Many of you will know the sad and sorry story of DAB digital radio broadcasting in the UK. The main stations were simulcast at 192kbps mp2 from 1995, but in order to introduce new stations without using more radio spectrum, bitrates were cut to 128kbps or below for most stations by 2002. With mp2, and copious amounts of transcoding on the commercial stations, the result is a horrible mushy sound.

However, the BBC have plumbed new depths during the current "summer of sport". They have decided to keep their second (part time) sports channel "BBC Radio 5 live sports extra" broadcasting on DAB, even when the station itself is off air! This means there is a 32kbps repeating advert for the station playing in a continuous loop on DAB. To make way for this, their flagship classical station, "BBC Radio 3" has had its bitrate cut to 160kbps discrete stereo (previously 192kbps discrete stereo). Meanwhile, other BBC stations on the BBC national DAB multiplex remain at 64kbps - 128kbps mp2.

BBC Radio 3 listeners aren't too happy. Previously, they sometimes had the bitrate cut to 160kbps joint stereo during the day to make way for sporting events, but it was usually put back to 192kbps discrete stereo in the evening in time for live classical music concerts (in the evening, the bitrate was "borrowed" from BBC Radio 4, switched from 128kbps joint stereo to 80kbps mono). The current 160kbps discrete stereo, even during the important "Proms" live concert season, doesn't sound very good.

Here is a diagram of the BBC national DAB multiplex configuration...

http://www.wohnort.demon.co.uk/DAB/uknat.html

There are various threads on the BBC Radio 3 message boards, and on alt.radio.digital (Usenet / Google Groups).

If you care about audio quality, it might be worth complaining to the BBC, though be warned: they already have their pre-written response ready...



Some factual inaccuracies, and some wishful thinking, but no mention of the broadcasts on DTT (Freeview) and DSat (Sky / FTA satellite) where the station can still be found at 192kbps.

It is a shame the BBC are so desperate to defend their indefensible DAB service!

Maybe if enough people complain, at least this latest change will be reversed.

Cheers,
David.

I don't know much about MP2, but why do they sometimes use discrete stereo? Isn't Joint-Stereo better always? Or does that only apply to MP3?

Absolutly terrible. I stopped listening to DAB a while back due to poor quality.

On another note, "The Chris Moyles Show" a few weeks back when in Germany, on one particular day, was very poor quality, and I'm talking over FM (97.9MHz) I can't prove the quality or provide ABX, but the broadcast was deffinatly lossy at some stage, I'd approximate the kind of quality you get with lame at 64 kbps!!!! I turned it off! The vocals were annoying and the music was even worse!!

Shame these standards take so long to ratify and adopt. Imagine what you could do with more modern encoding approaches with that avaliable bandwidth!

Krsitian

I'd say from limited experience that 128k mp2 is around 80k mp3 and should never be used. Worse case should be 160k, decent at 192k, optimal at 256k. I don't see much reason to use L/R over M/S coding , even if an mp2 encoder doesn't use adaptive m/s.

With mp2, Joint Stereo is "intensity stereo" only = frequency selective panned mono! As you might expect, it's audible. Sometimes it's preferable to the artefacts that would be heard if you used discrete stereo at low bitrates.

So, at a constant bitrate, it can make a bad situation (i.e. not enough bits) slightly better. Unlike mp3 (where mid-side joint stereo encoding and variable bitrate encoding can be manipulated to deliver exceptional results at the optimum bitrate), it cannot be a genuine positive benefit when aiming for transparent coding. You could argue that, when the source is near mono, intensity stereo will allow a more accurate encoding than discrete stereo at a given bitrate. This is true, but if you rely on this trick to deliver reasonable quality, then you probably haven't got enough bits to encode wide stereo sources properly at all.

Therefore, if the aim is transparent encoding, then with mp2 you need to have enough bits available to ensure that intensity stereo is never required.

EDIT: That said, if you're running at 160kbps mp2 (as Radio 3 is doing now), you really need intensity stereo. The reduced stereo sound stage is usually preferable to increased coding noise and lower bandwidth.

Cheers,
David.

