We check out the new hardware incarnation of Antares' award-winning microphone-modelling software.
The Antares Microphone Modeller started out as a software plug-in, which we reviewed in SOS April 2000. It has since won a TEC Award and has become a highly popular and useful tool in a broad range of recording and post-production applications. The Antares AMM1 performs exactly the same task as its plug-in sibling, namely transforming the sonic characteristics of one microphone into those of another. It also operates in pretty much the same way, albeit with dedicated hardware controls and I/O facilities — firstly, it removes the sonic characteristics of the source microphone; then it imposes the characteristics of a different, more desirable, microphone; and finally it models the variable saturation distortion of a typical virtual valve preamplifier. The microphone models incorporate appropriate variables such as low-cut filters, selectable polar patterns, and proximity effects, allowing the best possible match with reality.
The rear panel of the 1U, 150mm-deep AMM1 case boasts a promising selection of interfaces. Starting at the right-hand side, a seven-pin DIN socket accepts three low-voltage AC power rails from the supplied mid-cable transformer unit. Moving left, a quarter-inch tip-sleeve socket provides a footswitch connection used to bypass the signal processing. While powering up, the machine automatically detects whether a normally-closed or normally-open switch is attached.
A pair of five-pin DIN sockets provide MIDI In and Out, although the latter is not currently enabled. Next along, a pair of XLR sockets provide AES-EBU digital I/O interfacing and a software menu determines which AES channel is processed (left or right) and what happens to the unused channel (passed through unchanged or muted).
The final set of connectors are for the analogue interface. Balanced line-level inputs are accepted through XLR or TRS quarter-inch jack sockets, the latter taking priority, and the sampling rate can be selected through a menu (44.1 or 48kHz). The analogue outputs are provided simultaneously on an unbalanced quarter-inch socket and a balanced male XLR, as well as through the digital output, which carries the processed signal on the left channel with an inverted (opposite polarity) version on the right channel.
Moving around to the front, the AMM1 is simply styled, with only a handful of controls and a clear, 20-character, two-line LCD screen. The leftmost switch powers the unit, although it does not provide isolation from the mains as only low-voltage AC reaches the box. A red LED illuminates when a digital input is present, and the adjacent rotary encoder adjusts the digital signal level in conjunction with a five-segment bar-graph meter (scaled from -3 to -30dB). When the level control is moved the LCD briefly shows the input level in dB.
A wide button below two red LEDs selects either the source or modelled microphone for editing and two similar buttons to the right, labelled Preserve Source, allow the treble or bass portion of the source signal to be passed through the unit without being processed. The lightly-detented rotary encoder, labelled Data Entry, and the LCD provide the means for parameter adjustment, and the next four buttons access parameters for the source or destination microphone setup. The next control is another detented encoder which alters the amount of emulated valve saturation, and the LCD shows the nominal drive level in decibels whenever the control is moved. A pair of wide black buttons to its right engage the System Edit mode and impose an overall bypass — both have red LEDs to warn when these functions are active. Bypass can also be activated by footswitch or via MIDI, the LED illuminating in all three cases. The final control adjusts output level attenuation, again with a detented encoder. Consistent with the other level controls, moving the encoder displays the amount of output attenuation on the LCD panel.
Setting up the AMM1 is a logical enough procedure. First you determine the signal input source, sample rate and level. An important note here is that the level control is after the A-D converter, so optimising the conversion process rests with the send level from the mixer. This should not be a problem if using aux sends, but the insert points of many mixers may well under-drive the AMM1's analogue inputs — the specifications state they are calibrated for peak levels of +17.9dBu.
Once the input has been optimised, the next stage is to select the source microphone model — this highlights the necessity to log which microphones are used during a tracking session, along with polar pattern and low-cut filter settings, and their proximity to the sound source. There is a considerable variety of mic models to choose from (see 'Available Models' box for further details) and a lot of these include variations (such as later versions, or with windshields attached). There is also a source modelling bypass mode available, for when a DI box or pickup has been used, and an A/B compare function enables the comparison of two selected source models. The latter is an extremely powerful function as it recalls not only microphone makes and models, but also relevant low-cut, proximity and pattern settings, and is very useful when trying to find the best match of settings if your specific model of microphone is not in the list of those supported in the software.
