Mike’s Blog

24-bit audio saviour or con?

Thu, 24 Feb 2011 11:07:27 +0000

Some time ago I wrote a blog piece (here) on digital audio, explaining the significance of bit-depth and sample-rate in PCM (CD, WAV and AIFF) audio. Recently, rumours began to fly about that Apple was in talks with labels about encoding music as 24-bit instead of 16-bit. Since 24-bit has 256 (28) times more accuracy than 16-bit audio this is a good thing, right?
A quick look through my Twitter feed however and you might think differently. The 24-bit audio revelation was being torn apart as a con. Engineers who I respect greatly were describing the rumour as if Apple was opening a new iHomeopathy store. These engineers aren’t content with the current 256kbps AAC format, they want CD-quality (note my inherent reluctance to refer to it as lossless) 16-bit/44.1kHz .WAVs.
So what’s going on here? Why do people who want better audio file quality seem so unhappy at Apple’s attempt to improve audio file quality?

Dynamic range
One of the first arguments presented by the naysayers was that since music uses such a small amount of the dynamic range (the difference between the loudest and quietest levels it is possible to encode) available in 16-bit that the enhanced dynamic range of 24-bit would be useless. This is true. 16-bit audio has a dynamic range of roughly 96dB, compared to 144dB for 24-bit (human hearing has a dynamic range of roughly 140dB). However, the dynamic range in music is nowhere close to this figure. And even if it wasn’t for the problems highlighted by Dynamic Range Day, music would still be very unlikely to use even all the dynamic range available in 16-bit audio files. The problem with this argument however, is that the extra bits don’t simply add more room for quieter audio, they increase the resolution of the waveform through it’s full cycle. In fact the increased dynamic range is a product of the increased dynamic resolution, and not as many seemed to think in spite of dynamic resolution.
How can this be? How can you get both greater resolution (the smallest possible amount the level can be increased or decreased by) AND greater range. Surely one is a trade-off against the other, and you have to fix one to have any concept of the other? If the 16-bit number represents the height you have 65,000 possible measurements, but you need to know how big each one is (the resolution) before you can know the maximum height (range). And doubling the distance between each measurement would give you twice the total range, but decrease the accuracy. So adding extra measurements would let you measure the same range more accurately, or measure a greater distance with the same accuracy.
The reason this metaphor doesn’t work however, requires a better understanding of the physics at work and of how waves are encoded in PCM audio. The loudness of a wave is the distance between it’s highest and lowest points, with silence being a flat line. The bit depth of a PCM audio file encodes the height of the wave between -1 and 1, whether it’s in 16-bit, or 24-bit (with 16-bit audio having 65,536 steps between -1 and 1, and 24-bit having 17million steps between -1 and 1). So an increase in bit-depth effects only the resolution of the waveform, not the range. In both a 16-bit and 24-bit file the loudest wave you can encode (known as 0dBFS) will go all the way from -1 to 1 and back to -1. However the quietest wave you can encode in 16-bit will oscillate up and down 1 step from 0 to 1/32500 and back to 0, while the quietest wave you can encode in 24-bit audio will oscillate between 0 and 1/8,500,000 and back to 0. So you get not only greater dynamic resolution, meaning more detail in loud material, but also a bigger dynamic range.
24-bit as an Mp3
The other argument presented was that it doesn’t matter what file Apple has, if the end result is the same low-quality AAC file (AAC is actually based on mp4 not mp3, but for all purposes here, they’re much the same). Again, this is true, currently Apple sells 256kbps AAC files, encoded from 16-bit .WAV files.
The encoding/decoding process is quite complex, and rounding errors during reconstruction mean that the reconstructed 16-bit samples can sometimes be distorted with quantisation errors. The amount of error being dependent on the accuracy of the encoder and decoder and the bit-rate (NOT bit-depth) of the AAC file. If a 24-bit waveform is created on decoding (even if you started with a 16-bit file) the end result could be perceptually closer to the the original waveform than a 16-bit reconstruction. Both will have rounding errors, but where the 16-bit version will round up or down to the wrong bit, the 24-bit version will round to a 24-bit approximation between the two 16-bit positions. Imagine if after a complex sum for which the answer is 2.76 you get the answer 2.76664 to two decimal places the answer would be 2.77, but to three decimal places you get 2.767 which despite still being wrong is closer to the number you wanted, by 0.003.
So if Apple use 24-bit encoding and decoding the end result will be closer to the original waveform, whether the original waveform was 16-bit or 24-bit. And if the original was 24-bit then we’re getting better approximations of a better audio file. Everyone’s a winner!
Well, not quite. We’re still not getting PCM audio, and even encoded at 24-bits the decoded AAC file is only an approximation of the original WAV not a like for like copy. But we’re moving in the right direction, and as the MPEG audio format improves and bit-rates get higher there is no reason why in the future the humble AAC (or derivative) couldn’t be capable of giving us both smaller files and equal audio quality. And with 24-bit WAVs as the starting point, we could very quickly get to a point where the AAC is better than the 16-bit WAV, even if it is still some way off the 24-bit WAV - just as now an AAC is better than an 8-bit PCM audio file. In fact some tests show that the perceived quality of high-bitrate AAC files is equal to and in some cases higher than PCM audio (and this is where I wish I was a proper academic, with access to the original papers, did they encode the two files from a DSD file, or did they start with 24-bit files and compare decodes to AAC and 16-bit, in which case, did they dither it?).

