Comparison - Voyetra Turtle Beach OEM Montego sound cards

(Montego A3DXStream, Montego II and Montego II Quadzilla)

Review date: 4 January 2000.
Last modified 03-Dec-2011.

 

Digital output and multi-speaker compatibility used to be features only of high-end professional sound cards, meant for studio use and no good for games. But both features, and positional audio (see the sidebar to the right), too, are now available in audio boards that work with all current PC games and cost under $AU150 ($1 Australian, by the way, is worth about $US0.65).

If you're shopping for a no-holds-barred PC sound card at the moment, a few more options are opening up. We here in Australia haven't got Creative's new low cost Sound Blaster Live! cards with S/PDIF output quite yet, but they'll be along very shortly. In the meantime, people who want S/PDIF have the option of a high-priced full Live!, which is a great piece of gear but, at around $AU450, is well outside the price range of most buyers. The full Sound Blaster Live! models have S/PDIF input as well as output, and more analogue inputs than pretty much any other mainstream card as well, but you'd bloomin' well want them to, for the money.

If you want a big old home studio setup with all of those inputs and outputs, a Live! is probably your best choice. But you probably don't. Most people can settle for fewer plugs and sockets, and that's where the various Aureal Vortex and Vortex 2 chipset sound cards come in.

I've checked out the plain OEM (or "bulk pack") versions of Voyetra Turtle Beach's Aureal-based cards. These are the ones provided to computer assemblers, with minimal software and no fancy packaging, and a correspondingly low price. You can buy OEM cards yourself from various dealers including, for Australian and New Zealand buyers, my site sponsors Aus PC Market.

Montego I A3DXStream
The plain Montego A3DXStream (the Montego II cards are physically almost identical)

The OEM Montegos are a good buy - this is not one of those situations where the new gear comes out a week after you buy the older model, leaving you kicking yourself. High end sound cards, like high end video cards, get cheaper all the time, but the next Great Leap forward is far enough away that it's a no more than normally financially foolish idea to buy a Montego now.

Voyetra Turtle Beach's (they used to just be Turtle Beach, until Voyetra bought them in 1996) Aureal-based OEM cards come in three flavours. The original Montego, now selling as the Montego A3DXStream, is a Vortex board. The Montego II models use the more advanced Vortex 2, which, among other improvements, supports the new "wavetracing" features of the A3D 2 sound API in hardware. The OEM Montego sells for $AU120, the Montego II is $AU149, and the Montego II Quadzilla, with S/PDIF output and four speaker support, is $AU235.

What you get

Montego II Quadzilla
The Montego II board that comes in the Quadzilla package is exactly the same as the standard Montego II, but it comes with the "daughter card" expansion board that's an extra-cost option for the cheaper Montego II. The daughter card has the extra speaker and S/PDIF connectors on it. The Montego A3DXStream is, physically, very similar to the more advanced boards.

All of the cards have three standard 1/8th inch stereo jacks on their back panel, colour coded. There's a microphone input, a line input and a line output, which can be pressed into service as a headphone connector as well but doesn't have the grunt to drive speakers. You also get a joystick connector, which can as usual be used as a MIDI in/out connector with an optional adaptor cable.

The Quadzilla edition of the Montego II adds, on the daughter card, another speaker output and the S/PDIF RCA connector.

These might be OEM cards, but they still come with the standard manual, which is well written and comprehensive.

Setting up

Installing the Turtle Beach cards is simple enough, as long as you're running a late version of Windows 95 or, preferably, Windows 98. Linux and BeOS support is apparently on the way, and the cards will work for ordinary stereo sound under Windows NT, but not for positional audio until NT makes the leap to Windows 2000 - WinNT 4 and earlier don't support any kind of positional audio.

Assuming you've got Win95/98, you just shut down, pop the lid, remove your old sound card, insert the new board into a spare PCI slot and you're in business. The Quadzilla requires you to screw the daughter card into a slot somewhere close to the main card and hook them up with the provided cable, but this is easy enough - provided, of course, that you have enough spare slots for both boards. If you've got two spare slots, but they're not near each other, you'll have to shuffle your cards.

As is the case with many better sound cards, there are three extra internal MPC2 standard connectors on the Montego card, for CD audio, modem audio and "auxiliary" audio input. The AUX connector could be used for the audio from another CD-ROM or DVD-ROM drive, or anything else inside the computer that delivers line level audio.