Once a source model has been selected, you select the degree of proximity effect to correct for if a pressure-gradient mic has been used, and you also select which filter setting and polar pattern were used on the source mic. The latter setting can appear to have little effect with some mics, as their on-axis performance remains fairly consistent with all patterns, although others exhibit wildly different on-axis responses with alternative pattern settings. Once the source microphone's details have been entered correctly, the same process is repeated with the microphone to be modelled, though here the characteristics are being added rather than removed.
Once the microphone modelling has been set up, valve saturation can be dialled in to emulate the effects of a high-quality valve preamplifier. The Saturation control caters for a range of effects from complete linearity, through gentle 'warming' to fairly aggressive distortion. However, the saturation drive control is limited to +10dB of boost and if the input signal peaks at less than -10dBFS then no distortion will be produced. It is therefore necessary to juggle the AMM1's Input Gain and Saturation Drive controls together to achieve the desired effect — in some cases you may also have to adjust the send level from the mixer. The quality of the valve distortion may be desirable as an effect in its own right, so it is useful that both source and modelled microphones can be set to bypass, leaving the AMM1 to serve purely as a valve preamp modeller instead.
Worth a brief mention is that there is a small processing delay through the machine (particularly via the analogue interface), which may need to be addressed in some circumstances. However, this is far from excessive and will probably not cause a problem for most users.
The AMM1 produces impressive results, but it is not magic and needs to be used carefully. For example, if the original signal was recorded with a distant omnidirectional microphone, picking up a lot of room sound along the way, using the AMM1 to model a directional microphone will not help remove this room sound! Similarly, if the source mic employed a low-cut filter, and the modelled mic does not, the overall bass boost applied through the two modelling processes may result in an appreciable noise contribution!
To assess the AMM1, I experimented with an AKG C414 ULS and a Neumann U87 mic, both mounted close to the same sources (voice, guitar, and a loudspeaker replaying a variety of more complex signals). I set the machine to replicate the Neumann using the AKG, and vice versa, in each case comparing the output directly with the real thing. The results were not identical, but shared enough of a resemblance to the primary characteristics of each mic to be useful — bear in mind, of course, that the natural variances of individual microphones with age and abuse would affect the accuracy of the modelled mics anyway. I'd say that, where recordings have been close-miked to minimise off-axis sound, the AMM1 will prove acceptably convincing in simulating a number of the qualities of a more desirable microphone.
Although the AMM1 is intended to be an accurate mic modeller, it strikes me that the vast majority of owners will probably use it more as a device for spectral shaping and tonal alteration of a signal source — to create interesting sonic signatures rather than replicating the sounds of specific mics. Certainly, reading through the expansive list of endorsees on Antares' web site, most describe using the machine as a means of adding colour and interest to recordings and mixes.
The AMM1 will appeal to a broad range of users: first, those who have a few decent mics already but dream of owning a more resplendent mic cupboard; secondly, engineers who want an unusual method of adding colour to their tracks and mixes; and finally those who would find the ability to replicate the sound of specific microphones very useful. The bottom line is that the AMM1 works in these applications, and works well.
As shipped, the AMM1's internal EPROM stores 100 different microphone models, each with Source and Model variants. FLASH memory is also provided to store additional models downloaded from the Antares web site and transferred as standard MIDI files containing SysEx data. There is a good selection of esoteric and high-end mics, as well as more familiar mid- and budget-priced models, though the selection still only includes about a tenth of the possible mics available.
