Re: [PATCH 3/4] v4l-utils/libdvbv5: add support for ISDB-S scanning

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



Em Wed, 08 Oct 2014 21:09:40 +0900
tskd08@xxxxxxxxx escreveu:

> From: Akihiro Tsukada <tskd08@xxxxxxxxx>
> 
> added NIT scan for ISDB-S,
> fixed wrong/too-close frequency of the scanned entries,
> when freq_offset/bandwith was not set properly.
> 
> ISDB-S/T specific charset conversion is separeted off
> as an iconv module, and not implemented in this lib.
> 
> Signed-off-by: Akihiro Tsukada <tskd08@xxxxxxxxx>
> ---
>  lib/include/libdvbv5/dvb-scan.h |   2 +
>  lib/libdvbv5/dvb-fe.c           |   6 +-
>  lib/libdvbv5/dvb-file.c         |  21 +++++--
>  lib/libdvbv5/dvb-scan.c         | 125 +++++++++++++++++++++++++++++++++++++++-
>  lib/libdvbv5/parse_string.c     |  23 ++++++++
>  5 files changed, 170 insertions(+), 7 deletions(-)
> 
> diff --git a/lib/include/libdvbv5/dvb-scan.h b/lib/include/libdvbv5/dvb-scan.h
> index fe50687..e3a0d24 100644
> --- a/lib/include/libdvbv5/dvb-scan.h
> +++ b/lib/include/libdvbv5/dvb-scan.h
> @@ -387,6 +387,8 @@ int dvb_estimate_freq_shift(struct dvb_v5_fe_parms *parms);
>  
>  int dvb_new_freq_is_needed(struct dvb_entry *entry, struct dvb_entry *last_entry,
>  			   uint32_t freq, enum dvb_sat_polarization pol, int shift);
> +int dvb_new_ts_is_needed(struct dvb_entry *entry, struct dvb_entry *last_entry,
> +			   uint32_t freq, int shift, uint32_t ts_id);
>  
>  struct dvb_entry *dvb_scan_add_entry(struct dvb_v5_fe_parms *parms,
>  				     struct dvb_entry *first_entry,
> diff --git a/lib/libdvbv5/dvb-fe.c b/lib/libdvbv5/dvb-fe.c
> index f535311..93c0b9b 100644
> --- a/lib/libdvbv5/dvb-fe.c
> +++ b/lib/libdvbv5/dvb-fe.c
> @@ -372,6 +372,8 @@ int dvb_set_sys(struct dvb_v5_fe_parms *p, fe_delivery_system_t sys)
>  	parms->p.current_sys = sys;
>  	parms->n_props = rc;
>  
> +	if (sys == SYS_ISDBS /* || sys == SYS_ISDBT */)
> +		parms->p.default_charset = "arib-std-b24";

Actually, the best would be to add a Country property at the library
for this kind of adjustment. If you take a look at page 12 of:
	http://www.dibeg.org/techp/aribstd/harmonization/090918_Harmonization_volume8_Receivers.pdf

You'll see that the charset definition is different between JP and BR.
Also, there are several other things that are different among the
descriptors used in Japan (ARIB specs) and South America (NBR specs).
I don't doubt that some other Countries will end by deciding to apply
some different definitions.

So, the best is to add a Country field and use it to handle those
differences. We'll need a new parameter at dvbv5 tools to handle the
Country code. Btw, wscan has an interesting logic of guessing it
from the locale vars. We could use something like that at the userspace
tools as a hint (that could, of course, be overriden).

>  	return 0;
>  }
>  
> @@ -683,8 +685,10 @@ int dvb_fe_set_parms(struct dvb_v5_fe_parms *p)
>  			dvb_logdbg("LNA is %s", parms->p.lna ? "ON" : "OFF");
>  	}
>  
> -	if (dvb_fe_is_satellite(tmp_parms.p.current_sys))
> +	if (dvb_fe_is_satellite(tmp_parms.p.current_sys)) {
>  		dvb_sat_set_parms(&tmp_parms.p);
> +		parms->freq_offset = tmp_parms.freq_offset;
> +	}
>  
>  	/* Filter out any user DTV_foo property such as DTV_POLARIZATION */
>  	tmp_parms.n_props = dvb_copy_fe_props(tmp_parms.dvb_prop,
> diff --git a/lib/libdvbv5/dvb-file.c b/lib/libdvbv5/dvb-file.c
> index bcb1762..0a0e41a 100644
> --- a/lib/libdvbv5/dvb-file.c
> +++ b/lib/libdvbv5/dvb-file.c
> @@ -1125,12 +1125,25 @@ static int get_program_and_store(struct dvb_v5_fe_parms_priv *parms,
>  		entry->props[j].cmd = parms->dvb_prop[j].cmd;
>  		entry->props[j].u.data = parms->dvb_prop[j].u.data;
>  
> -		if ((!channel || !*channel) &&
> -		    entry->props[j].cmd == DTV_FREQUENCY)
> -			freq = parms->dvb_prop[j].u.data;
> +		if (entry->props[j].cmd == DTV_STREAM_ID &&
> +		    entry->props[j].u.data == 0 &&
> +		    parms->p.current_sys == SYS_ISDBS)
> +			entry->props[j].u.data = dvb_scan_handler->pat->header.id;

