Difference between revisions of "Private:mtv issues"
From NMSL
| Line 6: | Line 6: | ||
*The MPE and/or MPE-FEC sections missing also seem to be randomly determined. | *The MPE and/or MPE-FEC sections missing also seem to be randomly determined. | ||
*Unfortunately a different TS file gets generated even if you run the program with the same configurations. This means that dvbSAM shows different #of packets per burst, different #of bytes corrected by MPE-FEC, etc. I produced at least 5 different TS files using the same configurations, none of which contained identical burst sequences. | *Unfortunately a different TS file gets generated even if you run the program with the same configurations. This means that dvbSAM shows different #of packets per burst, different #of bytes corrected by MPE-FEC, etc. I produced at least 5 different TS files using the same configurations, none of which contained identical burst sequences. | ||
| − | + | Perfect burst: | |
[[{{ns:image}}:GoodFecAnalysis.png|685px]] | [[{{ns:image}}:GoodFecAnalysis.png|685px]] | ||
| + | Problematic burst: | ||
[[{{ns:image}}:BadFecAnalysis.png|685px]] | [[{{ns:image}}:BadFecAnalysis.png|685px]] | ||
| + | Section preceding missing section: | ||
| + | [[{{ns:image}}:SectionNumber.png]] | ||
| + | |||
| + | Section proceeding missing section: | ||
| + | [[{{ns:image}}:NextSectionNumber.png]] | ||
'''Theory of why this is happening: ''' | '''Theory of why this is happening: ''' | ||
Revision as of 15:49, 9 July 2010
MPE, and MPE-FEC sections are not being picked up by dvbSAM
The Bug:
- When the program is run using the STAND algorithm and the TS file is generated, dvbSAM has determined that there are entire MPE and/or MPE-FEC sections missing inside some bursts. (Maybe some are half missing as well, but I have not verified this claim)
- There does not seem to be a pattern in determining which burst will be missing sections. Some bursts will be 100% complete whereas some bursts will be missing at least one section of data.
- The MPE and/or MPE-FEC sections missing also seem to be randomly determined.
- Unfortunately a different TS file gets generated even if you run the program with the same configurations. This means that dvbSAM shows different #of packets per burst, different #of bytes corrected by MPE-FEC, etc. I produced at least 5 different TS files using the same configurations, none of which contained identical burst sequences.
Perfect burst:
Problematic burst:
Section preceding missing section:
Section proceeding missing section:
Theory of why this is happening:
- There is some semantic error in the code, possibly very small.
- I believe that dvbSAM is throwing out some TS packets which happen to include the MPE/MPE-FEC header information which causes the entire section of data to be discarded.
- The fact that there are some bursts which are perfect suggests that all the MPE and MPE-FEC sections are being generated correctly. And this in turn would also suggest that packing MPE, MPE-FEC, and TS packets are done correctly.
Steps Taken:
- Analyzed Burst::enc_mpe(), all mpe sections seem logically generated
- Analyzed Burst::enc_mpe_fec(), all mpe-fec sections seem logically generated
- Analyzed Burst::enc_ts(), all ts sections seem logically generated
- Created file containing hex dump of the burst being passed to Transmitter using Burst::dump_ts(). I viewed the TS file using dvbSAM to pick out missing sections in the burst. Then I looked into the hex dump for the TS packets containing the MPE/MPE-FEC section preceding the missing section, the missing section, and the section proceeding the missing section. All sections where found in the hex dump. There was nothing unusual with the TS packets for the missing section.
Current Status:
- Bug remains unsolved.
- However, the good part is that MPE-FEC corrects these missing bytes of data.
