Danish designer Jakob Erland's quest to design an innovative audio processor using only 1950s technology has created a very special device indeed...
Gyraf Audio's gloriously titled Gyratec XXI Magneto Dynamic Infundibulum (G21 for short!) was one of the most unusual products announced at this year's Musikmesse show in Frankfurt — and it turns out also to have been one of the most interesting and innovative. Like so many great inventions, it started with the simple question: "What if?”. In this case, Gyraf's mastermind Jakob Erland asked himself, what if he used 1950s analogue electronics to do something new, rather than simply to imitate or refine popular classic gear. After two years of design and development work, with Danish mastering engineer Emil Thomsen helping to fine-tune the product, Erland has his answer — and it really is something special.
The mystical name gives little away, so what exactly is this 'Infundibulum' — a Latin term that means 'filter'? Well, it's certainly not a simple filter! Its working title during the early stages of development was Audio Shaver, and that gives us a stronger clue: essentially, the G21 is a multi-band soft-clipping processor. It was originally conceived as a device that could be used during the mastering process, the aim being to reduce the crest factor by several decibels, thus helping the final brickwall limiter to operate in a much more comfortable range.
That sounds pretty straightforward, but having had the final product on test in my studio for over a month now, it's abundantly clear that this is by no means a simple processor to deploy! Indeed, there's a label on the front panel which screams 'USE WITH CAUTION' (Gyraf's capitals, not mine), and I can understand why. It's also clear that while this is, indeed, a very useful mastering tool, its range of possible applications extends far beyond mastering.
Unusually, the G21 is made entirely from passive analogue circuitry, which means there's not even a mains inlet on the rear, only the analogue input and output connectors. The construction of the solid-looking enclosure, as well as the selection of parts employed under the lid, meets the most stringent standards, but, due mainly to the lack of need for a PSU (and, therefore, a heavy toroidal power transformer), the G21 weighs surprisingly little.
All knobs and switches feel rock-solid, and the meters, with their bakelite bezels, add another beautiful detail to the staggering visual appearance of the G21. The controls, though, hint at greater complexity. Not only are there enough controls and meters to fill up the front panel of the chunky, 4U, rackmount enclosure, but many of those controls are very flexible, and there's also interaction between some of them.
The front-panel layout comprises three distinct sections. In the first, there's a bypass switch with a stepped trim pot for the processed signal. There's also a switchable low-cut filter, with corner frequencies at 25Hz and 35Hz and an 18dB/octave slope, which may be used to remove any unwanted infrasonic rumble.
The next section is a full-range clipper, which affects the entire spectrum of the program material. This is controlled via a threshold switch, with 11 positions and a stepped level-pot, and it can be switched off when the threshold switch is turned fully clockwise. This full-range clipper also boasts a clip-range control, which activates various pre-clipping emphasis curves: White (linear), Pink (3dB/octave) and Red (5dB/octave). These curves are used to make the clipping more sensitive to higher frequencies, thus effectively leaving all the low-end transient 'punch' intact.
The third department is the most complex. Here, the G21 provides three individual, frequency-specific clipping bands for the low, mid-range and high frequencies. Each band employs its own stepped level-pot, just like the full-range clipper, and also a rotary frequency-selector switch with 11 positions. The latter control allows the user to define the frequency range for each band. The low band offers a shelving curve, with corner frequencies between 35 and 245 Hz — thus covering the lower mid-range down to the fundamental octave of the audible spectrum — while the high band is based on a bell filter, with a selectable range between 3.3 and 15 kHz. The middle band can be toggled between bell (270 Hz to 2.7 kHz) and shelving (235 Hz to 18 kHz) curves.
Each band also offers a three-position mode selector, which controls the basic operation of the respective band. The three bands can each individually be hard-bypassed or switched into the clipping mode, and there's a cut mode, in which each band acts as a subtractive, passive EQ filter. So the cut mode can simply be used as an EQ if you wish, but it's also a very useful means of determining the frequency range(s) in which you want clipping to be performed.