Hmm... I don't like very much the idea of having a code here accessing the
pat->header.id, just for ISDB-S.

All the standard-specific issues should have already being solved here.

It probably makes more sense to fill it at dvb-scan.c, while handling the
standard-specifi stuff, for example, at the function
__dvb_add_update_transponders(). It will likely need a parser for
the ISDB-S descriptor too, at least when we start to have some code that
would allow to force the demod to not use the TMCC data.

> +
> +		if (entry->props[j].cmd != DTV_FREQUENCY)
> +			continue;
> +
> +		if (dvb_fe_is_satellite(parms->p.current_sys) &&
> +		    entry->props[j].u.data < parms->freq_offset)
> +			entry->props[j].u.data += parms->freq_offset;
> +
> +		if (!channel || !*channel)
> +			freq = entry->props[j].u.data;
>  	}

>  	if (!channel || !*channel) {
> -		r = asprintf(&channel, "%.2fMHz#%d", freq/1000000., service_id);
> +		r = asprintf(&channel, "%.2f%cHz#%d", freq / 1000000.,
> +			dvb_fe_is_satellite(parms->p.current_sys) ? 'G' : 'M',
> +			service_id);

Good catch! better to put it into a separate patch, as it is actually
unrelated and also affects DVB-S.

>  		if (r < 0)
>  			dvb_perror("asprintf");
>  		if (parms->p.verbose)
> diff --git a/lib/libdvbv5/dvb-scan.c b/lib/libdvbv5/dvb-scan.c
> index 3b70f5a..470ef61 100644
> --- a/lib/libdvbv5/dvb-scan.c
> +++ b/lib/libdvbv5/dvb-scan.c
> @@ -635,7 +635,7 @@ int dvb_estimate_freq_shift(struct dvb_v5_fe_parms *__p)
>  		rolloff = 115;
>  		break;
>  	case SYS_DVBS:
> -	case SYS_ISDBS:	/* FIXME: not sure if this rollof is right for ISDB-S */
> +	case SYS_ISDBS:
>  		divisor = 100000;
>  		rolloff = 135;
>  		break;
> @@ -672,11 +672,14 @@ int dvb_estimate_freq_shift(struct dvb_v5_fe_parms *__p)
>  		 * purposes of estimating a max frequency shift here.
>  		 */
>  		dvb_fe_retrieve_parm(&parms->p, DTV_SYMBOL_RATE, &symbol_rate);
> +		if (parms->p.current_sys == SYS_ISDBS)
> +			symbol_rate = 28800;

Better to add it at the switch, splitting DVBS and ISDBS, like:

	case SYS_ISDBS:
		symbol_rate = 28800;
		/* fall though */
	case SYS_ISDBS:
 		divisor = 100000;
 		rolloff = 135;
 		break;

Btw, we should likely be adding a similar code at the Kernel, as it
provides already a symbol rate estimation for other Satellite and Cable
standards.

>  		bw = (symbol_rate * rolloff) / divisor;
>  	}
>  	if (!bw)
>  		dvb_fe_retrieve_parm(&parms->p, DTV_BANDWIDTH_HZ, &bw);
> -
> +	if (!bw && parms->p.current_sys == SYS_ISDBT)
> +		bw = 6000000;

No, this is wrong, as the specs allow other bandwidths.

Btw, according with: http://www.soumu.go.jp/main_sosiki/joho_tsusin/eng/Releases/Telecommunications/111019_c.html
(link obtained from http://en.wikipedia.org/wiki/ISDB-T_International),
Maldives is using ISDB-T with 8MHz.