You only get one MPC2 audio cable, though, so if you want to connect more than your CD-ROM's audio output to the sound card you'll need to pick up extra cables. The other header on the Montego boards is an eight-pin connector for the optional daughter card, included with the Quadzilla package.

There's also a connector on all of the Montego boards for another kind of daughter card - a WaveBlaster compatible MIDI expansion card, which plugs straight onto the Montego in piggyback fashion. Given the software MIDI capabilities of the Montego boards, though (of which more later), not many users are likely to care.

Software setup

When installing the software for the Montego boards, you have the choice of setting up Turtle Beach's driver and utility package, or going with the stock Aureal reference drivers.

The Turtle Beach version of the Montego II application installer isn't one of those namby-pamby Windows-standard jobbies that asks you where you want the software to go and whether you'd like to restart now or later; when you run it, it installs the utilities in C:\Program Files\VOYETRA\, the one and only place it knows where to put them, and then reboots your computer whether you want it to or not.

All of the Montegos come with Turtle Beach's somewhat ludicrous AudioStation 32 control console.

AudioStation 32 - behold!

Goofy interface

I remember when every second sound card came with some damn fool interface that thought it was a stereo system. Ah, nostalgia.

Voyetra's "AudioStation 32" is, at least, functional; it plays CDs and WAVs and MP3s and MIDI files and so on, and lets you set all of the EQ and effects and volume options.

But the Turtle Beach installer replaces the simple Windows Volume Control with this monster. You can disable the extra modules so it doesn't eat your whole screen, but it's still dumb. This is a whole lot of interface to have to load every time you want to twiddle the volume.

You also get the quite good Voyetra MIDI Orchestrator package, which lets you play with the instrument assignments of MIDI tracks. It's not a proper sequencer, it's just an instrument rearranger, but it's handy nonetheless. The AudioView 32 sound editor is... well, it's another bundled sound editor that does the basic jobs OK. There's also the SoundCheck diagnostic package, which can run tests on all of the sound cards' functions.

The OEM Quadzilla comes with a slightly more salubrious bundle than the other OEM Turtle Beach cards.

The only full app in the Quadzilla package is Hip Hop eJay, one of those mix-up programs that anybody with Beastie Boys fantasies will probably like. You also get a demo of the Digital Orchestrator Pro package, for audio recording to hard disk.

The Quadzilla package also has an "Audio by Turtle Beach" sticker, to satisfy the Datsubishi Grapefruit Racing Team types who can't use any product unless they also get to advertise it.

Changing drivers

As is the case with graphics cards that use the same chipset, all basic Vortex 2 cards are essentially the same. Some may be better made than others, and they come with different software bundles, but you can use the standard Aureal reference drivers with all of them. They work fine with the Montego and Montego II.

Customised cards, though, are different. The Quadzilla will work with the reference drivers, but all you get is plain Vortex 2 performance; no S/PDIF, no four-channel output.

If you've only got a two channel Montego or Montego II, though, there's a lot to be said for the reference drivers. You can get them from here, but be warned - the full package is 34Mb!

A less startling download, from the same page, is the current version of the A3D drivers (2.25, at the time of writing), at a bit more than a megabyte. These make sure you've got the latest A3D API, so games that try to use A3D 2 don't strike an incapable old API (one of which seems to come, for instance, on the standard Montego II driver disc) and crash.

How's it sound?

The Montego II supports A3D, A3D 2 and DirectSound 3D for positional audio. Aureal have been promising Vortex 2 drivers which support EAX for rather a while now, and should have working drivers out soon. A leaked v2.048 driver set with EAX is already out, but buggy.

For plain stereo applications, the Montego cards sound great, but then so does every other decent card on the market. The magic happens when you start using software that supports positional audio, in particular A3D 2.

With only two speakers, the positional audio effect is good, but not fabulous. With four speakers from the Quadzilla, it's not as much better as you might expect; in four speaker mode, the Quadzilla only does the fancy HRTF tricks with the front speakers, doing plain stereo panning with the rear ones. This can create odd transitional effects when something's supposed to zoom past you.

Back in two channel mode, headphones are the way to go. Two speaker positional audio is, of necessity, a compromise, because the programmers don't know the specs of the speakers or their location relative to the listener. They also have to include "crosstalk cancellation", to try to minimise the amount of sound meant for one ear which makes it to the other.