As he was in germany, they most likely used a satellite or a secure low bandwidth wired connection to transmit the audio back. MAybe, they would of had to resort to an ISDN/POTS transmition codec which are below than your normal 128kbps MP3, especially in POTS. Its most likely the case that the low quality is down to the BBC not being able to use a high quality codec to transmit back to the BBC. This is probably not BBCs fault entirely. In most cases, quality is deceased when on an Outside Broadcast. Nothing to do with the FM transmittion to your radio.

On the point of DAB, I have always refused to move over from FM / Internet Streams. Being an audiophile, I can't see how this is acceptable. But you think most of the country being the average joe, they probably wouldn't care or not notice the quality decrease. So this is why stations can get away with this.

All my fav stations are on the internet at reasonable quality (XFM/Virgin Radio/Captial) as 128kbps WMA streams. Nice to see it improving all the time and stations coming away from 32kbps Stereo WMA.

Bring on Digital Radio Mondial (DRM) and scrap DAB OFCOM, you silly buggers!

The best stations (IMO) broadcast online anyway. Not at 192kbps, but still good enough for me.

MP2 joint-stereo is intensity stereo, which is a lossy coding method that ignores the phase difference between the channels. It is used for the higher frequency subbands only because humans don't usually notice phase differences at frequencies above some certain value (2 kHz?).

MP3 joint-stereo coding can be intensity stereo or mid/side-stereo, with the latter being a lossless transformation and therefore suitable for high quality encoding.

edit: OK, I'm too slow...

Sadly, it's all rather incestuous. Some of the people in charge at OfCom are the same people who developed and authorised DAB in the first place. They don't want to admit that they messed up.

DRM+ and DAB+ are both on the way, and will probably be used in other countries (not the UK). However, I know first-hand that people inside the BBC have slowed down DRM and DRM+ standardisation to allow DAB to get a hold first - they don't want to give DAB any serious competition until its irreversible entrenched in the UK. They were hoping to pull the same trick off across the whole of Europe, but some European broadcasters are too smart to fall for this - they know DAB v1 with mp2 is a daft choice when DAB+ / DAB v2 with AAC+ is just around the corner.

If only they'd stick everything at 256kbps mp2 on satellite (some stations are currently 192kbps, which is nearly good enough!), most people would be happy, and those who care about audio quality could just write off DAB as a bad joke.

Or we might see 128kbps AAC via the net for all BBC stations eventually.

Unfortunately, the person in charge of digital radio in the BBC sees DAB as the way forward, and isn't keen on letting anything else compete. Hence no DRM(+), no high bitrate satellite service, etc etc.

Cheers,
David.

It also depends on the "switch point" (the point at where the above spectrum is IS coded and the lower par is still L/R stereo). The theory that led to this design is the following: At high frequencies phase differences don't play the major role in determining a sound source's azimuth for us humans anymore. Thus IS (intensity stereo) solely concentrates on level differences (frequency selective panning).

Possible switch points (for MP2 IS) at 48 kHz are
3000 Hz, 6000 Hz, 9000 Hz and 12000 Hz.

I once read an article about this kind of azimuth perception that stated a switch point of above 10 kHz is "pretty safe" (not sure about its validity) -- you need to take special care while downmixing, though. BTW: Vorbis at -q5 does "point stereo" (aka IS) at above 10 kHz! <irony>someone please tell Monty "That's not how we do things for high bitrates."</irony>

So, MP2 intensity stereo with a smart encoder could sound better (in most cases) than full stereo (given that the thing about "humanoid azimuth detection" is true)


Sebastian

Try this. Generate a pulse train of gaussian pulses, about 1/2 millisecond sigma, carrier at 10kHz.

Generate the pulse train at about 2 per second, starting with left preceeding right by about 25 milliseconds, and , and make the rep rate of the second such that it moves at about .5 millisecond per pulse until it's right preceeding left by the 25 milliseconds.

Put THAT through your intensity stereo coder. I dare you!

Someone somewhere really misunderstood this stuff, and hard coded it into intensity stereo. Or else they knew full well it was a compromise, but the marketing people tried to pretend that it wasn't!

Here's the truth: if you have a high frequency sine wave (actually above 1.5-3kHz), it is perfectly true that humans simply cannot detect inter aural phase differences in the sine wave itself. It is impossible to detect, because the processing does not exist within the ear. There's a damn good reason for this: the size of the human head and the wavelength of sound at this frequency make it a pointless thing to try to detect - it won't help you to figure out where the sound is coming from.