At the budget end, there are microphones from such manufacturers as Alesis (the AM61), Behringer (the ECM8000 and XM8500), CAD (Equitek E-series), and Shure (SM57/58 and Beta 57/58 models), to name but a few. Moving up to the mid- and high-range mics, offerings include AKG's C1000S, C3000, C4000B, C12A and four different types of C414; a wide variety of Audio Technica mics; Neumann U87 and TLM103; Brauner VM1; CAD VSM1; Coles 4038; B&K (DPA) 4007; and Shure SM81 and 98A. The list also includes the Sony C800G; Microtech Gefell's UMT800; and esoterica like Telefunken's U47, the Soundelux U95S and a selection of vintage Tannoy ribbons. Other manufacturers included are Audio Engineering Associates, Audix, Beyerdynamic, Earthworks, Electrovoice, Groove Tubes, Lawson, Manley Labs, Oktava, RCA, Rode, Royer, and Sennheiser.
Although it's easy to think of favoured mics not in the list, the variety is impressive and even if a specific model can not be matched, it is possible to get pretty close by selecting an alternative which is similar in operating principle, diaphragm size, vintage, and so forth. New additions are also appearing on the web site, with three new microphone models on line at the moment: the beyerdynamic M500, Electrovoice RE55 and Neumann M147.
Given the current state of the art in digital modelling, the inner workings of the AMM1 are relatively straightforward. It uses a technique based on impulse measurements which has been around for quite a while in various guises — the current ‘big boys’ in this technique being the Sony DRE-S777 sampling reverb and the Sintefex FX8000 Replicator. For the AMM1, on-axis impulse response measurements were derived from a broad selection of microphones, including any relevant variations resulting from alternative polar patterns, high-pass filters, and the proximity effect associated with each design. From this data it is possible to apply the inverse of a particular impulse response to neutralise the characteristics of any original source microphone within the machine’s library — and, once an idealised ‘flat response’ has been obtained, the characteristics of any other desired microphone can be imposed by convoluting the signal with the appropriate impulse response.
The limitations of this approach are worth going into. Firstly, information cannot be recovered if it was not present in the first place. A cheap and nasty microphone may not even capture some elements of the frequency range, and expecting the AMM1 to then replicate the sound of a Neumann U87 from this inadequate source material is simply unrealistic. Secondly, it is only practical to model the on-axis response of a given microphone and, whilst this is a perfectly acceptable compromise for many applications, it also has significant drawbacks which limit the accuracy of the modelling under certain circumstances. For example, any appreciable level of off-axis sound arriving at the source microphone will be neutralised as if it is arriving on axis — the modelling cannot determine the direction of any incoming sound. However, as we all know, the off-axis response of a microphone varies according to the degree of incidence, the polar pattern, and the physical attributes of the particular microphone involved (diaphragm size and shape, for example). Consequently, a ‘neutralised’ and remodelled sound containing off-axis spill will not match the sound produced by a real example of the replicated microphone in the same location.
The AMM1 is therefore at its most accurate if the source microphone is used in such a way that the wanted sound arrives exactly on axis and with negligible contributions from any off-axis components. In practice this means using close-miking techniques, ideally with directional microphones. Fortunately, this technique happens to be precisely the kind of thing employed for most recordings in semi-professional and home studios without significant acoustic treatment, as well as on the live concert stage.
Where it is of less validity is in remodelling microphones used in distant placings and in reverberant rooms, where there will be a much greater contribution from off-axis sound, and therefore a large divergence from the ability of the modelling software. That is not to say the machine won’t impose new characteristics which may be sonically interesting — just that they will bear rather less of a similarity to the modelled microphone had the real thing been used under the same conditions.
- Creative and unusual effects are produced easily.
- Reasonably realistic modelling in ideal circumstances.
- Intuitive user interface.
- Small processing delay may cause some users problems.
- Modelling cannot compensate for poor source quality.
- Only about 10 percent of available mics are currently modelled.
A hardware version of a popular software plug-in. Within the inherent limitations of microphone modelling, the AMM1 does a reasonable job of imitating specific mics, but is even more useful as a creative tool in its own right.