Correcting the frequency response of a monitoring system is one thing, but Dirac claims to correct time-domain response too. Can it possibly work?
Paul Dirac was a British mathematician and physicist who made huge contributions to the theories of quantum mechanics. He served 37 years as the Lucasian Professor of Mathematics at Cambridge University (a post that both Isaac Newton and Stephen Hawking held), and his 1930 book The Principles Of Quantum Mechanics is still relevant today. Dirac shared the 1933 Nobel Prize for Physics with Erwin Schrödinger (he of the infamous live/dead cat thought experiment), but is probably best known for his eponymous equation and work on the mathematics of impulse signals.
So why am I telling you all this? Well it perhaps hasn't escaped your notice that the subject of this review is also called Dirac. And that's not a coincidence. The founders of Dirac the company chose the name because Dirac the physicist's work plays a fundamental role in their monitor and room acoustics correction algorithm.
That's the thing about theoretical physics: ideas that, at the time of conception, seem almost beyond the esoteric and of absolutely no practical value can turn out years (or even decades) later to have real-world applications. If you're interested, I've explained, or at least tried to explain, a little about how Dirac's work found its way into a room-correction plug–in in the box.
The Dirac Live 'room correction' system was originally conceived in 2001 by a group of PhD students in the Signals and Systems group at Uppsala University in Sweden. To date, Dirac has been found primarily in the consumer and automotive audio sectors where companies such as Arcam, Focal, NAD, BMW and Volvo have built Dirac functionality into their hardware. Now however, with the launch of the Dirac Live For Studio plug‑in, Dirac is joining Sonarworks, ARC and Trinnov in the pro audio room and monitor acoustics correction market.
Live In The Studio
There are three elements to Dirac Live For Studio: two software and one hardware. The hardware element is an omnidirectional measurement mic, and Dirac recommend the MiniDSP UMIK‑1 USB from www.minidsp.com/products/acoustic-measurement/umik-1_ as a suitable option, so that's what I used. Other measurement mic options are available, of course, but they need to have an appropriate calibration file; using an uncalibrated mic would be akin to using a ruler with no length scale markings. The UMIK‑1 is a USB mic so it connects directly to the Mac or Windows PC running Dirac rather than via an audio interface, although a conventional mic and interface connection option is also available.
The two Dirac software elements are the Dirac Live analysis and EQ filter generation application (this is common with the consumer flavour of Dirac) and the Audio Unit, VST, VST 3 or AAX-format plug‑in. They are downloaded and installed separately which, in my case, went very smoothly. Dirac Live requires Windows 10, or Mac OS 10.14 or later. The only slight oddity was that when I opened Pro Tools to look for the Dirac Live plug‑in, I found it listed under 'Dynamics'. Dirac Live may be many things, but a compressor it isn't.
On launch, the Dirac Live application searches for a Dirac 'device'. Not being aware of any new device, I found this slightly confusing. However, the application is more normally to be found working with network-enabled hardware in which the Dirac functionality is embedded, so actually what it means in the studio context by a 'device' is the Dirac plug‑in. So the first thing to appreciate when using Dirac Live in a studio context is that it needs the plug‑in host application to be running, with the plug‑in instantiated on an output channel, otherwise it won't find a Dirac device.
Once the Dirac Live application has found the Dirac plug‑in you're up and running. The first stage is microphone selection. My system identified three possible input devices (my USB interface, a webcam mic and the UMIK‑1 mic) and the first move was to select the UMIK‑1 by clicking on its icon and uploading the calibration file I'd previously downloaded from the MiniDSP site on entry of the specific mic serial number. With the input mic selected, clicking the 'Proceed to Volume Calibration' icon opens the next window. Volume Calibration lets you feed pink noise to the monitor channels and adjust the output level and mic input gain so that the input level meter hits the green range and you're not deafened. I found this stage somewhat quirky initially because what I'd failed to appreciate was that not only does Dirac Live need the Dirac plug‑in host application open and the plug‑in instantiated, it also needs the host application in Play mode with an output signal present. Once I'd appreciated that, however, setting levels was a breeze.
Moving on, the Select Arrangement monitoring format selection page enables a choice to be made between a typical studio DAW monitoring setup and a variety of domestic listening arrangements. It'll come as no surprise that I went with the DAW setup. The next step of the process, Measure, is where the fun really starts. For the Dirac Live DAW monitoring format, measurements at nine positions around the monitoring position, including one at the primary monitoring location, are suggested. The accuracy of mic positioning is not hugely vital, say Dirac, because the system is not so much building a 3D map of the room acoustics but using multiple measurement positions to analyse the time-domain characteristics of the monitors and the room — more on this later. Before that, however, clicking on Measure Selected Position kicks things off, after a suitable delay that provides time to get out of the way of the measuring mic.