However, if that sine wave is amplitude modulated in any way (e.g. it has a defined start and end point! or it gets louder/quieter/louder/quieter very quickly), and there is some time delay in this amplitude modulation between the two ears, then people will very easily detect this delay and use it to localise the sound.

So, while you can trash the phase of the fine structure (i.e. the time delay of the individual cycles of the high frequency waveform itself) and this is inaudible, the overall envelope of the high frequency waveform is important - if you have a signal where the envelope is different on each channel, and you make the two channels identical (or identical apart from an amplitude difference) then this is detectable and un.

Interestingly, Dolby understood this in the design of AC-3 - there's a mechanism for trashing the fine structure inter-channel delay, but maintaining the envelope inter-channel delay. With mp2 intensity stereo, both are trashed AFAIK.



I can't remember how sensitive we are to envelope phase at 12kHz. It almost certainly depends on the characteristics of the signal, and how you listen. If we're sensitive to it at all, then it's a problem. However, I agree that there must be a point, or situations, where use selectively it isn't a problem.

The point I was making is that, with CBR mp2, if you have to rely heavily on IS, then you probably don't have enough bits for transparency most of the time.

Which brings me back to UK DAB: have you heard 128kbps mp2 JS?

Cheers,
David.

I may be able to grab a Freeview radio brodcast direct form my TV card, and upload an mp2 sample from that, no transcoding etc. Thats if anyone is interested

Would it be similar to DAB broadcast? It also is mp2 and has some low bitrates!

Kristian

Can someone let me know where I should complain about this? Is there an email address I can write to, or an online complaints form anywhere?

It's a real shame that people abuse technology like this... like my freeview box - some channels are really poor quality, and I even get blocky pictures on the main channels sometimes. "Digital" used to be synonymous with high quality, but the overuse of compression in MP3s, Freeview and DAB mean this just isn't the case any more.

Ed

The only difference I see is the ability to invert the signal (180° phase shift) on selected channels for the "coupling channel" (which is also possible for AAC's IS).


Not that I remember. I encoded a movie soundtrack once in 160 kbps JS MP2 for an SVCD via tooLame. Quality was "okay" IIRC. (tooLame used to dynamically switch between 5 variants in its JS mode: 32/32 stereo bands (full stereo), 16/32 stereo bands, 12/32 stereo bands, 8/32 stereo bands, 4/32 stereo bands) -- Dunno how bad 128 kbps JS gets.

Sebastian

http://www.bbc.co.uk/complaints/

Cheers,
David.

It's even worse on Sky Digital believe it or not - ITV1 for example is reduced in resolution to about 580 x560 iirc. Pointless.

They do this on Freeview also, generally on repeated stuff on channels like ITV3. I noticed this when recording stuff on my Freeview card then burning to DVD and noticing it had to be reencoded because it's not a DVD standard resolution

I assume by "gaussian pulses" at "carrier 10 kHz" you are referring to amplitude modulation.
Woudn't a bad outcome just indicate that the temporal resolution of the "panning side infos" is too low?

I think it'd be fun to check how current Vorbis coders perform at -q5 (usually transparent, usage of IS at above 10 kHz) in this situation

Sebastian

Complained! DAB sounds worse than FM!

This looks like good news. It was argued the BBC was too deep into this DAB stuff to make the switch to DRM but apparently not. AAC instead of MP2 as a codec would be a great improvement.

BBC R&D have already carried out lots of DRM test broadcasts, but this looks quite interesting.

That article quite unhelpfully mixes up the new DAB tx sites (where the BBC is playing catch-up with Digital1 - the national commercial DAB mux) with the DRM test from one tx site.

Maybe they're sick of waiting to simulcast existing BBC local FM stations on many as yet nonexistent local commercial DAB muxes which they are forced to use.

Cheers,
David.

Yes, on channels such as ITV3 or More4 resolution will change to 544x576 - I think for ad breaks. You can use DVDPatcher to fool apps into treating the stream as valid DVD format.