>  	/*
>  	 * If the max frequency shift between two frequencies is below
>  	 * than the used bandwidth / 8, it should be the same channel.
> @@ -758,6 +761,87 @@ struct dvb_entry *dvb_scan_add_entry(struct dvb_v5_fe_parms *__p,
>  	return NULL;
>  }
>  
> +int dvb_new_ts_is_needed(struct dvb_entry *entry, struct dvb_entry *last_entry,
> +			 uint32_t freq, int shift, uint32_t ts_id)
> +{
> +	int i;
> +	uint32_t data;
> +
> +	for (; entry != last_entry; entry = entry->next) {
> +		for (i = 0; i < entry->n_props; i++) {
> +			data = entry->props[i].u.data;
> +			if (entry->props[i].cmd == DTV_STREAM_ID) {
> +				if (data != ts_id)
> +					break;
> +			}
> +			if (entry->props[i].cmd == DTV_FREQUENCY) {
> +				if (freq < data - shift || freq > data + shift)
> +					break;
> +			}
> +		}
> +		if (i == entry->n_props && entry->n_props > 0)
> +			return 0;
> +	}
> +
> +	return 1;
> +}
> +
> +static struct dvb_entry *
> +dvb_scan_add_entry_isdbs(struct dvb_v5_fe_parms *__p,
> +			 struct dvb_entry *first_entry, struct dvb_entry *entry,
> +			 uint32_t freq, uint32_t shift, uint32_t ts_id)
> +{
> +	struct dvb_v5_fe_parms_priv *parms = (void *)__p;
> +	struct dvb_entry *new_entry;
> +	int i, n = 2;
> +
> +	if (!dvb_new_ts_is_needed(first_entry, NULL, freq, shift, ts_id))
> +		return NULL;
> +
> +	/* Clone the current entry into a new entry */
> +	new_entry = calloc(sizeof(*new_entry), 1);
> +	if (!new_entry) {
> +		dvb_perror("not enough memory for a new scanning frequency/TS");
> +		return NULL;
> +	}
> +
> +	memcpy(new_entry, entry, sizeof(*entry));
> +	if (entry->channel)
> +		new_entry->channel = strdup(entry->channel);
> +	if (entry->vchannel)
> +		new_entry->vchannel = strdup(entry->vchannel);
> +	if (entry->location)
> +		new_entry->location = strdup(entry->location);
> +	if (entry->lnb)
> +		new_entry->lnb = strdup(entry->lnb);
> +
> +	/*
> +	 * The frequency should change to the new one. Seek for it and
> +	 * replace its value to the desired one.
> +	 */
> +	for (i = 0; i < new_entry->n_props; i++) {
> +		if (new_entry->props[i].cmd == DTV_FREQUENCY) {
> +			new_entry->props[i].u.data = freq;
> +			/* Navigate to the end of the entry list */
> +			while (entry->next) {
> +				entry = entry->next;
> +				n++;
> +			}
> +			dvb_log("New transponder/channel found: #%d: %d",
> +			        n, freq);
> +			entry->next = new_entry;
> +			new_entry->next = NULL;
> +			return new_entry;
> +		}
> +	}
> +
> +	/* This should never happen */
> +	dvb_logerr("BUG: Couldn't add %d to the scan frequency list.", freq);
> +	free(new_entry);
> +
> +	return NULL;
> +}
> +
>  struct update_transponders {
>  	struct dvb_v5_fe_parms *parms;
>  	struct dvb_v5_descriptors *dvb_scan_handler;
> @@ -987,6 +1071,36 @@ static void add_update_nit_dvbs(struct dvb_table_nit *nit,
>  				     SYS_DVBS2);
>  }
>  
> +static void add_update_nit_isdbs(struct dvb_table_nit *nit,
> +				 struct dvb_table_nit_transport *tran,
> +				 struct dvb_desc *desc,
> +				 void *priv)
> +{
> +	struct update_transponders *tr = priv;
> +	struct dvb_entry *new;
> +	struct dvb_desc_sat *d = (void *)desc;

Hmm... this looks wrong... please add a FIXME here, as we'll need to handle
it better, as the Satellite descriptor at ARIB STB-B10 is not quite the
same as the one defined at ETSI EN 300 468.

