Prom 32 - Audio Limiting (again) !

[johnb]

First, what effect on the sound will this limiting have? Will it be a form of distortion? Will it be very noticeable to the ear if you have not seen the waveform?

IMO you won't hear actual distortion but the dynamic range will seem wrong - in the same way as the dynamic range in R3 FM seems 'wrong' in comparison to CDs (or DAB/Freeview/iPlayer). Also the transients (timps, percussion, etc) will have a softer impact.)

Second, just wondering what software you use to look at the waveform. If I bring it up in Audacity, will that waveform be the same as the one you are looking at? The vertical axis in Audacity seems to be calibrated differently.

The audio was taken from iPlayer LA. If you looked at the same file in Audacity it would look pretty much the same. (I used Sound Forge.)

[Bryn]

Johnb

The top plot in your interesting post 6 is (excluding the last bit from the studio) exactly what I get on Cool Edit Pro (now Audacity I think) ...

More like Adobe Audition, I think. Audacity is freeware. Audition is anything but (though you can get a free trial version).

[johnb]

The top plot in your interesting post 6 is (excluding the last bit from the studio) exactly what I get on Cool Edit Pro (now Audacity I think) but I had always understood that that was the result of the Optimod sound compression system in operation on the FM broadcast. The lower plot is the sort of waveform I see from CDs, LPs etc.

You are right that the top graphic is very similar to the effect produced by Optimod - which, presumably is because a similar dynamic range compression/limiting was employed for the Bernstein Mass, but it seems to have been employed at source, (presumably) affecting all delivery 'modes'. Optimod, as we know it, is only used for FM - not for DAB, Freeview, Satellite or iPlayer. Both of the examples I used were taken from iPlayer LA.

I was under the impression that the Optimod system was on 24 hours a day. But, as I type just listening after the prom, and my Cool Edit waveform is pretty similar to your lower plot, so, in my mind, the Optimod has been turned off.

The Optimod parameters are probably tweaked for different times of day but I thought that it was always in operation (though I might be wrong). Its effect is worse for orchestral music but much less noticeable with chamber music or speech

Of course, I presume the Optimod limiting can be progressive as users of Audacity and its earlier names can easily control the sound volume input from a preamp by use of the Line In recording slider, the resulting waveform showing levels and any distortion, and adjust as required.

As I understand it, Optimod (and similar systems) derive from the need of pop music broadcasters to limit the dynamic range to a narrow band and to push the levels up to the maximum possible and automatically adjust the levels for the tracks in order to give the station 'impact'. The version used by R3 will have been specified for classical music, of course, and has a more subtle effect. But it is the same idea in principal. It destroys the dynamic relationships between ff and pp passages. On the other hand, to an extent our brains 'adjust' what we hear - we can tell that a passage for full orchestral is fortissimo by the timbre, the tonal quality, of the orchestral. So if the timbre of the orchestra suggests it is getting louder our brains believe it actually is getting louder, even when there is little or no change in the actual dB level.

How does one get a double dose of Optimod? By at source, do you mean the recording venue? I seem to remember from a series of posts with Jonathan Swain a couple of years ago that the BBC studio monitors broadcast the sound before the Optimod compression got to it.

I believe that dynamic range compression/limiting was probably done at source (or early in the chain) for the Bernstein Mass - certainly before the Optimod is applied to FM. Therefore Optimod would have been applied over the top of the existing dynamic range compression.

[johnb]

PS This is an example from a few years ago, comparing FM and DAB of the same broadcast of Haitink's recording of Shostakovich 12 broadcast. It clearly shows how Optimod mangles the relative dynamics. (The dynamics of the DAB broadcast were very similar indeed to those of the actual CD.)

[OldTechie]

Just a few thoughts..

OPTIMOD on FM
I'm sure the Optimod would stay in place on the FM feed. It's not just there because someone wanted to mangle the signal for fun. It is there to mangle the signal automatically so that it does not have to be pre-mangled by the sound engineer.

The FM system has pre-emphasis. So starting around 3 kHz the transmitted signal doubles for every octave above that. The reason is that in general music has much less high frequency content than low frequency content, and by boosting the generally low amplitude high frequncies on transmission and then reducing them on reception, the background noise is also reduced on reception. However, it means that it is impossible to transmit high frequency full amplitude sources without breaking all the transmission rules (or causing smoke to be emitted from the transmitter.) This is 1950's audio data compression!

