Audyssey XT32 vs. Dirac Live Listening Comparison

Audyssey XT32 vs. Dirac Live Listening Comparison


Introduction

Last year HTS published a review of a the miniDSP DDRC-22D, a two-channel Dirac Live Digital Room Correction (DRC) product. The review included a comparison to Audyssey XT. A number of readers requested a comparison of Dirac Live with Audyssey XT32. That comparison was recently completed during the Home Theater Shack High-End Amplifier Evaluation Event at Sonnie Parker's Cedar Creek Cinema in rural Alabama. This report provides the results of that comparison.


Test Approach, Equipment, and Environment; Abbreviations



Difficult Decisions: What To Compare and How

The most important decisions about making a product or technology comparison are determining just what you are actually comparing and exactly how to make a fair apples-to-apples comparison. It is not always as easy as you might think it should be. I am convinced that most casual comparisons made between products are fraught with unaccounted-for variables, and are therefore flawed.

These were not easy decisions for the Dirac Live / XT32 comparison. In the end, we opted to complete a comparison that was very limited in scope but gave clear and meaningful answers as opposed to one that gave broad but vague results. Some readers will be disappointed in this choice, but in retrospect I stand by it as the right one, for the reasons which follow.

-----------------------

When I was writing the HTS Audyssey MultEQ FAQ and Setup Guide two years ago, it quickly became apparent that, as a somewhat particular two-channel listener, the results achieved by following the conventional guidelines for setup gave consistently unusable results for music and for movies. The soundstage was vague and image clarity was soft to nonexistent. Frequency response was not that great either.

With a little experimentation it became clear that the technology was very capable of creating great sound stage and imaging, but the emphasis was entirely upon trying to improve frequency response (FR) over a broad area with little regard for Soundstage and Imaging (SS&I).

The only logical conclusion I could draw from this was that most listeners had never had a really great SS&I experience, and therefore did not recognize how immersive and completely engaging it could be. Since then I've become more and more convinced that this is true, even among serious and experienced listeners.

-----------------------

Why put more emphasis on SS&I versus FR, and why is this not done already?

Let's start with the second question.

• Good quality SS&I are not as easy to achieve as good FR. It may be considered unachievable to the novice, and therefore automatic low priority.
  
• SS&I are not directly measurable. There are those who will argue that because of this fact they do not exist, and therefore are not worth pursuing. It is clearly describable, however, and there is a common response for all who listen to the same SS&I setup.
  
• But the biggest factor, I believe, is that most listeners have not experienced it. While not universally true, most audio lovers with whom I have talked who put enough priority on good sound to even bother with good FR are completely blown away when they hear great SS&I, and quickly prioritize it above other audio qualities. Sonny Parker spent a year as a SS&I missionary of sorts while working for a prominent speaker manufacturer as a customer service representative. Almost without exception, once he convinced a customer to "just try it" - take a few steps to prioritize SS&I in their speaker setups - he and customer service and company leadership received glowing letters and semails of thanks, that they were getting the best sound they had ever heard.

Now back to the other question. Why put more emphasis on it on SS&I?

• In finer terms, human hearing is relatively insensitive to FR. We do not have a good internal reference system for flat FR. Our hearing response curve varies with sound pressure level, and is constantly adapting by accepting the status quo profile as normal and tuning it out, staying sensitive to change. This is an instinctive response. Tune out the sound of the waving grass so you might hear the footsteps of the stalking tiger. FR IS probably the most obvious quality to be noticed in a speaker or system, and often feeds our first impressions strongly, but it is far from being the defining "ultimate" quality.
  
  You will never hear me argue against good FR. Horrible FR is very easy to perceive, and to be annoyed by. But we are talking about refinements here, not gross qualities. Our ability to gauge flat response without help is rather poor, and our tolerance for variations is fairly high, and even when it can be achieved, personal preference reigns supreme. Some like extra bass, some like less treble. Perhaps it is not the best listening quality to put at the top of the priority list.
  
