Mastering engineers are notoriously fussy about EQ, and until now, plug-in manufacturers have struggled to compete with the best hardware designs. Can Algorithmix's Red and Blue change hearts and minds?
Until fairly recently — at least until the end of the last century(!) — most professional mixing and mastering engineers had 'old-timer' ways of working that meant they were wary of in-the-box software processing. Bobby Owsinski's The Mastering Engineer's Handbook, published in 2000, barely even mentioned plug-ins, and though it was less elitist in its examples, Bob Katz's Mastering Audio, published two years later, still concentrated on hardware (both analogue and digital), giving only qualified approval of software processors. Common complaints were about the compromises forced by insufficient computing power for top-notch DSP, about the superior ergonomics of outboard equipment, with one physical knob per function, and, of course, the intrinsic superiority, for certain processing, of analogue components.
All this has changed dramatically in the past three or four years. As a replacement for digital outboard there is now plenty of native DSP in desktop computers for high-resolution processing, which can also be augmented by shunting audio to dedicated DSP devices such as the SSL Duende, TC Powercore, Universal Audio UAD1, Weiss Powerhouse and Waves APA32/44; and as a replacement for analogue outboard, new emulation techniques are coming frighteningly close to the originals. Plug-ins also have intrinsic advantages including the ability to save (and share!) unlimited user-generated presets, the possibility of using multiple instantiations in a mix and, not least, freedom from clock-routing issues. Even the question of ergonomics, which seemed on one scoring system to have been answered decisively in favour of the hardware, has changed. Plug-ins that tried to emulate the appearance as well as the sound of vintage analogue equipment have often been ergonomically poor — rotary pots were obviously invented to be turned by hand, not mouse — so programs that were easy on the ear, and even more so on the eye, were horrible on the hand that was trying to mix with them. But new arrivals such as McDSP's range of plug-ins (see review in last month's issue) show that software design has grown up, and that new processors no more have to look like an analogue box than new electric guitars still have to look like either a Les Paul or a Stratocaster.
The software processors under review here, Algorithmix's Linear Phase PEQ Red and Classic PEQ Blue, seem set to take the digital plug-in advantage to a new level, going well beyond what is feasible (or even possible) in any kind of outboard, digital or analogue. According to Algorithmix, their Red EQ offers not only linear phase purity of sound, but also a unique proprietary technology which allows for continuous slope adjustment of shelving filters: for details, see www.algorithmix.com/en/peq1.htm. Blue, meanwhile, is a modelling design offering "the most complete collection of classic equalisers ever in one plug-in", importantly including two designs which have analogue sonic characteristics, but which are "impossible to implement in the analogue domain".
We installed both EQs onto our studio workstation running Sequoia 9 on Windows XP Professional, and encountered no problems or irritations except the necessary addition of yet another Wibu USB dongle which authorises the software and has to be present in the port for it to operate. Both Red and Blue are available for the Windows platform only, and they can work as either VST or Direct X plug-ins.
Let's get some grumps out of the way to begin with. There have been complaints on various Web boards that although Red and Blue may have an unsurpassed inner beauty, on the outside they are simply pug-ugly. I'd not go quite that far, but they're certainly not the prettiest around, with their eponymous colours and grey pretty much exhausting the palette, and maybe the customer could have expected something a bit more eye-catching. On the other hand, they are very affordable relative to their hardware counterparts and comparable software competitors, and if that means a sacrifice of beauty for function, I'm more than willing to accept it.
However, there are also darker mutterings about the overall effectiveness of the Graphical User Interface, and anything that affects absolute functioning has to be taken rather more seriously. The main complaints here revolve around the layout and possible errors of operation that can arise from this. Red and Blue are identical in layout except for the function of one small box. On the left of the GUI are four filter parameter boxes for EQ Type, Frequency, Gain and Q, plus a fifth box. In Red, this is where you choose Resolution, while in Blue, it's where you choose from the 12 EQ types. On the right is a row of 10 red Band Select buttons, and immediately below them are 10 green Band On/Off buttons. Below that are four boxes which let you choose to Store and Recall A and B settings, a Bypass button, and a Flat button that resets all the parameters. These buttons are a bit cramped, and unfortunately, spreading the GUI across the full width of the screen just opens up a further expanse of grey between them, rather than giving them more room to breathe. Perhaps even worse, although the GUI is completely scaleable, and the graphics provide plenty of visual information, neither EQ can give you the numeric values for more than one band at a time. Also, turning on a band does not actually select it, so you can find yourself making changes to the wrong band; and more than one user has complained that it's all too easy to reset the plug-in when you think you're storing a preset.