So far, then, you should have a feel for the basic controls and principles of operation, but this tells you little of what makes the G21 unique. To reveal that, I'll need to explain in greater detail how its circuitry works.
The first thing to note is that, being an entirely passive device, the G21 has no active amplifier stages, and the signal path, which largely comprises inductors and transformers, can accurately be described as 'minimalist'. It's of the highest calibre, though, both in terms of the cleverness of the design and the quality of the components used.
Unlike a traditional multi-band compressor, in which the bands are separated with filters (which inevitably affect the audio signal, often causing phase issues), the G21 works more like a parallel dynamic EQ. In fact, its frequency-selective bands employ passive, inductor-based LC filters of the kind Gyraf have long used in their Gyratec XIV Stereo Parallel-Passive Equaliser.
Each channel's single audio transformer acts simultaneously as both the input and output transformer, but it also takes care of the lion's share of the soft-clipping/saturation. All of the clipping stages and bands are tapped via a tertiary winding on the transformers, meaning that the clipping circuits present an impedance to the transformers' magnetic fields. This, in turn, results in varying degrees of collapse in the magnetic flux inside the transformers, and it's this action which attenuates the program material passing through. It's extremely simple in concept, yet absolutely effective — and that's what makes this design so brilliant.
Being passive, the clipping networks are inherently bi-directional: the tertiary winding on the transformer not only feeds the clipping bands, but the rather complex impedance load that's generated in the clipping circuits is also fed back to the transformer, via the same winding. Putting this into simpler terms, one could conceive of the load on the tertiary winding as a kind of side-chain signal, which affects the way the audio signal is transmitted from the input (primary winding) to the output (secondary winding). But this tertiary winding acts simultaneously as the source and destination of our 'side-chain'!
This may seem like wizardry already, but Erland's design goes a step further. Each band of the clipper opens a frequency-selective transmission line, which is shaped by its LC filter. Clipping generates harmonics, but because these transmission lines are frequency selective, at least some of the harmonic by-products have a hard time getting back into the audio signal path.
In case this sounds confusing, here's a simple example using a 1kHz band-pass filter. Frequency content in this range is transferred and clipped in the transmission line, but the harmonics at 2kHz, 3kHz and higher cannot 'escape' back through the band-pass filter. At least to some degree, then (the slope of the filters is not infinitely steep), the harmonic content generated by the saturation process remains 'trapped' inside the clipping circuit. This is where that word 'Infundibulum' comes from. As Erland puts it: "We passively funnel audio through to the clipper stages, giving the distortion products a hard time to escape.”
The most obvious impact of this is that the G21 sounds, at least up to a certain point, more transparent than other saturation devices. But another is that Erland had the opportunity to use the residual energy from this process to drive the needle movements of the G21's ballistic meters without requiring any external power — and if this can't be considered ingenious, I don't know what can! Speaking of the meters, each band, as well as the full-range clipper, has its own. The meters display the actual current that's being removed from the audio, but they're not calibrated to display relative values — so a low reading on the meter for a low input signal can actually be describing stronger saturation than a higher reading on a hot input signal. (That's another sound reason to obey Gyraf's use-with-caution directive, then!)
But why did Erland choose to make the entire unit passive? Wouldn't there be benefits in using valve-based stages, as in other Gyraf units? When I asked Jakob Erland that, he laughed and simply said "because we can!” On a more serious note, though, there are advantages in not employing any active circuitry. The G21 generates solely what could be called 'first-order non-linearities'. Most active stages are controlled by negative feedback, and in this way clipping is piled on top of clipping. In other words, leaving the entire unit passive means that both the length of the signal path and sonic artifacts can be kept to a minimum. From an audiophile perspective, this integrity is a very good thing.
