[thedoc (OP)]

The big disadvantage of a DAB is the delay.  How can we we get the timing right for digital radio?
Asked by Anonymous, via text


                                        Visit the webpage for the podcast in which this question is answered.

 

[thedoc (OP)]

We answered this question on the show...

 John -   There's no magic answer to this one because there's so much intelligence going on in the audio coding system in order to compress it to 256 or less.  In fact, the DAB standard for most stations now is 128 kilobits.  There's a lot of sampling of the audio to begin with, to analyse how much can we throw away and still make an acceptable signal.  That takes time at the transmission end, but also, it takes time at the decoding end where your set has got to understand what’s going on there.  So, fundamentally, DAB is going to be behind analogue, in the same way as digital television is behind analogue television, when it existed of course.

[techmind]

There are two principal sources of delay:
A) the audio coding to bit-rate-reduced digital audio data and subsequent decoding ("audio codec")
B) the coding and interleaving used for transmitting over the radio waves

A) The digital audio standards are block (or "frame") based. The MP2 standard used by DAB breaks the audio down into chuncks of 1152 samples (a little over 40ms, depending on sample-rate). The encoder at the transmitter has to receive and buffer one block of ~40millsecond, process and encode that block (taking a maximum of 40milliseconds), transmit the block (taking ~40milliseconds) ... then the receiver decodes the block (taking a maximum of 40milliseconds) then it can start to 'play' that block.
Owing to the initial buffering and then the transmission time, there's a bare minimum of a two-frame (~80millisecond) delay, achieveable only if the coding and decoding computation happened instantaneously. Realistically, the coding and decoding will probably each add roughly a 1-frame delay (so 4-frame total).

B) For the radio-leg of the transmission, the data is again sent in discrete blocks (typically different sizes to the audio-codec blocks) which have to be buffered, encoded, transmitted and received, then decoded to digital data (and then eventually passed to the audio decoder). For DAB, the block size (strictly "symbol period" including guard interval) is 1.246ms for the UK system. Using similar arguments to (A), this will likely add around 4 block-delays (~5ms) - though this is quite negligible compared to the audio codec. But there's another issue: to improve resilience to impulsive interference, the data is "time-interleaved" for transmission. This means the data is re-ordered over a half-second to a second so that if a little bit gets corrupted due to a very brief burst of interference, after reordering back into the correct sequence instead of one big dropout, you get lots of tiny dropouts spread over a second or so which are more easily corrected by digital error correction, or just concealed.

So although we can't really eliminate the delay compared to "live" or analog broadcasts (at least not without losing many of the benefits of digital transmission)... in principle I'm quite sure the overall delay could have been standardised so that there wouldn't be an annoying mismatch between different radios in the house tuned to the same programme. This could have been done for one technology (e.g. DAB).
It wouldn't be realistic to expect Freeview-radio, DAB radio, radio over freesat etc to all remain in step - partly as the standards were set at very different times, but also because you'd end up with a lowest-common-denominator problem - where every system had to hold itself back for the slowest.

[CliffordK]

Many "live" programs have a 5 second delay to allow inappropriate content to be scrubbed.
Presumably if there is a 1 second (or so) overall encoding delay, an abort signal could bypass that, and still abort before transmission even after encoding has begun.

I agree with techmind that the delay, both on the transmission and receiving side should be standardized for multiple media forms.

How common is it to have both analog and digital radios playing the same station in a house?  At some point, I presume most radios would be upgraded to digital capable...  Well, I'm not sure they've hit the market here yet, but perhaps sometime...

However, I could imagine multiple media forms playing government debates, or something simultaneously in different rooms.  TV, different radios, etc.

[evan_au]

An increasing amount of content is "time-shifted": transmitted at a time convenient to the content producer or distributor, but consumed at a time convenient to the content consumer.
The print media (newspapers & magazines) have always operated in this mode.
This now also applies to video content (with personal video recorders and Youtube) and audio content (with podcasts).
Since the time differential is often hours or days, time alignment is not critical.

The main things we seem to want in "real-time" now seem to be the news (perhaps just minutes old), Twitter (seconds old) and Stock prices (computerised traders want them milliseconds old).

Real-time interaction with an audience is still desirable when there is a studio guest, and listeners can ask questions as the show proceeds.

Traditionally, the time-beeps were the most accurate way of setting your watch, and so a known delay was important. However, these days your smartphone is probably synchronised to GPS satellites or a Network Time Protocol server somewhere on the internet, so it is always "on time".

[RE.Craig]

Stereo drift is another problem with DAB broadcasts so much so that genuine live stereo broadcasts are no longer transmitted. The stereo we hear at our end is usually the result of software processing which takes time which results in a considerable delay. If you listen to a genuinely live stereo broadcast on DAB it has the habit of momentarily going mono at regular intervals, it has I believe something to do with the way bit rates are fed from the the mics to the  AD converters on the DAB transmitters. 

[techmind]

Stereo drift is another problem with DAB broadcasts so much so that genuine live stereo broadcasts are no longer transmitted. The stereo we hear at our end is usually the result of software processing which takes time which results in a considerable delay. If you listen to a genuinely live stereo broadcast on DAB it has the habit of momentarily going mono at regular intervals, it has I believe something to do with the way bit rates are fed from the the mics to the  AD converters on the DAB transmitters. 

First I've heard of 'stereo drift'.... can you give any link to an original source about it?

Definitely programmes can drop to mono if the broadcaster is trying to squeeze the transmission bit-rates too much - but I don't think they'd want to do that dynamically mid-programme - that would be distracting.