• Beyond that, SS&I engage us with three sensory systems instead of one: the auditory, the visual, and the kinesthetic. With SS&I, we can see where an instrument or vocalist sits, can gauge how far away and how large the source appears to be. Is that voice the size of a basketball? a golf ball? A pea? How big is the room the drums are being played in? What is that voice doing way over there? All three of the sensory systems with which we navigate our life experience are engaged in the interpretation and enjoyment of a music or movie experience. Audio becomes more than the series of sounds and can even convince the psycho-acoustical brain that instruments and people are right there in the room with us, challenging what we know to be real. As one HTS member put it, "It messes with your head." Flat FR, while nice, does not have that effect.
  
• Here is where some will argue that great SS&I are qualities that occur at only one spot in the room, so why put such emphasis on qualities that only one person in the room can appreciate at a time? The answer? Cases where there are more than one person in a listening room or home theater at a time who really care about the audio quality are about as rare as the muon neutrino. And when that does occur, they are not interested in pretty-good sound across two or three seats, they are far more excited about taking turns in the one seat where the sound is truly phenomenal.
  
  It is a simple acoustical fact that averaging FR over an area, as advanced DRC products are designed to do, can only make FR at any one position worse. (To the reader who feels compelled to make a fuss about Schroder's work, hold on, we'll get to that.) The one person who cares gets mediocre sound quality so the other listeners, who probably care very little, can have mediocre sound as well. It is a totally fair auditory compromise and is pretty much a waste of time and effort.
  
• The final reason for emphasizing SS&I is that the DRC setup process is actually easier, and can in fact be extremely easy. And, as it turns out - after all that discussion minimizing the importance of FR - Dirac Live and Audyssey MultEQ and other DRC technologies can help us get great SS&I AND flat FR at the same time.

The Setup Mic Pattern

The work of Schroeder and Toole and others can be used to argue that above a certain frequency, EQ is pointless because the speaker's response predominates in the sound achieved. The key word is predominates, and FR variations of 6 dB and more are easily possible at closely-spaced points. THAT is enough to matter. So, within a certain area, there tends to be a "room sound" which can be corrected by DRC to enhance the listening experience.

The intent of the averaging process employed by advanced DRC products like Audyssey XT32 and Dirac Live is to determine that "room sound" over a certain area and correct the sound for all who sit in that area. Using widely-spread mic calibration mic patterns, these products try to make as much of the room sound good as possible. It is a nice idea.

But the result, regardless of your views on room or area FR characteristics and averaging techniques, is a compromise. Even with fuzzy logic throwing out or minimizing the importance of the worst data points (seating positions), the result will still be a compromise, making all positions sound equally mediocre and allowing none to sound really good.

All of this effort to please listeners who simply do not care. A nice democratic idea, but it kinda makes you wonder who got us going in this direction, and how differently these products might have been structured and marketed if the right people had only been exposed to a transcendent soundstage with precise image clarity in their formative years. I jest, of course. These are superbly engineered products by innovative, industry-leading companies. We simply see their best use from a different perspective, all due respect.

It is our good fortune that using these products we can attain both great SS&I AND great FR together. Which brings us, via the long route, to the listener's primary means of influencing the behavior of a DRC product like Audyssey XT32 or Dirac Live, and our decided emphasis for this product comparison: the setup mic pattern.

I wrote The Audyssey setup guide recommending setup mic patterns that emphasized SS&I at the main LP and gave great FR there at the same time. There were those who argued and those who tried the recommendations. Those who tried the recommended tightly-spaced mic setup patterns said that it sounded better. There was, however, one mic pattern that I did not include in the published guide, because I knew the amount of arguing against it would be immense. That is the single-point setup.

Here is why it works, and why it was ideal for our comparison:

• As discussed, the main LP is the only one that matters 99.9% of the time.
• At the main LP, the chair predominates in FR influences. Typically, a reflection and cancellation between the chair back and the ears causes a 10 to 15 dB FR dip above 1 kHz. This is not seen anywhere else in the room.
• The listener's ears sit equidistant from the chair back, so any measurement point along the ear-to-ear line close to the center of the chair will measure the influence of the chair. Centered, the mic will measure speaker-to-ear timings in a way that reinforces SS&I performance and gives good FR performance at the same time.
• And it fills a vital need for this exercise that was mentioned earlier: elimination of variables. With multiple mic points, how many uncontrolled variables are added? What pattern do you use? A favorite? A recommended one? Symmetrical or random? Use the same pattern for both products? How precisely can it be repeated? How do you accomplish that precision? How many times do you repeat the process to be able to trust that your result is repeatable. And how do you measure the result effectively? So many questions. With one mic point we eliminated those questions.