Footnote: If anyone does have any academic textbooks that cover this in more detail than I’ve gone into here, please feel free to let me know what they are in the comments. But academic texts only please - the ones with full citations etc, not texts written by engineers/journalists, I’m one of them, and I already know how little we know.

Room Acoustics, Part 1

Wed, 24 Nov 2010 16:52:37 +0000

Getting acoustic treatment has always been on my wish list, but what with renting a succession of small dives and landlords not being so keen on people gluing foam to the walls I’ve never done it before. Even were I to still be working at Metropolis I was going to treat this room, but now that I’ve gone freelance it became a priority. Just from my own experience with the room I knew I had bass issues, especially with sub-bass build up towards the back of the room. On top of the standing waves, the square shape of the room means that there are a lot of reflections and this was going to make mixing quite problematic.
You can do all sorts of calculations to figure out what needs doing to your room. You can calculate room modes from the room dimensions to discover which frequencies are going to be problematic, you can use measurement equipment and software to calculate the frequency response and impulse response. And being something of a geek, this is exactly what I did using FuzzMeasure... But this level of geekiness isn’t strictly necessary, the cure for your problems, whatever the results, is almost certainly going to be to stop the sound waves from bouncing around the room quite so liberally.
But I wanted to know how bad things were, so I fired up FuzzMeasure, and using the measurement mic which comes with the KRK Ergo (more on which later, suffice to say it wasn’t plugged in at this point), I measured my room.

Fig. 1: Listening position with sub woofer before treatment.

Fig. 2: Room acoustics with sub woofer before treatment.
As you can see from these graphs, the frequency response is hardly smooth across the spectrum, with as much as 20dB difference between some the peaks and troughs. Above 1kHz, things start to even out a little, with a high-end roll-off towards the back of the room (second graph). But below 1kHz it’s a worrying mess - how can you expect to make balanced mixes in this type of environment. You can clearly see the resonant spike around 40Hz which I’d noticed at the back of the room, and very worryingly the peak at 70Hz and trough at around 100Hz in both the listening position and rear of the room. Since my sub-woofer crossover is set to 80Hz, I wondered if this was a side-effect of the sub, and did an extra set of measurements with the sub disengaged.

Fig. 3: Listening position no sub woofer before treatment.
As you can see, the trough at 90-100Hz is still there with no sub and the low-end below 60Hz is significantly lacking.

Part Two here.

