There's been an interesting discussion about UK digital radio recently.

DAB digital radio has been compromised by the use of low bitrates (e.g. 128kbps mp2, often transcoded!).

Digital Radio via the FreeView TV service had maintained the four main national stations(BBC Radios 1-4) at 192kbps.

For this reason, a lot of people have bought FreeView to enjoy high(er) quality digital radio. However, it turns out that BBC Radio 3, when it's broadcast at 192kbps on DAB, sounds better than BBC Radio 3 at 192kbps on FreeView (DTT).

So, here are two samples, both direct stream captures from the digital broadcasts. They're not quite in sync, and they won't stay in sync even if you sync them. There's also a 0.5dB level difference, so ReplayGaining them is advised. Still, given this, can you ABX them?

The original discussion is in alt.radio.digital - take a look in Google Groups.

I'll upload the DAB version, then the DTT (FreeView, i.e. FV) version.

What do you think they sound like?
Can you ABX them?
What do you think caused the (audible?) difference?

Cheers,
David.

DVB-T version might be a second generation digital copy (e.g. viciously transcoded from DAB to same bitrate). It also could be that the digital signal is extracted at different point from the stream i.e. DVB-T version might have more compressing etc. than DAB version, or the encoder used in DVB-T is crap.
I'm sure that some technician from BBC would be more than delighted to answer this question if you knew the right e-mail address to ask this question.

Possible, but it's unlikely the BBC would choose to source their DTT feeds (in general) from DAB, given the higher DTT bitrate (on the other stations) compared to DAB, and the fact that the other 192kbps stations on DTT are obviously not transcoded from their 128kbps and 80kbps DAB relations! So, there's a separate feed - no transcoding.




The amount of compression (at least on this station) appears to be the same. The encoder certainly could be crap! Radios 1-4 on DTT were the last digital radio services to be launched (DAB 1-5 came first, then DSat, then the newer BBC channels launched across all platforms, and finally Radios 1-4 were added to DTT) - maybe they re-used some older encoders for this, and installed newer encoders on their (gnerally) lower bitrate DAB services? This would certainly make sense.




I know a few people to ask, but they're not keen to talk about BBC digital radio - the BBC have come in for a lot of stick over it - often from me!

(There was another "DAB doesn't sound very good on most stations" quote from me in The Sunday Times this weekend!)


Anyway, the files are up now, so you can listen for yourselves and let me know what you think.

Cheers,
David.

First time I listened to both files (nice music!), they sounded ~the same for me on my laptop+headphone. I suspected something like additionnal ringing with FV_1, but it was a very tiny and fragile feeling.
After replaygaining the files, the sound was much louder (+15 dB). And FV1 obviously suffers from ringing/chirping, especially audible on the beginning (violins). DAB is far from perfection, but sounds better to my ears.

On the middle, both have chirping problems. DAB is nevertheless better again.

=> DAB sounds much better for me, but is not free from artifacts. I must add that quality is not very good for me.

Shostakovich's Leningrad Symphony - recent BBC Radio 3 broadcast.



I found the difference to be subtle, and (like you) more obvious after ReplayGain. I ABXed 8/8 (which surprised me!) and preferred the DAB sample, being more natural. The DTT one is more metallic, because the high frequency cut off is changing, and is always a little too low.


192kbps isn't enough for mp2. So you can imagine what the 128kbps broadcasts sound like!

UK DAB digital radio has been a great advert for FM! If you have a good FM aerial, then BBC Radio 3 FM is still very fine - especially in the evening during live concerts when the dynamic range compression is switched off. (A few years ago, Radio 3 had no DRC at all).


Still, the question remains: why does one 192kbps stream sound worse than another? Different encoders? Settings? Source? etc?

Cheers,
David.

First I had problems to work with that *.mp2 fromat. I had to do an unknown conversion process with winamp, I hope it didn't have much influence on the following.