The Setup Process

Attaining the best soundstage and imaging with good FR is basically a two-step process:

• Position the speakers for best SS&I without regard for FR (details beyond the scope of this article). The best SS&I at this point usually involves some off-axis listening angle, so high-frequency response will usually be down somewhat. But it will usually match fairly well between L and R channels (or else the Image Clarity would be poor), and will be improved by the second step.
• Use DRC to equalize the system, correcting the FR and improving the SS&I further.

The Oppo HDMI output was run to the nanoAVR DL, which was run to the AVR's HDMI input. This series setup was used throughout testing, and either XT32 or Dirac Live or neither could be active at any one time. As a side note, at one point early on, Dirac Live was accidentally set up with XT32 active and some preliminary listening was done with both processors active. The resulting soundstage seemed very unnatural and fragmented. There may have been other variables involved, too, so this piece of data is offered as information with limited value.

It took us awhile to get everything working properly and to reach the point where a valid comparison was possible. Some of that detail is included to show how difficult it can be to accomplish such a comparison as we hoped to.

The first difficulties were with interfacing with the Denon AVR, choosing a configuration without disturbing that which Sonnie already had in use.

The ripple-upstream method was used to get HDMI configuration data to all the right places in the system, starting with the AVR, then attaching it to the nanoAVR DL and powering it on, then attaching that to the laptop used for Dirac Live configuration. It took a number of attempts to get all this to configure property, including windows 7 device settings. Then an accidental configuration change in the AVR might ripple back through the system and disrupt the configuration, not resetting to normal as as it should with the setting corrected, so the whole setup process would have to be repeated. This is not the fault of any individual product, just the way things go with digital audio sometimes.

We carefully determined the proper mic position with Sonnie sitting in the chair and then centered it after he stood up, further verifying equal distance from the two side walls with the laser distance meter. The Denon Audyssey setup mic was positioned pointed straight up and the UMIK-1 for nanoAVR DL setup was positioned pointed straight down, tip-to-tip. One would be swiveled out of the way, then moved back into place before the other was swung away, so the two setup processes were completed and measurements were taken with Room EQ Wizard without losing the measurement reference point. Once we accidentally did lose that reference point before we meant to, and we started the process over. XT32 setup was done first. then Dirac Live setup. This all happened numerous times (we lost track of the number of attempts) before we got everything running properly.

For both Dirac Live and XT32, we ran four repetitions of the measurement routine without moving the calibration mic. This was intended to give us better noise immunity. With Sonnie's system, we ran the calibration in 5.2 mode for XT32 and 5.1 mode for Dirac Live. Our listening tests involved only 2-channel material, and the system was set to Stereo with subwoofers active.

Twice we did the Dirac Live setup with Audyssey activated thinking we had turned it off. Several times we lost the AVR configuration completely and had to start from scratch. This happened twice just as evaluators were preparing to sit down and listen to the comparison. Once the laptop display froze up and it had to be re-booted, along with everything else. There were numerous reboots, restarts, resets, and a couple of exasperation breaks. Once we got completely through the configurations and it was not until we sat down to listen and compare that we realized we had forgotten to check AVR delay settings determined by the XT32 process, and there was a 1/2 foot difference between settings for the L and R front speakers. Needless to say SS&I were abysmal. we went through the whole process again.

This is all reported to remind the potential user of the many ways that such an effort can be thrown off track, and the many ways that a user evaluation can have something go awry in the setup process, erroneously leading them to conclude that a product is unacceptable.

It is worth noting that miniDSP's instructions for achieving all of this are incredibly detailed. Our troubles were largely the nature of the often unruly digital audio system beast, of working under pressure when tired, of some quirky bad fortune, all of which I am already laughing about. With experience, much of this would go more smoothly. That said, that beast sometimes just likes to bite! Note also that the final listening was done during afternoon hours after a break and a good nap when ears and brain were fresh.

At this point I have to confess that all the other evaluators were gone before we had a working comparison, and it was only Sonnie and me completing this work. I had really hoped to have those other fine sets of ears to help out with this comparison. But I am still confident in our results.


