[Mp4-tech] [HE-AAC]understanding of AAC+ conformance test

Fri Apr 18 11:02:12 EDT 2008

Hi,
First off: Yur conclusion that you are 'not fully compliant' with an RMS of -100.95 dB is not exactly correct. AAC-LC conformance gives you one degree of freedom, which is the accuracy you claim to achieve. For an accuracy level of k=16 bit, you need to achieve an RMS of -101.10 dB. Since this can be pretty hard to achieve even for the completely accurate 16 bit decoder output (cf. your experiment where you check 16 bit reference software output against itself), you may use an intermediate signal if you want to test for k=16 bit accuracy.
Regarding your questions:
Scenario one is the procedure you would use, if you don't have access to the time domain data before rounding to 16 bit. In this case testing for k=16 bit accuracy is bound to fail, but testing for k=15 bit accuracy can very well work.
Scenario two is what you would do, if you want to check for k=16 bit (or more) accuracy. You may call your decoder 16-bit compliant, even if you round the output to 16 bit and fail conformance after rounding.
Scenario three is a nono. The reference must remain unchanged.
As you have observed: When testing for k=N bits accuracy, you test that the noise that is introduced by your decoder is equal to or less than the quantization noise when quantizing to N bit.
Best,
Andreas
> Scenario 1:
> My Decoder output is 16 bit.
> Append 8 zeroes at LSB to make it 24 bit and compare with 24 
> bit reference output.  In such a scenario, always my lower 8 
> bit will definitely differ.  In which case, my decoder may 
> not be fully conformant.
> With this experiment, difference w.r.t RMS level as compared 
> to reference is -100.95dB (meaning not fully
> compliant)
> Please let me know is this correct?
> 
> Scenario 2:
> I will tap 24 bit output from my decoder before converting it 
> 16 bit PCM. Then compare with 24 bit reference output.
> With this experiment,  difference w.r.t. RMS level as 
> compared to reference is approx. -115dB.
> But, for my application I will be still using only 16 bit out 
> of these 24 bits.
> Please let me know is this correct? If correct, is it right 
> to have 16 bit PCM out of 24 bit for my application?
> 
> Scenario 3:
> I will convert 24-bit reference to 16-bit reference using 
> audio editing tools like Goldwave.  Then compare my decoder's 
> 16-bit output to 16-bit reference.
> In this case, difference w.r.t. RMS levle as compared to 
> 16-bit reference is approx -103dB.
> Please let me know is this correct?
> 
> I did some more experiments to understand the conformance 
> criteria.  I converted 24 bit reference itself to 16 bit with 
> rounding.  In this case, RMS difference w.r.t RMS between 
> 16-bit converted reference and 24 bit reference is -101.09dB 
> (which is not fully compliant).  This means, reference itself 
> if converted 16 bit is not fully compliant.
> 
> Please note, I have used SSNRCD tool provided by ISO for all my tests.
> 
> Thanks in advance for clarifying my doubts
> 
> Regards,
> Dattaguru
> 
> 
> 
>  
> 
> 
> --- "Schneider, Andreas" <andreas.schneider dolby.com>
> wrote:
> 
> > Hi Andy,
> > 
> > You are not forced to use 16 bit PCM data, but contrary to what is 
> > required for 16 bit accuracy conformance testing of AAC, 16 bit PCM 
> > data will work for SBR conformance testing.
> > Regarding your second question: MPEG doesn't do critical changes to 
> > documents when they are published, e.g. when creating the 
> AMD from the 
> > FDAM.
