Dirac Room Correction Interview With Mathias Johansson
Like Audyssey MultEQ, Yamaha's YPAO, and Pioneer's MCACC, Dirac Live is a room correction system designed to counter the ill effects of placing loudspeakers in a real room. While we know Dirac's products are used on high end products like the Emotiva XMC-1 and Theta Digital Casablanca, we can't really claim to know all the fine details of how Dirac Live actually works. Fortunately, we know somebody that does: Mathias Johansson, CEO of Dirac. Of course, Mathias isn't just a business man; he's got a Ph.D. in Signal Processing. If you want to know the full scoop on Dirac Live, keep reading!
Audioholics: What kind of filters do your room correction products use, and at what resolution (i.e. 1/3 octave, 1/12 octave, etc.)?
Mathias Johansson: Dirac Live uses a proprietary structure that is neither plain FIR nor plain IIR. The downsides of using parametric filters (IIR) include that you are limited in resolution by the number of filters employed and they have inherent limitations in optimizing phase/impulse response properties of the loudspeaker. If you don’t have sufficient word lengths, you also get problems in maintaining the resolution in the bass region. FIR filters, on the other hand, are also limited but in slightly different ways: First, with a finite number of filter taps (FIR = Finite Impulse Response) you lose resolution in the low-frequency region. The high-frequency resolution may instead even be too high, unless you smooth your response by other means. Too high resolution seems like an oxymoron, but then don’t forget that a filter acts both in the frequency domain and in the time domain. If you do very detailed corrections in the high frequency region, you also introduce ringing in the time domain. That may not always be easily audible, but a typical artifact of too detailed high-frequency compensation is a more diffuse sound when you are using stereo or multi-channel systems. The reason is that the time-domain ringing on one channel is different from another channel, and then the sound becomes de-correlated. Second, they cost more to implement than IIR filters, that is, they require more CPU power. On the up side, an FIR part is required in order to do correction of the phase/impulse response properties of an acoustic system. FIRs are also less cumbersome to implement than IIRs, as they are non-recursive and therefore less susceptible to round-off errors, etc.
Dirac’s solution is to use the best of the FIR filter (mixed-phase correction, i.e. time-domain correction) and IIR’s for what they are best at. Dirac Live does not simply use an FIR in cascade with a number of IIR sections, but a more complicated structure. The result is that our filter resolution is not limited by the filter structure, but by the measured properties of the acoustic environment. In essence, by using multiple measurements you can find the optimum resolution that does not lead to overcompensation. In a small listening area, you can do more detailed correction than in a large listening area.
Audioholics: What is the maximum boost / cut your room correction products will apply? Is it possible for users to set a limit in this respect (i.e. no more than 3dB of boost)?
Mathias Johansson: Normally, Dirac Live does not allow you to boost more than 10 dB. You can lower this by adjusting the target curve. However, narrow dips (nulls) are never compensated, as they are always position dependent in real acoustic spaces.
Audioholics: How do your room correction products help to address the typical issues a room causes (resonances, modal peaks/nulls, etc.)? Does room correction largely negate the need for room treatments for consumers, or would you suggest using both room correction and physical treatments?
Mathias Johansson: No room correction system negates the need for room treatments. What makes Dirac Live different from other systems that we have tested is how we make use of many measurement points to be able to do the best optimization of time-domain and frequency-domain properties. It is important to understand that only by using multiple measurements is it possible to understand the large differences in the response in the room, and thus to make the best optimization and compromise between different measurements. For example, making a time-domain optimization based on a single measurement in a reverberant room is a very bad idea as the impulse response varies so much from one position to another. With a single measurement you will always be guessing what happens in other positions. Room acoustics is cruel. Just inches away, the correlation drops so much you can’t infer what happens just based on one measurement. Only by using several measurements is it possible to get enough information to make a good correction which does not imply pre-ringing or de-correlation just inches away (and don’t forget that we have two ears. Just inches apart…) See the images below to understand some of the difficulties faced in making a good tuning of a speaker to a room.
This picture below shows 64 different measurement points in a room (one loudspeaker only). You might think that you don’t care about the results in that many points. Correct, but then you should be aware that these measurements were taken in an area spanning a very small volume of only 0.3 by 0.3 by 0.3 meter in the normal listening position of a chair! If you would just pick one of these measurement positions, you would be very lucky to get anything near the average (bold) line below. And if you are not that lucky, your “correction” will only improve some aspects, and actually deteriorate others.
Here is another image. It shows impulse responses from 9 different measurement positions within the sweet spot of a very well damped room using Genelec monitors. The impulse responses have been time-aligned here to make them easier to compare. For reference, we show how the same impulse responses look after application of a Dirac Live filter and another filter with identical magnitude response but no phase correction (a so called minimum phase filter)
Our approach gives an optimal compromise between multiple measurement positions, both in the frequency domain and in the time domain. Out of the room correction systems that we have been able to test so far, no other system yields a time-domain correction which is useful for many positions (for example, two ears). From the multiple measurements, we infer which time-domain and frequency-domain properties are common to all measurements, and therefore can be robustly compensated. In mathematical terms, we can derive the optimum filter by evaluating how the poles and zeros of different measurement positions are clustered. It is a cautious strategy in line with the maxim that we should optimize as much as we can, but not more.
Most other room correction systems are minimum-phase, which means that they try to not add any time-domain distortion, whereas Dirac Live actually reduces time-domain distortion. The results are audible in terms of imaging, clarity and bass tightness.