Erland has taken considerable care over the selection of components, and you can tell this from the way in which many of them have been scribbled on in his own hand! I've already discussed the proprietary transformers, which Erland explains are "made by an ultra-nerd friend of mine, who works in the power-transformer business... but his heart is in audio!”. The inductors, of which the G21 employs no less than eight (six multi-tapped and two single-tapped), are wound in-house by Gyraf. There are plenty of more conventional high-quality components too, such as WIMA capacitors. To get the stereo tracking right, both the absolute values and the temperature coefficients of the components are matched — a labour-intensive task, which is one obvious reason for the asking price. The threshold settings of the full-range clipper employ an array of diodes of various Schottky, Tunneling and Germanium types, all of which offer different forward frequencies and response curves, resulting in a gradually harder 'knee' of the saturation curve as the threshold is lowered.
Mastering the G21 (pun intended), requires plenty of practice, good ears and a precise listening environment, but, despite the learning curve, I found the results extremely rewarding. It definitely takes more time to get an idea of the capabilities and the sonic scope of the G21, and to learn to use it effectively, than when learning to use other, less complex processors. I found that it required me to think about and listen to audio in a very different way than I usually would. At first, it sometimes seemed to act rather unpredictably, and the highly interactive nature of the multi-band section only added to the initial feeling of chaos. But all that time and effort was well worth investing.
The low-cut filter and the full-range clipper are the easiest sections to understand, but even the clipper presents a huge variety of tonal options, depending on different threshold/level settings, and, of course, the source material itself. I found the clip-range control really useful, particularly when employing the 'Red' pre-emphasis curve, as I found that it was possible to retain all the low-end punch even when using pretty drastic settings. This came in handy whenever the body and the groove of the program material relied heavily on low-end elements, such as bass and kick drums.
Using the multi-band section requires that you pay attention to the tiniest of details. This section can be viewed as a three-band dynamic EQ without any time constants. In fact, I can't think of a better way to describe it, really: it allows the user to shape, or 'shave' transients in selectable frequency ranges without creating any of the pumping artifacts that might occur when using a regular multiband compressor. Because of this, it offers sonic sculpting effects that are reminiscent in some ways of analogue tape, but with a much greater degree of user-control over the result. You can read more of my thoughts on that in the 'An Alternative To Analogue Tape' box. I'm unaware of any other device currently on the market which can perform quite the same function, and while I don't see the G21 as a direct replacement for analogue tape (it wasn't intended to be), the two means of audio processing share far more similarities than differences.
Unsurprisingly, then, I found the G21 to work exceptionally well on virtually all of the sources where I could also see the benefit of printing the final mix to tape. This included most acoustic material and genres such as jazz or rock — or, for the lack of a better explanation, productions with a substantial number of microphone recordings of 'real' instruments.
In many cases, I was able to use the G21 to clean up excess transients that seemed to be 'buried inside' the mix — and that's a task I find extremely difficult using more conventional tools. Once set up appropriately for this task, the device seems to bring out a really rich low-mid warmth, and the spatial information in the material becomes much more apparent, with rooms seeming a little larger and more 'three-dimensional'. I suspect that some of those buried transients tend to mask low-level spatial information, and that by taming those transients, there is — to put it very simply — much more space inside the music that allows the rooms to breathe.
On the other hand, the G21 seemed far less appropriate when used on EDM mixes. These form a large part of my daily work, and with most such material I always had a hard time finding a sweet-spot setting on the G21. Either it wasn't doing enough to justify keeping it in the signal chain, or it was sucking out too much of the punch and low-frequency energy that's essential in this genre. That said, I wouldn't normally use analogue tape as the master medium on those productions either, so the G21's unsuitability in these roles came as no surprise.
In most cases, I found that the G21 could be used to reduce the crest factor by 1-3 dB without unwanted side-effects. This reduces the load placed on the final limiter and may already be reason enough to justify employing this unit in a mastering role. But there are many more possibilities, especially when working with acoustic drum sounds. In other words, the G21 is a unique mastering device, but there's such a lot of room for experimentation that it is certainly not limited to mastering applications!
