'HD' audio web stream - why is Radio 3 setting levels so low??

[Guest]

Anyone have any ideas on this?? Radio launched their 'High Definition' audio web stream last year, with audio in AAC at 320kbps, which is very high quality by normal broadcast standards. But then they use such low levels on all their broadcasts, at least of the Proms. Peaks typically occur at -10dB (-8dB if you're lucky), and I'm referring here to peak transients in 'loud' and large-scale orchestral works. Why launch a so-called 'high definition' audio stream but then set levels such that you're making use of only a portion of the resolution and dynamic range that's available? It seems self defeating.

Elion

[Bryn]

How are you making these measurements, Elion? Though there was severe a problem with the audio in a couple of the early Proms, in recent times the dynamic range has been pretty wide, with one or two Proms giving peaks within 3dB of saturation level. In general they appear to aim for around 6dB of safety headroom to reduce the possibility of clipping. What I am critical about is the way, all too often, the peaks during the presentation chatter are higher than any during the music, but that is, I suppose, more predictable than transients during the performance itself.

[Chris Newman]

It is a constant complaint with digital and HD for many Prom concerts. Mysteriously Wednesdays come off much better for the difference between peaks and troughs. The announcers at the end of a concert are often higher to my ears that the climax of big symphonies!! The powers that be argue that there is a limit to the bandwidth they are allowed to use. We must keep complaining to Radio 3. The general sound was better in the 1960s and 1970s.

[Chris Newman]

Hi Bryn, we agree about the presenters. Last night the Mahler suddenly flattened out as the last movement approached its end. The moment I guess when the brass stood up. Even if the concert is not televised the guys in the SIS and BBC vans have a TV link to the stage. I have watched them. They usually chicken out if they see anyone move. Yet they are inconsistent: the volume was allowed to keep going right up in the jolly audience choice party an hour later. I do not have anything to measure peaks and troughs on so I go by my ears which my doc says are very good.

[Elion]

How are you making these measurements, Elion? Though there was severe a problem with the audio in a couple of the early Proms, in recent times the dynamic range has been pretty wide, with one or two Proms giving peaks within 3dB of saturation level. In general they appear to aim for around 6dB of safety headroom to reduce the possibility of clipping. What I am critical about is the way, all too often, the peaks during the presentation chatter are higher than any during the music, but that is, I suppose, more predictable than transients during the performance itself.

(i) Computer converting the AAC/320 to wav in digital domain in such a way as to leave levels unchanged. Output (digital) to DAT recorder and CD recorder give same measurement of peaks and 'margin'. (ii) Friend who was PC software suite that allows for direct measurement of peaks from the AAC/320 stream - this gives the same results at (i).

The engineers use the rehearsals to position mics, get the mix they want and then set the levels. Obviously, some degree of safety margin is wise given that adrenaline usually leads musicians to produce more dB live in front of an audience than in rehearsal (a well documented effect), but it's the degree of margin that perplexes me. I haven't come across any Proms performances that have had peaks at -3dB. I've probably listened to far fewer than you have, but the ones that I have heard have included major romantic/late-romantic orchestral works with copious brass/percussion etc.

Elion

[johnb]

Elion,

When measuring the peak levels are you using iPlayer as the source of the audio? If so, have you set the iPlayer volume slider to maximum? (You need to do that to avoid the iPlayer attenuating the sound.)

I agree with Bryn - the peaks are usually -6dB, e.g. last night's Mahler 1 was -4.9 dB and tonight's Kodaly was -5.6 dB.

Even so the ambient noise level in the hall during a silent passage in each work was -59 dB RMS and -66 dB RMS respectively (sampling other 'silences' would almost certainly produce different results) which gives quite a respectable dynamic range.

[Bryn]

I managed to miss the start of the Elgar tonight, but for the rest of the concert, the highest HD Sound peak I have found during speech was -5.3dB (left channel), and during the music I found a -5.5dB (again the left channel) during the Kodaly. Pretty much as usual. The RMS levels of the left and right channels were within 0.4dB of each other, with a bias towards the left.

[Elion]

[QUOTE=johnb;80550]Elion,

When measuring the peak levels are you using iPlayer as the source of the audio? If so, have you set the iPlayer volume slider to maximum? (You need to do that to avoid the iPlayer attenuating the sound.)

