Tuesday, 2 August 2016

Bluetooth audio without the delay

I often get people asking my questions about Bluetooth headphones and speakers.

Typically they'll say something such as "My new headphones sound great, but there’s a noticeable processing delay, which means I see ‘lip sync’ issues – when people talk their words don’t quite match up with their mouth, it’s very frustrating and makes watching TV quite difficult" or "Can you recommend a wireless speaker which doesn’t delay the sound". People seem just as interested in reducing that delay as they do in the overall sound quality.

Pretty much any digital signal is going to suffer this effect to a great or lesser extent. As those of you who’ve worked in networking or electronics will know, this delay is known as ‘latency’. It’s caused by a number of factors including the original analogue to digital conversion, buffering, compression and other signal processing, processing the data into packets and transmitting it, and then pretty much the same at the other end, in the reverse order.

If you’ll allow me a slight diversion for a moment, the biggest digital latency problem that I can think of is the Greenwich Time Signal, also known as ‘the pips’ when played out across DAB radio. They are often wrong by several seconds, meaning that as a ‘time signal’ they are pretty much useless. You can hear this delay yourself if you place a DAB radio next to an FM one, and compare the output on the same station. What you might not also realise, though, is that for a few years now most FM stations have had a delay too, due to the way that the signal is delivered to the transmitters using NICAM encoding. FM will typically have about one second delay, which is wrong, but probably just about usable for most people. It would work in an “OK chaps, let’s synchronise watches” scenario. DAB, though, can be around 6-8 seconds out – the delay is variable, depending on the brand and age (and thus processing speed) of your DAB receiver. Any such delay obviously makes the time signal pretty useless. I once made an official suggestion to the BBC that they add a ‘warble’ to the time signal on DAB (and indeed internet streams) so that listeners would know that the signal isn’t accurate, but this fell on deaf ears.

Anyway, back to wireless headphones and speakers. As Roy points out, most of these seem to be Bluetooth driven these days. In fact, I just did a quick search for Wireless Headphones on the Currys website: At the time of writing it lists 88 products, of which 84 use Bluetooth. I’m not going to get into arguments here about the audio quality of Bluetooth audio – especially as my cloth ears seem to max out at around 192kbps VBR MP3 these days – I really can’t detect any improvement at higher signal rates, or when compression is removed. So for me at least, quality wise, Bluetooth based headphones or speaker can sound plenty good enough, although I can appreciate the better sound quality of some models over others. Although by ears are poor I still hate tinny, or over-boomy sound. But that’s more about construction, speaker size, etc. than the transmission medium or compression used. For me it’s all about balancing good quality sound and convenience/portability. With headphones in particular some can feel very uncomfortable when used for extended periods.

Let’s look again at audio latency: With Bluetooth speakers or headphones the delay will typically be in the 100-200ms range, although some products have a delay of a whopping half a second. But if we take a typical UK video source running at 50Hz, each frame lasts 20ms. Incidentally, it’s a common misconception that TV is captured at 25Hz because of interlacing. That may have been the case some years ago, but these days many of us watch 1080P (or better) sources, which aren’t interlaced and which have a proper 50Hz or even 60Hz frame rate. Hardly anything is filmed for TV at 25Hz these days, except when people are aiming for a deliberate ‘archive’ look.

So at 20ms per frame, audio which is delayed by 200ms will be ten frames out. That’s very noticeable.

There have been various studies into how bad an audio delay needs to be before it causes so-called ‘lip sync’ problems, but one of the most respected studies is from the American ‘Advanced Television Systems Committee’ and it recommends an absolute maximum of 45ms – see http://www.webcitation.org/60UbU5Ziv. An earlier study by the European Broadcasting Union had suggested that 125ms was acceptable. The point at which the delay becomes unnoticeable appears to vary from person to person – there’s no fixed point, so all of these studies try to adopt a sensible “most of the population” recommendation.

