then we could specify a well-defined behavior for each country, hiding
>>>>>> the firmware-related problems to the user (which will have problems
>>>>>> understanding parameters like force-DTV7-firmware-in-VHF-band).
>>>>>>
>>>>>> All I need to know is what is the best behavior for each country.
>>>>>
>>>>> That's the hardest part ;) We would need someone on each possible
>>>>> Country,
>>>>> in order to test. Also, a per-country setup like that sucks.
>>>>> Ideally, the
>>>>> driver should use the bandwidh and the other information at the
>>>>> standard
>>>>> DVB
>>>>> parameters, in order to select the right firmware. This works with all
>>>>> other
>>>>> frontends. Not sure what's broken on xc3028 design that it requires a
>>>>> per-country
>>>>> hack. I suspect that it is not a pure per-country hack, but it is
>>>>> also per
>>>>> demod.
>>>>>
>>>>> As we don't have much complains about it nowadays, I assume that the
>>>>> current
>>>>> behavior is ok for most users. So, a parameter would be used only
>>>>> for those
>>>>> where the default behavior doesn't work.
>>>>>
>>>>> Btw, we already have a similar parameter to force the audio
>>>>> demodulation
>>>>> standard,
>>>>> due to the same reasons.
>>>>>
>>>>> Regards,
>>>>> Mauro
>>>>>
>>>>
>>>> Probably there are no complains about the firmware because in most
>>>> countries VHF is not used at all, or is only used for marginal TV
>>>> stations.
>>
>> Makes sense. Well, it is not hard to detect VHF. If the DTV bandwidth is
>> properly filled, detecting between 8/7/6 MHz is also easy.
>>
>>>> In Italy instead the main DTT mux (RAI mux 1) is broadcasted
>>>> in VHF band for historical reasons, so many Italian users are concerned
>>>> about it. I've already faced problems related to DTT VHF reception, that
>>>> were never noticed before by users from other countries.
>>>>
>>>> Moreover, another problem is that there are 2 firmwares types: D2633 and
>>>> D2620;
>>
>> AFAIKT, this is related to the demux sensibility: D2633 is normal level,
>> D2620 is lower output level.
>
> In fact I tried to force the load of the D2620 firmwares on my stick but
> there was no signal.
>
>>>> also there are 2 completely different files: xc3028-v27.fw and
>>>> xc3028L-v36.fw.
>>
>> The v3.6 is for xc3028L (low power). It was reported that using v2.7 on a
>> xc3028L will cause it to overheat.
>
> Thanks for the info.
>
>>>> In my case, my stick uses firmwares of type D2633 from file
>>>> xc3028-v27.fw; I cannot test anything else. It is also possible that the
>>>> VHF part of the DTV78 firmware of type D2633 in file xc3028-v27.fw is
>>>> simply broken.
>>
>> Maybe. From my remarks, Australia would use only 7MHz firmware on both VHF
>> and UHF bandwidths. B/G countries would use 7/8 MHz. There were several
>> discussions at the ML by the time we've added support for this tuner, and
>> later on. The logs for it can provide some useful info about the changes.
>>
>>>> As far as I know, this firmware is older than than the
>>>> first DTT mux in VHF band here in Italy, so maybe it was optimized for
>>>> analog reception and has never been tested with digital modulations.
>>>>
>>>> What about a module parameter "alternative-VHF-firmware" which can be
>>>> either on/off (default=off)?
>>>>
>>>> By default, we keep current behavior; when enabled, DTV7 firmware is
>>>> always loaded for VHF channels, for all firmware types and files.
>>
>> Maybe vhf_firmware, being 0 (default behavior), 6, 7, 8 or 78?
>>
>
> In a simpler way:
> vhf_firmware = 0 (default, allows 78 under certain conditions), 1
> (select 6, 7 or 8 depending on bandwidth);
> or with a shorter description:
> vhf_firmware = 0 (universal), 1 (bandwidth specific);
>
> In the end the problem is not the VHF itself, but the firmware loaded
> for channels with 7MHz bandwidth. In Europe 7MHz=VHF, but this is may
> not be the case in the rest of the world.
>
> By the way, in June 2009 Italy adopted the European VHF canalization (GE
> '06) abandoning the old national canalization (channels
> D-E-F-G-H-H1-H2). I think this is the new standard in Europe. So I don't
> see how things can be significantly different in Germany or Spain.
>
> As you pointed to me, in Australia all channels (UHF and VHF) use 7MHz
> bandwidth, so this explains way nobody is complaining: in this situation
> the DTV7 is the only firmware ever loaded. And the VHF canalization in
> Australia seems identical to the European one.
> So the patch will never affect Australian users, no matter how they set
> the module parameter.
>
> It would be great if someone from Germany or Spain could test a device
> with the xc3028 tuner and give some feedback about VHF reception.
>
> In the meantime I will post a new patch.

(C/C Christoph, as he is maintaining the DVB-T tables at both dvb-utils and Kaffeine).