That explains it. Thanks for your help. I don't know whether I was being dim but I was clicking on the 'Listen in HD' icon on the Radio 3 program page, then on the 'Click to play' icon that comes up with the HD audio window. I'd noticed the feint grey volume symbol in the lower right of that window but had assumed this was for muting the sound (given its feintness). In fact, the default appears for it to be set at 9/11. So now I've increased it to the maximum of 11/11 (how odd for them to use a scale of 1 to 11 and to have the default set at 9...why would they assume listeners would want to attenuate the volume in this way when every PC and Mac has umpteen easy controls for volume??). With it set to 11/11, doubtless I'd have achieved the same peak levels as you. Nevertheless, I feel a margin of say 5dB still to be playing unnecessarily safe...there are now very clever psycho-acoustic algorithms to manage clipping in the rare event that, say, a timpani crash hits 0dB - these don't affect the sound in any way until 0dB is reached and, at that point, they re-shape in a way that gives the impression that no clipping has taken place. In so doing, they allow an engineer to be more aggressive in setting levels. The upside is that they allow listeners to benefit from greater resolution and dynamic range. The downside is that, perhaps once or twice in a Proms season, a cymbal or timpani crash might invoke a few milliseconds of this type of clipping...but nobody would hear it on those one or two occasions.

Elion

[Bryn]

I think 0 to 11 has become the standard since Spinal Tap.

[Elion]

I think 0 to 11 has become the standard since Spinal Tap.

Bryn - sorry to mention this in the public forum but I didn't receive a reply to the second e-mail that I sent you on 10.8; I did receive an initial reply to my first e-mail but then my second e-mail (which was sent to both the e-mail address you gave me and as a PM to you at Radio 3 forums) didn't draw a response. Doubtless various gremlins got involved. But I now have something that I'd like to send you and that I think you might find of interest - a screen shot that establishes beyond any doubt that Radio 3 were applying hard limiting at -11dB in the Liszt 2 in the First Night Prom (so you were entirely right about this). You may want to use this screen shot (which derives from some respected software) to send to the BBC - I don't see how they could argue with such evidence, and would then need to explain why they saw fit to apply hard limiting at -11dB to the HD AAC/320kbps broadcast stream on that occasion.

Elion

[johnb]

Elion,

You might be interested in these threads:

Prom 1 - HD Sound Severely Limited/Clipped

BBC Reply to complaint about limiting on Prom 1 sound

The excessing limiting (during the music only) was almost certainly a test run for the Havergal Brian Prom a few days later, where the same thing happened.

[Elion]

Elion,

You might be interested in these threads:

Prom 1 - HD Sound Severely Limited/Clipped

BBC Reply to complaint about limiting on Prom 1 sound

The excessing limiting (during the music only) was almost certainly a test run for the Havergal Brian Prom a few days later, where the same thing happened.

Thank you for pointing this out - I hadn't been aware of either thread. Depressing that the BBC should be incompetent and/or disingenuous in their response (when the facts are undeniable from the various dB graphs). I assume no further response has been received from the BBC on this?

Elion

[Bryn]

Bryn - sorry to mention this in the public forum but I didn't receive a reply to the second e-mail that I sent you on 10.8; I did receive an initial reply to my first e-mail but then my second e-mail (which was sent to both the e-mail address you gave me and as a PM to you at Radio 3 forums) didn't draw a response. Doubtless various gremlins got involved. But I now have something that I'd like to send you and that I think you might find of interest - a screen shot that establishes beyond any doubt that Radio 3 were applying hard limiting at -11dB in the Liszt 2 in the First Night Prom (so you were entirely right about this). You may want to use this screen shot (which derives from some respected software) to send to the BBC - I don't see how they could argue with such evidence, and would then need to explain why they saw fit to apply hard limiting at -11dB to the HD AAC/320kbps broadcast stream on that occasion.

Elion

Sorry Elion, somehow I missed you email and message at the time. I have just found and read it. I think it might be useful for you to post the comments on the aac codec and the levels it 'expects' to deal with here. I was certainly not aware of the important issues you raised in that message. I would be most grateful if you were able to reference the information. It appears to be something that the iPlayer team need to be challenged on.