With headphones, the speaker position and crosstalk problems go away, and you hear remarkably good actual honest-to-goodness 3D audio, with height, width and depth. The included A3D demos are remarkable...

heli320.gif (3305 bytes)

...although I feel safe in saying that the best thing about them is not their graphics.

 

A3D 2.0 demos

Incidentally, if you install the full reference driver suite, you get the most up-to-date A3D demos, including the Rooms demo which is missing from the plain Turtle Beach CD.

A3D's height cues are the most elusive, but they're still definitely there; you can close your eyes and tell with considerable accuracy when a sound is meant to be above or below you. It's not perfect, but it's very good.

Sound quality

Turtle Beach have a tradition of making high-powered sound cards aimed at home studio enthusiasts. Their Aureal-based boards are cheaper and add the Vortex's positional audio features, but they're still great for studio work. Minuscule distortion and superb linearity are actually pretty common among decent sound cards these days, but the Aureal Vortex boards certainly don't let the side down.

The Montego boards, like every other good sound card these days, deal very well with the hugely electrically noisy environment inside a PC case. Older sound cards commonly provided accidental feedback on the status of the computer, with crosstalk from umpteen different internal systems clearly audible through the speakers whenever something loud wasn't happening. That's in the past, though; Turtle Beach has been making very clean cards for years, but they're far from alone these days. One nice feature of the Montego II that helps it keep the hiss down is automatic muting for unused inputs - if a connector doesn't have a signal applied to it, it's disabled.

The Vortex chips are also good at keeping sound quality up when they're mixing lots of samples, because of their 96 parallel high quality sample rate converters. Just mixing sounds algorithmically at their original sample rate causes more and more loss of accuracy as you pile more and more sounds into the mix. Raising the sample resolution before the mix doesn't make the original sounds any better, but it lets smaller amplitude differences be included, instead of just rounding off the results after each mix stage and reducing the effective amplitude resolution. The Vortex 2 does sample rate conversion even better than the original Vortex.

Performance

OK, that's why you should use A3D. Why shouldn't you?

Well, because it sucks up a significant amount of CPU time, that's why.

A3D 2.0's fancy "wavetracing" - the audio equivalent of graphics "raytracing" - tracks where sounds are going in quite detailed ways, allowing realistic wall reflections and occlusions and so on. The price of this, though, is a big CPU load.

A3D 2.0 also includes "A2D", which is A3D emulation for sound cards that don't have a Vortex chipset. Using A2D can elevate the CPU load from large to enormous, but not necessarily; it can use DirectSound 3D hardware, if it's present, and not elevate the CPU load at all, unless it's asked to do things the hardware can't handle, like reflections.

A couple of quick Quake 3 Arena benchmarks tell the story. Q3A supports A3D 2.0, but not EAX or A3D 1.0, and its A3D support is less than perfect. You have to turn on A3D in the setup menu every time you run Q3A, and it still seems prone to crashing now and then, with the current v2040 drivers. But it works, within the definition of the term.

You can reduce the A3D performance hit in Q3A, by the way, by turning off the fancy wavetracing features. Reflections are turned off with the
/s_reflect 0
console command, and you can disable wavetracing entirely with
/s_geometry 0
The
/s_occlude 1
command, by the way, enables occlusion support (it's turned off by default), which will of course slow the system further.
With the geometry features turned off, A3D might become a significant multiplayer advantage, since the "X-ray ears" effect you get lets you know in advance what direction things are happening in, even if there are walls in the way.
There's lots more information on Q3A A3D-relevant console commands at Aureal's page on the subject here.

With Q3A set to graphics settings which make the CPU the limiting factor - which, on my NVIDIA GeForce 256-equipped test box, means any graphics settings at all up to 1024 by 768 and a bit beyond, since the superpowered graphics card can turn CPU-generated geometry data into an image as fast as the 500MHz Celeron CPU can send it - turning A3D on chopped the frame rate on the standard Q3A demos from around 55 frames per second to around 40. This is about a 30% loss, and it can become quite significant in hectic deathmatch games, when there's a whole lot of geometry goin' on.

The more frantic the action, the more sounds the system will have to deal with as well, so the sound load scales pretty evenly with the graphics load.

The Vortex chipsets have 16 hardware 3D sound stream channels. When a game exceeds the number of hardware sound streams a card supports, the most economical strategy for the programmers to adopt is to dump the extra sounds to the plain stereo mixer - giving them no position, but at least making them audible - or ignore them completely. The alternative is performance loss.

