Prom 1 - HD Sound Severely Limited/Clipped

[johnb]

I thoroughly enjoyed Janacek's glorious Glagolitic Mass.

BUT

When I look at the HD version of the concert in Sound Forge I see that the music was severely limited/clipped at ~-12dB. Curiously the announcements and the interval feature weren't limited. As the announcements weren't limited I assume that the limiting was done by the sound "engineers" on site or at Broadcasting House and has nothing to do with the iPlayer/HD itself.

The sound quality of the HD player has usually been very good indeed so I hope this isn't a portent of what we can expect from the rest of the Proms.

[Stunsworth]

John, the tv sound - BBC HD channel - sounded fine, and was in 5.1 surround sound.Usually the radio sound is 'limited' to allow for unexpected peaks (as I'm sure you know). Could it have been something similar with the online HD sound?

I see you've posted a screenshot. Isn't the limit more like -4 db?

[johnb]

Steve,

Radio 3 has been recording/broadcasting live concerts since the days of yore. Recently, as we all know, all the evening concerts have been live - but this kind of thing is new. Perhaps there is some new sound engineer in charge of the Proms. It smacks of incompetence.

(Up until the middle of the Proms last year the iPlayer stream was always permanently limited/clipped but that was a constant 'feature' do to with incorrect settings. This is something different.)

[Bryn]

I thoroughly enjoyed Janacek's glorious Glagolitic Mass.

BUT

When I look at the HD version of the concert in Sound Forge I see that the music was severely limited/clipped at ~-12dB. Curiously the announcements and the interval feature weren't limited. As the announcements weren't limited I assume that the limiting was done by the sound "engineers" on site or at Broadcasting House and has nothing to do with the iPlayer/HD itself.

The sound quality of the HD player has usually been very good indeed so I hope this isn't a portent of what we can expect from the rest of the Proms.

That really is disgraceful incompetence on the part of the engineers used by the Beeb for this job. It calls far a compensation claim to be launched by RW I would say.

[Stunsworth]

John, apologies, I didn't see the timescale for the metering. I agree, it looks clipped during the concert.Sorry for my mistake.

[Bryn]

I have just checked out the on demand (192kbps aac) version on the iPlayer. It has the same severely compressed dynamics as the HD Sound version did. Neither version, however, is actually clipped. The excessively heavy limiting prevented that.

[PJPJ]

I recorded from the iplayer HD....

Here's the first 10 mins of the Liszt.



Uploaded with ImageShack.us

[Bryn]

There's something wrong there. The HD Sound stream uses 44.1kHz sampling, not 48kHz as shown on your image.

[The results I get by extracting the unaltered 320kbps aac-lc file from the FLV stream made available live as 'HD Sound' by the iPlayer, resolving that back to 44.1kHz sample rate 16 bit quantisation PCM and using the same software as johnb to display it give exactly the same image with the tell-tale 'flat top' sections graphically revealing the severe dynamic limiting applied by the engineer(s) hired by the Beeb (it's no longer an in house operation).]

[[Just a though. When you wrote "iPlayer HD ...", were you referring to Radio 3's HD Sound Stream or BBC television HD's audio aspect? BBC2's 256kbps mp2 stream does indeed use a sample rate of 48kHz.]]

[Bryn]

I have just checked the results from the BBC2 256kbps mp2. They are pretty much identical to those shown the image of the HD Sound profile posted by johnb. It would seem that the dynamic limiting was imposed at the mixing desk. Not exactly professional.

[PJPJ]

There's something wrong there. The HD Sound stream uses 44.1kHz sampling, not 48kHz as shown on your image.

That's because the recorder was set up to convert to flac at 48kHz. It's the HD stream recorded into Goldwave......

[Serial_Apologist]

If this sort of "post-production" meddling with dynamics is practised on The Gothic on Sunday, it will make a total mockery of the work...

[Gordon]

I It would seem that the dynamic limiting was imposed at the mixing desk. Not exactly professional.

This is appalling audio engineering!! It does look like a mixer desk limiter set far too low and too aggessively too. A clipper would distort audibly and is there as a last resort transmitter protection mechanism.

The strange thing is the annoucements and interval excerpts. These could be studio sourced in the case of the interval and presentation announcements but local RAH announcers would have a channel into the desk. Maybe the desk allows input grouping thus separating processing the music from the speech? Maybe the channels were swapped so that the limiting was meant to apply to the speech!

Continuity at BH should have seen this on their level meters but perhaps they weren't looking or perhaps they don't use level indicators in this digital age. On the old BBC peak reading meters this -12dB level limiting would have been obvious which is why they were there.

[Serial_Apologist]

Perhaps some clever dick thinks, hmm, one person announcing is less loud than an orchestra.

Why for goodness sake don't they just lower the levels for quieter passages and leave louder passage inputs as they are - thus leaving it to the listener to decide when his or her neighbours arer deserving of equal consideration? This used to be the case for certain. I have many reel-to-reel recordings off R3 from the 1960s to the late 1980s that prove it! Back then, broadcast decibel ranges were far broader, no question; and you appreciated and listened more attentively to pianissimo and adjusted the volume control as necessary: the Choice Was Yours.

S-A

[johnb]

PJPJ,

The 10 minutes you selected isn't the best section to illustrate the limiting but there is something very strange going on.

When I saw your illustration I realised that there were noticeable differences to what I had in Sound Forge. I was so concerned that I downloaded Audacity to see how my file appeared in that - but it was the same as with Sound Forge (of course). I actually captured the concert two different ways (simultaneously): capturing the 'native' aac stream (in its flv 'container') and also capturing the audio via Total Recorder (so I can detect if there any dropouts in the aac stream). Both versions show exactly the same limiting, etc.

This is the same passage that you displayed, using Audacity to make comparisons easier:



As you can see there are significant differences:

- The level of your audio is higher. Presumably somewhere in your processing gain was added.

- Your version has far, far more transients, etc.

Three possibilities:

1) There is something wrong with the audio file that I have. But that is derived from the original aac stream, which should mean that it is about as accurate as it could be. Also the second version captured by Total Recorder (no gain, no resampling, saved as PCM 44.1/16) agrees with the aac stream file.

2) The process you used to capture the audio was (cough) less than ideal. (Actually it would have to be pretty bad for the differences to be so very apparent by sight.)

One of the problems with capturing the audio with various software packages is that you are never sure at what point in the Software->Windows Mixer->Soundcard->Analogue Output chain the audio is picked up. The very worst is after the PC soundcard, the ideal is to capture it in the digital domain between the software (i.e. iPlayer) and the Windows Mixer. (In XP Total Recorder can pick up the digital output from the software. In Windows 7 that is problematic and it picks up the digital output of the Mixer.)

3) Both of our audio files are inaccurate!

[Serial_Apologist]

Sorry to seem thick, being non-technical, but am I right in thinking that the peaks on those graphs represent points in the broadcast when announcements were being made?