The G21 sounds exceptionally transparent for a saturation processor. Active input and output stages usually account for the lion's share of sonic coloration imparted by an audio processor, and their absence in this design means that the actual processing is completely out of the way of the audio signal unless the unit is clamping down on transients. As it's passive, there will be an insertion loss of 2-3 dB when putting the G21 in the processing chain (the actual amount depends on the equipment it is connected to), but this is easily restored later, and only serves to prove the clever design of the proprietary transformers, because passive devices typically result in a much higher insertion loss than this.
The Gyratec XXI is an extremely versatile and powerful processor that combines two key qualities: it sounds rich and wide when appropriate, but absolutely pristine and clear when necessary. The device offers ample transient control, and its unique circuit design genuinely pushes the envelope of the audio processing technology that's familiar to most of us.
No matter how you look at it, the G21 is no set-and-forget tool. It requires a lot of attention because it offers so much control and, sometimes, because the sweet spot of the processing is so small. I strongly recommend that new users start with the multi-band section in bypass! The three frequency-selective bands can come in handy in many cases, but one does not necessarily have to activate all of them at the same time. A little of the G21 really can go a long way!
While it may not be suitable for everyone, the G21 can prove an astonishingly versatile and powerful machine when controlled by the right pair of hands. The concept might seem difficult to understand at first glance, but having used and abused so many conventional compressors, EQs and saturation devices over the years, I'm thankful that surprise processors such as this come along from time to time! It might not fully replace any one of those tools, but it can greatly reduce the job you ask them to do in many situations.
Finally, there's the question of cost. The G21 isn't cheap by any stretch of the imagination, but in this small, boutique, high-end sector it certainly isn't overpriced either, particularly when you consider factors such as the small market for such a device, the long development time (two years seems long to me, even for such a boutique company), and the highly specialised parts and their matching. By way of comparison, an Anamod ATS1 analogue tape emulator, with a few additional cards, is in a similar price bracket.
There is no direct alternative. At a similar price point, analogue tape simulators such as the AnaMod ATS1 can achieve similar results in some respects. Following the same logic, a vintage tape recorder made by a manufacturer such as Studer, Telefunken or Ampex can offer a comparable amount of transient control without applying time constants. But it should be pointed out that the Gyraf G21 is in no way conceived just as a tape simulator: while it can achieve similar-sounding results, it offers the user far greater control over the result.
The author created a number of audio examples during the course of this review, and these are available on the SOS web site.
Analogue soft-clipping is the practice of driving an analogue circuit so hard that it behaves in a non-linear fashion, and it's an important tool for anyone interested in smoothing out dynamically unruly audio signals. It's also a technique used simply to increase the RMS level of an audio signal. The sonic effect of slightly overdriven analogue circuits has been both beauty and burden to engineers for many decades: a flattened-out transient response can lead to dull and boring recordings, but adding just the right amount of saturation is one of the more important reasons why so many productions from the era of transformers, tubes and magnetic tape recording sound so appealing, even today.
For many years, I've been running a well-maintained Studer quarter-inch tape recorder in my studio, and — when sonically appropriate — I always use this as the main master medium for my mixes. I have always seen the sonic effect of tape as some kind of multi-band dynamic shaper; tape saturation can act almost like a multi-band compressor that has an infinite number of bands, and without time constants — and that is the very nature of soft clipping! There are some ways to control this effect —for instance, the level going to tape, the tape speed or the tape formula being used — but there are clear limitations: either it can work in a given situation or it can't, because there is only so much possibility for fine-tuning.
In contrast, the Gyratec XXI offers ample control over the frequency-dependent transient response, and it does not impart the inevitable artifacts that tape and tape machines do, such as wow/flutter and the much higher noise floor. The G21 was not conceived as — and I don't consider it to be — a perfect analogue tape emulator, but in practice it can yield some very similar effects without the inevitable sonic sacrifices, while affording the user a much greater degree of control over the results.