Positional audio can be done partly or entirely in software - DirectSound 3D implementations commonly do this, when they run out of hardware 3D channels or are running on hardware that has no built-in 3D capability. It's analogous to the way that regular sound cards that only support a certain number of plain stereo sounds at once can play more, by mixing the sounds in software.

But the huge CPU load caused by pure software positional audio makes it of no use for games, and physically impossible for slower computers. The performance hit can be reduced with clever driver coding - with each driver release, Aureal and the other companies that have Vortex boards speed the cards up a little - but the basic problem remains.

Vortex 2 based cards can deal with up to a very respectable 76 simultaneous DirectSound or A3D 1 streams. But when they're doing A3D 2.0, which has reflections, they can only handle 16 streams in hardware - the other 60 hardware streams are a "reflection pool" for the A3D wavetracing effects. You can't have more than 16 A3D 2.0 sound sources without doing software mixing, and A3D 2.0 loads up the system enough with its geometry demands anyway. The original Vortex chipset has only 8 hardware 3D sound channels, and no reflection pool at all - the extra 60 channels make the Vortex 2 the only option for A3D 2.

Creative's Sound Blaster Live! models have 32 hardware 3D sound channels, but no reflection pool. They cause less of a performance hit than the Aureal-based cards. This is partly because their drivers seem to be a bit tighter, but mainly because the Live! boards are almost definitely not using an A3D 2.0-type sound API, so there's no need for reflections anyway. EAX 2.0, which has similar features to A3D 2, is supported by close to no games, so far. So the Live! cards are stuck with EAX 1 and DirectSound 3D, which is much less demanding.

EAX 1 and DS3D, though, can still sound very good - the difference between genuine A3D 2 reflections and well picked and neatly changed pre-set EAX reverb settings can be hard to pick.

Sound mixing can present a significant problem, pretty much regardless of the number of hardware channels available. In games like 3D action titles, for instance, there are commonly tons of sounds coming from all over the place, thanks to rapid-fire weapons bouncing projectiles off walls. In a hectic deathmatch game, it's quite possible that plenty more than 32 3D sound streams will be needed whenever a whole lot of shooting happens at once, and this is the time when you least want 40% of your CPU time to be eaten by the sound system! Of course, in this sort of bunfight you're less likely to notice if most of the sounds have dropped back to plain stereo.

Deathmatch games are less about immersing yourself in the illusion than they're about kicking the snot out of your friends. Single player games are different.

When you're playing single-player, turning on 32 bit rendering and maximum quality everything and 3D sound helps you lose yourself in the game, the star of your own little studio-quality movie. Single player 3D games tend to be rather less demanding on the processor, so you can keep up a more respectable frame rate even with all the pretty-options turned on. If you want to be a winner in a multiplayer game, though, everything that slows down your response times is a handicap.

If you're just after hey-wow fun, of course, and have a resilient enough self-image that getting trampled by people with fewer fancy options turned on isn't a problem for you, then by all means use A3D and every other option you like in multiplayer. If you don't mind the frame rate loss, there's no problem.

DVD

If your playback software supports it, four speaker sound cards can "downmix" 5.1 channel surround sound from DVDs to output through their own speakers. The centre channel's mixed between the two front speakers, and the sub channel's usually discarded.

For this to work, you need a DVD player solution that supports your particular four speaker sound card.

To get surround sound from your PC into your home theatre system, though, you have to send a suitable signal to a surround decoder, which in most home theatres lives in the surround receiver.

Plain old analogue Dolby Pro Logic can be created on the fly from the Dolby AC-3 5.1 channel audio encoded on many DVDs, and output through a standard single analogue stereo y sound card for decoding by anything that supports it - I don't know of any home theatre receivers that don't. The results from this downmixing are usually disappointing, though. Full digital AC-3 is the way to go, and the only way to get that from your computer to something, like a surround receiver, that can properly decode it is via a digital connector. You need either S/PDIF, using a plain RCA connector as used for analogue home audio connectors, or the more upmarket "TOSLINK", named for the kind of connectors it uses. TOSLINK is just S/PDIF over optical cables, and carries exactly the same data.

Unfortunately, playback software won't necessarily be able to send AC-3 data out of the Quadzilla's S/PDIF connector; it might not be aware of the existence of that connector. DVD playback software that understands other Aureal-based cards with an S/PDIF connector isn't necessarily good enough, because Turtle Beach have done their own thing with the daughter card.

