Cloud DAB+ encoding for small-scale radio. The same bitrate, sounding considerably better.
Three things. Everything else on this site is explanation.
No change to your capacity, your carriage bill or your processing. The same 32 or 48 kbps, with far less of your sound damaged on the way to the listener.
One encode, one clock, every multiplex that carries you. Drive from one coverage area into the next and nothing jumps, skips or changes character.
Station text and slideshow images normally take a fixed slice out of every frame of your audio, whether the audio can spare it or not. Ours sends them through the quiet moments instead. At 32 kbps and above you cannot tell whether images are running or not – so run them.
It is a reasonable assumption, and it is wrong.
The DAB+ standard defines two things: the format of the data sent over the air, and what the receiver in your listener’s kitchen must do with it. It says almost nothing about how an encoder should decide what to put in that data. Those decisions – which sounds to keep, which to discard, how to spend a limited number of bits across a constantly changing signal – are left entirely to the encoder. That is where all the quality lives, and every encoder makes those decisions differently.
The clearest proof comes from Fraunhofer, who invented AAC. They give FDK-AAC away free, and they separately sell a commercial AAC encoder which they market on sounding better than the alternatives. If the standard determined the result, they would have nothing to sell.
FDK-AAC is a genuinely good piece of engineering and it deserves its reputation. Fraunhofer originally wrote it for Android handsets, and it has since become widely regarded as the best free open-source AAC encoder available. It is well suited to low-cost DAB+ encoding on modest hardware – a Raspberry Pi, or a general-purpose shared virtual machine in a datacentre – which is exactly what small-scale DAB was designed around, and it is a large part of why carriage is as affordable as it is.
It achieves that by making fast decisions rather than exhaustive ones. That was the right engineering trade for the hardware it targets, and it is why it will run on a computer the size of a credit card.
We made the opposite trade.
We have spent six years on an encoder that does the opposite. Ours analyses before it decides, tries alternatives and measures the results, and works to a precision that only makes sense when processing is not the constraint. It is many times slower than FDK-AAC, deliberately. It will not run on a Raspberry Pi, and it will not run usefully on a general-purpose virtual machine: it needs a modern multi-core processor with vector extensions, doing nothing else.
That is one of two reasons we run it for you rather than selling you software. The other is licensing: we are not in a position to distribute it. It runs in our own secure datacentre on hardware chosen for it, with redundant connectivity, and all of that is included in what you pay.
We are asking you to believe that encoders differ, so here are two things you can verify without trusting a word we say.
We have monitored over 100 UK small-scale, regional and national DAB multiplexes off air. Not one of them varies its bit allocation within the superframe – every frame gets the same share whatever the audio is doing. You can confirm this yourself in a few minutes with the free etisnoop tool. Then look at one of ours.
Send us a sample of your output and we will encode it at a range of bitrates through our encoder and through ODR-AudioEnc – the open-source encoder built around FDK-AAC that is the de facto standard across UK small-scale DAB, and almost certainly what you are on air with today. The same audio, the same rates, both sets of files returned to you along with the settings used for each, so the comparison is one you can reproduce. Your own ears can settle it.
Seven things, in plain English. Each one is explained in full on How it works.
Ours analyses a complete superframe before encoding a single bit of it, then gives the capacity to the moments that need it. As far as we have found, nothing else on UK DAB does this.
The two need opposite treatment, and a fixed setting is a compromise by definition. Ours detects which it is hearing and changes its approach, around the clock, with nothing for you to set.
Ours decodes its own output, compares it against your original, and keeps only the best result – hundreds of times a second.
Convincing stereo from 24 kbps upwards, at rates where most encoders give up on stereo altogether.
Station text and slideshow data ride through the quiet moments instead of taxing every frame. At 32 kbps and above you cannot tell whether images are running or not.
True floating point throughout the analysis, so quiet detail sitting alongside loud material stays visible – and an encoder can only protect what it can see.
One encode, one clock, every multiplex. Drive between coverage areas and nothing changes.
A FLAC or MP3 stream address from your playout – or use the free Icecast mountpoint we give you, so you can send the best quality your connection allows without paying anyone for a second feed.
Which multiplexes, at what bitrate. We contact each operator and arrange the handover of your service. You do not need to.
We encode, deliver, monitor and maintain. Your text and slides come from files your playout already produces. You get an email when you are live, and you never touch it again.
Any multiplex running a standard EDI input can take our feed, which covers essentially every small-scale multiplex in the UK.
Get startedSend us a sample of your output and we will encode it at a range of bitrates through our encoder and through ODR-AudioEnc, the open-source encoder built around FDK-AAC that is the de facto standard across UK small-scale DAB. Both sets of files come back to you with the settings used for each.
Send us a sampleMaxxwave has been building radio systems since 1953, when the company started by converting ex-military transmitters for commercial use.
Audio coding is not a new departure for us. We designed and built the MW-CODEC, a four-layer resilient audio codec for studio-to-transmitter links: a high quality main stream, interleaved error correction spread across twenty-one seconds, a redundant low-rate stream, and a synthesiser that reconstructs intelligible audio when everything else has been lost. It keeps a station on air through ninety per cent packet loss. We wrote every line of it.
The narrowband voice codec we built for our Ambitalk radio network won the Gerald David OBE Innovation Award at the Business Radio Awards in 2015, an International Critical Communications Award in 2016, and two Comms Business Awards in the same year.
Alongside that we have designed an advanced DAB exciter from the ground up, including the FPGA code that carries out the modulation, pre-distortion and peak-to-average power reduction, along with a thirty-two band audio processor, trunked radio logic controllers, radio modems for the water industry, and guidance software for autonomous agricultural vehicles – where an error of a few centimetres, accumulated through long chains of arithmetic, puts the sprayer over the crop instead of between the rows. That problem and the one this encoder solves are closer than they sound.
We are the UK’s largest independent radio site owner and operator. We have designed, built and run national radio networks carrying safety-of-life traffic across more than a hundred transmitter sites, so operating at scale is not new to us. This encoder was not built to serve one or two multiplexes.
And when we found and fixed a limitation in the open-source multiplexer that most small-scale DAB runs on – one that was capping multiplexes at around fifteen to thirty services – we contributed the fix back to the project free of charge. It is in use across the industry now, including on multiplexes run by our competitors.