Time to revive this thread...
Jan Didden gave a presentation at BAF yesterday, about the loudness wars. And it got me thinking. Today, I have an idea I'd like to share on what might be done in future digital recording formats to address the problem:
I think a new very high bit-depth digital format is needed (like 32 bits). Certainly not because more downward dynamic range is needed, but because more upward dynamic range is needed! If we have a new format, then I think it must incorporate measures to remove the temptation for recording engineers to squash music up against the hard ceiling of 0dBFS.
Of course, there would still be an ultimate 0dBFS ceiling, however the new standard could dictate a nominal "0dB" level which leaves a very large amount of headroom (perhaps 20dB or more), and simultaneously require mandatory meta-data tags containing a specified average loudness metric for the album and song (think replaygain). This is the key - sufficient headroom to handle peaks, combined with a mandatory loudness metric, which is universal and has a specified algorithm so that it is consistent on every recording. The playback equipment would also be required to decode this meta data and normalize the playback level according to it. If the above conditions are not met, then the recording or equipment will not be approved by the licensing agency and cannot bear the trademark of the format.
Of course, there are some complications. For example, how would one go about building a DAC for this format? The playback end of the chain would need to perform loudness normalization based on the metadata. Mild peak compression or limiting in order to fit the data within the dynamic range of the actual DAC would be the second part, but then this will be within the control of the end user and/or equipment makers who have an incentive to make the music sound it's best. The end user can clip and compress as they see fit, and differently for different purposes.
32 bit fixed point seems like the natural choice to me - 8 bits of peak headroom over a 24 bit "base" sample depth, and most computer and mixing equipment can already process 32 bit data.
What do you think?
Jan Didden gave a presentation at BAF yesterday, about the loudness wars. And it got me thinking. Today, I have an idea I'd like to share on what might be done in future digital recording formats to address the problem:
I think a new very high bit-depth digital format is needed (like 32 bits). Certainly not because more downward dynamic range is needed, but because more upward dynamic range is needed! If we have a new format, then I think it must incorporate measures to remove the temptation for recording engineers to squash music up against the hard ceiling of 0dBFS.
Of course, there would still be an ultimate 0dBFS ceiling, however the new standard could dictate a nominal "0dB" level which leaves a very large amount of headroom (perhaps 20dB or more), and simultaneously require mandatory meta-data tags containing a specified average loudness metric for the album and song (think replaygain). This is the key - sufficient headroom to handle peaks, combined with a mandatory loudness metric, which is universal and has a specified algorithm so that it is consistent on every recording. The playback equipment would also be required to decode this meta data and normalize the playback level according to it. If the above conditions are not met, then the recording or equipment will not be approved by the licensing agency and cannot bear the trademark of the format.
Of course, there are some complications. For example, how would one go about building a DAC for this format? The playback end of the chain would need to perform loudness normalization based on the metadata. Mild peak compression or limiting in order to fit the data within the dynamic range of the actual DAC would be the second part, but then this will be within the control of the end user and/or equipment makers who have an incentive to make the music sound it's best. The end user can clip and compress as they see fit, and differently for different purposes.
32 bit fixed point seems like the natural choice to me - 8 bits of peak headroom over a 24 bit "base" sample depth, and most computer and mixing equipment can already process 32 bit data.
What do you think?
I don't believe it can help;
one famous sound engineer said, "24 bit / 96 KHz gave me ability to record without worrying of clipping"
It means that he intentionally looses some bits in order to be avare no clipping happens. But later during mastering he squeezes everything wildly.
Why?
Again according to his explanation, because artists want to be squeezed. Period. Once a band asked him to record them as clean as possible, as natural as possible. He did. They were satisfied until few days later they run in with their own disk and a disk of another band, heavily compressed: "Hey man, they sound louder! We want the same loudness!"
one famous sound engineer said, "24 bit / 96 KHz gave me ability to record without worrying of clipping"
It means that he intentionally looses some bits in order to be avare no clipping happens. But later during mastering he squeezes everything wildly.
Why?
Again according to his explanation, because artists want to be squeezed. Period. Once a band asked him to record them as clean as possible, as natural as possible. He did. They were satisfied until few days later they run in with their own disk and a disk of another band, heavily compressed: "Hey man, they sound louder! We want the same loudness!"
Well, OK. I will not argue that some artists might still prefer squashed dynamics over greater contrasts (hopefully this preference would diminish with time as the 'norm' returns to something other than the maximum compression possible). This is where the recording engineer's judgment becomes important in moderating an artist's desires, just like any good producer will help guide the artists' performance, based on many considerations, some of which are not artistic, but commercial criteria - things that artists probably don't often care about, or view from a different perspective.
But the objective of the normalization system is to eliminate average loudness differences regardless of compression, making it moot to try and squeeze up against the 0dBFS ceiling. That will only make the recording sound cruddy as the waveforms clip and so on. The loudness meta-data plus playback normalization guarantees that every recording is reproduced at the same average level, and that there is headroom in the format for transient peaks.
But the objective of the normalization system is to eliminate average loudness differences regardless of compression, making it moot to try and squeeze up against the 0dBFS ceiling. That will only make the recording sound cruddy as the waveforms clip and so on. The loudness meta-data plus playback normalization guarantees that every recording is reproduced at the same average level, and that there is headroom in the format for transient peaks.