Powers Of Nine
Despite the fact that Dirac uses a noise signal to set levels, it captures data using a logarithmic sine-wave sweep. It measures the left speaker first followed by the right and then, presumably for a measurement validity check, goes back to the left. The system will work with fewer than the recommended nine measurements, but Dirac say it then produces significantly fewer useful results, and the reason for that is fundamental to the way the system is conceived. So, rather than continue describing the measurement process I'm going to pause at this point and explain a little of the philosophy behind Dirac.
The basic concept of room compensation is not really rocket science. You measure the frequency response of the monitoring system at or around the listening position, and then create an inverse equalisation curve that's applied to the monitor signal to pre-shape the frequency response so that the audio integrated by the ears at the listening position is nominally flat (or of some other desired shape). As demonstrated by the other commercial systems available, it can work pretty well, but it's not a complete solution to either room acoustics or monitor response errors.
Dirac simply made a dense and complicated mix I was working on less of a struggle. Mix elements were just easier to hear, untangle and make judgements about.
Firstly, there are some room acoustic characteristics that it simply isn't feasible to correct. If a dip in the frequency response at a particular location in the room is a result of the direct radiation from the monitors being cancelled by a reflection at the same frequency that happens to be 180 degrees out of phase (ie. delayed by a time equivalent to half the wavelength), no amount of correction gain at that frequency will fill it. For example, there's usually no point in trying to flatten, through applying gain, the response dip that's commonly caused by reflections from the wall just behind the monitors.
Secondly, both the monitor and the room acoustics can produce time-domain errors that vary with frequency, and a correction system that operates only in the frequency domain does nothing to fix them. And this is where Dirac believe their approach is slightly different to that taken by some other systems. Dirac's argument is that unless the time-domain response of the system is corrected (as far as it is possible to do so without either contravening the laws of cause and effect, or imparting unacceptable overall system latency), correction in the frequency domain alone can only achieve so much. This is one of the reasons why multiple measurements are needed (the other being to ensure that the EQ-based correction is not entirely focused on one listening position): the Dirac system looks for time-domain errors that are common to multiple positions and infers that they then must be either inherent to the monitor itself, or to the way the monitors and room acoustics interact over a range of locations. Either way, Dirac will generate its own frequency-variable time-domain correction to put the impulse response right as far as is possible (and that's where Paul Dirac's maths joins the party).
Design Me A Filter
Meanwhile, back on the Dirac measurement trail, the screenshot shows the result of the measurements of my KEF LS50 monitors positioned on wall brackets either side of the DAW screen. The curves displayed are the averages calculated from the nine measurement positions (the big peak on the right speaker at 15Hz is almost certainly noise that I didn't notice at the time — there's a railway line not far away). The measurements show pretty typical in-room speaker performance and tie in well with those I've previously seen generated by Sonarworks, IK Multimedia's ARC, Trinnov and Room EQ Wizard.
The next stage with Dirac's measurement procedure is to click on Proceed To Filter Design, which results in the Filter Design window appearing. For clarity, I've chosen to show just the left monitor channel in the screenshot. The curves displayed are the Dirac measured frequency response, the response target and the frequency response after correction. The target is perhaps the most interesting aspect because it illustrates that rather than simply correcting the monitor and room so that the frequency response is nominally flat for the listener, Dirac enables non-flat, voiced targets to be created by dragging nodes attached to the target curve. But, I can hear you asking, "Surely the whole point of monitor and room correction is to produce a flat frequency response?" Well, yes and no. From Dirac's perspective, the point is to correct as far as possible the major flaws in the time and frequency domains, and when that's done, it's perfectly valid for a user to tweak the overall balance to suit personal preferences. The default Dirac target curve is a gentle 5dB downward slope from LF to HF, but I immediately found that too dull for my tastes, so I modified it to something like the target shown in the screenshot.
The final act of the Dirac Live measurement procedure, once you have tweaked the target to your satisfaction, is to export and name the filter. Once the export is complete it will be available for use in the Dirac plug‑in. There are eight filter tabs available in the plug‑in, so multiple filters can be exported and loaded — either different targets for one monitor, or perhaps targets for multiple monitors measured on different occasions.