In the main I am only reporting here the troubles of others, and not anything more than mild irritation from my own experience during the review, but then reviews can perhaps be a tad more leisurely than real life, and these are issues that definitely need to be addressed. Given the speed at which Algorithmix responded to my request about bundles (see below) and their general sensitivity to the needs of their customer base, I would expect to see these issues corrected in the very near future — ideally with a free upgrade for existing users.
What is most odd is that getting the interface right would seem to be the easier part of the task. Luckily, when we come to the harder part — the absolute sound quality — Algorithmix haven't put a foot wrong.
Algorithmix seem to be marketing the Red EQ as mainly for mastering, or at least for master buss processing, and Blue as a (multiple) channel insert for mixing. So although both can do double duty — within the limits of Red 's CPU hunger — that's mainly how I used them. Red is CPU-intensive, and introduces substantial latency, but these are features of all well-designed linear-phase EQs, so although they need to be mentioned, they're not really failings as such.
Red was given an immediate trial by fire: I had to master a CD of Shostakovich cello sonatas that had some marked room-induced nasties at the low end: in a number of places the joint excitation of the recording venue by the cello and piano had led to a messy indistinctness in the piano left-hand figures and in the cello lower string phrases. Because the problem was episodic, with the 'fullness' of the room at all other times quite desirable, I would normally have dealt with this kind of issue using our hardware Weiss EQ1 Dyn/LP MkII in dynamic mode — which, as its name suggests, only operates when the problematic frequencies reach a certain threshold (it's something like a multi-band compressor in theory, but in my experience far more transparent than most in practice). However, phase shift plays havoc with timbral accuracy, and even the best minimal-phase equalisers can introduce changes which stand out on natural instrument sounds. Mastering being the art of the compromise, I decided to trade a little overall weight for tonal reality and brought in the Red EQ to see if its linear-phase reduction of the offending frequencies would do the trick. The result was very satisfactory — in fact, surprisingly so. The low-end 'lump' seemed to be well defined, so a high-pass at 28Hz and mild cut with a moderate width at 200Hz cleaned it out entirely, uncovering the instrumental detail 'hidden' underneath.
At the other end of the spectrum I was working on a three-track demo by a local folk-rock band. This was nicely recorded but suffered from what used to be called 'NS10 syndrome' — that is, it sounded like it was mixed using Yamaha's once-ubiquitous nearfield monitors, whose high-end screech revealed great detail, or tore the hair out of your ears, depending on your point of view. In sheer self-defence, mixing engineers using these tended to pull back the highs (or tape tissue over the tweeters), and I suspect that this is what had happened here, as the mix had a certain overall dullness that was especially evident in the vocal choruses. We were charging a minimum (local talent) rate for this work, but even so, it was just so easy to inject some life into the mix using the Red EQ that it felt faintly improper to get paid for it at all. I simply added a gentle rising shelf from just under 4kHz, and then tweaked the Continuous Slope parameter until it sounded sufficiently bright and perky. The acoustic guitars jangled nicely just a little more, the vocals came forward just enough, and there was suddenly some air around the music as though a curtain had been drawn aside. This CS parameter is unique to Algorithmix, and means that as well as choosing whole filter orders between one and four poles, you can also continuously shape the slope with 'imaginary' orders. This allows very fine adjustments to be made between, at one extreme, a very steep start-up, or at the other, a curve so gentle that with an appropriate boost it 'reaches back' below the chosen frequency.
It was using this same material that John Watkinson's claim (in The Art Of Sound Reproduction) that "the sound of an EQ is the sound of its phase response" hit home. With a non-linear-phase EQ, a boost of just 3dB or so at 3kHz is immediately obvious — in fact many engineers have learned to hear EQ, or hone their EQ skills, by doing blind 'boosting and guessing' exercises. But with Red this extra was simply not there. Looping a passage where the female vocal had minimal backing I did just that: I set a bell filter at 3kHz, with a Q value of 0.8 and started to add gain. Even at +6dB, although the sound was obviously no longer right for the track, it still sounded relatively natural: the same voice, but brighter, rather than a brighter voice.