Going solo

Sat, 6 Nov 2010 18:50:35 +0000

A little over a month ago I was asked into a meeting at Metropolis, where I was basically told that since I’d made myself far more useful than they’d imagined (or budgeted for) that they could no longer afford to keep me on as a member of staff. At roughly the same time Anthem Publishing phoned to ask how involved I’d like to be on the next Guitar Tech issue, so I took on as much as I could handle and dived into a full month of writing. This gave me a handy time-buffer to think over my next moves.
I’m glad I got the chance to work at Metropolis, it’s one of the biggest and best equipped studio complexes in the world, and I not only had the opportunity to work with and learn from some of the best engineers in the world, I made some great friends and I now have quite a few great stories to tell while I’m down the pub. I’ve also learnt the problems associated with working in a big studio like this, and the problems these studios are facing to stay in the game.
Moving forward it seems to make sense to me to give freelance engineering a proper shot. There’s still a place for big studios, but it’s not the place I want to be. The bands I want to work with aren’t signed to majors, they don’t have massive recording budgets, and for the most part they aren’t even interested in the old million dollar record deal - they’re going alone, retaining control and paving their own way. Bands I’m interested in recording aren’t disillusioned at the state of the music industry, they’re excited by it. They aren’t scared of the internet, they’re embracing it. They’re marketing themselves by engaging directly with fans on social networks, they’re bypassing the distribution wings of major labels by selling straight to the fans via sites like Bandcamp, and instead of worrying about file-sharing as lost revenue, they’re encouraging it, and offering “Pay What You Want” deals to reach more listeners and build a fanbase. These bands can’t afford to go to big studios, like Metropolis, they’re recording in small local studio spaces or in last-minute downtime slots and mixing in cheap project rooms, or doing it themselves at home.
This isn’t actually news to me, I’ve been working freelance on top of my Metropolis commitments the entire time I worked there, and the projects that I’m most proud of have all come from this work.
I’m lucky enough to already have a few projects on the go, but I’m going to have to keep that going to maintain the momentum going freelance. And since I no longer have quick access to mix/mastering rooms at Metropolis I’m going to have to make sure my own project studio is up to scratch. The first step on that path arrived about a fortnight ago, in the form of several boxes of Auralex acoustic tiles, much of which is now adorning my walls. I’ve made some detailed measurements of the room before and after treatment, and will soon be delving into that analysis in a blog. Unscientifically though, the room is sounding considerably better already. I’ve been listening to records I know really well, and a few of my own recordings to try and get an idea of what the treatment has done, and I’m very happy with it. The bass sounds much more controlled - particularly towards the rear of the room, where it used to be quite out of hand. And the stereo spread is much more well defined, all of which should make mixing here much simpler.

The Green Room has moved

Thu, 24 Jun 2010 15:46:29 +0100

So I’ve been quiet on here for a while. In no small part, because a) I’ve been very busy writing up reviews, workshops, interviews, etc. b) I’ve moved house.
The second part of this has been a long heavy struggle, but now I’m in I’m glad I did it, and the effort is certainly being repaid. Not only is my new flat bigger, but I also no longer share it, and to top things off I can now get back to having a room for my studio. That room might be the lounge, but who cares, it’s not like I needed a living room for anything else...
This means my project studio now has a dedicated space, one which thankfully sounds much better than my last room, although I’m having to get used to the enhanced low-end, now that I have space for a sub.
In the meantime I’ve been working on a redesign for this site, which should be coming soon, and this weekend I’m going into Metropolis to sample the balafon I bought in Mali which I mentioned I’d sample up for you all. I’ve not much experience with making sample libraries, nor in fact with recording balafon’s, but I’ll experiment away until it sounds good and have something for you all to play with soon enough. I’ll even try and write up my experience with making the sample library for anyone interested.

