The sound of magnetic tape stems from a hugely complex set of interacting factors. Slate Digital's VTM is the most detailed attempt yet to capture these in digital form.
Few companies are brave enough to try to model the behaviour of a professional analogue tape recorder in software. In fact, Steven Slate tells us that their Virtual Tape Machines is the most complex plug-in they've ever coded. Slate sought out two particularly good-sounding tape machines to base their emulation on: NRG Recording's 16-track, two-inch Studer A827 machine and Howie Weinberg's two-track Studer A80 RC half-inch mastering deck. Apparently, Steven told Slate Digital's Chief Technology Officer Fabrice Gabriel that unless the end result was indistinguishable from the real thing, he wasn't prepared to release it at all. VTM turned out to be Fabrice's greatest challenge so far, and we're told it took around a year of optimising, listening and tweaking to get it to the point where they were happy to allow it out to the world.
So why so much fuss about modelling a tape recorder? Surely a bit of distortion, a touch of compression and some EQ would do the trick? Well, no. The audio path in a tape machine is surprisingly complex, and sophisticated though the later tape recorder designs were, they still suffered compromises and technical limitations that gave them a distinctive sound. And that begs the question: if tape machines are so inherently imperfect, why do we like the way they sound?
Part of the answer has to be historical. Tape was used to make so many great records over a period of so many decades that we have come to associate the sound of these records with the way recorded music should sound. Digital audio, by comparison, still carries the stigma it acquired when it was an emerging and immature technology competing with analogue tape machines — a mature technology that had evolved almost to the point where no further improvement was possible.
Digital may be way more accurate these days, but many people still prefer the sound of tape. I don't think anyone can tell you exactly why analogue tape sounds so 'musical', at least not if you base your reasoning purely on spec sheets and measurements, but there's no denying that terms such as 'warm', 'smooth', 'big' and 'musically involving' sum up the way that many of us feel about the sound.
This sound is the result of numerous complex, interacting factors. For a start, the tape signal path involves an analogue input stage and an amplifier section to drive the tape head. In vintage machines, this path often includes audio transformers which introduce complex phase shifts, saturation effects and frequency-response perturbations, and that's before the signal gets anywhere close to the tape. Once it does, things start to get very complicated. For a start, the transfer of audio to tape and back via a pair of magnetic heads doesn't produce a linear frequency response, so EQ has to be applied both while recording and during playback, in order to flatten the response. In addition to the user-tweakable calibration EQ, tape machines also have fixed NAB or CCIR 'pre-emphasis' equalisation that boosts mainly frequencies above 3.5kHz to achieve the best headroom and noise performance across the audio spectrum. This has to take into account the fact that a replay head functions as an integrating device and so also acts as a kind of equaliser. In practical terms, this means that the repro side of the circuit must have a filter response that is complementary to the actual recorded magnetic flux, not just the simple opposite of the record pre-emphasis EQ.
Unlike digital systems, where everything stays clean until you hit full scale, an analogue tape recording becomes progressively more non-linear as the magnetic saturation limit of the tape is approached, much in the way that valves also distort gracefully. A professional machine with bias and all the correct EQ trickery applied will offer an almost linear response over quite a wide dynamic range, but above a certain level, magnetic saturation starts to set in, and both positive and negative peaks of the waveform become progressively more squashed or compressed. This gently limits the dynamic range of the signal, and also adds additional distortion in the shape of mainly odd harmonics. In moderation, this type of distortion can flatter drums and bass guitar, and to create more saturation you simply record further into the red and use your ears to tell you when enough is enough.
Because of the pre-EQ built into the recorder, however, high frequencies tend to saturate earlier than low frequencies, so the frequency response of tape actually changes in a dynamic way according to the signal level being recorded. The harder you hit it, the more the highs get squashed: another complication to model. The use of pre-emphasis EQ also explains why things like hi-hats can sound horribly distorted once recorded to tape, unless you keep their level well below the red zone.
As if that wasn't enough, there's also the high-frequency bias signal to be taken into account. The magnetic hysteresis (and hence gross non-linearity) exhibited by the metallic particles that make up recording tape means that simply feeding a signal to a tape head and then running tape past it would result in unusably distorted playback. It was discovered back in the pioneering days of tape recording that adding a very high-frequency tone, or bias (typically 60kHz to 250kHz depending on the machine and manufacturer, with 100kHz being a common choice), to the audio produced a far more linear response. This frequency is then filtered out of the playback signal. I guess that in some ways the bias frequency is as close as analogue recording comes to having a sampling frequency.
When optimally adjusted, bias allows the dynamic response of the tape to behave in a linear way up until the point where the tape itself starts to saturate — there's only so much magnetic charge a piece of metallic oxide can hold. Lowering the bias level below optimum has the effect of increasing the dynamic range slightly, and also tends to boost the high frequencies to some extent, while making the lows sound a little 'looser'. Conversely, over-biasing (relative to the optimum level) reduces the dynamic range so the tape saturates at a lower level, and means the high frequencies will be pulled back. Engineers often make small bias adjustment above or below optimum to tailor the sound of their machines, but stray too far from the recommended settings and the sound can get quite nasty. As you might imagine, these effects pose a significant challenge to anyone trying to model them — and there's still a lot more more to come.