> > 
> > Best,
> > 
> > Andreas
> > 
> > Andreas Schneider
> > Senior Research Engineer
> > 
> > Dolby Germany GmbH
> > Deutschherrnstr. 15 - 19            
> > 90429 Nuernberg, Germany           
> > http://www.dolby.com
> > 
> > + 49 911 928 91 - 26 (phone)
> > + 49 911 928 91 - 99 (fax)
> > mailto:Andreas.Schneider dolby.com
> > 
> > HRB 17557, Amtsgericht Nürnberg,
> > GF: Dipl.Ing. Martin Dietz (Managing Director)   
> > 
> > > -----Original Message-----
> > > From: Andy Quan [mailto:zhengyuan.quan gmail.com]
> > > Sent: Montag, 14. April 2008 12:34
> > > To: Schneider, Andreas
> > > Cc: mp4-tech lists.mpegif.org
> > > Subject: Re: [Mp4-tech] [HE-AAC]understanding of
> > AAC+ conformance test
> > > 
> > > Thanks, andreas.
> > > It is nice to know there is new comment of
> > conformance point
> > > for AAC testing in this spec. I have some more
> > question below.
> > > 1. According to ISO/IEC 14496-4:2004/FDAM 8:2004,
> > this
> > > flexibility of conformance point applies to LC RMS
> > testing
> > > only, right? (And for SBR test, input to RMS
> > comparison
> > > module has to be 16-bit?) 2. I dont have any
> > corrigendum/spec
> > > later than ISO/IEC 14496-4:2004/FDAM 8:2004, so is
> > there any
> > > critical correction/supplementary to SBR
> > conformance
> > > criteria/method described in amd8?
> > > 
> > > --
> > > Regards,
> > > Andy Quan
> > > 
> > > On 4/14/08, Schneider, Andreas
> > <andreas.schneider dolby.com> wrote:
> > > > Hi Andy,
> > > >
> > > > There are two aspects involved here:
> > > > One is that the AAC conformance spec allows you
> > to define
> > > against which criteria you want to test your
> > decoder. This
> > > can be done by selecting the parameter 'k' in the
> > equations
> > > that are used to determine the conformance
> > thresholds. So one
> > > way to approach this is to set the k-parameter to
> > 15, which
> > > results in conformance thresholds that can be met
> > with 16 bit
> > > PCM output. The value of k that was used for the
> > conformance
> > > test must be mentioned when claiming AAC
> > conformance (,
> > > unless you set k to the default value of 16, in
> > which case 16
> > > bit PCM output won't suffice for conformance
> > testing).
> > > > The other aspect is that the conformance point,
> > i.e. the
> > > signal that is considered for conformance testing
> > is the
> > > actual output of the AAC decoder after the
> > overlap-add of the
> > > inverse transform but before a quantization to 16
> > bit PCM. 
> > > MPEG added a clarifying statement regarding the
> > location of
> > > the conformance point to the AAC conformance
> > standard in the
> > > same document that specifies SBR conformance
> > (ISO/IEC
> > > 14496-4:2004/FDAM 8:2004).
> > > >
> > > > Best,
> > > >
> > > > Andreas
> > > >
> > > > Andreas Schneider
> > > > Senior Research Engineer
> > > >
> > > > Dolby Germany GmbH
> > > > Deutschherrnstr. 15 - 19
> > > > 90429 Nuernberg, Germany
> > > > http://www.dolby.com
> > > >
> > > > + 49 911 928 91 - 26 (phone)
> > > > + 49 911 928 91 - 99 (fax)
> > > > mailto:Andreas.Schneider dolby.com
> > > >
> > > > HRB 17557, Amtsgericht Nürnberg,
> > > > GF: Dipl.Ing. Martin Dietz (Managing Director)
> > > >
> > > > > -----Original Message-----
> > > > > From: Andy Quan
> > [mailto:zhengyuan.quan gmail.com]
> > > > > Sent: Donnerstag, 10. April 2008 04:50
> > > > > To: Schneider, Andreas
> > > > > Cc: mp4-tech lists.mpegif.org
> > > > > Subject: Re: [Mp4-tech] [HE-AAC]understanding
> > of AAC+ conformance
> > > > > test
> > > > >
> > > > > Hey Andreas,
> > > > > The explanation is crystal clear. Thank you.