The problem with these studies is that they allow broadcasters to work within these tolerances, and so in many cases the material you are watching will already be pushing at the limits. So even if your own equipment is also within these limits, the compound effect of both delays added together can push things into the detectable range. This means that you need to do everything you possibly can to keep it as low as possible.

So what can you do? Well, for starters there’s equipment out there which doesn’t use Bluetooth for audio. In fact it doesn’t use digital processing at all. I mentioned above that 84 of the 88 headphones on sale at Currys use Bluetooth – well, that means there’s four that don’t. I’m pretty sure they all use 868Mhz analogue RF channels. That means they essentially have a zero milliseconds delay.

When it comes to wireless speakers there are also a few that use analogue RF channels, but they tend to be at the lower end of the market, and are pretty dire. More expensive wireless speaker systems usually have ‘better than Bluetooth’ latency. For example, with Sonos speakers it’s around 70ms. That’s still enough to sometimes notice lip sync problems, but it’s better than most.

So what’s the answer? Well, the best one I’ve found so far is Bluetooth. But not normal laggy old Bluetooth. What many people don’t realise is that Bluetooth Audio isn’t always the same. For starters there are a number of different ‘profiles’ which can be used to transmit audio. In the early days these were mainly designed for telephony applications, but then along came A2DP or Advanced Audio Distribution Profile which offered much better sound quality. By default, A2DP uses a codec called SBC, or Low Complexity Subband Coding (no, I can’t work out how you get from this to SBC either). In fact there are several different versions of SBC, but like most things Bluetooth the transmitter and receiver do an initial handshake and then choose the highest standard acceptable to both.

But the spec allows for other codecs to be used too, and this is where we can start to shave a shedload of time from the standard Bluetooth latency. In particular, there’s a commercially licenced codec called aptX Low Latency which comes from a company called CSR. Using aptX LL (as it is usually called), the latency is reduced to 32ms. That’s just about the best you are going to get for wireless digital headphones or speakers, without stumping up for products at the megabucks end of the market, using custom audio protocols.

The aptX LL codec has to be specially licensed by product manufactures, so you won’t find it available on a massive range of kit. In fact there’s a fairly up to date list on CSR’s website : http://www.aptx.com/products-low-latency/browse/categories. Oh, and please don’t confuse aptX LL and normal aptX. The latter is all about getting sound quality from Bluetooth, but it doesn’t address the latency issue in the same way that the LL version does.

In terms of speakers, I really love Denon’s Envaya range. There’s a ‘mini’ version, and I reckon that in many respects it actually sounds better than its bigger and more expensive sibling (although bear in mind what I said about my hearing). You can pick up the Envaya Mini for around £70-£80, and that has to be something of a bargain for such a quality product. Mine is probably one of my most used gadgets, and I don’t only use it for low latency applications. It also regularly plays music streamed from my phones and tablets.

Bear in mind that if you want to banish lip sync problems, for aptX LL to work it needs to be supported at both ends. This means that if you’re going to use one of these speakers of headphones for listening to an audio-visual source such as TV you’ll also need a transmitter which supports aptX LL too. As you’ll see from the CSR website there are quite a few of these, but I’m a huge fan of the Avantree Saturn Pro. It’s a small puck-like device – one will easily fit into the palm of your hand. Powered by a rechargeable battery it’ll go around 10 hours between charges. Or if your telly has a USB socket you can use this to keep the device permanently powered. It’s important to get the Saturn Pro – the non-pro version only supports the normal latency version of aptX.

The great think about the Saturn Pro is that it’ll work as both a transmitter or a receiver (there’s a little switch on the side). And that’s a really good thing because if you take a look at the range of headphones with built-in aptX LL on the CSR website you’ll see that they are all quite high-end, with price tags to match. But with the Saturn Pro running in Receiver mode you can simply plug your favourite wired headphones. If you need them, you can get a pack with two Saturn Pros, ready paired. The packs are usually about 20% cheaper than buying the devices individually, which is always nice.


No comments:

Post a Comment