Digital room correction cannot change the response variations across different listening positions. This is where room treatments help tremendously, as they eliminate the reflections, etc, that cause acoustic variations across the room. However, Dirac recently showed a preview of an altogether different kind of digital room correction, which we prefer to call active room treatment. This approach, which we have named Dirac Unison, aims not just to combat average room behavior, but also changes the whole wave pattern in the room. This is accomplished by using multiple speakers. For example, if you are listening to stereo material on a 5.1 speaker set-up, Dirac Unison can use the remaining speakers to actively treat the room, rather than just do a virtual surround up-mix or simply shut them off. A speaker may be a primary source or a support source. The support sources are used to improve the impulse responses and frequency responses of the primary sources in all measurement positions. This way, we can control the wave pattern in the room, and speakers in corners start acting as bass traps, etc. We showcased this system at CES 2014, but we have not yet launched it commercially within HiFi. The first commercial system with Dirac Unison will actually be the brand new Volvo XC90 with a B&W loudspeaker set-up. In premium car audio systems, you may have up to 20 speakers and beyond. Without Dirac Unison as a ”conductor” that makes them play in perfect synch, it is very difficult to make that many speakers act coherently as a team.
Audioholics: Do your room correction products utilize multi-point measurements? Why or why not? If so, how many measurement points are available?
Mathias Johansson: Please see above. We recommend 9 measurement positions spread out in the listening region as randomly as possible, to capture all relevant sound field variations. In large listening venues such as cinemas, we of course use more measurements.
Audioholics: Do your room correction products correct subwoofer response? How do your products calibrate and correct the response of multiple subwoofers? Please describe the benefits of your method.
Mathias Johansson: Yes, we correct subwoofers. Regardless of the number of subwoofers, one of the main benefits of Dirac Live is the time-domain optimization. A lot of room acoustic behavior is non-minimum-phase and only a proper mixed-phase correction using multiple measurements can maximize the tightness of the bass.
It should be mentioned that the coming Dirac Unison technology will improve first and foremost the bass interaction of multiple-woofer systems through its unique joint time-domain optimization of all speakers. The first launch for the HiFi segment is not yet scheduled firmly, but will likely happen in 2015.
Audioholics: What is the default frequency range corrections are applied to? In other words, is there a frequency ceiling or floor above/below which correction isn't applied? If correction isn't applied full band, please explain.
Mathias Johansson: Dirac Live RCS uses a full-band correction as a default. We however offer the user a low-frequency limit and a high-frequency limit if the user wants to compensate only a certain frequency region. We have just started offering a ”bass-only” version of Dirac Live, which applies correction below 500 Hz only. But you can still adjust the target curve freely within that range.
Audioholics: Does your room correction software show users "before", "target", and "after" response curves? Is it possible for end users to adjust the final response curve such that they can flavor the sound to taste?
Mathias Johansson: Yes. We also suggest a target curve, and the user can change the response by simply editing the curve.
To continue our investigation of the various room correction products on the market, we got in touch with Mathias Johansson of Dirac who was kind enough to answer a few questions. For those unfamiliar with the company, their room correction solution is used commercially (i.e. in cinemas, recording studios, etc.) as well as in home audio products like the Theta Digital Casablanca and Emotiva XMC-1. Want to know the details of how Dirac Live works its magic? Keep reading!
Audioholics: What do you feel are the important differentiators between your room correction solution and competitors?
1. Correctly accounting for the spatial domain, that is, variations between the ears and between listening positions.
2. Working time-domain optimization that improves imaging, clarity and bass tightness
3. Compatibility when listening from a computer. Our ”virtual soundcard” processes all sound output from your computer, not just certain music software.
4. User friendliness. Advanced technology does not have to be difficult to use.
Audioholics: Does your room correction software run on an AVR/Pre-Pro or require an external computer for processing? If it requires an extra computer, what platform(s) does it run on? Can consumers purchase your solution as a stand-alone product/solution or only bundled as part of an AVR/pre-pro?
Mathias Johansson: Dirac Live Calibration Tool, which is used for taking measurements and creating the filters, run on Mac and Windows. To process sound, the Dirac Audio Processor, processes all sound output from the computer and acts just like a normal sound card. Sonic Studio also offers a version of Amarra that has built-in Dirac Live support up to 384 kHz.
There are a number of hardware units that use Dirac Live. These include Datasat RS20i, AP20 (for professional cinema), Theta Digital Casablanca, Emotiva XMC1, and the MiniDSP Dirac series. Users include broadcasting stations such as Swedish Radio, cinemas across the world, music producers, studio engineers, artists, as well as a wide variety of HiFi and home theater enthusiasts. You don’t have to be a professor to understand how to use Dirac Live or enjoy the benefits of a time-domain correction.
Dirac is available as a software solution for your computer in addition to being integrated on a variety of hardware units.
Audioholics: Do users have the ability to set multiple profiles--for example, setting a separate "music" and "movie" room correction profile where one can be tailored for a different frequency response range, bass response, different crossovers, etc?
Mathias Johansson: Yes. The number of presets depends on the particular hardware implementation. For Dirac Live RCS you can create as many presets as you like.
We'd like to thank to Dr. Johansson for sharing his insights on Dirac Live with us.
Confused about what AV Gear to buy or how to set it up? Join our Exclusive Audioholics E-Book Membership Program!