Remixing Carry Your Heart for the PEACE compilation

Wed, 14 Apr 2010 13:00:05 +0100

In January, shortly after I got back from Mali, I was asked by Catherine A.D. if I could help her out with a remix for a compilation she’d been asked to contribute to. Being the generous sort I leapt on the idea, grabbed the stems to “Carry Your Heart”, figured out the chords, and then chucked most of the stems out and got started on a remix.
Considering the short turnaround required for the remix I possibly took on a little much by throwing out all the original parts other than the vocal, but the last thing I wanted to do was throw a house beat and a bass line under the track, and every time I approached the song’s original parts individually I felt like they worked so well as a whole that I didn’t want to split them up in any way or remix it in any conventional way. I also wanted to put something of my recent trip to Africa into the remix. Thankfully I had my partner in crime Phil Joannides along for the ride.
We started the process with me adamantly insisting that the chorus was going to be in 6/8 using a calabash loop (a calabash is a West-African percussion instrument made from the dried out and hollowed shell of a gourd), that our kick drum sound was going to be a heavily treated calabash sample, and that the chords would be picked out by three acoustic guitar samples I’d made that morning, each lasting only one bar and comprising of a single arpeggiated chord. I expected Phil to put up a little more of a fight, but he seemed happy with the suggestion, or at least he didn’t just leave the studio immediately... and between us we got stuck in chopping up the three acoustic guitar samples and pasting them onto the timeline under the vocal, paying little regard for the initial chord structure of the piece.
Initially we just built up the guitar chords over the intro and first verse following this with an instrumental chorus built-up from the calabash rhythm. We toyed with various sounds to add to the chorus but both agreed a cello sample-instrument seemed to lend itself to the atmosphere we were creating. While I layed down a cello line using the London Solo Strings library in Kontakt, Phil toyed around with expanding the drum sounds using Battery - this actually got swapped out later, but provided us with a rough idea to jam around. I layered the cello part with a viola part and Phil got stuck into Massive to develop a new bass line for the second verse. At this stage everything is a little sketchy, no production work had been done and while the song was coming together the arrangement was being constantly swapped around, with parts flying backwards and forwards an considerable amounts of tea being drunk.
The bass line and calabash under the vocal of the second verse seemed to be working but going back into the chorus seemed a little lack-lustre so we brought up another bass sound in Massive and layered this under the chorus. The other problem we had in the second verse was the line “I folded your heart into mine to make a paper dove”, which for some reason we kept mis-hearing as “make a paper towel”. This isn’t something anyone else would probably spot until you tell them (don’t go looking for it please, Catherine will kill me!), but since it was winding Phil and I up rather than try to mask it I decided to make a feature of it, and delayed the word “dove”, which is clearly enunciated in the solo’d vocal, adding MoogerFooger ring modulation in increasing amounts to each delay. Of course with the word “dove” now delayed we no longer heard “towel” and all was well again.
For the third verse we decided to keep the momentum going by keeping the bass line from the second verse but adding back the acoustic guitars from the first verse, with a new even more choppy arrangement. To prevent the progression of the song from being to obvious though we moved the bass line from Massive to the cello library in Kontakt, this also meant we could drop the strings out from the second chorus, which seemed to us to be a good idea. The strings were the hook in the instrumental chorus, so keeping them out of the second chorus meant more room could be given to the vocal, and by holding the hook back you increase anticipation. Bringing the cello back in reminds you of that hook without requiring us to use it again. The cello actually got moved onto double bass in the end, because we wanted a lower register, but the association with the strings is still there.
For the final chorus we decided to go all out, hitting it with almost everything we had to make it sound huge before dropping away for the final held note. The last thing we added before taking a break for a few days was the banjo part, ably played by my housemate, James ‘Pag’ Pagliero. Phil and I had agreed from the start that we wanted a banjo part to bring out the folky vibe a little more, but as we went on we started hearing more eastern themes in the vocal, so we weren’t entirely sure that was the way to go until Pag picked the banjo up and almost immediately played the riff that went onto the track. It seemed to capture the spirit perfectly, especially when I dropped a bit-reduction plug-in over it so that it fitted with the glitchy cut-up feeling of the acoustic guitars.
After a few days break to clean our ears, Phil sent over the synth pads and new drum beats and I cut and pasted audio and MIDI regions from throughout the track to create the outro, adding some of the new drums Phil had provided.
The arrangement was finally taking shape so I started to work on a mix, which I bounced back and forth with Phil. A couple of parts were added, removed or just moved but on the whole we were happy with it, the crackles and heavily automated bit-crushed delays were added to try and build around the atmosphere of the chopped up guitars and 8-bit banjo. We didn’t have time to head into Metropolis to get the mix done, so everything was done in the box before I handed it over to Phil who mastered it himself at Metropolis.
The final remix can be heard on the PEACE compilation, which is available here along with 180 other tracks, including artists like Mogwai, Patrick Wolf, Fennesz, Amiina and Vieux Farke Touré for a €5 donation to Amnesty International. For comparison, the original single is available on the Carry your Heart EP here, or for streaming on Catherine’s SoundCloud here.
My thanks to Catherine for letting us turn her track inside-out. We were quite worried when she came for a playback, but she must have liked it, because she asked me back to mix her new EP (and she asked Phil to master it at Metropolis); Skeleton Songs, which is available for pre-order now here. Full release 7 June, 2010.
Skeleton Songs was recorded as a raw as possible by Catherine, using her home studio to capture the first take of every part and was mixed in a single day at Metropolis. However, you wouldn’t know to listen to it.
Phil and I had done a couple of things together before this remix, but since doing this one we’ve really hit the ground running and we’ve got a bunch of stuff in the pipeline and a few things already in our “done” pile, of which the only one you can hear so far is the Tunng remix up on our SoundCloud account, but we’ve got a few remixes and some original material we hope to unleash soon. But we’ve still got our ears to the ground so please feel free to get in touch with us by leaving a comment on here, or e-mailing me here, if you think you’ve got something that you’d like us to take a look at.