Fine Tuning and Results

We were set up to switch quickly between Dirac Live and XT32. XT32 could be activated or deactivated with about a two-second silence gap as the AVR switched modes from a single-button-push on the automated system remote. Dirac Live was turned on and off using a universal remote programmed to control the nanoAVR DL.

Here are the uncorrected and corrected FR measurements at the LP, as measured by Dirac Live and by Room EQ Wizard, respectively.

Measured FR at LP, no correction.


Measured FR at LP, Dirac Live with Default Target Curve, XT32 Reference, and XT32 Flat.


It was immediately apparent to my ear that there were significant differences between the corrections resulting from the XT32 and Dirac Live target curves. XT32 produces corrections using two target curves, named Reference and Flat. Dirac Live has a default target curve, which is infinitely customizable and can be saved and recalled at will. XT32 has a similar capability with the Audyssey Pro Kit, with AVRs that are Pro Kit capable. The Pro Kit was not available for this test. It was easy to identify each of the three resulting corrections and the no-correction state. The XT32 Flat and Dirac Live corrections were fairly similar in the high frequencies and a little harder to differentiate. Bass response for the Dirac Live correction was quite different from that of the two XT32 corrections. As seen below, the Dirac Live FR contour involved a steady rise from HF to LF, and both XT32 corrections had a small step-down in the LF range. Sonnie had dealt with numerous Audyssey users who were disappointed with initial Audyssey LF response, but there is no reason why subwoofers cannot be turned up to give hotter LF response if desired after calibration is complete.

This target curve difference might have some users picking Dirac Live over XT32 because of its higher default bass level without realizing that subwoofer bass levels can be increased to compensate. Modifying target curves is much more convenient and flexible. Again, the Audyssey Pro Kit makes this possible for XT32 with some AVRs.

We decided to modify the Dirac Live target curve to match the XT32 Flat target curve. It only took a few mouse clicks to make the Dirac Live curve mimic the LF response of the XT32 curves. Likewise, it only took a couple of clicks to roll off Dirac Live to match the top end of the XT32 Flat curve. Then the Dirac Live correction was reprocessed and loaded into one of the four nanoAVR DL program slots, all in less than a minute.

Dirac Live Corrected, Default Target Curve


Dirac Live Corrected, Subwoofer


Dirac Live Corrected, Target Curve Modified to match XT32 Flat


Finally, we had achieved an apples-to-apples comparison base. I listened through a number of familiar test tracks as Sonnie switch at will between the Dirac Live and XT32 Flat, every ten or fifteen seconds or so. Sometimes I would repeat parts of tracks, sometimes I would ask him to switch at certain points while I was listening for certain differences. The tracks included full-range rock, an orchestral piece (primarily strings), a classical piano piece, a bluegrass instrumental, and a male/female vocal track with acoustic guitar and piano backing.

The first contrast I listened for was any FR difference. Although the two target curves did not end up precise matches, they were close enough that I could not pick out any shifts in frequency content between Dirac Live and XT32 Flat. Perhaps with a track specifically chosen with content in the right frequency range, the difference might be discernible. But with the tracks chosen, I was not able to hear a FR difference.

I listened also for differences in Soundstage and Image Clarity. This is where I had expected to be able to hear a difference if there was any. I was listening for any shifts or differences in image placement or size or clarity, for any apparent differences in image definition or stability, for changes in instrument definition and detail, for changes in the overall cohesiveness and clarity of the soundstage, and for contrasts in soundstage characteristics relating to how natural and easy they were to engage with, to accept as reality.

Here is the bottom line. I could hear no differences whatsoever. Extended listening sessions might have exposed some difference, but I believe it would have been extremely subtle, if any. The resulting performance from each of the products was completely engaging, completely natural sounding, gave us pinpoint imaging and a huge, deep soundstage with decent depth acuity, and very good clarity and detail.

Sonny Parker spent time in the evaluation seat as well, and reported the same findings, that he could hear no difference.

