is there any contemporary reason for leaving headroom on a track???

[Doug1978]

...

you'll never convince all of 'em!

[Compyfox]

I guess not...

[4damind]

There is no point on having headroom on a finished master.

This is not true. As you explained it by yourself this headroom is good to avoid intersample peaks (often it can also be avoided with higher oversampling). Besides ISP some Mp3 converters having a problem if it goes near 0dB. There is (or was) also a difference between different implementations eg. Lame vs the Fraunhofer reference codec. So it was always a good hint to leave some space... with testings I made (but some years ago) a good value was between -0,4 to -0,6. Other MEs came with similar testings to a bit higher values (Afaik -0,8dB)
But we don't speak about 6dB headroom, so I think this means a not mastered mix. Anyway, I cannot remember that MEs asking for such a big headroom.

Btw. I've heard also about problems with some D/A converters on cheaper CD players if the mastered audio had no headroom. But I have no idea if this true or a known behaviour.

[Edit]
I see you mean headroom _after_ the finished master... so yes, you are right for this. The finished master doesn't need an additional headroom, this makes no sense.

[PurpleSunray]

The "System Mixer" is what I would actually consider as "post master editing" (not the bitstream decoder).

But isn't that written in meta tags and then read out on the fly, according to the setup of the player? It's is still post editing, no doubt. But not destructive to my knowledge.

In fact, i did mess around with various AC3 CODECs, and they can only take a certain average signal strength per channel. In Stereo, it was only about -12dB RMS. Everything else happens "post" encoding.


Same with MP3 - I still don't know why this is happening (I'm not a nut at this, CODECs are not my pet peeve), but try to encode a -8dB RMS file with a "headroom" of -0,2dBFS. Then analyze the MP3 after you did the encoding. Suddenly it's not clipping free anymore. So I take it that the CODEC in question does either of two things: 1) normalizing (even if that feature was turned off), 2) it creates ISP incidents

And the stronger the average signal strength on input, the worse this phenomenon is getting. Yet most software players ignore(!) the clipping incidents.

This thread is very detailed and useful. One of the best for a while

I'm honored - glad that at least "some" people think that I don't talk out of my rear end...

So you do not care about channel mix during mastering at all? Sry for asking, but i'm the c++ geek at the lower end that codes the codecs. Really interesting to talk to a guy at the studio end
I mean, for a DVD-movie: do you master 5.1 and then you are done? Or are you testing / tweaking mixing coefficents to sound good on stereo, 2.1, 7.1 .. ect systems too? (hmm it might more of an authoring than a mastering task if think again about it..)

(we can have spin-off about the non-clipping mp3 and exact reasons for ).

There are various metadata layers involved. I gues you refer to the transport container. That is where stuff like the artist, title or replay-gain info is located (ID3 tags are on the transport container). I was referring to metadata that is emdedded on the elementary bitstream. On DVD there is yet another dvd navigation metadata layer on top... but think we move offtopic now
.... short version: dolby, dts & co work with -3/-6dB because such formats are designed to adapt automatically to the speaker setup of listener. Speaker setup of the listener is not known to ME, but only to the decoder. So they splitt the work and same like an ME, the Dolby decoder als want to have headroom

[PurpleSunray]

~wrong button~

[Compyfox]

This is not true. As you explained it by yourself this headroom is good to avoid intersample peaks (often it can also be avoided with higher oversampling). Besides ISP some Mp3 converters having a problem if it goes near 0dB. There is (or was) also a difference between different implementations eg. Lame vs the Fraunhofer reference codec. So it was always a good hint to leave some space... with testings I made (but some years ago) a good value was between -0,4 to -0,6. Other MEs came with similar testings to a bit higher values (Afaik -0,8dB)

See my last posts please - I can confirm this.
In fact, this is why the "Fraunhofer ProCodec" system exists now - to show "on the fly" if things go awry.

But we don't speak about 6dB headroom, so I think this means a not mastered mix. Anyway, I cannot remember that MEs asking for such a big headroom.

Yes, the -6dBFS headroom is before mastering.

Though there are masters these days (again) where -1dBTP or even lower values (up to -3dBTP - due to the low loudness) are common. But of course, no further post-editing is needed.

Btw. I've heard also about problems with some D/A converters on cheaper CD players if the mastered audio had no headroom. But I have no idea if this true or a known behaviour.

I can confirm this. I still keep my old Phillips Disc man for this particular purpose.
I've also seen/used desktop players that started to "spit at you" (which is the resulting effect, skipping/spitting sounds) if the headroom was lower than -0,5dBFS for CD releases.

This is really down to the ADC, and what it's limitations are. Or how good it "masks" the ignored overs.