Judging by newsgroup posts, RAVISENT's CineMaster software works properly with the Quadzilla as long as you've used the Region Selector hack to set the playback software to Aureal output. Make sure that the DVD playback system you intend to use will work with the Quadzilla before you buy.

Incidentally, if you've got some other Aureal board with an S/PDIF connector on it and you're using the current Aureal 2040 reference driver set - or some other driver set closely based on it - the S/PDIF connector will not, by default, work.

To make it function again, you have to install this little registry patch and then check the new box that says "Send AC3 stream to S/PDIF", in the Digital Out tab of the Vortex control panel. And you'll have to re-check the box every time you restart your PC, because it doesn't stay checked. Elegant.

MIDI

The Montego MIDI interface is the usual MPU-401 emulator, which works, if you choose to use it, via an adaptor cable that plugs into the joystick port. This lets you hook your computer up to outboard synthesisers or control keyboards. Most computer MIDI, though, is an internal-only deal - software playing tunes with the MIDI voices of the sound card itself.

Unless you've bought and plugged in a WaveBlaster daughter board, the Montego II stores its MIDI voices in system RAM - it doesn't have any ROM sample storage on board, and uses a DLS 1.0 format sound bank for its MIDI instruments. With the standard sample set, it can use up to 4Mb of RAM when MIDI's being played, although it'll probably use a lot less, since it only loads the instruments it's actually using. It's possible to use a lot more RAM, if you switch to a larger DLS sound bank - many DLS sets are available for free download from sites like this one. The size of the sound banks you can load is only limited by your RAM.

The standard Aureal MIDI voices are all right, and of course far better than the crummy FM-synthesised MIDI still offered by bargain cards, but they're nothing to write home about. It's easy enough to grab a better DLS file, though. With 64 hardware MIDI voices and up to another 512 with software mixing (only 256, from earlier driver versions such as come with the Quadzilla), the Vortex 2 can handle arbitrarily complex compositions. Each simultaneously played note takes up one voice, so a six-note chord, for instance, uses six voices; if there's a snare drum and a cymbal crash at the same time, that's another two. These ludicrous maximum voice numbers should be taken with a grain of salt; the more mixing's done, the lower the sound quality becomes. There's no need for more than 64 voices in the vast bulk of MIDI tracks, though.

MIDI effects settings

Like many other current cards, the Montego lets you put up to two effects on top of your MIDI playback, so you can have one of several reverb settings as well as chorus, flange, delay and so on. Each effect has a configurable level, as well.

Many users don't give a toss about MIDI either way, as they only hear it when they visit some dodgy Web page that insists on playing "Copacabana" in the background. MIDI for game music is less popular than it used to be, though it persists in games with big musical scores, like Ultima Online and other adventures. The smallness of the MIDI format makes it possible for whole new scores to be squirted down the phone line in seconds, and huge amounts of music can be used in a game without using monstrous amounts of space for digitised samples.

Who needs it?

There are three classes of user that should look into a better sound card: game lovers, home theatre enthusiasts and people who want to make a home sound studio.

If you're into single player games, you want positional audio. No, don't tell me what you think you want, I'm telling you - you want positional audio. By the use of cunning algorithms that play with the spectral content and phase of the sound, positional audio allows two transducers (speakers or, better, headphones) to give you the firm impression that a sound or sounds are coming from anywhere around you, including above and below. Perfect it is not, but dramatically better than plain stereo it is. Far from all games support positional audio so far, but enough do to make it worthwhile already.

The two most popular positional audio systems available for PCs at the moment are Creative's Sound Blaster Live! cards, which use E-mu Systems' EMU10K1 processor, and boards from various manufacturers, including Aureal themselves, which use Aureal's Vortex (AU8820) and Vortex 2 (AU8830) chipsets. In a nutshell, the Aureal boards sound better, but use rather more CPU time. CPU usage is the main downside of positional audio.

A great deal of the processing for positional audio is done by the custom processor on the sound card, but the CPU has to help, especially with more advanced systems like A3D 2.0. If a system includes occlusion (sound being blocked by objects) and proper reflections, it has to know where walls and objects are, so the CPU has to provide geometry information to the sound card driver, as well as to the video card driver.