Other points to consider:

• HDMI audio was passing through the nanoAVR DL during all testing. With Dirac Live inactive, miniDSP informs us that only the delay blocks would be active, that all of the processing would be bypassed and effectively straight-wire. I do not know at this point if resampling takes place with Dirac Live inactive, but I believe it would. I will report on this in the upcoming nanoAVR review.
• The Nano AVR gain was set at -12 dB throughout testing. Gain matching as we switched between XT32 and Dirac Live was important, so this value had to remain constant. Sonnie complained about the signal loss, that he would not be able to achieve the maximum volume that he sometimes likes to use. The -12 dB gain setting was chosen arbitrarily. Depending on the equalization required, this value might be set higher. With Dirac Live inactive, it is my understanding that it can safely be 0 dB. I will verify this for the upcoming nanoAVR DL review.
• XT32 gets points for ease of use, being built into the AVR.
• The Dirac Live configuration program gets points for flexibility and for ease of use, too, once everything is communicating properly. It is very intuitive. Individual measurements can be retaken at will, and the target curve can be modified at will. An infinite number of correction programs can be saved and recalled and reloaded to the nanoAVR DL unit, which holds up to 4 correction programs at a time (a clear advantage over XT32 for some). The Audyssey Pro Kit gives much of the same capability, but only with Pro Kit capable AVRs. Loading a configuration into the nanoAVR DL takes around 30 seconds, much faster than loading an alternate configuration to an AVR from the Audyssey Pro configuration utility, according to Sonnie.

Subwoofers and Bass Correction

We did not have time to focus on subwoofers and bass correction. This is an area that could take days of work on its own. We realize it is of great interest to many home theater owners.

XT32 has separate outputs for two subwoofers and corrects their delays and levels independently, then applies MultEQ correction to them together. Dirac Live has a single subwoofer control channel.

The three-curve corrected response diagram above shows us the Dirac Live FR was very flat below 100 Hz with a small amount of variation between 50 and 80 Hz. Both XT32 curves have a sizable dip between 70 and 80 Hz.

In later work Sonny was able to tame the dip with XT32 by changing delay times independently for the two subwoofers in the front left and right corners of his room. He also increased subwoofer levels and ended up with a slightly increasing bass response below 100 Hz. This was all done manually. It has been his experience that some manual variations are usually needed to get the bass response that he likes with XT32.

The Direct Live bass response, on the other hand, would have been totally acceptable to him "out of the box."

This is a small amount of data as it pertains to one room, and I am not suggesting it should be projected to form a universal statement about XT32 vs Dirac Live as it pertains to bass correction. We simply offer the data available for your consideration.


Impulse Response

Dirac Research emphasizes that their product performs impulse response correction. Audyssey XT32 certainly does some degree of phase correction to achieve the SS&I results which it does, but "impulse response correction" per se is not mentioned, for what that is worth. Looking at impulse response measurements taken with Room EQ Wizard, it is clear that Dirac Live cleans up impulse response far more effectively than XT32. What this means to the listener is not immediately obvious. I can imagine arguments in either direction, that Dirac Live gives a more perfect correction, or that XT32 does all that it needs to and nothing more. I have done some research which leads me to believe that a cleaner impulse response can yield greater clarity with dense program material at high volumes, but it is an unproven theory with little data at this point. In reality, there is no conclusive answer to this question that I am aware of.

Impulse Response Graphs



Conclusions

While I hoped to get additional expert ears involved in this comparison, I am still quite confident in our results. Given the conditions outlined above, with a two-channel listening focus, we ultimately achieved what we believe was a true apples-to-apples comparison between Dirac Live and Audyssey XT32 and concluded at both did an excellent job and that under the chosen conditions any differences in performance were not audible. Usability differences are many, and are likely to be driving factors in a product choice.

Thanks to miniDSP for supplying the nanoAVR DL hardware and software for this exercise, and thanks to the executives at Audyssey, miniDSP, and Dirac Research for their suggestions and help.




EDIT:

A complete review of the miniDSP nanoAVR DL has been posted HERE. Where appropriate, comparisons are made to Audyssey XT32, focusing on the end-user experience.

A post has been added to this thread titled Audyssey XT32 (without Pro Kit) vs Dirac Live End User Experience Comparison Summary of Audible Characteristics. Read it HERE.




EDIT:

A post has been added to this thread titled Further Conclusions Regarding the Comparison of Audible Results Between Audyssey XT32 and Dirac Live. Read it HERE.