Wednesday, July 25, 2007

Technical Highlights of the XM/Sirius Opposition

XM/Sirius presented some interesting ideas that they hope will allow them to expand the number of channels, such that a subscriber for one service can receive the "best of" the other service. It was an interesting discussion on how the two encode their respective signals, the differences, and how these differences can be combined to create an even more efficient coding algorithm. The purpose of these Exhibits are to dispute the study by NAB that indicated that it could not be done.

There were to Exhibits: one on behalf of Sirius:

A Technical Report Regarding Coding Efficiency and the Sirius-XM Merger

Prepared by Dr. Deepen Sinha, ATC Labs

July 24, 2007

and one on behalf of XM:

Neural Audio Corporation

Report Regarding Certain Technical Aspects Of the XM-Sirius Merger

July 24, 2007

The Sirius exhibit is worth the read if you have the opportunity. The XM exhibit is not all that enlightening, nor does it really explain how channels will be added, other than the continued evolution of the various techniques. It did little to refute the NAB report, other than to point out a few errors of marginal significance.

Highlights

* CODEC efficiencies have evolved over time and will continue to do so

* Sirus has increased its efficiency by approximately 40% since 2002

* Sirius and XM technologies have progress through different technologies

* Sirius has focused on Psychoacoustic Modeling and statistical multiplexing, while XM has focused on optimizing the pre-processor configuration

* Combining the above "like would yield improved efficiency on both the systems"

* Sirus uses the PAC audio compression scheme (no mention of AAC or AAC+)

* XM uses AAC+

* Appling "an Enhanced Psychoacoustic Model (EPM) developed for Sirius and used in PAC" to the AAC CODEC (not AAC+) can improve its efficiency. On average, there was "a bit demand reduction" of about 8.5% without optimization when tested at 51 kps)

* XM is believe to have installed "hooks" in its firmware to implement statistical multiplexing

* "Statistical multiplexing operates by encoding multiple channels instead of a single channel such that channel capacity can be adaptively steered from one channel to the other. This improves channel utilization by allowing the bandwidth to be divided based on need."

* Statistical Multiplexing can increase efficiency by 30 to 40% in video broadcasting (oddly, the exhibit does not give statistics for audio)

* Efficiencies should allow an increase of 7 music channels and 4 voice channels (assuming an average bit rate of 40 kbps for music channels and 16 kbps for voice.

* HD Radio can't benefit for statistical multiplexing

* For XM, Neural Codec Load Analysis System (N-CLAS) is applied before encoding

* Channels can be added based on seasonal substitution

Our Thoughts

We notice that the EPM was applied to the AAC CODEC and not AAC+, which XM apparently uses. Also, it seems to be trying to accomplish the same results as the neural processing, but it a different way. Both seem to try to take advantage of how the human ear perceives sounds. It is questionable whether it would yield any more efficiencies if applied to the XM stream after the neutral processing. We don't know how it would react with the AAC+ CODEC combined with Spectral Band Replication (SBR).

Statistical Multiplexing would seem to hold great promise. Whether or not it can be combined with XM's other techniques is a big question. If XM indeed has "hooks" in its code, it may well have planned on this possibility. Let's hope so. It wouldn't surprise us to find out that its current technique only works on static bit rates. If so, it could be a significant development cycle.

I have first hand experience with "hooks". They can be useful, but one rarely anticipates everything that is needed. They are usually secondary and often don't evolve with the rest of the code. I was recently project manager for a development project in which we developed our application around vendor code that left "hooks" in its code for its customers to customize the application. How did that work out? Not too good. It turned out to be a fairly major development with the vendor making significant changes to these "hooks" to make them work. We discovered numerous errors it the code using these "hooks". I would venture that 50% or more of the "hooks" had some sort of problem. And this was for a worldwide distributor of equipment.

xm sirius satellite radio stocks patent patents trademark copyright experimental licenses wcs eas invest investing fcc

Jump to :: Satellite Radio Techworld Home Page ::

4 comments:

PocketRadio said...

Anything is better than crappy HD Radio:

http://hdradiofarce.blogspot.com/

Anonymous said...

Hey, These so called Bad hooks.... did they cause a system wide shutdown for aproximatly 24 hours? hint hint.. Nudge nudge wink wink

Ned Hughes said...

I don't know... "best of" CDs and that sort are usually okay, but there's always those few things that you want to listen to that isn't there. It makes you miss out on that.

Bert said...

I think folks will be really disappointed if they can't access Stern, NFL, Martha Stewart on their XM or O&A, MLB, Oprah, and college conference games on their Sirius radios. There is no indication yet that these will be one of the 11.

Personally, I have no interest in any of those. I might be interested in some of the non-conventional music channels or public radio, but something tells me these won't be available.