Last month, Paul Wiffen explained how to get MIDI into and out of your new‑style Macintosh. This month, he turns his attention to the tougher problem of audio... This is the second article in a four‑part series.
I would like to begin this month by thanking all those who contacted me after last month's instalment was published. Many of you offered more information on the whole MIDI/USB scenario, both from a technical point of view and from empirical experience. Some of this is so useful that I have summarised it in the 'Other Voices' box on page 104. Keep those comments coming — they are most welcome and will prove very useful, particularly as I have just been asked by the Editorial team at SOS to take over writing the regular Apple Notes column from next issue. As a result, I will now have a monthly forum to air your experiences and discoveries and increase our shared knowledge pool. If you wish to send me email comments, you can now start to do so at applenotes@soundonsound.com.
USB or PCI For Audio?
Most SOS readers will be aware that the amount of data in a digital audio stream is vast in comparison to anything that ever flows down a MIDI connection. You might therefore expect the kind of bandwidth‑sharing problems which affect USB for MIDI to hamper the smooth passage of digital audio as well. Sadly, some people's experiences of USB Audio have indeed been negative thus far — like SOS's very own Vic Lennard, who detailed the problems he encountered running USB audio via the Apple Sound Manager on last month's Crosstalk page (see SOS November 2000, page 254). I will also be citing a couple of examples from my own experiences with USB audio installations later on in this article.
Having said that, the USB audio situation is far from static. It feels as though new USB audio interfaces are being announced every week, and most Apple owners are eagerly awaiting the arrival of Mac OS X to see whether it improves the implementation. Nevertheless, as a result of my experiences so far, my advice at the time of writing is to use USB for audio I/O only if there is no other way to achieve this in your setup (rather like my advice on MIDI last month!). In other words, I think the best way to get multi‑channel audio in and out of blue‑and‑white G3s or G4s for the time being is the traditional one — an audio card in one of your Mac's PCI slots.
It's good to know that purchasers of new G4s still have recourse to this option, but there are many other Apple computers which do not have any PCI slots, such as PowerBooks (more on these in a moment) and of course even G4s only offer three slots. Moreover, I think new computers like the PCI‑less Cube provide a clear indicator of the direction Apple would like to take. All of this makes the need for a new standard way of getting audio in and out of Macs rather pressing. As detailed in the 'Introducing mLAN' series which finished in last month's SOS, it is possible that a digital, FireWire‑based protocol such as mLAN may prove our salvation, but nothing is definite yet, so PCI cards remain — for the moment — the best audio solution for most G‑series Mac owners. The only caveat to this concerns the new multi‑processor G4s — I have heard rumours of compatibility problems between these and certain PCI cards. I have requested an evaluation dual‑processor Mac to allow me to conduct tests with various PCI cards, and will of course report on any problems I encounter, either in a future article or in Apple Notes. In the meantime, I would suggest you take my standard advice if buying a new card with one of the new multi‑processor G4s — ensure that you get a cast‑iron guarantee of at least a swapout if there is a compatibility problem (indeed, I recommend this strategy in all such joint purchases of computers and peripherals).
There are so many PCI‑based audio solutions on the market that it is impossible to go into them all here — and indeed, since the basic idea of using a PCI card for audio I/O remains unchanged if you move from a Power Mac or beige G3 to the translucent G3s and G4s, you could argue that further detail on the subject falls outside the scope of this series. Consequently, I will be writing a separate SOS article on what to look out for when buying a PCI audio card. Without wanting to digress too far into subject matter more rightly covered in that article, one problem that hampers the card method of audio I/O is worthy of mention here, as it is also a concern when getting audio in and out of modern computers via USB — latency.
Routing audio from the inputs to the outputs through the CPU on a computer requires a certain amount of processing which in itself takes time — only fractions of a second, but often enough to become noticeable if you are recording new audio into a computer while listening to it playing back tracks you've already recorded, as you have to when overdubbing.
Many musicians are unaware of the problem of latency until they have already bought some form of audio expansion card. Fortunately, if some care is taken with the audio driver design, it is possible to reduce the latency introduced by many audio cards to a level where it is no longer noticeable. This is why Steinberg came up with the idea of the ultra‑efficient ASIO driver (and Emagic later followed up with their equivalent, EASI). Most soundcards show a major improvement in latency when using such drivers. But what of the other Macs which do not allow the use of PCI cards?
Non‑USB Solutions For Non‑PCI Macs
G3 PowerBooks, although not capable of taking a PCI card directly, do offer various other ways of expanding their audio capabilities without turning to USB. The most obvious of these is the PCMCIA slot, which shares many of the same characteristics as PCI. Sadly, the only game in town in the PCMCIA card stakes so far is made by French manufacturer Digigram. Their stereo VXPocket is the only shipping PCMCIA card which works on all the G3 PowerBooks (Wall Street, Lombard and Pismo series). It does give you up to 24‑bit I/O with ASIO drivers in both balanced analogue and S/PDIF formats. The sound quality of the card is also excellent — if you're recording in the field or collecting sounds for later use, there is no better solution. But sadly, as a means of multitracking and overdubbing audio, it is less than ideal. Even using an ASIO driver, the VXPocket has a latency of 46 milliseconds, making it impossible to use for overdubbing unless you monitor your overdub instrument independently of your Mac.