You may have heard of 'head bump', a phenomenon that adds a little bass boost at certain tape speeds, specifically when the record head's pole-piece geometry coincides with a half wavelength of the audio signal. At a tape speed of 15ips, this has the effect of fattening up drums and bass instruments in a musically pleasing way. There's also a corresponding dip an octave higher and often at multiples of that frequency, causing small ripples in the low-end response. As you might expect, different tape speeds and different head designs place the bump at different frequencies. The characteristics of the playback head, including its electrical inductance, also add to the complexity of the record/playback relationship.
Have you ever recorded a sine wave on a tape machine and then listened to the playback? If you have, you could be forgiven for thinking the machine was broken, as your nice pure sine wave comes back accompanied by a very unpleasant, gritty kind of noise known as modulation noise. It is only audible in the presence of signal and tends to be masked by real-life complex audio waveforms. Alhough it may pose a real problem for those whose hobby is recording sine waves, it is thought to be a contributory factor to what we know as analogue warmth.
Also vying for places in the warmth department are our old friends wow and flutter. These terms refer to low- and high-frequency speed variations in the tape transport caused by mechanical tolerances and wear (for more detail, see Hugh Robjohns' review of Celemony's Capstan wow and flutter removal software in last month's SOS). These variations should be so minimal as to be unnoticeable most of the time on a good-quality machine, but even at low levels, they can contribute to this mystical analogue warmth we keep banging on about. Oh yes, and let's not forget the sworn enemy of all of us who used to work with tape: hiss, and the various technologies that were developed to reduce it...
Slate Digital's plug-in, which looks gorgeous with its revolving tape reels, is available in VST and RTAS formats for Mac and PC, and as an Audio Units plug-in on the Mac, with AAX support coming soon. You'll need a second-generation iLok dongle.
Having modelled all the key factors, and numerous other more subtle elements, such as crosstalk, to their satisfaction, I guess the next choice for the Slate design team was how much control to place in the hands of the user. All the VTM's key controls are on the main plug-in window, where there's a choice of the two tape machines, two tape speeds (15 or 30 ips), two tape types (Ampex 456 or Quantegy GP9) and a pair of input and output gain controls. These default to being linked, so that if you turn up the tape input, the output is adjusted downwards by the same amount, meaning that you hear only the effect of tape saturation without being distracted by level changes. There's also a bias switch giving three positions, so in addition to optimum bias, you can also try under- and over-biasing relative to normal. The old-school 456 tape has more 'character' than the modern GP9 formulation, adding more body to the low end and smoothing the highs. GP9 has a little more headroom and helps maintain definition, giving it a little more edge and clarity than 456.
What isn't modelled — thankfully — is the way that tape recordings deteriorate with repeated playing, and there are no modelled drop-outs due to faulty tape coatings or bad edits either! The outside tracks of tape machines were often left unused, as they tended to be the least reliable, and who can remember having to leave a blank track between the audio and SMPTE timecode track to prevent the code bleeding through to the audio?
Like Slate Digital's earlier Virtual Console Collection, VTM is intended to be used across many or all of the tracks in a multitrack session, so a grouping function is included to enable multiple instances of the plug-in to be adjusted from a single set of controls. Each plug-in instance can be assigned to one of eight possible groups, and adjustments made to any instance within a group will be reflected by all the group members. The multiple groups allow you to apply different settings to different sets of tracks, such as multitrack drums and guitars, to best suit the material. Grouping applies to the Process/Bypass switch, so you can disable VTM processing for an entire group of tracks, providing a quick way of evaluating its cumulative effect.
In addition to the obvious controls, the user can open a settings panel to access some additional parameters. These include the VU meter ballistics plus a Noise Reduction fader, which adjusts the level of the modelled tape hiss, and Hiss Automute, which cuts off the hiss when there's no signal present. Bass Alignment allows the low-end response of the tape itself to be adjusted, while the amount of wow and flutter can be controlled to taste (a setting of around 25 percent corresponds to a well-maintained machine). These advanced parameters are all global, and are thus applied to all instances of the plug-in used in the DAW project. This is fair enough, as every track of a 'real' tape machine should behave consistently in these areas. It is also possible to change the level calibration if you want to record hotter, for example, although I found the default setting to be perfectly adequate. However you adjust the Calibration Level, a 0dB VU reading will always correspond to the same amount of tape saturation, where a red warning LED comes on when you run out of headroom. A different calibration can be set up for each of the eight groups via the Settings panel. There is no Settings file as such for the VTM plug-in, so all adjustments and settings are stored in the song session.
I could make this section really short by saying that, to my ears, Slate's VTM sounds just like tape — end of story. As you might expect, I don't have the specific models of tape machine here for comparison, and if I did they probably wouldn't be set up the same way as those modelled here, but I still have my old Tascam two-track and have owned and used more multitrack tape machines over the past few decades than I care to think about!