Before I leave Dirac screenshots behind, there's one more worth showing and that's the Impulse Response tab on the Filter Design window. This shows the before and after position–corrected and averaged impulse response of the system (the 8ms or so extra latency of Dirac is also very clear). Again, for clarity, I've chosen to display just the left channel. Despite its compressed time scale it's possible to see that the processed step response looks cleaner and tighter but I was intrigued to make a few more measurements to see if I could get a clearer look at what was happening.
I launched Room EQ Wizard (REW) and used its remote measurement features; these allow the sine-wave sweep stimulus signal to be output as a WAV file, played and recorded in an offline application, and the recorded data re-imported into REW for analysis. The offline application in this case was Pro Tools with both the Dirac Live and an alternative room-correction plug‑in inserted on the output bus. I recorded playback of the REW stimulus with no room correction, the alternative room-correction system and Dirac Live in turn. The resulting impulse responses from one speaker at a single central mic position are shown in Diagrams 1, 2 and 3. They show the initial impulse from the monitors followed quickly by the reflection from the rear wall at 1.4ms.
The uncorrected data and that produced by the alternative room-correction system are shown in Diagrams 1 and 2. They're not hugely different in character, demonstrating that the room-correction system is not significantly manipulating things in the time domain. By contrast, Diagram 3 — the Dirac corrected curve — definitely shows a less blurred impulse response. These admittedly ad hoc measurements do appear to confirm that Dirac Live is able to manipulate things in the time-domain performance of monitors and rooms, as well as in the frequency domain.
Appliance Of Science
But what does it sound like? Perhaps just as much as with a subjective response to monitors, your mileage may vary, but the results I achieved with Dirac were very satisfying. I have, in my KEF LS50s, a pair of extremely capable small monitors and I use them in a medium-sized room that I find to be relatively free of major vices, but still Dirac Live brought a significant improvement to my monitoring. One aspect of the improvement was a 'flatter' tonal balance, apparent through a more natural tonality to acoustic instruments and voices, but for me, a more significant aspect of Dirac correction in terms of mix work was an increase in the focus of separate mix elements and clearer audibility of plug‑in processing. Dirac simply made a dense and complicated mix I was working on less of a struggle. Mix elements were just easier to hear, untangle and make judgements about.
So are there any down sides to Dirac? From a mix point of view, I uncovered no negatives beyond its slightly quirky nature and a feeling that it's not as mature and well sorted a package as its direct competition. Having written that, though, one must of course observe the usual health warnings associated with room and monitor correction. Firstly, it's undoubtedly sensible to have a benign listening environment and sensible monitoring setup before resorting to correction, and secondly, all monitors have their limits and it's possible for the EQ applied by correction systems to exceed them. Dirac is no different in these respects.
Outside the mix context, the likely 10ms or so of latency might well be a problem for tracking. However, Dirac hinted to me that a low-latency mode is in the pipeline, so that potential problem may well fade away. Lastly, I was so impressed with Dirac that I was disappointed not to be able to have it running full-time in a stand-alone mode on my workstation. When asked about the possibility of a stand-alone version the answer was a definite 'no' — combined with a gentle steer towards a Dirac-equipped power amplifier (such as the MiniDSP SHD Power — www.minidsp.com/products/streaming-hd-series/shd-power— for example). Can't say I'm not tempted...
IK Multimedia's ARC v3 and Sonarworks Reference 4 are direct competitors to Dirac at around the same kind of price. I'd suggest experimenting with all three options.
Doing The Math
Paul Dirac's work in the 1930s concerned efforts to understand and integrate the newly developed theories of quantum mechanics (the otherworldly, irrational behaviour of particles and waves at the atomic scale) with Einstein's framework of gravity and space-time. It's work that goes on today because still no complete and unambiguous solution has been found. The two theories, while faultlessly successful independently, are fundamentally and intractably at odds with each other.
Dirac, however, conceived of and developed a mathematical description of the quantum behaviour of electrons moving at near the speed of light. The Dirac Equation was among the very first hints that solutions to the intractable incompatibility of quantum mechanics and Einstein's relativity might be found, and it won Dirac his Nobel Prize. Dirac's insights, however, didn't only help describe the esoteric fast-moving electron (and at the same time introduce the concept of anti-matter), it also produced a new mathematical framework for the analysis of impulse signals. It's this, known as the Dirac Impulse (or the Dirac Delta Function), that plays a role in the Dirac room correction algorithms.
- Effective time and frequency domain monitor and room correction.
- Impressive results.
- Slightly quirky setup procedure.
- Inherent latency.
- No stand-alone version.
I've felt slightly uneasy about room and monitor correction in the past, but Dirac is a technically impressive achievement that appears to bring something genuinely unusual and worthwhile to the concept.