Are stations still broadcast at 192kbps on Freeview? I have a card in my PC and a Humax PVR9200T downstairs, but I've only used them for video so far. I have the demuxed streams of Gangster No.1 on my laptop at the moment and the audio is 160kbps CBR MP2. I'll have to try recording some radio.

I think I'd rather take 128kbps IS MP2 over the ultra low bitrate HE-AAC (gotta be less than 32!) that XM radio is using.

Actually I think its 64kbps. In any case its not that great I agree. And sounds even worse if you don't have your XM hooked up by your aux input in your vehicle.

It doesn't sound like 64kbps. 64kbps actually sounds decent. What I heard was absolutely terrible. It really, and truly sounded like a 24/32k stream, much like di.fm's 24k stream.

64kbps HE-AAC should sound better than 128kbps mp2, though neither is really transparent so it could depend what artefacts annoy you most.

If 64kbps HE-AAC sounds really bad, then the feed into the encoder is probably really bad!

Cheers,
David.

Well, I'm very sympathetic. In years of working on coding, I've found that that which is good at 256 kb/s (for stereo) gets used at 128. That which is good at 160kb/s gets used at 96. That which is (barely) good at 128 gets used at 48...

And so on.

And to think the number of times that people told me "in 5 years nobody will need to do any coding".

I recorded some radio on my Humax PVR9200T last night. The Humax lets me transfer the TS (transport stream) file, via USB, to my laptop, which I then load into ProjectX to demux the stream(s).

Here's the results, extrapolated from the ProjectX log:


Files were recorded between 19:07 and 19:45 last night.

A bit off topic, but in the Beeb's defence and for Radio 3 in particular, their website is broadcasting the complete works of Tchaikovsky and Stravinsky this week. With content like that, even the shortcomings of 32kbps RealAudio can be overlooked! (at least, sound quality becomes less of an issue until regular programming resumes) I've been busy time-shifting it all for listening later at leisure.

Having searched for the PVR9200T, listed as a Freeview set-top-box with 160GB hard disc, am I right in assuming that those bitrates refer to the radio channels as broadcast on Freeview (the UK's terrestrial digital TV service) in the South West, rather than another digital TV service such as cable or satellite.

If it is, that's hopeful for reasonable quality (about as good as FM) for the most popular channels all at 192 kbps full stereo.

Incidentally, does anyone have any idea what audio bitrate is used for the major terrestrial TV (video)channels - presumably also MPEG-1 layer II stereo audio alongside MPEG-2 video?

Correct on all accounts.

I tend to mainly take films from Film4 off onto the laptop. I have just checked a few and they are all MPEG-1, Layer2, 48000Hz, stereo, 160kbps. I did have one stream a while back that had two audio streams, but can't remember the bitrates.

BBC1/2/3/CBBC/4/CBeebies (TV) audio on Freeview and satellite is 256kbps mp2 stereo.

Freeview audio and video bitrates are listed here:
http://www.aums30.dsl.pipex.com/home.htm

The various digital radio bitrates on DAB and the digital TV platforms are listed here:
http://www.digitalradiotech.co.uk/dab/digi...o_bit_rates.htm

There are some strange (stupid!) encoder settings in use on some services, and some transcoding going on too - you need to listen to find out which sounds best, rather than just trust the bitrate as an indicator of quality (though I think HA regulars will know that ).

BBC Radio 3 @ 192kbps mp2 on DAB is still the highest quality broadcast of any BBC Radio station. The IP multicasts may be 128kbps AAC, but they're transcoded at present.

The 256kbps TV audio is generally higher quality than all the radio stations (though I suspect there's still transcoding), and of course the 320kbps you can receive in the UK from German radio stations via satellite is much better still (and nothing to do with the BBC!).

Cheers,
David.

Ah, looks like I should have looked around a little first.

Thanks for the tables 2Bdecided, very useful. As someone who burns to DVD the resolution is of particular interest also.

Are you sure? BBC 3 at nighttime is streamed through the Norwegian 'NRK Alltid Klassisk' as Ogg Vorbis @ vbr 172 kbps, which should be noticable better. Are you saying this is transcoded from mp2@192? Btw, all national channels in Norway are streamed as ogg 172.

No, I wasn't aware of that broadcast.

If it is sourced from Radio 3, I have no idea how it's sent over to Norway for broadcast.



Nice!

Cheers,
David.