If multiplayer games are more your thing, positional audio can give you an edge - you can tell what's going on a bit better, thanks to sharply localised sounds. But the performance hit you'll take by enabling the funky sound hardware may be too high, if you have a less than cutting edge processor and/or video card.

In 3D games played on accelerated 3D cards, cranking up the resolution increases the load on the video card, not the CPU. This means that the CPU hit from 3D sound in a given game should be much the same no matter what resolution you're running at, but if your graphics settings are advanced enough that your video card's the limiting factor, you may see much less speed loss if you enable positional audio; the CPU is already waiting for the graphics card, and has some cycles to spare. Exactly what balance of options suits your system depends on the CPU, the graphics card and the game you're playing, though.

Other options

The Sound Blaster Live! X-Gamer has an S/PDIF output, and is presently selling in the USA for less than $US80 (list price $US99.99), but it's only available there; it, and its sibling Live! MP3+, are not distributed outside the Americas. The rest of the world gets the Live! Player and the Live! Value Digital, both of which seem to have the exact same hardware as the US Lives but different bundles - the Value Digital is the cheap one, and should cost around the same as the Quadzilla, if not less. The Value Digital is currently selling from discounters for less than $US70, and its local RRP is $AU249. The same discount outlets have Montego II boards for less than $US60.

Then again, Vortex 2 boards will sooner or later support Creative's open EAX standard via a driver upgrade, while Creative's cards will never support Aureal's proprietary A3D APIs. And the Vortex 2 hardware is clearly technically superior. And, at present, A3D 2 has wider game support than the younger EAX 2 standard - although even A3D 2's support is nothing very exciting.

This is par for the course in the 3D sound world at the moment, unfortunately. It's getting better by the day, but 3D sound is still, generally, quite poorly supported. Different driver versions work, or don't work, in drastically different ways; games that allegedly support one or another positional audio standard actually don't do it properly, or at all; "fixes" to positional audio problems simply involve turning it off, and so on.

Direct3D disable option

Incidentally, the standard Vortex drivers have the ability to turn off DirectSound if it's causing problems.

This is depressing, because positional audio is really cool, and it's strange that after more than two years it still isn't a mainstream proposition.

But it is, now, worth bothering with; quite a few games support it well enough, and the cards aren't stupidly expensive.

Overall

With EAX drivers due shortly and super-fast CPUs and graphics cards getting cheaper by the day, Vortex 2 based sound cards are looking like a better and better idea. Their technology ain't new, but neither is it outdated, and a sound card that supports positional audio is worth a bit extra. With the OEM versions of the cards you get the hardware you need and none of the software you (probably) don't, and Turtle Beach's take on the Vortex and Vortex 2 has a lot to commend it.

Of the three, I prefer the plain Montego II, because I don't need S/PDIF and the panned-rear four speaker mode of the Quadzilla is less than amazing - it's headphones all the way for me, baby. If DVD is your thing, though, the Quadzilla is a good choice, provided your playback software likes it.

If all you want is A3D 2 and four speaker output, an OEM Diamond Monster Sound MX300 (I review the original retail version here) gives you that for about $AU145.

The money you'd save by buying the old A3DXStream is, I think, false economy; A3D 2.0 makes the Vortex 2 significantly better than the original Vortex.


Turtle Beach software download page

Aureal's reference A3D drivers

Vortex 2 Technical Support FAQ at 3DSoundSurge


Buy stuff!
Aus PC Market no longer sells these cards (which is just as well, as there aren't any surround drivers for them for Windows 2000 or later); click here to see their current audio range!
(if you're not in Australia or New Zealand, Aus PC Market won't deliver to you. If you're in the USA, try a price search at DealTime!)

Understanding positional audio

"3D sound" is a tarnished term. Most of the gadgets that claimed to create it in years past did nothing of the sort. They were stereo "expanders" of differing levels of quality, or reverb units, or stranger creatures - I once reviewed a gizmo that panned computer sound left or right depending on which way you moved your joystick!

Today, though, you can get real 3D sound, also known as "positional audio". That's three real dimensions - forward and back, left and right, AND up and down. A plain surround system, with speakers behind you, gives you 2D sound, but with the help of fiendishly cunning HRTF (Head Related Transfer Function) algorithms, it's possible for even dual-transducer systems (a single pair of speakers, for instance, or headphones by preference) to give a genuine impression of properly enveloping sound. A dual-transducer system still can't actually create sound sources above or below the listener, or indeed anywhere else where there isn't a physical transducer, but careful study of the way hearing works has made it possible to convincingly fool the listener into thinking that sounds really are coming from anywhere around them.