Although a much bulkier (and more expensive) option, you will get much better performance on a G3 PowerBook from a PCI expansion chassis. The only range I am aware of which work on PowerBooks at all is from Magma. Unfortunately, the compatibility issues are much more complex; of the three series of G3 PowerBooks, Magma have only been able to get their chasses working on two. The first series, known as Wall Street (the heavier, chunkier version that still has a serial port and SCSI) work with Magma's Expansion Bay connector, which can in turn be used with their two‑, seven‑ or 13‑slot systems. I have been using the two‑slot chassis with various PCI cards for the last 18 months, including Korg's 1212 and OASYS cards, the Sonorus STUDI/O, and RME Hammerfall, and I have seen it working fine with the MOTU 2408. Sadly, if you have the Lombard (the middle series G3 PowerBook, lighter but still with SCSI) then you are out of luck, as Magma have been unable to successfully link either via the Expansion Bay or PCMCIA. As a result, you will have to choose between the Digigram VXPocket or the USB devices covered below. The most recent Magma two‑slot chassis, the CB2, has a much nicer moulded case with the same footprint as the PowerBook, and uses a PCMCIA card to connect to the latest Pismo G3 PowerBooks. Although the CB2 has only been shipping a short while, compatibility signs are good. So far, the Korg cards and Digi 001 have been proven to work, although Magma are still working with Digidesign on getting larger Pro Tools systems up and running. As always, until more documentation is available on compatibility, insist on the option of swapping your card if it will not work in the chassis.
Making USB Audio Work
If you have an iMac/iBook or the new G4 Cube, you don't have the option of installing either a PCI or PCMCIA card under any circumstances, and USB will be your only option for audio I/O (or MIDI, for that matter, as detailed last month).
Although Apple have officially been supporting USB audio via the stereo‑only Apple Sound Manager since the release of Mac OS System 9 late last year, the actual results have been variable. Things seem to have improved in System 9.0.4, but there are still the kind of problems which Vic Lennard identified in Crosstalk last month, namely occasional noisy playback until you reboot and/or unstable performance. Latency has also been a problem for USB audio I/O. Although you can now get all the USB devices I have tried working via the Apple Sound Manager, the reliability and fidelity of their playback leaves a lot to be desired. Curiously, it seems it is only playback that is affected — even damaged‑sounding, jittery low‑fi files are actually fine when you examine them carefully in a waveform editor — and this even applies to files I have recorded over USB, suggesting that whatever it is that affects playback, it does not dog the recording process in the same way. Nevertheless, this way of working could not be described as professional even in the most flippant sense. I think the worst problem I've experienced is Logic refusing to open at all when USB was selected in my Mac's Sound Manager — a message stated that the Inputs and Outputs to the program were different, even though they weren't.
Long before things got to this stage, Swissonic, makers of the USB Studio D line mixer, got tired of the situation and commissioned a USB ASIO driver for the Studio D. As with ASIO drivers for PCI cards which bypass the Apple Sound Manager, this solution seems to have a very positive effect on the I/O latency, but additionally, there seems to have been a benefit in terms of stability and playback fidelity. Steinberg's worldwide demonstrator Rodney Orpheus has been using the ASIO driver with the USB Studio D for some time, and even uses the mic preamps on the unit to mix in his commentary with the music. The good news for those with other USB devices who are still struggling with Sound Manager‑related problems is that the people who developed the ASIO driver for Swissonic (a German company called Propagamma) have just started selling a more general version of the USB ASIO driver for around £30 via their web site (www.usb‑audio.com). In theory, you should now be able to get the same low latency and reliable perfomance with other USB devices on the market. As yet, I have not personally had the chance to try this with any of the other USB devices out there, but these have been requested for a thorough testing session in time for the next instalment of this series.
Ch‑Ch‑Ch Changes
Of course, things rarely stand still for long in our business, and there are now USB audio interfaces shipping or at the planning stage which also integrate USB MIDI busses. Tascam's new USB428 is one example that has just been released at the time of writing, and others, such as Event's EZbus, are on the way. This may seem like foolhardy behaviour in view of the unreliability which has been observed in MIDI‑ or audio‑only USB devices — you might be forgiven for thinking that the problems can only get worse if you put MIDI and audio down the USB from one device — but there are two sides to this argument.
The other way of looking at it is that it might be better to have both MIDI and audio interfacing in one box, so that the single device can arbitrate both data formats and optimise their transmission to and from the computer. Having said that, keeping MIDI and digital audio in two separate USB devices allows them each to be connected to a discrete USB port, preventing them from treading on each others turf and stealing bandwidth from each other. However, arriving at a proper conclusion on this is going to require some serious comparative testing, which both copy deadlines and the late arrival of some of the newest USB devices preclude me from doing this month. I am in the process of testing several of these devices, and I hope to have arrived at some definitive answers by next month. In the meantime, at the risk of sounding like a broken record, if you need to buy a USB audio solution this month, ask for an exact latency figure in milliseconds before you buy, and as always, get the guarantee of a swapout if it doesn't live up to the promised performance.