And here we come to a bit of a paradox. Throughout the review process I regularly referred my results using Red back to our standard linear-phase EQ: the outboard Weiss. Given that this box, in its full Dyn/LP version, costs roughly six times as much as Red, I had been expecting the latter to show some slight drop in quality by comparison. In fact, Algorithmix Red equals, and in some ways surpasses, the performance of the Weiss. Bob Katz and I are now working on the second edition of his Mastering Audio, and one of the many new issues to be discussed in that book is this very comparison between Red and the Weiss. After I'd done my review sessions, Bob sent me the results of some impulse response tests he'd done which showed how the filters in the Red EQ have textbook transparency compared to the Weiss, despite also suffering minute phase anomalies that the Weiss did not. The difference is to do with design rather than implementation, and the sonic result is absolutely minimal — far less than between Red and PSP's Neon, for example — but nonetheless real. The paradox I mentioned is this: even though I now know that Red is actually less invasive than the Weiss, for some material I still have a definite preference for the Weiss. On an exposed female vocal, or high strings, for example, whatever the Weiss was doing less transparently still sounded slightly more natural to my ears. It's not that it was euphonically smoothed or coloured, just — bizarrely — more natural.
Whereas the linear-phase Red glories in its transparent processing of the material, the Blue EQ wears its colouring proudly. Algorithmix's marketing material doesn't exactly crow about this plug-in, but they do know it's quite special. Although they are being extremely coy about revealing the inspirations for their EQ types, Algorithmix drop a few hints in the manual (for example, there are "three constant-Q equalisers with characteristics mostly used by some American brands") and, anyway, it's pretty easy to guess what they might think they should include from the analogue models that find their way onto the most engineers' wish lists: SSL, Neve, Pultec, Massenburg, Manley and so on.
According to Algorithmix, the filters that are most significant in determining the character of the EQ are the bell or parametric bands, and in particular, the way the bandwidth of the bell changes (or not!) when boosted or cut. As suggested by their names, Constant-Q filters have, or strive for, unchanged bandwidth whatever the gain changes, whereas in Proportional filters, bandwidth generally varies inversely to gain: the more gain, the narrower the Q.
With so many models to play with, taking up so little CPU power, you could spend hours tweaking and experimenting; but as my reviewing job sits alongside day-to-day working, I had to find a more rational way of proceeding in order to both investigate the possibilities and get the projects finished in time. What I came up with was exceedingly simple. Given an EQ task in mixing, for example, a problematic electric bass on another nouveau folk-rock track, I started at the top (Model 1), dialled in what I thought was a ball-park remedy, and then clicked through the model types (a right mouse click in the box takes you to the next model number, looping back to the beginning again when you get to 12), watching the curves update as the different ways of implementing Q kicked in, until I found the one that seemed best suited for the purpose. I then concentrated on that, making tweaks and changes as required.
The only time this is not advisable — a warning for those in a hurry who decide not to read the manual first — is when changing between serial and parallel model types. Because the serial and parallel models have totally different DSP structures, the result could possibly be a noise burst. In action this didn't happen to me, but probably because I was using fairly gentle curves. I didn't do any destructive tests to see if I could provoke a noise burst.
Overall I found the Blue EQ to be quite amazing. It's very easy to dial in some non-fatiguing sparkle or to add those hints of warmth that make for extra tension or listenability. I found myself concentrating on just four of the EQ types: two vintage parallel models, one proportional, and one — the Const-Q New — which I assume is unique to Blue. The first two gave warmth with varying degrees of accuracy, while the second had more attitude, especially when pushed harder, in a way that worked very well, for example, for giving some grit to relatively clean electric guitar. The last was the most controllable and found its way into my mastering workflow as well.
With the Red and Blue EQs, Algorithmix have produced some very special plug-ins which, in terms of function and relative affordability, look bound for success. When I received the review copies from Algorithmix, they quoted a bundle deal for Red plus Orange (another linear-phase EQ) but only a stand-alone price for Blue. As it seemed to me that Red plus Blue would make a very desirable package for mixing and mastering, I suggested they bundle them too, and I'm very happy to report that they agreed.