At DVB, we have (lib/include/libdvbv5/desc_sat.h):

        uint8_t modulation_type:2;
        uint8_t modulation_system:1;
       	uint8_t roll_off:2;

However, for ISDB-S, this should be, instead:

	uint8_t modulation:5;

The fec field for ISDB-S is also 4 bits, but has a different meaning:
On EN300468, 1000b means 4/5 conv. code rate, while it means ISDB-S 
at STD B10. Upper codes there are also different.

> +	uint32_t ts_id;
> +
> +	if (tr->update)
> +		return;
> +
> +        ts_id = tran->transport_id;
> +	new = dvb_scan_add_entry_isdbs(tr->parms, tr->first_entry, tr->entry,
> +				       d->frequency, tr->shift, ts_id);
> +	if (!new)
> +		return;
> +
> +	/* Set NIT ISDB-S props for the transponder */
> +	/* modulation is not defined here but in TMCC. */
> +	/* skip setting it since no "AUTO" value in fe_modulation_t */
> +	dvb_store_entry_prop(new, DTV_POLARIZATION,
> +			     dvbs_polarization[d->polarization]);
> +	dvb_store_entry_prop(new, DTV_SYMBOL_RATE,
> +			     d->symbol_rate);
> +	dvb_store_entry_prop(new, DTV_INNER_FEC,
> +			     dvbs_dvbc_dvbs_freq_inner[d->fec]);
> +}
> +
>  
>  static void __dvb_add_update_transponders(struct dvb_v5_fe_parms_priv *parms,
>  					  struct dvb_v5_descriptors *dvb_scan_handler,
> @@ -1045,6 +1159,13 @@ static void __dvb_add_update_transponders(struct dvb_v5_fe_parms_priv *parms,
>  				satellite_delivery_system_descriptor,
>  				NULL, add_update_nit_dvbs, &tr);
>  		return;
> +	case SYS_ISDBS:
> +		dvb_table_nit_descriptor_handler(
> +				&parms->p, dvb_scan_handler->nit,
> +				satellite_delivery_system_descriptor,
> +				NULL, add_update_nit_isdbs, &tr);
> +		return;
> +
>  	default:
>  		dvb_log("Transponders detection not implemented for this standard yet.");
>  		return;
> diff --git a/lib/libdvbv5/parse_string.c b/lib/libdvbv5/parse_string.c
> index 3750d04..6a4de2b 100644
> --- a/lib/libdvbv5/parse_string.c
> +++ b/lib/libdvbv5/parse_string.c
> @@ -366,6 +366,22 @@ static void charset_conversion(struct dvb_v5_fe_parms *parms, char **dest, const
>  				     parms->output_charset);
>  }
>  
> +static void isdb_parse_string(struct dvb_v5_fe_parms *parms, char **dest,
> +			      const unsigned char *src, size_t len)
> +{
> +	int destlen;
> +
> +	destlen = len * 4;
> +	*dest = malloc(destlen + 1);
> +	dvb_iconv_to_charset(parms, *dest, destlen, src, len,
> +			     "ARIB-STD-B24", parms->output_charset);

Hmm... I don't think that iconv supports it. We'll need to add a handler
for it, just like we had to do for EN 300 468 table.

Anyway, you don't need a separate function here. Just 
parms->default_charset needs to be filled properly.

> +	/* The code had over-sized the space. Fix it. */
> +	if (*dest)
> +		*dest = realloc(*dest, strlen(*dest) + 1);
> +	return;
> +}
> +
> +
>  void dvb_parse_string(struct dvb_v5_fe_parms *parms, char **dest, char **emph,
>  		      const unsigned char *src, size_t len)
>  {
> @@ -386,6 +402,13 @@ void dvb_parse_string(struct dvb_v5_fe_parms *parms, char **dest, char **emph,
>  	if (!len)
>  		return;
>  
> +	/* Strings in ISDB-S/T do not start with a charset identifier, */
> +	/*  and can start with a control character. */
> +	if (!strcasecmp(type, "ARIB-STD-B24")) {
> +		isdb_parse_string(parms, dest, src, len);
> +		return;
> +	}
> +
>  	if (*src < 0x20) {
>  		switch (*src) {
>  		case 0x00:	type = "ISO-6937";		break;
--
To unsubscribe from this list: send the line "unsubscribe linux-media" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html




[Index of Archives]     [Linux Input]     [Video for Linux]     [Gstreamer Embedded]     [Mplayer Users]     [Linux USB Devel]     [Linux Audio Users]     [Linux Kernel]     [Linux SCSI]     [Yosemite Backpacking]
  Powered by Linux