Other Voices — Update On USB & MIDI
Several readers contacted me with useful news of their own USB and MIDI experiences after the last part of this series was published. As you may recall, I spent some time last month explaining that, in my experience, MIDI interfaces that require OMS to operate on new‑style serial port‑less Macs are particularly prone to instability. I posited that one of the reasons for this is that much of OMS is still in 68K code, placing an intolerable burden on modern Macs trying to manage MIDI data flow and still preserve timing accuracy. Many people contacted me to confirm that they had had similar troubled experiences with OMS. Perhaps the most comprehensive response came from Graham Hinton of Hinton Electronics, who you may remember bring the MegaWolf Romulus and Remus PCI cards into the UK. I mentioned these cards last month as a good way to add multiple MIDI I/O to a modern Mac, assuming you can spare the PCI slot they need to work. The cards were initially supplied with FreeMIDI and OMS drivers, but the OMS ones used a workaround for multiple‑port interfaces which did nothing to improve OMS's reputation for reliability. However, for the last year or so, the OMS driver Graham has been supplying has been a different one for accessing two ports directly. This is a rewrite of the original OMS modem/printer driver, and as such, works more reliably, directing OMS to the MegaWolf card in one of your Mac's PCI slots instead of the now‑non‑existent built‑in serial ports. This is much closer to the gPort/Stealth port way of working I covered last month, where MIDI data is directed to the G3 or G4's built‑in modem serial port instead of the now‑missing old‑style serial ports.
Graham also voiced some concerns he had about these 'modem‑replacement' devices, averring (with various data transfer measurements of his own as evidence) that the vestigial old‑style serial port of the G3 and G4, while fine for modem data transfers, is actually not as reliable for heavy data transfer as the built‑in serial ports used to be on old‑style Macs. He is of the opinion that while users with single‑port MIDI interfaces will probably be OK, you might come unstuck if trying to run a multi‑port interface such as a MOTU MTP or Opcode Studio 64 or 128. I can honestly say that I've yet to experience problems of this nature with the modem‑replacement solutions in any of the system installations I have performed or attempted to troubleshoot, but I would love to hear from anyone with a multi‑port MIDI interface who has.
Graham was also one of many who contacted me to report much greater success with MOTU's FreeMIDI and USB than with OMS, and in particular the use of FreeMIDI's OMS emulation mode, whereby the routings and configurations set up in an OMS Studio can be run by FreeMIDI without all the 68K emulation necessary under OMS. From various readers' experiences, it would appear that FreeMIDI provides a neat workaround when using Digidesign's Pro Tools software or Emagic's Logic, but doesn't work so well with Steinberg's Cubase or Opcode Vision.
New USB interfaces are being released all the time which means that any series like this is a set of snapshots of the time each one was written. At the time of writing last month's part, I was not aware of Emagic's MT4 (shown above), which offers two Ins and four Outs, but I am happy to report that I have since done a couple of installations with it. One was in direct connection to Logic Audio, in which it performed admirably, thus lending further weight to my belief that it's best where possible to use software with interfaces from the same manufacturer, and thereby take advantage of companies' direct drivers for their proprietary hardware. With the MT4, access to a proprietary interface for Logic is possible for a couple of hundred quid less than it used to be, which can only be a good thing. Interestingly, the other MT4 installation I oversaw was with a system built around Cubase VST, which meant that OMS had to be used. Despite my reservations about OMS for USB, there were no obvious problems with this installation either!
Another USB MIDI interface I hadn't had a chance to try out before the first part of this series was printed was Yamaha's UX256, reviewed by Derek Johnson in last month's SOS. I still haven't had hands‑on experience of the UX, but I nevertheless think I can answer Derek's question about whether Yamaha will be able to update its firmware to accommodate an additional eight Outs and even eight Ins via the 'To TG' port. It's important to remember that the To TG port is acting as a conduit to the USB interface and that adding another eight I/Os to the six already on the UX256 is really asking for trouble. I suspect Yamaha stopped at five Outs and one In as the most they could reasonably hope to work with without major data loss. I don't want to bore you all with the maths of exactly how much MIDI information the bandwidth of USB can support, especially if you are using continuous controllers and SysEx info, but multiplexed MIDI channels will use it all up even more quickly. That may well be the reason why Derek's Mac crashed when he tried to do the automatic OMS identification of all the attached MIDI equipment. I can see how a good few devices all responding with their identities simultaneously would cause the OMS/USB combination to fall over. I would suggest any UX256 purchasers do the same manual configuration that Derek found worked for him.
Finally, a quick correction to last month's concerns about the lack of support for Emagic Logic users with old serial‑port dongles who move up to a newer‑style Mac. UK Emagic distributors Sound Technology have asked me to point out that a 'Serial‑for‑USB' dongle‑exchange program has in fact been in place for some time, although it does incur a charge of £25 per dongle exchanged.