So you do not care about channel mix during mastering at all? Sry for asking, but i'm the c++ geek at the lower end that codes the codecs. Really interesting to talk to a guy at the studio end (we can have spin-off about the non-clipping mp3 and exact reasons for ).

To be honest, I barely even touch surround mixes. I mostly work in the 2.0 realm. And here I go via time tested metering tools.

Everything else in this realm, is bonus knowledge for me. However... if I'd work in surround (5.1 and co), I'd use the ITU-R BS.1770-x metering tool and shoot for -16LUFS max loudness wise. The per-channel peaks never exceed -1dBTP (if they even reach up until -3dBTP).

There are various metadata layers involved. I gues you refer to the transport containe. That is where stuff like the artist, title or replay-gain info is located. I was referring to metadata that is emdedded on the elementary bitstream. On DVD there is yet another dvd navigation metadata layer on top... but think we move offtopic now .... short version: dolby, dts & co work with -3/-6dB because such formats are designed to adapt automatically to the speaker setup of listener. Speaker setup of the listener is not known to ME, but only to the decoder. So they splitt the work

But it's still meta data, not actual raised signal strength per WAV file (bitstream per channel), no? It's like a "Playback Gain" setting, only way better than what Playback Gain (Normalization) for MP3 and FLAC ever was.


Like I said - I barely scratched the surface of Dolby and DTS codecs. This was never really my work area. I only looked a bit more closer out of pure curiosity. In days of HD-AAC (multi-channel capable, and Wavelab FINALLY catching up with surround streams!), affordable surround mixing environments, and the age Pure-Audio Blu-Rays... things turned drastically simple.

But I'm not in the finalizing section of the audio realm... I record, I mix, I master... but mainly 2.0. Though I'm not against diving into that topic/section of work if the chance might occur in my area of living.

[Chris-S]

btw. Just out of interest:
Do you know if modern DAW master mixer include any hard- or soft-clip or this gone completely?
I mean, if I bounce two overdriven 6dB signals in for mixing (2.0f sample), is the result a dump 2+2=4.0f / 12dB? even if it is way above full scale? or does it have any build-in hard or soft-clip that limits to a max?

Reaper does auto-mute when a channel goes over +20 db or whatever. When bouncing to integer-WAV hard clipping takes place, no clipping when bouncing to float-WAV.

[PurpleSunray]

But it's still meta data, not actual raised signal strength per WAV file (bitstream per channel), no?

It is something in between. It's not really related to this thread, so very high level example:
On the .ac3 file are 6 audio channels stored. FrontLeft/FontRight/LFE(bass)/ReadLeft/RearRight . The codec decodes this 6 channels from the bitstream, with a set of metadata. The channel signals and the metadata then go into the mixer.
The mixer looks at the attached speaker configuration and then mixes the 6 channel onto the speaker channels, following dolby spec.
Like, there could be following rule:
If no custom mixing coefficent are suplied, for stereo downmix of a 5.1, mix FrontLeft to Left with -3db, RearLeft to Left with -6db, FrontLeft to Left with -3db, .....
By applying the given mixing coefficent, you can listen to this 5.1 ac3 file on a stereo system and still hear all of the 6 channels because dynamically mix down from the 6-channel config stored in the file to the attached 2-speaker setup.

[Compyfox]

Okay, okay... but this is in the "chipset" of the DECODER (hardware), controlled via the UI of the player in question. The stream itself isn't rendered multiple times before it will be put on to disc. The source material is therefore delivered "prepared" (read: at a specific signal strength).

But you're right - this would go highly OT.

[PurpleSunray]

Okay, okay... but this is in the "chipset" of the DECODER (hardware), controlled via the UI of the player in question. The stream itself it's rendered multiple times before it will be put on to disc. The source material is therefore delivered "prepared" (read: at a specific signal strength).

Not necessary. Think on up-mix. You have a 2 channel file and want to play it on a 5.1 sytem. So there are 4 channels that are not persent at all on the source material. The ME never thouched it, but the decoder has to create this additional channels on-demand. Nobody prepared it (unless the ME was tweaking the mixing coefficents - that was the background of my previous question *g* ).
But I will stop now.. let give this thread back to the headroom guys

[Baddo]

I remember a long thread by Paul Frindle about this on GS. I'll see if I can dig it up.

[Compyfox]

Was that "The Reason Most ITB mixes don’t Sound as good as Analog mixes" or something? Oh wait, that was by Skip Burrows.

Or do you mean the "Myth or Truth Q&A" thread with Paul Frindle... which was not a good thread IMO. A lot of ego's taking it out on each other who is ultimately right...

[Googly Smythe]

...give the ME some breeding room...

Whoa! Where's this thread heading!

[Aleatoriac]

...give the ME some breeding room...

Whoa! Where's this thread heading!

First they want more head... now breeding room?

When will the DEPRAVITY END?