> > > > > And here comes another question. As I
> > remember, some early
> > > > > discussion of LC decoder in this forum says
> > that there is
> > > no way for
> > > > > a 16-bit output PCM to pass RMS conformance
> > test in LC
> > > suite as the
> > > > > reference is an independent 24-bit sound file.
> > If this is
> > > true, how
> > > > > can we judge the compliance of this kind of LC
> > decoder? As we all
> > > > > know, most LC decoders available today provide
> > 16-bit
> > > output rather
> > > > > than 24-bit ones.
> > > > >
> > > > > On 4/8/08, Schneider, Andreas
> > <andreas.schneider dolby.com> wrote:
> > > > > > Hi Andy,
> > > > > >
> > > > > > If you want to call your decoder ISO
> > compliant, it must
> > > > > produce an output signal that satisfies the
> > criteria for max.
> > > > > diff and RMS as defined by ISO. The
> > conformance document actually
> > > > > tells you to apply these thresholds in the
> > column "test
> > > procedure" 
> > > > > of the table that defines the conformance
> > bitstreams.
> > > > > > So any decoder that claims that it is fully
> > ISO compliant
> > > > > effectively claims that it passes all these
> > conformance tests.
> > > > > >
> > > > > > Andreas
> > > > > >
> > > > > > Andreas Schneider
> > > > > > Senior Research Engineer
> > > > > >
> > > > > > Dolby Germany GmbH
> > > > > > Deutschherrnstr. 15 - 19
> > > > > > 90429 Nuernberg, Germany
> > > > > > http://www.dolby.com
> > > > > >
> > > > > > + 49 911 928 91 - 26 (phone)
> > > > > > + 49 911 928 91 - 99 (fax)
> > > > > > mailto:Andreas.Schneider dolby.com
> > > > > >
> > > > > > HRB 17557, Amtsgericht Nürnberg,
> > > > > > GF: Dipl.Ing. Martin Dietz (Managing
> > Director)
> > > > > >
> > > > > > > -----Original Message-----
> > > > > > > From: mp4-tech-bounces lists.mpegif.org 
> > > > > > > [mailto:mp4-tech-bounces lists.mpegif.org]
> > On Behalf Of Andy
> > > > > > > Quan
> > > > > > > Sent: Dienstag, 8. April 2008 15:36
> > > > > > > To: mp4-tech lists.mpegif.org
> > > > > > > Subject: [Mp4-tech] [HE-AAC]understanding
> > of AAC+ conformance
> > > > > > > test
> > > > > > >
> > > > > > > I am recently working on AAC+ ISO
> > conformance test(SBR,
> > 
> === message truncated ===
> 
> 
> 
>       
> ______________________________________________________________
> ______________________
> Be a better friend, newshound, and
> know-it-all with Yahoo! Mobile.  Try it now.  
> http://mobile.yahoo.com/;_ylt=Ahu06i62sR8HDtDypao8Wcj9tAcJ
> _______________________________________________
> NOTE: Please use clear subject lines for your posts. Include 
> [audio, [video], [systems], [general] or another apppropriate 
> identifier to indicate the type of question you have.
> 
> Note: Conduct on the mailing list is subject to the Antitrust 
> guidelines found at 
> http://www.mpegif.org/public/documents/vault/mp-out-30042-Anti
> trust.php
> 

Andreas Schneider
Staff Customer Support Engineer
Dolby Germany GmbH
Deutschherrnstr. 15 - 19            
90429 Nuernberg, Germany           
http://www.dolby.com
+ 49 911 928 91 - 26 (phone)
+ 49 911 928 91 - 99 (fax)
mailto:Andreas.Schneider dolby.com
HRB 17557, Amtsgericht Nürnberg,
GF: Dipl.Ing. Martin Dietz (Managing Director)   
-----------------------------------------
This message (including any attachments) may contain confidential
information intended for a specific individual and purpose.  If you
are not the intended recipient, delete this message.  If you are
not the intended recipient, disclosing, copying, distributing, or
taking any action based on this message is strictly prohibited.