I found that a practical way to work was to put instances of the multitrack VTM emulation on key tracks within a mix, such as bass, drums, guitars and maybe vocals, usually as the first insert effect, then place the stereo mastering machine at the end of the chain in a master mix insert slot. I preferred the results using the multitrack at 30ips and the mastering machine at 15ips, with the tape hiss turned down a little. This is a great treatment for classic rock and also adds the necessary beef to dance tracks. In fact, everything just sounds that little bit more solid and polished, with better integration of the individual sound sources. Electric guitars take on a subtly fatter perspective, while their highs are rendered more smoothly, but without losing any essential bite. And drums just sound bigger, more confident and more 'rock & roll'. The fact that you feel a profound sense of disappointment when you bypass the plug-in says it all, really! You really need to switch the plug-in in and out to hear what it is doing, as listening to an untreated file and then a treated one, with a delay between, makes it harder to spot the difference. The effect of tape can be quite subtle on some sound sources, and it could be argued that the better the tape recorder modelled, the more subtle the effect will be. As you'd expect, you hear the most difference with the meter pushed well into the red.
VTM is also a great way of polishing mixes at the mastering stage. Applying the two-track mastering recorder emulation to a mix, the most pronounced effect is audible at the 15ips speed setting using 456 tape: the low end somehow sounds bigger and fuller without necessarily sounding louder. The Bass setting can be adjusted to change the amount of bass 'flattery'. For mastering applications, it may be useful to back this off a little.
The grouping option is very simple to set up and really helps manage multiple instances of the plug-in, and I like the way that the main page shows a fairly simple set of controls that allow you to do just about anything necessary. The advanced settings are only a click away if you need them, and even then are quite straightforward. If you also have Slate's Virtual Console Collection, you can use that in combination with VTM to create incredibly authentic-sounding analogue mixes without any of the hassles associated with the actual hardware.
Inevitably, questions will be asked as to how VTM compares with the Universal Audio and Waves offerings, as these are currently the main players in the software tape emulation market. There isn't a simple answer, though, as all three model different tape machines set up by different engineers. I think that when it comes to dogged persistence in modelling the true detail of a tape machine, Slate probably get as close to the real thing as is practical but, on the other hand, the only thing that really matters at the end of the day is how successful the process is in a musical context. I've listened to a lot of pieces of audio processed via all three, and it's easy to fall into the trap of believing that whatever seems to make the most obvious difference must be the best. In reality, though, the effect of going via a really well set-up tape machine can be quite subtle.
I've used the UAD Studer A800 plug-in on several mixes and have been really happy with the results, but to my ears VTM sounds a little more realistic — very musical, but also less obvious. It adds just a hint of weight, density and smoothness to the sound without veering off into excess. There are some convincing videos on the Slate web site comparing how waveforms are distorted when passing through a real tape machine and then through the VTM plug-in, as against their (unnamed) competitors. On the other hand, UA's approach gives you deeper access to the alignment controls and more tape types, and can be coaxed into producing a much more blatant effect than either the Waves or Slate plug-ins, which although it may exceed what you'd expect from a well-maintained tape machine, can be very useful as an out-and-out effect. Both the UAD and Waves offerings also allow tape echo to be set up. Slate have opted for a more focused approach, concentrating on replicating two basic tape machines and tape types in as much detail as possible, and providing enough user control to get the job done, but without over-complicating things. In some cases, a less faithful emulation may produce an appropriate musical result, but if accurate tape emulation is what you're after, I think you'd be hard pushed to get closer than VTM.
By way of CPU load, VTM doesn't seem unduly greedy, although it still takes up around four times the processing power of the Waves Ampex tape emulation. Running 16 or so on a modern machine alongside your other plug-ins should be practical — on my Mac Pro Dual Quad, 16 instances took up around half of one of the cores, according to Logic's CPU meter — but if your machine is struggling, a few instances placed on key buses and groups will provide the sound. In fact, on many of my mixes I found that putting one stereo mastering machine over the whole mix and perhaps one 16-track over the drum bus was enough to add the necessary polish.
Overall, I really love this plug-in and can't really find anything negative to say about it, other than that some users may be surprised at just how low-key the effect of a real tape recorder can be. On the other hand, if you're going to insert multiple instances in a mix, you don't want the effect to be overblown, as it is cumulative. For me, VTM supplies the essential magic of analogue tape in a manageable and utterly convincing manner, without the cost of maintaining a tape machine, buying tape, spending time aligning the thing and then having the tape run out just before the end of the only perfect take of the day! Having used it, I'd find it very hard to go back to mixing without it.
The obvious alternative is Universal Audio's Studer A800 plug-in, which requires their UAD2 hardware platform to run. Those who are mainly interested in master-bus processing should also investigate Waves' Kramer Master Tape.
To accompany this review, we've prepared a number of audio examples using the multitrack recording of Esqueleto's 'Empty Pockets' kindly provided by drummer and studio owner Bugs, who runs Half-Ton Studios near Cambridge. You can hear Slate VTM in action here: /sos/sep12/articles/vtmmedia.htm