High Fidelity Sound

HD Audio – myths, facts and Tidal

Beginning with Neil Young’s Pono system, the last couple of years have seen a few attempts to sell ‘HD Audio’ to consumers. Usually, technical specs on what ‘HD’ actually means are pretty hard to come by. Instead, these products are typically accompanied by a press release boasting that it ‘sounds just like the master tapes!’

I call snake oil. Here’s why.

Selling a hi-fi in the 21st century

Hi-fis haven’t got much better since the ’70s. Quality comes cheaper as components become cheaper to manufacture. So you can buy better for less. That’s about it. Why? Amplifiers and speakers are pretty simple pieces of electronic kit, bound by physical laws. There’s not really a lot else you can do.

As listening habits have shifted from the living room hi-fi to laptop speakers, earbuds and worse, manufacturers have hit upon two things:

  1. They need to sell hi-fis as a premium product to older consumers with disposable income
  2. This demographic already loves the HD video their blu-ray players gives them

HD is a recognisable tag from video. And many blu-ray home cinema systems already talk about HD audio – correctly, as blu-rays typically contain lossless audio, as opposed to the 192kbps stream usually found on a DVD.

So not HD as such. There was no great leap forward in digital audio quality. It’s just that what the guys in the post-production suites were listening to could now fit on a disc alongside the video rather than being heavily data compressed. Not-Shit-D rather than HD.

Better than CD!

There are two ways in which audio manufacturers are trying to sell high definition audio to consumers. The first is sample rates higher than CD quality, the second is in thinly veiled re-packaging of higher quality DACs (digital to analogue converters). The former is snake oil, the latter is welcome.

Let’s dive into sampling theory for a second. CD is at 44.1kHz, 16 bit. That means there are 44,100 audio ‘slices’, each containing 16 bits of information on a CD. Given this frequency, that means CD quality audio could technically reproduce a tone at 44.1kHz as loud as 0dB and as quiet as -96dB.

The way audio is read from a CD actually means it can reproduce half that frequency (look up Nyquist curves and enter the world of esoteric digital mathematics). Which is fine because the human ear can only hear up to 20kHz at best. This drops off with ageing. I’m 30 and can hear up to 18-19kHz which is great for my age.

All higher sample rates – 96kHz say – are doing is reproducing higher frequency information that we can’t hear. That means pumping out audio that’s putting more strain on your amplifier etc. for no reason. It might sound different to you but what you’re responding to is distortion further down the spectrum caused by the unnecessary frequencies at the top.

It’s worth working at higher rates (as I do) because certain processes – particularly non-linear dynamics processing like compression – works better at higher sample rates. But for listening? Nonsense.

As for 24 bit? Possibly. Again, all it means is that the dynamic range extends to -144dB instead of -96dB and no one can hear a signal that quiet. But it does remove the need for quite such heavy handed dithering, and not all dithering algorithms are equal. So there is potentially benefit in listening to material that hasn’t been dithered to 16 bit.

Digital-to-Analogue-Conversion for ‘norms’

Higher quality DACs really do make a difference. My Benchmark DAC-1 cost me £900, and you don’t spend that money for nothing. But it’s a very nerdy thing to buy for a hi-fi; you’d have to be a serious audiophile to even think about it.

A friend of mine brought round his Pono player (‘just like the master tapes!’) I listened to it. Could hear how the transients were sharper and the top end smoother than say, listening directly from the output of my 1980s CD player. Then we played the same song going through my mid-level Arcam DAC. The Arcam won. But by being better at the same thing.

Pono, and other systems like it, are very cagey about what they’re actually selling. I’m convinced they’re packaging higher quality DACs in prettier, non-nerd friendly packaging. I’ve no evidence for this, but given the physical limits of human hearing, can’t see any reason why they would be doing anything else. And that’s fine by me.

Part of the reason people prefer the sound of vinyl is because it doesn’t go through digital to analogue conversion. It doesn’t have to. So it’s not muddied by cheap electronics. When people wax lyrical about how ‘solid’ and ‘real’ vinyl sounds – they’re describing the same effect a high quality DAC has on lossless digital sources.

Tidal – ‘high fidelity’ streaming

Which brings me to Tidal. As clumsy and embarrassing as the star-studded relaunch was, there is something there. I’m an early adopter of Tidal. Why? Because it’s ‘high fidelity’.

What do they mean by ‘high fidelity’? Lossless. Pure and simple. It’s CD quality FLACs in streaming form. They’re not selling snake oil, they’re selling the real deal. Ignore all the guff about ‘curated content’, paying artists more royalties (this is also nonsense, but that’s for another post) and how the software is still a bit shonky at this early stage, and what you get is Spotify without the horrible, lossy audio.

Tidal will probably get washed away [sic] by all the negative press its received, but it’s paving a way forward. Or rather, back to what we had.

What’s real HD audio? A CD quality audio played through a good DAC. Been the same since the ’80s.

Yes, You Should Dither to 24 Bit

Bit-meterThere’s a lot of noise around whether or not you should dither when bouncing down to 24 bit. But the answer is simple – yes you should.

