So, after dropping the average level by 4-6db, I do a bit of dynamic EQ and gentle multi-band expansion using a collection of software mastering tools.
So, would you please tell me the contents of the mastering tools ?
Are they open source ?
I've been meaning to post a followup about how we selectively process songs for airplay on RP, so here it is.
The recordings in our library fall roughly into 3 categories:
Well-recorded. Examples would be the first Norah Jones album, Rick Rubin's acoustic recordings with Johnny Cash & Donovan, most jazz, The Beatles, Steven Wilson's stuff, etc. In this case we do nothing other than an overall level adjustment, if needed. In order to provide smooth transitions between songs, we need to match the levels from one song to the next. Occasionally I'll encounter recordings that have a very low overall level, requiring a boost, but with occasional peaks that would go into clipping if I didn't apply a little gentle peak limiting.
Poorly recorded or mastered. There's lots of variety here: early stereo recordings by people who didn't really understand stereo, stuff that was mastered for vinyl (with most of the low bass energy removed and an emphasis added to the upper-mids & highs in addition to the normal RIAA curve) and never properly remastered for CD, anything with exaggerated bass. In these cases, I will adjust the spectral balance, tame some aggressive thump, and do what I can to bring the sound into alignment with the well-recorded tracks.
Loudness war victims. These are the tracks that I process most aggressively, in an attempt to restore some of the natural dynamic range. An unfortunate percentage of 90s+ recordings fall into this category. There's only so much that can be done. Some companies market filters that claim to "de-clip" audio, but I haven't heard one yet that doesn't also add distortion in the process. So, after dropping the average level by 4-6db, I do a bit of dynamic EQ and gentle multi-band expansion using a collection of software mastering tools.
For those of you read waveforms, I have an example. It happens to be a track from the recent A Perfect Circle album. Great song, but driven into clipping throughout most of its length.
You can see that the original dynamics of the track remain pretty much the same, but the overall level is lower (to match the good full-dynamic-range recordings in apparent loudness) and the peaks have a bit of definition to them rather than being chopped off the way they are in the original.
It also sounds more natural when listened to critically (I use a pair of Beyer DT 1770 pro headphones & a good amp) and it survives encoding into a MP3 or AAC stream a lot better than the original would.
It's ironic that these overly-loud masters are almost always listened to via MP3 or AAC streams or files â and yet those codecs depend on dynamics in the recording to do their job properly. Mastering for loudness makes absolutely no sense whatsoever in this day & age, but it's become so much the norm that very few people seem to want to break out of that rut. Things are getting a little better (a few years ago even the "quiet" passages in that recording would probably have been slammed up against the wall) but it's a painfully slow process.
The saddest part is that the "loudness" mindset often infects the recording process as well, so that individual tracks are compressed to hell on the multitrack recording or at least in the mix â so there's no way to go back & truly undo the damage.
If producers and sound engineers wanted to apply more compression, they would do so.
It's just not that simple. The recording and mixing process is separate from the mastering process. The person doing the mastering of the final mix has to interpret the master final mix and make it work for farther down the food chain, namely us the listener. They also have to fix things that didn't quite go as intended in the studio. Note Bill's comment. This is completely separate from what we call the loudness wars. Compression and EQing are the primary parts of that process. It's a synergistic process where the whole is greater than the sum of the individual parts.
We as listeners have no idea how good or bad a particular release of a particular recording is until we listen to multiple versions of the same song or album critically. We only know what we know, until we stumble onto a better mastering job of a particular recording, or a worse one for that matter.
I forget which song, but it is a Metallica song widely released and the consensus is that the version used on a video game has the best mastering compared to the vinyl and CD versions. There was a tremendous uproar over that discovery.
In vinyl, people argue about is the US first press of a particular 1967 album better than the 1967 first press of the Great Britain issue. And then there are differences between a mono mix and a stereo mix.
Music released in one country will sound different from another country because different people are involved in the mastering processes involved in the manufacturing of the physical media. While they will all use the same unadulterated master tape for the transformation to a physical copy, they all do it their own way. Some add a little more compression than others, use different EQ's. Depends on the medium and the equipment involved and the ears and the tastes of the mastering engineer. Even more interesting is how those engineers evolve and what they did 20 or 30 or 50 years ago is far from what they do today. Some get better and others get worse.
Bottom line ... everything is altered from the original mix master tape in some way. And when you are putting together diverse collections of music for a long linear playback list, there must be something done to make it all work together in addition to the segue. In broadcasting, judgment calls must be made in order for the product to sound as good as possible on as many playback devices as possible without compromising the music too much. And some do that better than others. Here is about as good as it gets.