Why are you guys complaining about the quality ? It is very difficult to judge the quality without direct comparing to a CD version; even 128kbps sounded almost fine to me on some mp3 test I did some time ago without direct comparison. But, compared to German FM (cable) radio, the samples sound far preferable:
-No crackling, I always notice crackling on FM radio, it can be caused by many things, though.
-Very low noise; I noticed some strange noise varying in its loudness together with the signal, though. Perhaps it's already in the original.
-nice dynamics
David, you've chosen a very quiet portion, so maybe at louder parts it sounds worse. The original from CD, that the station used, seems to have very good quality. Also I have not that much experience with classical music, so I don't know what the violins should sound like.

After converting into 48khz/16bit/1411,2kbps-wav I had to downsample to 44,1khz, because my soundcard doesn't like 48khz. I synced the samples sample-precise. It was easy, the sample structures look very similar. The FV version is 0,00102% constantly faster than the DAB version - is it that what you call out of sync, David ?
After wavegaining, I performed an ABX test - I abxed 13/14, p<0,1%. The test was very difficult and stressful, similar to 160kbps mp3<=>original wav. Did you others also synchronize the samples perfectly ? If not, the test is irrelevant. One could have made his decisions because of the time shift.
The difference in sound is so small that a general judgement is difficult; but the DAB version seemed to have less noise (the varying one) to me.

A friend of mine has DVB-Sat, receiving with his computer. Most of the only-audio stations have 192kbps. The sound is very noisy and crackles, almost the same like FM (again it may be already in the originals on CD). Very bad, considering that it's digital broadcasting. We are quite sure that there's no digital=>analog=>digital step during the recording process.
After all, I'd say you can be pretty happy with the quality of your samples, David. What broadcasting system were they actually ? DAB-T and DVB-T ? (instead of cable or sat) And how did you receive and record them ? For the first time, I've now an idea what DAB sounds like.

Because quality is bad.

Not always necessary. I don't need a CD to know how sound a real violin; therefore I don't need a comparison to the CD to accuse the lossy encoding process to be responsible of extra-ringing/warbling/chirping/etc... Simply because no real violin produces this sound, no microphone too...
Same for JPEG: I guess you could detect macro-blocs and then accuse the JPG algorithm to be responsible of them, without any other visual reference.

For you, probably. But other people could perceive things you don't hear

No offence, but I assume you are too fast in judging, guruboolez (as well of me as of the samples). 128kbps mp3 is far away from anything that I would call acceptable. It is enough that there is a difference, audible is not necessary. (with ABXing it's easy for me to tell the difference)
The basic thing is that you can't be sure you just imagine the artifacts you think you are hearing without having a reference. Did you actually perform an ABX test with David's samples ?
Your long-time membership commands respect, though.

Some kind of artifacts are not coming from imagination. You could be sure that they exists, without ABX and without the exact reference for comparison.

Could you see something comparable to compression problems in the following picture?


If yes, do you really think that you're dreaming? Or do you think that's the cathedral which suffers from macroblocs? Or couldn't you be sure that the photo is heavily compressed?

Well, if you, guruboolez, hear mostly classical
-you are more familiar to this music style and you rather know what it should sound like (the violins)
-you may be familiar to a generally higher quality of sound, which causes your judgement to be more severe.

The picture seems to have some parts that are not keen, but I wouldn't know the reason (you said it) and even with a assumption, I couldn't be sure of it. Anyway, I would probably say "Something is wrong with the picture."

I would, even without explanations know what is wrong with this picture: it has been passed through a block dct based compression scheme, with too much high frequency dc coefficients removed. Some macroblocks are exhibiting this clearly.

In the same way, I think that for heavy/bad audio compression, a trained person could hear wrong parts without any reference.

The DAB sample is DAB-T - quite common in the UK now (about 500,000 radios sold, about 85% of the population covered).

The FV sample is FreeView, i.e. DVB-T - very common in the UK now (about 3 millions STBs sold, 75% of the population covered).


They're not bad samples - if you want bad DAB samples, I'm sure I've posted some somewhere...

http://www.hydrogenaudio.org/forums/index....showtopic=19964

...they're beautiful!