Without going into the actual mathematics of it (partially because you’ll be hard pressed to find someone who really understands) here’s why.

Word length reduction

When we describe a file as 24 bit or 16 bit, we’re talking about the word length. This literally describes the number of decimal places allowed to mathematically describe an event.

What is common knowledge (in the digital audio community) is that when you reduce the word length i.e. convert 24 bit to 16 bit – you should dither. Why? Because dither helps to better make the transition from a longer word length to a shorter word length. If you don’t dither, your DAW will just indiscriminately chuck out the last 8 bits of data. With it, your DAW will still do that, but a bit more artfully.

Quantization distortion

The reason dither is important is quantization distortion. To put it simply, when chucking out the last 8 bits, your DAW doesn’t know what to do with the last bit – the one that describes the quietest section – and this causes quantization distortion as a result of rounding errors. A gritty, horrible sounding distortion that can affect things much higher up in the dynamic range. It can cause a blurring of the stereo image, phase shifts, lack of detail, lack of warmth – despite occurring at the quietest end of the range, below most people’s hearing, the symptoms are dramatic.

Dither randomises the last bit, creating white noise along the lines of tape hiss. This is far more pleasant to our ears, smooths out the bottom end and prevents all the symptoms listed above.

OK, so you’ve been working in 24 bit all along, and you’re bouncing your mix out at 24 bit to send to a mastering engineer – who will dither when it comes to rendering the 16 bit file – you don’t need to dither right? Wrong.

32 bit float

Probably every DAW at this point works internally at 32 bit float. That means that although your individual track files and project are set to 24 bit, every process that is being undertaken by Cubase/Pro Tools/Logic etc. is happening at 32 bit float. Even a single fader move will mean the DAW is working at 32 bit float.

And that means when you bounce your mix to 24 bit, word length reduction is taking place.

And what do you do when you reduce the word length?

Dither.

It’s important to note however, that this is not the time for your fancy noise shaping dither. That’s definitely down to the mastering engineer. This is just for your garden variety ‘flat’ or triangular dither (commonly described as TPDF).

And word to the wise – your DAW may not automatically dither when you bounce. In order to dither correctly you need to insert a dithering plug-in into the last insert of your master fader (so it’s the last thing to touch your audio).

Hear for yourself

I was recently explaining this to a mixing engineer friend of mine. He was sceptical, having learnt (as most of us did) that dithering to 24 bit was unnecessary, and possibly even harmful. So I suggested a blind test. He sent me the same 24 bit mix – one with a flat dither and one without – labelled A and B.

It didn’t take long to pick out which had been dithered. The dithered mix was more focused, more solid and revealed more detail. He went back and re-bounced all the mixes for the EP for mastering.

Try it for yourself, you might be surprised.

The Art of Making Decisions

I generally prefer to take my time over things. Work on a track, leave it a day, make a few changes, leave it a day, change it again – each time the changes getting diminishingly smaller until I load it up one morning and go “yeah”.

I’m certainly not alone in that approach. But sometimes it’s good to be forced into making decisions quickly, and every now and again a job comes along that reminds me that even when you’ve got the time, sometimes it’s better not to use it.

The Mercedes effect

The biggest job with the shortest deadline I’ve had over the last year was mastering the online content for Mercedes’ flagship ‘Sound With Power’ campaign. Without going into details, I was given a day – maybe two – to master an enormous amount of material. It worked out to about 15 minutes per track.

Granted, the tracks were short, and most of them were very similar to each other, but that’s still an awful lot to get through in a very short time. Certainly no leaving it for a day, coming back to it… Even if I’d had the time, it would have been tortuous to go back and re-do every single track so many times.

So I worked on the first track for an hour or two, decided on a workflow, signal chain and general direction and just ploughed my way through it.

The lesson to learn

The Mercedes campaign wasn’t the first time I’ve had to work like that, nor will it be the last. Anyone who’s ever done any work for film, TV or advertising will know the excruciating deadlines and sky-high expectations commonplace in those industries. But as hard as they are, every time I have a job like that, I find it incredibly informative.

The key is making decisions. Anyone who’s ever mixed anything will know the torture of constantly switching plug-ins and techniques, flip-flopping between often minute variations until you can’t hear straight no more. And then you’re lost.

What jobs like these teach me is the value of making a decision and sticking to it. Treating it like what it is: engineering. Presented with a problem, what’s the solution? What are the best tools to use? And how shall we use them? Great, let’s do that then. Job done.

It’s easy to get caught up in the creative distinctions between different compressor models, different saturation algorithms… With so many digital tools at our disposal, perfection is just an oversampling option away. But the more you play around, the less perspective you have.

Know your tools

The real trick is knowing your tools. Pare down your options. I tend to use only one of two options for everything in my chain. After years of experimentation I’ve narrowed it down to what works best for what job. And when I get it wrong I try the other one. I very rarely have to reach for anything else.

Once you know your tools you can make informed decisions and stick to them. Mix blindness is a far worse enemy than choosing arguably the slightly less appropriate tape saturation.