[Mp4-tech] Re: [AAC] LTP for short windows

[Ralph Sperschneider]

Tom Colls wrote:
> Hi experts,
>  
> I was little confused about the way LTP is done for short windows(for 
> AAC decoder). The things that have been bugging me include
>  
> 1. how ltp_short_lag is combined with ltp_lag ? The standard doesnt 
> say anything about this , however the reference software does throw 
> some light on this. That is, it it has some DPCM kind of encoding(like 
> used for spectral data). Is it supposed to be handled the way its done 
> in the reference software 
> ?(http://www.mp3-tech.org/programmer/sources/refsoft990809.tgz)
>  
> 2. how is ltp done for short windows ?
> hows the total lag (computed as above) used for short windows. Is the 
> MDCT performed in the LTP loop a 256 point MDCT , or do we still use 
> a 2048 point MDCT over the whole frame(current frame + past frame) o! 
> nce ? , like the way its done for long windows.
>  
> Any help would be very much appreciated,
>  
> Regards
>  
>
Dear Tom,
LTP is specified for long blocks only. For a while, LTP in combination 
with short blocks was specified in conjunction with AAC scalable. 
However, no conformance test sequences were provided by the proponent, 
and hence this extension was removed from the standard. It might be that 
some (unused) code is still present in the reference software.
Best regards,
Ralph
-- 
Dipl.-Ing. Ralph Sperschneider  | Phone: +49 9131 776 344
Fraunhofer IIS                  | Fax:   +49 9131 776 398
Am Wolfsmantel 33               | mailto:ralph.sperschneider iis.fraunhofer.de
D 91058 Erlangen                | http://www.iis.fraunhofer.de/amm/