Hmm... looking at w_scan, we see:

	switch(txt_to_country(country)) {
                 //**********DVB freq lists*******************************************//
                 case    AT:     //	AUSTRIA
                 case    BE:     //	BELGIUM
                 case    CH:     //	SWITZERLAND
                 case    DE:     //	GERMANY
                	case    DK:     //	DENMARK
                 case    ES:     //      SPAIN
                	case    GR:     //	GREECE
                 case    HR:     //      CROATIA
                	case    HK:     //	HONG KONG
                 case    IS:     //      ICELAND
                 case    IT:     //      ITALY
                 case    LU:     //      LUXEMBOURG
                 case    LV:     //      LATVIA
                	case    NL:     //	NETHERLANDS
                 case    NO:     //      NORWAY
                 case    NZ:     //      NEW ZEALAND
                 case    PL:     //      POLAND
                 case    SE:     //      SWEDEN
                	case    SK:     //	SLOVAKIA
                         switch(*dvb) {
                                 case FE_QAM:
                                         *channellist = DVBC_QAM;
                                         info("DVB-C\n");
                                        	break;
                                 default:
                                         *channellist = DVBT_DE;
                                         info("DVB-T Europe\n");
                                         break;
                                 }
                         break;

...



         case DVBT_AU:  //AUSTRALIA, 7MHz step list
                 switch (channel) {
                         case  5 ... 12: return  142500000;
                         case 21 ... 69: return  333500000;
                         default:        return  SKIP_CHANNEL;
                         }
         case DVBT_DE:  //GERMANY
         case DVBT_FR:  //FRANCE, +/- offset
         case DVBT_GB:  //UNITED KINGDOM, +/- offset
                 switch (channel) {
                        	case  5 ... 12: return  142500000;
                         case 21 ... 69: return  306000000;
                         default:  	return  SKIP_CHANNEL;
                        	}

...

int freq_step(int channel, int channellist) {
switch (channellist) {
         case DVBT_AU:  return  7000000; // dvb-t australia, 7MHz step
         case DVBT_DE:
         case DVBT_FR:
         case DVBT_GB:  switch (channel) { // dvb-t europe, 7MHz VHF ch5..12, all other 8MHz
                               case  5 ... 12:    return 7000000;
                               case 21 ... 69:    return 8000000;
                               default:           return 8000000; // should be never reached.
                               }
}


 From the above, all w_scan-known European Countries are using 7 MHz/8 MHz,
including Italy and Germany.

However, looking at the DVB-T tables at dvb-apps:

  $ grep "T [12]" dvb-t/*|grep 8MHz
dvb-t/auto-Italy:T 177500000 8MHz AUTO NONE AUTO AUTO AUTO NONE
dvb-t/auto-Italy:T 186000000 8MHz AUTO NONE AUTO AUTO AUTO NONE
dvb-t/auto-Italy:T 194500000 8MHz AUTO NONE AUTO AUTO AUTO NONE
dvb-t/auto-Italy:T 203500000 8MHz AUTO NONE AUTO AUTO AUTO NONE
dvb-t/auto-Italy:T 212500000 8MHz AUTO NONE AUTO AUTO AUTO NONE
dvb-t/auto-Italy:T 219500000 8MHz AUTO NONE AUTO AUTO AUTO NONE
dvb-t/auto-Italy:T 226500000 8MHz AUTO NONE AUTO AUTO AUTO NONE
dvb-t/it-Aosta:T 226500000 8MHz 2/3 NONE QAM64 8k 1/32 NONE
dvb-t/it-Catania:T 226500000      8MHz 2/3 NONE QAM64 8k 1/32 NONE
dvb-t/it-Firenze:T 219500000 8MHz 2/3 NONE QAM64 8k 1/32 NONE
dvb-t/it-Genova:T 219500000 8MHz 2/3 NONE QAM64 8k 1/32 NONE
dvb-t/it-Milano:T 184500000 8MHz 2/3 NONE QAM64 8k 1/32 NONE
dvb-t/it-Milano:T 191500000 8MHz 2/3 NONE QAM64 8k 1/32 NONE
dvb-t/it-Milano:T 219500000 8MHz 2/3 NONE QAM64 8k 1/32 NONE
dvb-t/it-Pagnacco:T 226500000 8MHz 3/4 NONE QAM64 8k 1/32 NONE
dvb-t/it-Sassari:T 177500000 8MHz 2/3 NONE QAM64 8k 1/32 NONE
dvb-t/it-Sassari:# T 177500000 8MHz 2/3 NONE QAM64 8k 1/32 NONE

Several channels in Italy are marked as if they are using 8MHz for VHF (the auto-Italy is
the most weird one, as it defines all VHF frequencies with both 7MHz and 8MHz).

This may explain why, from time to time, we have complaints about tuner-xc2028.

The above bad tables may explain the issue. Are you sure you're using 7MHz for your
7MHz-bandwidth channels?

There is a hack inside tuner-xc2028, in order to allow it to work in Australia:

	switch (bw) {
	case BANDWIDTH_8_MHZ:
		if (p->frequency < 470000000)
			priv->ctrl.vhfbw7 = 0;
		else
			priv->ctrl.uhfbw8 = 1;
		type |= (priv->ctrl.vhfbw7 && priv->ctrl.uhfbw8) ? DTV78 : DTV8;
		type |= F8MHZ;
		break;
	case BANDWIDTH_7_MHZ:
		if (p->frequency < 470000000)
			priv->ctrl.vhfbw7 = 1;
		else
			priv->ctrl.uhfbw8 = 0;
		type |= (priv->ctrl.vhfbw7 && priv->ctrl.uhfbw8) ? DTV78 : DTV7;
		type |= F8MHZ;
		break;

The idea on the hack is to avoid firmware reload, when the standard changes, e. g. on
Countries like Australia where only 7 MHz is used, it will fix the frequency to 7MHz.
The same applies to Countries with just 8MHz. Countries that have both 7MHz and 8MHz
will load DTV78 firmware after the first switch from 7 to 8 (or from 8 to 7).

I think that the modprobe parameter could be to enable the above hack or to disable it.
something like "use_dtv78". If not, then it will just use DTV7 or DTV8, depending if
the frequency is 7 or 8 MHz. This will likely slow down a scan with auto-italy, as the
driver will need to switch the firmware each time a 8MHz is used for VHF.

Anyway, I think we should fix the Italy tables to remove the bad 8MHz VHF frequencies, if
you're sure that there's no city using it anymore.

Regards,
Mauro