Thursday, 21 April 2016

Science!

I love science.  I love the scientific method – the best thing we puny humans have ever devised to tell the difference between something that’s real and something that’s homeopathy.

There has long been a dual version of the world, with science in one corner and art, creativity and emotion in the other.  I’ve never thought this was a good idea.  I love both!  I’m equally proud of my rational scientific left brain, and highly irrational, artistic right brain.  I think they’re both just lovely, and all balancey.



Six weeks after first playing There’s Always Someone on Spotify and thinking “hmm, that sounds rather quiet”, I have gone completely insane. But on the plus side, I have also learned a great deal.  I have unfailingly walked into every known and unknown bear trap there is to walk into when it comes to these matters.  I like to think I’ve confounded the experts, revealing whole new ways to screw up that no-one even knew existed.  And I’ve learned an amazing amount of factual information about the whole scary-but-tedious subject of Loudness when it comes to music.  I’m now a veritable gold mine of boring Loudness facts.  I could have it as my specialist subject on Mastermind, if only John Humphries could stay awake long enough to ask the questions.



And yet, after all this time, I find myself finding words like “voodoo”, “dark arts” and the ever popular “black magic” when trying to work out what level a track will actually play at after it is unceremoniously spat out of the Spotify black and green box. In the past 48 hours, I have created two different new versions of There’s Always Someone, almost indistinguishable from each other to anyone except me.  And I’ve found two different ways to measure their Loudness, and predict its imminent change in level as according to Spotify – Dynameter, and Audacity’s ReplayGain plugin.  With Song Version A, both agree that Spotify will LOVE it.  With Version B, one says it will be twice as loud as the other.

TWICE AS LOUD


And here’s the really fun bit – the creators of both measuring devices say that they honestly don’t know which is the more realistic representation of what will actually happen.  Because – snigger - nobody actually knows what Spotify actually does.

Let’s just enjoy that for a moment.  Even the greatest experts in this field, who write the code to measure Loudness, don’t actually know how Spotify determines how loud something is, or how to accurately predict it.  So if you’re an artist or producer or mastering engineer, and want to know with any kind of confidence what something will sound like on the world’s largest streaming platform, then tough. You can't.  Ha!

It’s 2016 for crying out loud.  Has it really come to this?!!!

Look I’m getting all emotional.  That’ll be my right hand brain.  Time to employ the left handed brain instead.  Enough of this whining vaguerey. Science, and its adorable scientific method, to the rescue.

Spotify exists.  The original recordings exist.  We can measure what Spotify does to them.  And we can see what these meters THINK Spotify will do to them.  Well then… that’s all we need, right?  We can propose our hypotheisis – say, “The ReplayGain plugin in Audacity will accurately predict what Spotify will do to any given track”, and test it against real world data.  Well then, it’s just a case of some hardcore data logging and then some analysis.  And as world’s-greatest-TV-show Mythbusters* say, “the only difference between science and screwing around is writing stuff down”.



So I’ve been writing stuff down.  Til 3am.  Measuring, logging, looking for trends and causal relationships (that’s what people do on the internet at night, right?  Or did I mis-type?)  And in what will likely turn out to be my 27th false dawn, I may have found something.  Something to explain all this madness, actual light at the end of the tunnel, not hallucinatory light (damn stupid hallucinatory light).  A magic formula, with which I have successfully used to predict what Spotify actually did to actual tracks.  I rose up brandishing my new power wand like some mighty powerful wizard of opaque loudness.  I can’t possibly tell you what this formula forged at the very coal face of decibels is – it obviously has to be further tested, written up, submitted to a reputable academic journal and gone through peer review before it can be shared with a credulous world.

On second thoughts, I’ll just see what Ian thinks of it.  I’ll let you know.



*Mythbusters was only the greatest TV show ever for between about series 3 and series 5.  Series 1 and 2 have plenty of charm – and I did like Scotti and some of the other earlier build-teamers - but it was only really with the arrival of Kari, Tori and Grant that the series got into it stride, each bringing a ton of personality (and it Tori’s case, taste for spectacular personal injury) to their amazing skill set.  But before long – around series 5 or 6 - the dumb producers started scripting and staging more and more.  While still entertaining, it lacked that sense of eye-popping reality that made it such an essential watch.  They missed real reactions and substituted fake ones, edited round blind alleys and tangents.  It still felt scientifically valid in their bonkers way, but you felt shut out rather than included.  But then for series 10 the producers committed their ultimate sin by getting rid of Kari, Tori and Grant and turned it into a ghastly headache-inducing macho testosterone-fuelled joyless noise-fest.  The writing was very much on the wall and I didn’t mourn its recent passing as it was far from what it was. Besides, word on the street is Kari, Tori and Grant have a new science show in the works… just please don’t edit and script it to death…

Saturday, 9 April 2016

The April Playlist - SPRING!

The April playlist is up, and what better way to kick it off that the quintessentially spring-like and British Bluebell Morning by the fabulous Ooberman.  As befitting the After School Video Club, it's a ridiculously wide-ranging list - guitars, electronica, old and and new, well known and obscure from all four corners of the globe yet - in my head at least - coexisting rather splendidly.

 

Friday, 8 April 2016

There really, really IS always something...

Ha!

Ha ha ha ha!

Ho ho ho ho ha ha ha!

Etc!

The next song I'm going to write will have the lyric "Everything goes great for me it does/I don't have any technical hitches/I understand all technical things/And I've just won a million quid", and see if that has any magic real world properties.

So.

Sooooooooooo....

Well it took nearly 2 weeks for Spotify to actually get the new mastered versions up on their system (as opposed to 1 day for iTunes).  Then was time to measure the all new and improved version.  To discover.... it's almost exactly the same as the old one.

YAY!

I put the measuring devices across it, and it is one whole db louder (which equates to, if you're not so technically inclined, stuff all).  "How could this be", I hear you cry over the voices in my own head which are also screaming "How could this be"?



"But I thought this new version was going to solve this exact problem", you say - and yes, that's what the voices in my own head are cackling too.

I asked Ian Shepherd, the guru who developed Dynameter to take a look, and he came back to check I'd sent him the right version.  Now, never mind what these numbers mean or what the letters stand for, all you need to know is the higher the number, the better it is for Spotify.  High = good, Low = bad. I'd measured PSR 8, PLR 13.  Ian was getting PSR 5 from the same file.  Which - and I'm sure you're ahead of me here by now - is bad.

So I checked.  And guess what?  I was getting PSR 5 PLR 10 too!  Ha ha ha!  Ho ho ho!  So what had I done?  Had I lost my marbles?  I went back to the drawing board and loaded up the unmastered version, going through the identical steps to arrive at, essentially, the same file and measured again, Lo and behold - my magical and illusory readings came back - PSR 8, PLR 13, those figures that - not unreasonably it now seems to me - caused me to deduce that Spotify was just gonna lurve this file.  When in fact, for whatever reason, it just made the figures up.

Which has us up to date.  I've asked Ian to take a look and try to figure out what's going on.  And all this means that, yes, a third version will have to be made and uploaded, and I'll have to figure out how to measure things in the real world, rather than parallel magical one I'd accidentally stumbled into.