http://pc.watch.impress.co.jp/docs/2.../mobile357.htm
Quote:
* Though PSX (the DVD recorder) was too weak to decode SACD in software, it was expected that the PS3 would be able to. SCE began the development in a relatively earlier stage. But even at the beginning of this year the SACD decoder was 6 times slower than realtime to convert surround DSD sound into PCM.
* The manager of the Software Platform division at SCE who were in charge of software codecs and players estimated it'd be impossible to finish if they kept that pace. A developer from the Sony HQ who developed softwares for VAIO PC came to SCE and they assigned the development of an SACD decoder codec to him.
* SACD is stored in the DST format (= compressed DSD). DST changes compression methods and parameters every 1/75 seconds, which eats huge processing load when decoding. In the first 2 months the developer optimized it into realtime processing. Then he threw away the old source code and wrote a faster new codec in 3 weeks with the knowledge he had gained. It uses 5 SPEs - 3 SPEs for DST decompression, and 2 SPEs for DSD to PCM conversion.
* After that, he debugged it while consulting the professional audio equipments division of Sony that developed the DSD format itself. It took 6 months.
* Then the audio tuning specialist at the audio division of Sony began to use a PS3 to check the sound of the then unreleased HD receiver TA-DA3200ES. He pointed out where to fix in the PS3's digital audio processing to SCE. It was in the early October, and the dynamic range at that time was 140dB which was the initial goal set by another Sony developer who developed DSD. The PS3 firmware version 1.10 is this version with a few updates. The decimation filter of the SACD decoder outputs at 24bit/88.2kHz.
* However, last Friday, Honda went to a place for the interview with Ken Kutaragi for some audio/videophile magazine. They brought there an even newer version with a decimation filter at 24bit/176.4kHz. Also, the 64-bit DP internal data of the decimation filter is rounded down to 30-29-bit instead of 24-bit for the output. The dynamic range is over 170dB and the theoretical number reaches 180dB in 30-bit. Honda says its sound was really good even when compared with the sound of the 1.1 version firmware. The Sony developer attributes the goodness to the fact that the software SACD codec of the PS3 processes all data in 64-bit double precision. This new firmware version will be available when the BD remote is released in December.
* As for improving CD sound, upsampling is apparently easy by writing a FIR filter. But it is not yet in the PS3 software player as it takes some time to choose an appropriate upsampling function and noise shaping algorithm.
The #2 is about the BD video player.
http://pc.watch.impress.co.jp/docs/2.../mobile358.htm
Quote:
* Since the standards for BD-ROM such as BDMV and BD-J were delayed, particular player functions were developed later while 2 years has already passed for the development of software decoding function for codecs such as H.264.
* H.264 decoding itself was not very difficult for Cell with moderate optimization and they could play a movie in realtime at the first try unlike very difficult SACD optimization. However, because they began the development without knowing the final Blu-ray standard, they set the goal very high for decoding 2 full HD H.264 streams at 40Mbps simultaneously. Besides the clockspeed of the devkit was lower than the final product which made the development difficult. The current decoder can decode full HD H.264 with 3 SPEs.
* The current BD/DVD player for the PS3 does internal pixel processing at 16-bit per RGB. (The output depends on the display color depth.) Though the 1.0 firmware was 8-bit per RGB, Kutaragi ordered the change of the design at the final product review and it was upgraded to 16-bit in the 1.1 firmware.
* It's still not enough for TV and theater projectors that can display the original YCrCb stored in BD and DVD. For example the current PS3 player lacks gradation in brighter and darker parts of pictures when compared to Panasonic DMP-BD10. The reason why the internal processing is done in RGB is because RSX can't have 2 color spaces at the same time. Honda suspects they used RGB for the player because in the prototype design they planned to overlay XMB in RGB onto the picture of BD/DVD. In the final design XMB is not overlayed onto the player.
* Actually Kutaragi also ordered the development of the version with YCrCb internal processing at the final review. In the firmware version released in December it becomes possible for a user to choose RGB or YCrCb.
* The current player converts interlace to progressive only for SD movies. Right now SCE developers are working to improve DVD picture quality after they implemented YCrCb. After that it's planned to do something for 1080i to 1080p conversion. Though the RAM size can be an obstacle for it, SCE developers say they can do it by splitting the load between Cell and RSX.
* Currently there are complaints from users about non-existent DVD upscaling in the PS3. As for the interlace-to-progressive before upscaling, the current player has it already. There was a team at the division for broadcast equipments at Sony that does research for improving image quality by using the NVIDIA shader language. They ported it onto RSX. After Cell does noise reduction and other filters, RSX does interlace-to-progressive conversion. Though the current algorithm has bugs for some video sources, SCE says the conversion precision will be upgraded by future updates. Also the noise reduction filter is currently optimized for video sources with relatively many MPEG noises recorded with a video recorder. SCE knows lack of picture details because of it and will provide new default values for noise reduction by checking recording formats and media IDs for DVD discs. As for upscaling, they are doing the research to develop a high-quality multi-tap upscaling filter that balances the load between Cell and RSX. They also plan to use the non-original luminance information generated in upscaling for HDMI 1.3 deep color.
* Currently it's known that the audio quality of DTS is especially good compared to other audio codecs as heard in playing 480p DVD via HDMI with more bandwidth for the audio. SCE don't know the definitive reason. They just ported the reference source code provided by DTS with as high precision as possible.
* An SCE developer recommends trying 1.5x fast-forward playback in the PS3 BD player to see the power of Cell. When it's connected to a display via 1080/60p, it becomes very smooth as Cell has an enough margin for video decoding. In 1.5x fast-forward playback it decodes all frames then inserts them into 60fps with sped up audio.
* SCE declared that they'd support 1080/24p in the PS3.