EDIT: The first samples came from other people, but they, like me, are using direct stream capturing using a PC card. I used a Psion Wavefinder for DAB. There are various DVB-T and -S PC cards available that allow the same thing - for video too (if the source isn't encrypted - and sometimes if it is).

Cheers,
David.

Sounds like blade People are paying for that?! FM radio have maybe terrible flaws (noise, crackling, pops...) but it's still a "musical" signal. At least for me...

DAB is free-to-air, so no, they're not paying for that specifically!

The people who are buying DAB radios are buying them to listen to new services which are not available on FM. For the general consumer, some notice the quality is worse than the best that FM can deliver, while others notice that it's better than the worst that FM can deliver! The broadcasters have discretely dropped "near CD quality" from adverts because there were complaints. They've replaced it with the meaningless "digital quality sound" - which most people assume means "CD quality sound" until they hear DAB digital radio!!! For various reasons, sales of DAB digital radios have been pathetic - especially compared to successful products like DVD! Still, sales are rising slowly as prices drop, and the industry appears to favour low audio bitrates+quality, lots of stations, and no new spectrum to allow higher bitrates+quality in case it increases competition!!!


However, these new services are also available via the digital TV platforms in the UK, and listening via these is _much_ more popular. There are no separate listening figures, but my guess is about 5-10 times more people listen to digital radio via a digital TV platform than via digital radio - this doesn't meant they listen _through_ their TV - you can easily connect your STB to your hi-fi, and many people do.

That is what started this thread: It's cheaper to buy an STB to get digital TV _and_ radio than to buy a DAB digital radio tuner. The audio quality of the radio stations, generally, is better on the digital TV platform because the bitrates are higher. However, the samples which started this thread are an example of the only station which uses high bitrates on both platforms - and it sounds worse via digital TV.

Personally, I'd like to see all BBC radio stations at 256kbps on digital satellite - they already broadcast over 20 TV channels (at ~5Mbps each - which itself is too low for some content), so 11 radio stations at 256kbps is nothing to them in bandwidth terms. Yet their "premier" radio stations remain at 192kbps on DSat and lesser ones are as low as 128kbps. Still, it's better than the 64kbps mono they sometimes use on DAB! Yes, 64-80kbps mono on DAB for some drama, plays and comedy while the same content is broadcast at 160-192kbps in stereo on other platforms. DAB is something of a joke in terms of sound quality, and there is little sign of this changing.

Cheers,
David.

P.S. sorry for the rant, but the poor state of UK digital radio is enough to make anyone interested in audio quality complain!

I have to agree with a couple of things that precisionist said, in that, without hearing the source CD, it's hard to say whether or not the artifacting I think I can hear in the DAB version isn't a result of a poorly mastered CD (I do realise how unlikely this is though ). That being said, I think the FV is rubbish. That chirping on the violins is nasty and I think it's also audible (not anywhere near as bad) on the DAB version, but I don't find the DAB version particularly annoying*. I occasionally listen to classical, but it's not something I've ever tested a great deal for artifacting with lossy compression.
I've attached a picture of my ABX results, after replaygaining. I got 16/16 without any trouble at all by just concentrating on the chirping of the violins in the first few seconds.


On my ABX tests I didn't switch between tracks, during playback, once.

edit:

* perhaps I would if I was more familiar with the original piece

ABX 16/16
The encoding noise sound "white noisish" on the dab version, and "pure tonish" in the FV version. It seems that different encoders or different settings were used in order to compress the streams.

I confirm a speed difference of 0.001 %. This is the order of magnitude of the difference between two crystal oscillators. For example the clock references of my Yahama CD Player and my Sony DAT deck are 0.002 % apart from each other.

It means that either the signal went into a digital->analog->digital conversion, either through a resampling digital input (not slaved to its source). The later can happen with such devices as AC97 soundcards, with the sample rate fixed to 48 kHz (and these files are 48 kHz !), or with devices using asynchronous sample rate conversion on their input in order to completely eliminate line transmission jitter.