The fact that the immense majority of people listens to music with low-fidelity low-power stereos is what we have to blame.
Well, actually what whe should blame more is the fact that they like to listen to music that way.
Indeed, the only way to get some sound from that junk equipment is heavy compression.
No new format will change this.
It has nothing to do with bit depth, it has to do with TASTE. Despite its age and the conservative 44100/16, the CD format already has far enough resolution when it is used properly.
Well, actually what whe should blame more is the fact that they like to listen to music that way.
Indeed, the only way to get some sound from that junk equipment is heavy compression.
No new format will change this.
It has nothing to do with bit depth, it has to do with TASTE. Despite its age and the conservative 44100/16, the CD format already has far enough resolution when it is used properly.
Having metadata on disks to tell a system what the intended playback level is to be would be a good thing (it could even be just text printed on the disk label if any CD company wanted to do it immediately).
Coming up with an extra 8 bits of dynamic range (that's 48dB!) is more than a little problematic, though. If you want to give 48dB more at the top, that means you have to take 48dB off the bottom. Bear in mind that there aren't even any real 24bit audio A/D converters, that would need something like 144dB of dynamic range! 120-ish is more the real situation.
Coming up with an extra 8 bits of dynamic range (that's 48dB!) is more than a little problematic, though. If you want to give 48dB more at the top, that means you have to take 48dB off the bottom. Bear in mind that there aren't even any real 24bit audio A/D converters, that would need something like 144dB of dynamic range! 120-ish is more the real situation.
Eva said:
I'll sure agree with that!
And it has always been thus. Systems in days of yore were not all that dynamic, either. Some were, but the vast majority were not. There was a loudness war in the days of 45RPMs, too.
Having been to RMAF last week, I can tell you that there are a lot of non-dynamic sounding systems out there - some with rather "dynamic" price tags.
Listening to a pair of 18" coaxail speakers most CDs sound plenty dynamic.
Comments and links to more information on the over compression of music can be found on John Atwood's ClariSonus web page. Look for the "Turn me up" entry about half way down the home page.
ClariSonus
ClariSonus
Thanks Gary,.
FWIW, I was in a thift shop yesterday looking at used LPs. Came across a mono LP of the Singing Nun. Mid 1960s vintage. On the back it had pretty complete tech specs of the recording, including tape decks, mics, FR, etc.
One thing that caught my eye. They had used a compressor to "insure the loudest, clearest reproduction possible."
Nothing new under the sun....
FWIW, I was in a thift shop yesterday looking at used LPs. Came across a mono LP of the Singing Nun. Mid 1960s vintage. On the back it had pretty complete tech specs of the recording, including tape decks, mics, FR, etc.
One thing that caught my eye. They had used a compressor to "insure the loudest, clearest reproduction possible."
Nothing new under the sun....
hifiZen said:
Hi Chad,
I think for physically delivered media, you could do some extras bits on the master recorded file only. Before putting on s disk, it would be trivial to scan the file for max level and cut the disk a few dB below that. That way, you always would have an optimum disk compatible with existing equipment. And do level control in the analog domain, of course.
Come to think of it, a similar strategy could work for downloadable media, either as embedded metadata or as a one-time scan when you put the file in your lib.
Jan Didden
There will always be a large market for compressed lo-fi music. A lot of us will probably use an mp3 player when convenient - it's a bit problematic to drag a pair of Alephs along while jogging, to say nothing of 3-way boxes 
Fighting it is no use. What we need is a choice. For instance, a choice to download 24/96 wav or flac for $1.95 or the same track as mp3 for $ 0.39.
That way, everybody wins and everybody has a chance to make a buck.
Jan Didden
Fighting it is no use. What we need is a choice. For instance, a choice to download 24/96 wav or flac for $1.95 or the same track as mp3 for $ 0.39.
That way, everybody wins and everybody has a chance to make a buck.
Jan Didden
I find myself in the unfamiliar position of remaining the optimist. The cheapest portable flash player/ear bud combo trounces the sound quality of the cassette and 8-track decks of my youth. Absolute performance ceiling might be a debating point, what isn't is they do the same to the typical 45, cactus needle changer and big-wood consoles of the day by most any audible metric that matters. Contemporary portable electronics, which I suspects accounts for most of the world's listening, have arguably the best median sound quality in consumer history, a trend I don't see reversing soon.
Coming from a commercial radio broadcast background the perspective is much different. 'Louder' meant something in the 1960s when stations ran without on-air compression and levels were maintained 24/7 by an operator sitting at a console at the transmitter. Engineers have pushed back against the compress-it-to-death ethos for generations since. It's fashion coupled with over compensating programming departments. The music selection might be poor but WE'RE SO LOUD!!!! Now session producers have hopped on board, often using FM broadcast compressors to achieve their 'signature sound'.
Coming from a commercial radio broadcast background the perspective is much different. 'Louder' meant something in the 1960s when stations ran without on-air compression and levels were maintained 24/7 by an operator sitting at a console at the transmitter. Engineers have pushed back against the compress-it-to-death ethos for generations since. It's fashion coupled with over compensating programming departments. The music selection might be poor but WE'RE SO LOUD!!!! Now session producers have hopped on board, often using FM broadcast compressors to achieve their 'signature sound'.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.