Sampling Frequency and Constant Duration

[davroslyrad]

Greetings everyone,

I am developing systems using the ATSC-153 mobile broadcast. Part of the broadcast are the audio configuration tables which are to define parameters of the audio being broadcast.

Every broadcast I have worked with transmits the audio with 24Khz of sampling frequency, and the configurations tables indicate so. ATSC-153 defines the compression to be HEAAC-v2. So, for every AAC AU that is received gets an ADTS header, and everything works fine.

However, there is an exception. I have one broadcast in particular in which the configuration tables indicate the sampling frequency of 48Khz. The audio is actually 24Khz, and if I hard code the 24Khz index, the audio will play fine.

Competitor systems play the broadcast correctly. So, I can I can either think they are simply hard coding the index for 24Khz, or they figure out the sampling frequency by other means.

This particular broadcast also indicates a constant duration that other streams do not. By definition, the constant duration is the clock ticks the AU is. This deals more with the timestamps than sampling frequency.

So, the question is, is there a way of deriving the sampling frequency of the AU given the constant duration to go by? My thought is no, however I thought I would ask the community what their thoughts are.

Many thanks,

davroslyrad.

[davroslyrad]

Greetings everyone,

I have dug into this just a bit, and I discovered the real problem here is trying to wedge HEAAC-v2 parameters into an ADTS header. Please correct me if I am wrong, the idea is to audio work with an LC type of decoder. As a result, the SBR payload, which will double the sample rate, will be ignored by the LC decoder.

According to ATSC-153, sample rates of 24Khz, 22.05Khz and 16Khz will have SBR present, while sample rates of 48Khz, 44.1Khz, and 32Khz will not have SBR present. For this reason, dividing by 2 the 48Khz sample rate to 24Khz actually worked since no SBR was present. Would this actually work?

BTW: Does anyone know the level for the HEAAC profile for value 0x30? The table I have (14496-3_2005, Table 1.12) stops at 0x2f, which is level 5. It looks as though it should level, for adts profile value of 1 works (object type minus 1).

Many thanks,

davroslyrad

[kieranrk]

I can't comment about ATSC but in DVB you always signal the ADTS payload as plain AAC (i.e half sample rate SBR). SBR compliant decoders pick up the extension by themselves.