One of the broadcasts (or the two) probably uses a digital mixer, which resamples everything at 48 kHz, and causes the speed shift.

Thanks for the results.

So, no one prefers the FreeView version then?

Well done to the sensitive HA listeners. I wonder what percentage of the general population would be able to detect the difference?


I agree with your suggestions Pio2001, though I'd hope the BBC are not using AC-97 sound cards in their broadcast chain! I know a few broadcasters who do, but hopefully BBC Radio 3 has got more sense!

Cheers,
David.

The thing that guruboolez originally called chirping is probably what I called noise...


You may still have been influenced (not knowingly) by the different starting positions.
On my ABX test (like generally on my ABX tests) I've chosen a single 1or 2sec-beat and "click-click-click" compared.

I used to feel sad about German digital radio (DAB), lacking success. Digital radio and TV is not very common here (perhaps except DVB-S). People think digital=pay. But since I know that it's all pretty lossy in any case, I don't worry that much any more.

Please, allow me to reply that in many cases, an objective proof can be found.

In this case, using some differential edge processing the 8x8 squares become obvious:

ABX 15/16, DAB is less, ahem, woody, than the FV one... Although I'd have never noticed the difference in casual listening. Maybe I would have with some Trance or Aphex Twin, who knows?

Very interesting thread, btw... Is there an equivalent one on Doom9? Because the quality of Digital TV that I see... Well, it defies all belief. Macabre.

If you search the forums at digitalspy.co.uk or uk.tech.digital-tv you'll find plenty of threads.

Cheers,
David.

Sorry to take so long getting here. It was I who sent the original 1 minutes samples to David. I, in turn, had been sent the whole concert (it was a BBC Radio 3 recorded concert - Sir Edward Downes birthday concert, in fact) by another contributer to the Classical Music on Radio 3 Message Board. There was a six minute drop-out in the DAB broadcast, and I was interested in hearing the 'Leningrad' complete, so I was sent the raw MP2 of the Freeview version of the broadcast, along with a Wavefinder received MP2 of the DAB broadcast, which, as it happens, I already had from my own Wavefinder. So far, it appears that e3veryone who has heard the two samples has been happier withthe DAB sound.

I have little confidence my chances of getting an informative reply from the BBC these days, re. the basis for the difference in sound, but I will try.

I didn't say this. Without the original I can't tell if DAB is free of artifacts or not. There seem to be like a constant noise in the Dab version, that may screw up transients if there were some.

Listened to the 'bad' samples. They are very annoying, worse than FM. Are they from the same station ? It seems to have some ugly bass-boosting dynamics processing. I also noticed some typical mp3-ringing; without knowing the source of the samples and without auCdtect I would assume that the they had been mp3.
Did you replaygain/change the volume ? I'm surprised the samples aren't at maximum volume.

They're from two different pop stations on the UK national commercial DAB multiplex ("Digital1"). They're direct stream captures, no processing, no ReplayGain.

It's not uncommon to find heavy DRC, limiting, and audible clipping even though the decoded output peaks far below digital full scale. There's a simple cause: complete incompetence.

There are so few DAB listeners that it seems broadcasters just can't be bothered with the most basic quality control. I wonder if most of these stations even have an "engineer" - it's probably a "tea boy/engineer"!

Cheers,
David.

But the station could be much louder without even a shadow of lowering the quality!
In fact, on most clipressed album CDs the maximum samples already aren't +/-32768; the sample in my "clipped sample and its declipped version"-thread is a good example. Clipping occurs only at 0dB, but then they run some other processing; DRC, EQ, normalizing down, etc.

Just to add a little bit to this post, this webpage
http://www.digitalradiotech.co.uk/
seems to be using a Listening Test table I seen somewhere before to make their point.

That's my fault - I pointed Steve (a/k/a "DAB sounds worse than FM" a/k/a digitalradiotech author) to Roberto's excellent site.

I've also pointed the ASA there...

http://groups.google.com/groups?hl=en&lr=&...ting.google.com

Cheers,
David.