OK. The Tom Baxter song has been fixed. A very nice recording when listened to carefully — hadn't done that for a while.
Apologies for the previous inferior file. Please feel free to post any others that jump out at you. I know there are a few MP3 "masters" scattered in the library.
To clarify something: while we may use dynamics compression/expansion tools on some select tracks, that process is pretty subtle and produces no distortion or artifacts of the sort generated by the other kind of compression (data compression). With the aforementioned exceptions, the music goes through just one data compression step: the final stream encoding (or not, in the case of FLAC).
Dynamics compression is hardly evil. Every recording you've ever heard has utilized compressors, limiters, & expanders — most likely in all stages: recording, mixing, & mastering. Sometimes these tools are used well, and we don't do anything to a file other than basic level correction. With some older recordings, and really aggressively mastered newer recordings — plus some that were mixed & mastered by people whose perceptions were altered by drugs* — we do what we can to make the sonic balance and dynamics of the recording match the well-recorded stuff.
The subtleties of this are still a work in progress, and help is always welcome.
* listen to any random assortment of late-60s Stones and Beatles tracks & see if you can guess which studio had the mirrors & razor blades in the engineering area.
Thank you for fixing Tom Baster and providing clarification on how you process before streaming!
I would only say that, in my opinion, any kind of dynamic compression beside whats done on mastering stage of production, is evil. If producers and sound engineers wanted to apply more compression, they would do so. Ok, most songs today are over-compressed and all that loudness war makes a lot of good music almost unlistenable, but those rare few tracks with good dynamic rage should stay like that... If im listening in my car, or on headphones while commuting, i don't care much, but at home on proper stereo system its audible sometimes.
So RP has my vote to not do any kind of processing, or if its possible to leave original source untouched only for FLAC streaming.
OK. The Tom Baxter song has been fixed. A very nice recording when listened to carefully — hadn't done that for a while.
Apologies for the previous inferior file. Please feel free to post any others that jump out at you. I know there are a few MP3 "masters" scattered in the library.
To clarify something: while we may use dynamics compression/expansion tools on some select tracks, that process is pretty subtle and produces no distortion or artifacts of the sort generated by the other kind of compression (data compression). With the aforementioned exceptions, the music goes through just one data compression step: the final stream encoding (or not, in the case of FLAC).
Dynamics compression is hardly evil. Every recording you've ever heard has utilized compressors, limiters, & expanders — most likely in all stages: recording, mixing, & mastering. Sometimes these tools are used well, and we don't do anything to a file other than basic level correction. With some older recordings, and really aggressively mastered newer recordings — plus some that were mixed & mastered by people whose perceptions were altered by drugs* — we do what we can to make the sonic balance and dynamics of the recording match the well-recorded stuff.
The subtleties of this are still a work in progress, and help is always welcome.
* listen to any random assortment of late-60s Stones and Beatles tracks & see if you can guess which studio had the mirrors & razor blades in the engineering area.
Just to confirm this: there was Tom Baster Skybound on the radio (listening to 32kbit aac stream in foobar, configured for hifi playback, no replay gain, ASIO drivers, external dac). I switched to Spotify and Tidal, and difference was HUGE!
So it really doesn't matter if Radio Paradise streams in 128kbit or FLAC since compressors/expander filters are applied, and every songs is ruined. Such a shame.
First, I assume you mean 320k, not 32k.
Secondly, that track is indeed inferior. At one point, I acquired flac copies of some files that had actually been created from MP3 sources. Unfortunately, since they're labeled as having FLAC sources in our database, I can only find them by listening. There are just a few of these — mostly songs that are played infrequently, as the ones played more often have all called attention to themselves. I'll replace the file before it plays again.
We do not apply global compression to our stream (we used to, but haven't for quite a while). We do apply some selective dynamics processing to badly-mastered recordings.
Just to confirm this: there was Tom Baster Skybound on the radio (listening to 32kbit aac stream in foobar, configured for hifi playback, no replay gain, ASIO drivers, external dac). I switched to Spotify and Tidal, and difference was HUGE!
So it really doesn't matter if Radio Paradise streams in 128kbit or FLAC since compressors/expander filters are applied, and every songs is ruined. Such a shame.
I noticed on few songs (from Norah Jones, for example), that the same song played on RP (320kbit) sounds compressed compared to Spotify (Premium). Does RP uses some filters, like compressor/normalizer?
