Feature: Investigating interface protocols
With USB 2.0, Firewire and now Thunderbolt and USB 3.0 kit all viable options, what are the pros and cons of each?
It’s arguably now harder than ever to pick the right audio interface, with USB 2.0, Firewire and now Thunderbolt and USB 3.0 kit all viable options. But what are the pros and cons of each? Stephen Bennett asks a number of leading manufacturers.
One of the great advantages of the ‘digitisation’ of the audio industry is that those engineers designing audio hardware can take advantage of the advances in general computer connectivity and data transfer protocols.
It wasn’t so long ago that transferring a couple of channels of reasonably high-quality audio were only within the purview of high-cost and bespoke physical interfaces. Today, hot-pluggable interfaces enable the engineer to deal with many channels of high sample rate, 24-bit audio over USB, Firewire, Thunderbolt or Ethernet.
But, to paraphrase a great wig, the great thing about interface protocols is that you can’t have too many of them. So is there a requirement for faster and more capable ways of shifting data from microphone to computer and back, or do we already have all the capabilities that we require?
For many years, USB 1 and 2 and Firewire 400 and 800 (IEEE 1394) interfaces have been the staple connection protocols of the audio interface sector and there are still many devices that utilise these. The convenience of being able to unplug hardware without powering down and the implementation of bus power has made these types of interfaces popular with engineers, especially those who work with laptops. These protocols’ sustained data transfer rates have provided adequate bandwidth for the most common interface specifications on the market to date – that is up to eight channels of analogue audio input and output, eight channels of ADAT and two channels of digital audio at reasonable latencies.
However, with the advent of digital mixing desks capable of shifting hundreds of channels of audio over digital multicores and the need for post studios to channel terabytes of multichannel audio from workstation to workstation, manufacturers are looking at newer interface protocols to cope with this requirement to handle more data at lower latencies. The latest USB protocol, now at version 3.1, offers a significant upgrade to the data transfer rates of its predecessor, while Apple has championed the PCIe-based Thunderbolt protocol on its Macintosh computers. But, of course, specifications are not the only story, so what do these developments mean for the makers and users of audio interfaces and other digitally connected devices?
Still Got It
Wesley Smith, product manager at PreSonus, believes the older protocols still have a use even in a world that is demanding ultra high-speed interfaces. “Because Firewire 800 and 400 are fully compatible with Thunderbolt with just a simple adapter, these transports are still great for OS X users and offer a more affordable alternative to Thunderbolt interfaces,” he explains.
Matthias Carstens , RME’s head of development, says that the company no longer supply Firewire-based interfaces, but still finds the older USB 2 protocol useful. In fact, he expresses concerns about USB 3 becoming a ‘universal’ interface protocol. “USB 3 is certainly no better than USB 2 for our applications,” he comments. “It is known to have serious performance problems in many ways – there simply is no certification.”
He also believes that the USB 3 protocol “makes no sense at all as an audio interface format for the mass market, because the latency is exactly the same as what RME can achieve over USB 2”. According to Carstens, USB 2 is “much cheaper, easily fast enough, and can transfer enough channels – as proven by RME’s units which can transfer up to 70 channels both ways via USB 2”.
He continues by saying that the higher data bandwidth allows USB 3 to transfer more channels, but it does not affect the smallest buffer size and the reliability when using it with Mac or Windows computers. “This is why we have only implemented USB 3 for our 196-input, 198-output MADIface XT interface, designed to be used in specific professional situations,” he adds.
Carstens feels that despite all its technical advantages, Thunderbolt hasn’t made as much headway into the Windows world as Intel had hoped, which could limit the faster version 2’s uptake with audio interface manufacturers. Indeed, Intel appears to be championing USB 3 on its new motherboards, much to Carstens‘ concern.
‘As I’ve said, RME offers only one USB 3.0 interface today [the MADIFace XT, pictured above] because we can always guarantee the number of channels required, with the lowest latency figures and the highest reliability – so this means that our interfaces will work equally well on any computer, new or old, Mac or PC. Thunderbolt has only one specific advantage – it offers latency as low as PCI Express, because under the hood it is just PCI Express!” He also predicts that USB 3.1 will offer no advantages over version 3. ‘There’s no change in latency – just the ability to have more channels that no one really needs. I don’t expect any audio interfaces that specifically require USB 3.1 to appear within the next five years – simply put, no one needs it.’
Lev Perrey, director of product management at Universal Audio, is a fan of Thunderbolt. “We chose Thunderbolt as our main protocol for our Apollo interfaces on Mac due to the many benefits it provides – most notably high bandwidth, low latency and aggregatibility,” he reveals. “We’re distributing audio, clocking, and real-time UAD processing all over a single Thunderbolt cable. Thunderbolt also allows us to chain different Apollo models and allow all of them to work together in a single elegant system.”
Smith, however, champions the latest USB protocol, while also pointing out what Thunderbolt offers in comparison. “The biggest advantage of USB 3.0 is that it is virtually ubiquitous across all platforms,” he states. “While Thunderbolt offers some great latency advantages on OS X, not very many Windows machines provide this connection. USB 3.0 has the speed and stability to support high channel counts at low-latency. It delivers the capabilities we need now to the widest possible user base.”