Mixed up in the positional audio concept is simulation of particular audio environments - the different reverberation and occlusion characteristics of different places. Reverberation, or reverb for short, is all of the reflections of a sound that make it to your ears sooner or later after the sound itself. Reverb in even a simple, flat-walled environment can be very complex indeed, since the echoes can reflect again and again as the sound decays. Reverb characteristics also change as you move around the room; reflections from a nearby wall will reach you sooner than those from a more distant one. And irregular environments, like a cave, can be outrageously complex to model.

Things are further complicated by the fact that when you're faking an audio environment with, say, four speakers, the aim of the exercise is to feed the speakers the right signal to reproduce at the listener's ear'oles more or less the right combination of direct and "reflected" sound. If the listener is uncooperative enough to move, the mix will be out of whack; positional audio programmers therefore have to try to make the area of fairly realistic sound as large as possible.

Headphones get around the moving-target problem, but at the cost of making the soundstage move with the listener's head. Turn around and everything's backwards. Nifty virtual reality systems with head tracking can beat this problem, but don't look for one of them on the shelves of Clint's Crazy Bargains at any point in the near future.

The trick for positional audio designers is to model the behaviour of an environment as realistically as possible without exceeding the abilities of the equipment available. And the system they come up with has to be easy for other programmers to work with, too; it doesn't matter if it's the best sound modelling concept ever invented if nobody can put it in a game.

To this end, two main Application Programming Interfaces (APIs) have emerged in the positional audio world. Audio APIs work in the same way as 3D video APIs like OpenGL and Glide and Direct3D; they're a standard system that other programmers can use to accomplish a really complex task without, themselves, having to know how the processing is done.

The APIs

DirectSound 3D (DS3D) is Microsoft's DirectX component for positional audio. Games and other software can hook into DirectSound 3D and tell it where a sound is to be placed, how loud it is, how it changes with distance, whether it's directional, and so on. Then the DS3D sound card driver hooks into the sound card's specialised hardware to use rather complex algorithms, which are not Microsoft's work, to make the requisite signals. These algorithms are known as the "engine" used by the sound card, and cards with different engines sound different, even though all of them might support DS3D or A3D (see below).

If your sound card doesn't have positional audio hardware, DS3D can still do its thing, but in software, with all of the work done by the computer's CPU. The algorithm used for software DS3D is inferior to practically any 3D sound card's positional audio algorithms, but it's very processor-hungry and will cause a great big unacceptable performance hit, no matter how fast your computer is.

If your sound card has positional audio support built in, it may still use a lot of CPU time, depending on how much of the thinking it can do for itself. It'll also have to hit the CPU if you exceed the number of 3D sound streams - distinct positional audio sources - it can handle in hardware. If it has to start doing software mixing, down goes performance, and sound quality, again.

With DirectX 7, DS3D has hardware voice management support, which allows this nasty mixing to be minimised by feeding the important sounds to the hardware 3D sound streams. The less important sounds can go to the plain stereo mixer, which uses much less CPU power than software positional audio, or they can be dropped completely. Voice management was available for DS3D previously courtesy of QSound's manager software, made available as an extension to DS3D and supported by Creative's Sound Blaster Live!, among others. A3D (see below), though, has had hardware voice management for a while now.

EAX is Environmental Audio eXtensions, the PC environmental reverb standard created by Creative Labs, makers of the Sound Blaster line. EAX is an extension to Microsoft's DirectSound 3D, and creative have made it a public standard for anyone to use - anybody's sound card can have EAX compatibility.

The original EAX version 1 implements environmental audio effects by using 26 pre-programmed presets of much the same nature as those used by home theatre amplifiers for ambience enhancement - stage, hall and room reverb, for instance. The various presets can be further tweaked by changing the level of the reverb compared with the original sound, and the preset in use can be changed at any time, but EAX 1 has no ability to truly simulate how things would sound in a particular shape of room.

The difference between pre-programmed settings and proper modelling becomes more drastic when you start taking closed doors, obstacles and fancy reflections from funny-shaped rooms into account.

EAX actually doesn't have anything to do with positional audio - by itself, it's just reverb, and a game must also support DirectSound 3D if sounds are to come from particular locations. EAX 2 adds positional audio and room-shape support, including "occlusion" - blocking of sound by obstacles. But A3D 2.0 already has them, and is supported by rather more games than the couple that support EAX 2, to date.