Before the days of the Optimod, the pre-emphasised signal hit a hard limiter at the transmitter (there is probably still one there as a backstop.) So if someone hit the triangle very hard, the rest of the signal would be horribly attenuated for a short time.

Since the NICAM distibution went in, they have been using an Optimod, and it is capable of being set up much more cleverly to avoid the worst effects. Without it, we would be back to nasty really hard limiting triggered by various instruments.

Before Optimod the sound operators would compress the dynamic range manually. Good operators could hide this rather well. They would follow the score, and seeing a crescendo coming, they would gently reduce the level in advance so that the musical effect would not be lost too much. But they would not allow the signal to fall too low for too long because it would disappear below the noise for listeners towards the edges of the service areas. Also many people complain if there is too much dynamic range because of their listening conditions. Now the operator can leave most of the dynamic range in place and let the Optimod fix the level for car radios, portables and distant reception.

DAB has a control chain that allows the broadcaster to send dynamic range compression information so that the receiver can compress it just as the Optimod would (but without the need to handle pre-emphasis.) Some receivers allow the user to turn that on or off. I guess portables always use the compression.

I would prefer them to turn off all but the pre-emphasis features of the Optimod on FM, but I'm sure that would annoy many more listeners than it pleased (and it still would not work for me because of hum inserted at the local relay station.)

Compression at Source
I have a theory about how this may happen. I wonder whether anyone here knows the real facts about the operation at the RAH to comment on it:

I have not heard this particular transmission, but the Havergal Brian one I did listen to. It is normal to compress the presenter's voice and this is always audible (at least it always annoys me.) On the Havergal Brian transmission, the presenter's voice was definitely not compressed but the music was.

My uninformed theory is that there may be two stages at the RAH. The perfectly balanced music mix is probably fed into a second desk where the presenter's voice is added. The operator of the second desk is primarily concerned with cueing the presenter, contacting continuity and the stage manager etc. So if someone sets that up with the compressor in the wrong channel it will all go wrong. The person operating the second desk will not be listening critically to the mix, and the person doing the music mix will be listening before the compressor. The compressor intended for the voice would be set to compress whatever it gets to the normal speech level (peaking around 4-6dB below the maximum.)

[johnb]

Of course, I presume the Optimod limiting can be progressive as users of Audacity and its earlier names can easily control the sound volume input from a preamp by use of the Line In recording slider, the resulting waveform showing levels and any distortion, and adjust as required.

r&f,
I should have mentioned that Optimod doesn't work like a volume control, where the all the sound at the same level is treated exactly the same. Optimod employs a complex dynamic process. It looks ahead at the audio data and makes decisions about what to do based on that information. As you can see from the above comparison of FM and DAB/CD of the same recording the net effect is often to grossly distort the dynamic relationships in the music. This happens on the micro as well as on the macro scale.

Old Techie,

You might be right in your theory of how the compression/limiting occurred.

But I don't believe it is a coincidence that both the Havergal Brian and the Berstein Mass were Proms where there were unusually large dynamic ranges (JLW mentioned that she had read that amplification was involved in the Bernstein). I have no doubt that the compression/limiting that was employed for the Havergal Brian concert was deliberate - and it is arguable that the engineers were wise to do that. My gripe on that occasion was that it seemed to have been done rather crudely.

[johnb]

Going back some years I happened upon a good example of how Optimod can distort the dynamics over just a few bars. This is taken from a trail that was running at the time. It used a few bars of Handel's Zadok, which is useful because we can all hear it in our heads. (As I said earlier, our brains tend to compensate for the mangling of the dynamics, basing its perception on the timbre and texture of the sound as well as what we actually hear.)



By the way, I used DAB in the examples only because, at the time I did the comparisons, the iPlayer quality was extremely poor indeed.

(The graphical representation was taken from Total Recorder's display.)

Edit. Apologies for rambling on. (Sometimes hobby horses gallop away uncontrollably.)

[Serial_Apologist]

(As I said earlier, our brains tend to compensate for the mangling of the dynamics, basing its perception on the timbre and texture of the sound as well as what we actually hear.)

Sorry johnb, not in my case - and I'm not at all sure that I'm uniquely frustrated when, as the climax builds, the music seems to drift further and further away.