Roger Robindore, director of product evangelism at Apogee Electronics, agrees with the aforementioned advantages of Thunderbolt. “It’s built on a PCIe backbone, and this has always been the best way to connect audio interfaces to computers where high bandwidth and low latency are required. One of the PCIe protocol’s most important features is direct memory access (DMA), a function whereby the PCIe peripheral – in this case, the audio interface – can send and receive data directly with the computer’s memory, with little or no involvement of the computer’s CPU,” he notes. “This direct data connection is one of the primary reasons that Thunderbolt has such low latency and rock solid stability.”
And like Carstens, Robindore is of the opinion that USB 3 delivers little more than USB 2 where it counts. “One important thing about USB 3 is that it doesn’t offer a benefit over USB 2 with lower I/O count audio interfaces that are amply served by USB 2 bandwidth. USB 3 does not offer better latency performance than USB 2,” he says. “We use a simple airplane analogy to explain this – on the flight between Los Angeles and San Francisco, a larger 747 can carry more passengers than a 727, but it doesn’t get you there any faster.”
Furthermore, Robindore explains that lowly USB 2 maintains some advantages even when compared to Thunderbolt on interfaces with lower channel counts. “It’s significantly less expensive while still offering excellent latency,” he continues. “It also offers the possibility to connect to mobile devices like the iPhone and iPad.” Robindore also agrees that Thunderbolt connectivity is not ubiquitous on modern computers – a point that is particularly relevant to Apogee’s new Mac- and Windows-compatible Groove high-performance DAC. It appears that if you currently want cross platform capability, some flavour of USB is going to be the best way to achieve this.
Perrey concurs: “Although Thunderbolt is available on Windows, it isn’t integrated the same way as it is on a Mac. USB 3.0 provides native support in Windows OS 8, so we are exploring the benefits of this for Apollo on Windows,” he reports.
Another of Thunderbolt’s many plus points is that it is fully under Intel’s control, Carstens says. “It is guaranteed to work,” he proclaims. “On the downside, Thunderbolt is expensive to implement, usually allows only short – and expensive – cables runs and is (mostly) limited to Macs. The big Windows takeover never happened and right now it looks like Thunderbolt only has a little more time before it follows Firewire into obscurity – a situation not helped by Intel’s latest move to add it to a USB 3.1 C-type connector. That’s a pity, in some ways, because it means external PCI Express has never made it outside industrial applications and never gained real plug-and-play behaviour on Windows computers.”
Above: The Ensemble Thunderbolt interface from Apogee
'One Connector to Rule Them All?'
In a fast-moving technological world, manufacturers of high-quality audio interfaces are always looking towards future developments and how these might be useful in an audio context, as Robindore explains: “We’re watching the development of the USB Type-C connector, especially with Intel’s recent announcement that Thunderbolt 3 will employ it. This means that the audio industry could finally have ‘one connector to rule them all’ that could eventually be used in anything from mobile devices to high-bandwidth professional systems,” he reveals. “Hopefully users won’t be confused by the fact that protocols with vastly varying performance can run across a Type-C connection.”
This convergence of protocols could finally mean that engineers would not have to carry around a case full of leads and converters every time they leave the studio, while leaving open opportunities for manufacturers to innovate and create products that would meet the varying requirements of their various customers.
One connection protocol that is ubiquitous on both Apple- and Windows-based machines is Ethernet. Currently mainly in use for distributed audio and large-scale networked systems, the protocol has been championed by Merging Technologies. “There are a number of audio-over-IP protocols now established in the pro-audio market that use Ethernet as the connection – the main players being Ravenna, Dante and Livewire,” details Paul Mortimer, managing director of Merging’s UK distributor eMerging. “Compatible devices can be connected using a simple point-to-point connection or via an existing standard IT network infrastructure.
“The main advantages of Ethernet-based formats are the ability to run much longer distances between devices; being able to take audio signals from one source and route to many destinations; and sample accurate clocking from one master device on the network. Thunderbolt 2 and USB 3 offer the ability to connect to Ethernet, so would also be compatible with networked audio devices. With the introduction of AES67, all of these audio-over-IP formats will talk to each other, so enabling one harmonious compatible format.”
PreSonus could also make use of the Ethernet protocol in the future, states Smith. “Audio networking solutions hold great promise – especially for higher-end products. The ability to connect audio with Cat cables for long runs is a huge advantage for audio transports. Being able to pull any stream off the network to record makes them great for larger installations and super-high channel count situations,” he notes.
Perrey says that Universal Audio is also looking ahead. “Thunderbolt 3 was just recently announced, allowing for even more bandwidth over a USB Type-C connector, and it will be backwards-compatible with USB 3.0 as well as Thunderbolt,” he explains. “It feels like the natural progression of the protocol, and reinforces our decision to use it as the connector for our scalable system of Apollo interfaces.”
The prospect of a single protocol that provides all the capabilities that engineers might require in their audio interface is a beguiling one, but also one that is unlikely to bring to an end the heated discussions between manufacturers over the best way to implement the emerging underlying technology in future products.
Stephen Bennett has been involved in music production for over 30 years. Based un Norwich he spilts his time between writing books and articles on music technology, recording and touring, and lecturing at the University of East Anglia.