A3D is the positional audio standard created by Aureal, makers of a series of high powered audio chipsets, back when Microsoft hadn't yet figured out that hardware acceleration was a good thing, and so DS3D didn't support it.

Version 1 of A3D is much like plain DirectSound 3D; version 2 of A3D can more accurately simulate how sound sources in a complex environment behave, by using information on the shape of the environment provided by the game. A3D 1 and DirectSound 3D have no real idea what shape a room is - they only know where the sound sources are within it.

Glossary

5.1 channels: Surround sound systems with six speaker outputs (left, right, centre, left rear, right rear, subwoofer) are often referred to as "5.1 channel", because the subwoofer Low Frequency Effects (LFE) channel outputs only low bass sound, not the full range of frequencies.

7.1 channels: The 7.1 channel format adds extra left-centre and right-centre channels to the standard 5.1 format. It's supported in DVD only by discs with an MPEG-2 soundtrack, and even then is very rare. Which is just as well, because practically nobody has an audio-visual setup that can play it.

AC3: This was the original name for the Dolby Digital sound system used on the majority of DVDs, and it's still the name for the kind of compression it uses. Dolby Digital uses lossy compression, conceptually similar to the ATRAC encoding used by Minidisc and the MP3 (MPEG 1 Layer 3) computer audio format, to squeeze audio data into a much smaller space. AC3 sounds better for a given data rate than the older ATRAC and MP3 formats.

The normal data rate for plain two-channel stereo AC-3 is 192 kilobits per second (kBps; one kilobit is 1000 bits), and the format supports bit rates from 64 to 448 kBps. 384 kBps is the normal bit rate for 5.1 channel AC-3 surround, and this bit rate is used by many DVD movies and all laserdiscs (the old LP-sized analogue video discs) that have an AC-3 surround soundtrack.

For well-encoded MP3 audio, 64 kBps per channel is enough for near-CD quality (better than FM radio) sound reproduction. In 5.1 channel mode at 384 kBps, AC-3 uses this same bit rate per channel, but its superior sound quality at a given bit rate makes it about as good as CD for even critical listening. It's widely agreed that AC-3 surround at the full 448 kBps speed sounds only slightly better than 384 kBps. The AC-3 standard actually technically supports bit rates up to 640 kBps, but the implementation of it used in DVD doesn't. The general consensus is that 448 kBps AC-3 sounds as good as any other 5.1 channel system, including the much heftier DTS (explained, along with the other DVD movie audio flavours, here).

DMA: Direct Memory Access, the act of directly moving data from a device to memory without passing it through the processor. DMA is often used as shorthand for "DMA address", one of the limited resources of IBM compatible computers (see also IRQ). See my second Step By Step column for more information.

IRQ: Interrupt ReQuests are how IBM-compatible computers assign the CPU's attention to devices that need to talk to it right now. A device gets assigned a given IRQ and uses it whenever CPU time is required; if two devices have the same IRQ and try to use it at once, neither will work. Current PCs have 16 IRQ lines, but several of them are taken up by standard hardware. See my second Step By Step column for more information. See also DMA.

PC 99: Microsoft's latest standardised PC design specification, which like Microsoft's previous specifications will no doubt be adhered to in whimsical ways by different manufacturers.

For more information on PC 99, see the Design Guide here.

MIDI: Musical Instrument Digital Interface, the standard protocol for hooking together electronic instruments. PC sound cards have on-board MIDI music capabilities, and can also have a special cable connected to their joystick port to allow the connection of external MIDI devices.

MPC2: Another software and hardware standard, created by a computer company consortium led, again, by Microsoft, but rather older than PC 99. In this context, MPC2 just describes the kind of connector used by various sound cards for internal analogue audio, for instance from CD-ROM drives and internal modems.

MPU-401: An old dedicated MIDI interface card for IBM compatibles, which is emulated with varying degrees of success by various sound cards.

S/PDIF: Sony/Philips Digital Interface Format is an audio transfer protocol that can use either optical or electrical connectors - electrical, in this case. It's used to transfer either 16 bit stereo audio data between various components (CD players, Digital Audio Tape decks, some sound cards), or to transfer Dolby Digital (formerly known as AC-3) 5.1 channel audio for movies.



Give Dan some money!
(and no-one gets hurt)