Optimax seems technologically like two steps forward, three steps back; I have reel-to-reels of pre stereo R3 broadcasts from the late 1960s when compression does not seem to have been applied to anything like the degree it is today; they have greater clarity, even at top climaxes, and no distortion.

[rank_and_file]

More like Adobe Audition, I think. Audacity is freeware. Audition is anything but (though you can get a free trial version).

Woops. Yes you are quite correct. Thanks.

[Gordon]

[QUOTE=VodkaDilc;195497]So the over-powering climax which several of us have commented on towards the end of the Mass will be lost? Which will be the way of listening (FM. DAB radio, various TV systems) where we might hear it most "truthfully"?
QUOTE]

You won't, it's done at source. It does mean however that the dreaded Optimod will not be called upon to add more misery on top. I si necessary to protect the distribution chain and especially the FM transmitters from overload but what they have done again here is plain ridiculous. There are crowbars on the transmitters to deal with the worst effects of excess input level anyway.

I think that jb had it right as to the cause of this lazy production. Once upon a time there were enough people around to watch the meters and take care during rehearsal to mark up the points of concern so that they could be dealt with during live transmission. And they do make it worse for themselves by employing vast numbers of microphones which makes it even more important to watch the meters. Now there are probably too few people around to do that and it is left to automation to catch the peaks. In this case the limits are set far too aggressively. Lack of resource [ie cut budgets] and lack of attention to to detail. It is true that the issue is worse with large scale works but isn't that what professionals are for!??

[Gordon]

I have reel-to-reels of pre stereo R3 broadcasts from the late 1960s when compression does not seem to have been applied to anything like the degree it is today; they have greater clarity, even at top climaxes, and no distortion.

In those days engineers ran the network and well trained people watched the meters and rode the faders sensibly for difficult pieces - as did recording companies by the way during the analogue era. And R3 was not tailored for in car and portable listening either [Optimod]. I also have many recordings from that same period and can concur with your assessment, despite the greater noise levels and the more constricted bandwidth. Multimicing was also less prevalent back then when typically a single coincident pair with the odd spot or two were used in the RAH for most Proms.

[rank_and_file]

johnb

Thanks for your comprehensive replies.

The Shostakovich FM/Optimod and DAB waveforms prove a point - two lousy, even lossy, systems. I just wonder if the FM waveform looks “thicker” (for want of a better word) because all the quiet music is also being increased - to help car drivers hear quieter bits over the tyre noise!

Intrigued that Optimod looks ahead. Must be very responsive and quick, although I suppose the answer is that there must be a delay of x seconds for it to do its stuff before actual transmission.

I think Gordon in post 26 sums up the decline in the BBC’s once very high reputation for sound engineering to what we suffer now. Having said that, the iPlayer HD stream is getting very close to CD levels, with a good internet connection I know how I shall listen to R3 in future.

I was looking at the waveform of the afternoon prom repeat, and in the line in setting I used I am pretty sure that the Optimod system is on. Nothing got past 3db and whilst a forte passage with 60 odd musicians playing got up to the 3db, so did the presenters normal speaking voice at the end. Infuriating.

[Dave2002]

I believe that in those days they even employed musicians, or at least musically aware engineers, to look at the score and identify areas for special treatment to be carried out in real time beforehand.

[johnb]

r&f,

The "thickness" of the line is giving an indication of the fluctuations in the level which cannot be shown because 45 minutes worth of music is being squeezed into approx 6" - nothing more.

I'm not sure what you mean when you say the "Shostakovich FM/Optimod and DAB waveforms prove a point - two lousy, even lossy, systems". What is shown has nothing to do with the tonal veracity or lossy vs non-lossy qualities. It is purely about dynamic range compression (which is totally different from the data compression used in lossy formats).

About the prom repeat you are looking at - what was the source? Just because the peaks are -3dB doesn't indicate that Optimod has been employed - the engineers leave considerable headroom. From memory the peaks on DAB used to be approx -3dB and on iPlayer they can be anything between -10dB and -2dB!

[Dave2002]

Friday - R3 at 2pm
Thursday 6 Sept - BBC4 at 7.30

Information taken from the concert programme. I find that information about the afternoon repeats is not always easily accessible.

Don't miss it - stunning stuff!

Thanks.

A clever broadcasting company, having recorded the concert with as wide a dynamic range as the equipment could allow, could avoid most of the problems on repeat broadcasts and transmit with a much wider dynamic range. There would be time to do this between the first broadcast and the repeats.