Understanding the DAB clock of 2048 kHz is necessary to understand the 24 ms timing. In the following text there will be a lot of references to T (the period of the clock 0,48828125 µs). See “The magical number 2048“
How to fit different types of audio compression in the same 24ms frame?
To understand DAB/DAB+ audio transport, one must know that audio streams consist of a continuous sequence of so-called audio frames. The audio frames follow one after another to guarantee the continuity of the transported audio. Each frame hauls a chunk of audio of a specific duration. For the transport mechanism in DAB/DAB+, the duration time of an audio frame is 24ms. As a result these audio frames (with the audio bits) have to be transported every 24ms.
How does 24 ms fit in DAB timing?
24 ms is four times 96 ms (the length of the DAB transmission frame.
Or 196608 T (clock periods).
1 CIF being 196608 / 4 = 49152 T or 49152 T/2048 kHz = 24 ms
- Audio frames will be sent every 24ms by the multiplexer and will arrive at the transmitter input as ETI(LI) frames.
- Common Interleave Frames (CIF’s) are transmitted every 24ms
- The DAB/DAB+ transmission frame is always a multiple of 24 ms!
- DAB Mode I 96 ms (transmit 4 CIF’s per transmission frame)
- DAB Mode II 24 ms (transmit 1 CIF per transmission frame)
- DAB Mode III 24 ms (transmit 1 CIF per transmission frame)
- DAB Mode IV 48 ms (transmit 2 CIF’s per transmission frame)
A DAB Mode I transmitter will need to receive four ETI frames of 24 ms to transmit one transmission frame of 96ms.
It shows already the importance of the 24 ms timing in the DAB equipment.
What’s the origin of the 24ms timing?
This timing period originates from the audio encoding which is based on a conversion from time domain to frequency domain. Converting time to frequency domain can not be done sample per sample. To make the conversion you need a large number of samples (called a window) to be processed at once. The number of samples and their sampling rate define the timing of the window.
The original DAB uses the MPEG 1/2-layer II audio codec, which converts 1152 input samples per frame (PCM)from time to frequency domain. Layer II uses always 1152 samples per frame to make the conversion from time to frequency domain.
At that time two sample rates were available in DAB 48kHz and 24 kHz.
At a sample rate of 48 kHz it gives us a frame duration of 24 ms.
1152 samples / 48kHz = 24 ms
This is where we find the origin of the 24 ms.
At a sample rate of 24 kHz it gives us a frame duration of 48 ms.
1152 samples / 24kHz = 48 ms
So we will put one 24 kHz sampled frame in two consecutive DAB-frames of 24 ms.
What about transporting MPEG 2 Layer III (MP3) over DAB?
MPEG 2 layer 3 is a special case with 576 samples per frame. At 48 kHz sample rate it brings us at frames with 12 ms duration.
Calculation: 576 samples / 48kHz = 12 ms
It means you can transport MP3 frames with the DAB-system you just need two consecutive MP3 frames to fill the DAB frame of 24 ms.
However, it is unlikely that DAB(+) receivers can reproduce the audio, as this is not supported in the DAB(+) specifications .
What about transporting HE-AAC over DAB?
In order to make DAB ready for DAB+ (using HE-AAC) it became a bit more complicated. Depending on the sample rate HE-AAC frames have a different number of samples to convert to the frequency domain.
In AAC each block decodes to 1024 time-domain samples. 1024 samples at 48kHz sample rate result in a time lapse of 1024/48 = 21,3333… ms. Houston we have a problem!
The AAC frame duration of 21,3333… does not fit into the DAB 24 ms frame duration.
Solving the frame duration problem for DAB+
First the number of samples/frame for 48 kHz sample rate will be reduced from 1024 samples to 960 samples.
Standard AAC use to use 1024 samples per frame
DAB+ will use only 960 samples per frame!
HE-AAC coding we know is not the HE-AAC coding used for DAB+!
The number of samples per frame for 48 kHz results in a frame duration of exactly 20ms.
960 samples / 48 kHz = 20 ms
The number of samples /frame for 32 kHz sample rate will also be reduced to 960 samples. Which result in a frame duration of exactly 30ms.
960 samples / 32 kHz = 30 ms
The HE-AAC option SBR will divide the sampling rate by 2. It means that 48kHz with SBR the basic sample rate becomes 24 kHz.
960 samples / 24 kHz = 40ms
Also the sampling rate of 32 kHz will result in a basic sample rate of 16 kHz. Or a frame duration of 60ms.
960 samples / 16 kHz = 60 ms
Apparently the reduction of the AAC frame from 1024 samples to 960 did not solve the problem of the 24 ms frames used in DAB!
However, 120 ms is a common multiple of 24 ms and all the AAC frame durations 20, 30, 40 and 60 ms. This is the key to solve the problem.
DAB+ will work with superframes being 5 consecutive 24 ms frames or superframes of 120 ms. All the AAC frame durations will fit the superframe.
- 960 samples at 48 kHz will put 6 DAB-frames in one DAB superframe for DAB+
- 960 samples at 32 kHz will put 4 DAB-frames in one DAB superframe for DAB+
- 960 samples at 24 kHz (48 kHz + SBR) will put 3 DAB-frames in one DAB superframe for DAB+
- 960 samples at 16 kHz (32 kHz + SBR) will put 2 DAB-frames in one DAB superframe for DAB+
Nevertheless 24 ms remains an important the key to complete the DAB frame puzzle.
The 24ms is also related with the clock period of DAB (2048 kHz) being
24 ms = 49 152 * T (with 1 ms = 2048 T)