NOTE: Not for Dirac Live after all...
The initial measurements taken when this report was first posted seemed to support the use of a single mic measurement position for getting good results with Audyssey and with Dirac Live. More recent measurements and experiences have shown that this is not the case for most situations. With Dirac Live, it is always better to use multiple mic positions and is very easy because precision placement and careful physical measurements are not necessary. See this review for more details.
With Audyssey, a single microphone position calibration can be effective with low-backed chairs or sofas, but more recent measurements have shown that with high-backed furniture that is not the case. However, there is an approach with high-backed furniture which requires minimal effort and no precise physical spacing of the measurements. This is covered in Part 2 of this report.
Part 1 is left intact for the background info it contains. See Part 2 for the main takeaway info and conclusions.
Part 1
There has been some recent discussion about the use of a single mic position for the setup process for Audyssey or Dirac Live, or any Digital Room Correction (DRC) product. The reasoning for the single mic position is that, if properly chosen, it gives the DRC technology the best information possible for determining the filtering to apply through the auto-correction process, thereby giving the best sound possible where it matters most - at the listener's ears. The tradeoff is improved soundstage and imaging (SS&I) but with restricted sweet spot.
This recent HTS article explains in detail the main reasons justifying the single-point setup mic technique. The data below supports that article, particularly in how it relates to the Listening Position (LP) having a high seat back, by showing:
- There is a lot of variation in frequency response (FR) depending on closeness of the ear to the seat back. Therefore a single setup mic position will give best DRC results when located at the same distance from the seat back as the ear would normally be located.
- There is very little variation along the ear-to-ear line, even that close to the seat back.
The following graphs show the effect of chair back reflections when making measurements close to a high-backed chair. They have the vertical scale (dB) zoomed in more than we usually recommend for readability.
The first shows the effect of putting a blanket over the back of the chair to reduce reflections. In this case, the blanket was an inexpensive plush throw folded into quarters and draped over the back of the chair so that it extended down to below shoulder level. The first graph shows the difference between with and without the blanket with a mic 6 inches away from the chair back at Listening Position Center (LPC), the center of the listener's ear-to-ear line. Minimal smoothing was used to show the comb effect filtering at higher frequencies without the blanket in place. The dip at 1 kHz is the main mid-frequency effect of the reflection from the chair back, and it is only changed a little by the blanket because the wavelength is over 1 foot at those frequencies. (The 600 Hz dip is from the seat and would not exist with the listener in place.) At higher frequencies we see a lot of comb filtering effect from the reflection, which has a negative effect on image clarity. Adding the blanket improves that quite a bit, although not completely eliminating it. The difference is clearly audible.
Here is a third-octave-smoothed view of the same curves, and one can still see a significant difference between with and without the blanket.
This plot shows the change in mid-frequency response for distances of (from bottom to top) 4, 6, 8, 10, and 12 inches out from the chair back. I normally assume the 6 inch spacing. From ear line to back of head is 4 inches, and I usually sit with my head forward from the chair back a couple of inches while listening. I can hear the difference between that position and leaning my head all the way back, plus it is just more comfortable for me to have my head forward a little, more erect, most of the time. Even ignoring the more dramatic 4-inch curve, from 6 inches out to 12 inches there is significant change in mid frequency response.
Here is another view of the same data, with both left and right speaker curves shown, offset for easier readability.
As to the "We have two ears" argument which is sometimes used against the single mic point setup method, the plots below show the difference between LPC, left ear, and right ear positions for left speaker and for right speaker. The FR difference is very small whether the mic is at to left ear, right ear, or LPC position. The LPC mic location represents the ears very accurately.
The final plot shows, for each speaker, the average of the three "ear-line" curves versus the LPC curve. There is almost zero benefit in averaging the three curves vs just using the LPC curve to represent the ears.
Part 2
Except where noted, the following discussion is for Audyssey calibration. The tests below were done with Audyssey XT and should apply to all versions of Audyssey.
The measurement that made it clear that the single mic position calibration was not a good idea is in the first diagram below. The upper red trace shows the frequency response (FR) with the microphone 6 inches away from the chair back, and the cancellation dip we thought we wanted to have Audyssey compensate for. The measurement in green shows what happens when one sits down with the mic placed at the ear in the seated position. This is the measurement that changes it all.
The explanation is fairly obvious in retrospect. The side of the human head is hard enough at those frequencies that it becomes a boundary, much like a wall in the room. We know that at a solid wall or any solid boundary there is no air movement, only pressure variation. So the amplitude we normally associate with air movement does not exist there and the FR is very flat. Noticed that in the green plot the FR is much flatter at mid frequencies.
At high frequencies, reflections from the head and structures of the ear now start to cause new disturbances. The blue plot is a second measurement taken while seated with the mic by the ear supposedly in the same position, but see how much HF variation there is between the green and blue plots. Measurement repeatability is very poor.
The lower orange plot shows the Moving Mic Method (MMM), using pink noise as the measurement signal. The measurement mic is moved slowly through the measurement area with the RTA in infinite averaging mode. This plot shows the average FR in the area around the head without equalization.
How about selecting a mic measurement point above the chair back? The measurement taken straight up from Listening Position Center (LPC), the center point between the ears of the listener when ideally positioned, is shown next. This measurement position is 1 foot away from the closest point on the chair back, and shows much flatter FR above the bass frequencies. It is not the answer, though, as we will see shortly.
Several Audyssey calibrations were executed to try out various combinations of measurement positions.
First, the single LPC measurement position yielded the following MMM FR plot. As a reference, it involved three calibration steps all at LPC without moving the mic. FR and imaging were fairly good, but the soundstage was poor, very unnatural and disjointed. I found myself moving my head forward and back to try to get it to settle in and it never would.
As we continue, we will use a new term, the Center Plane. It is the plane that divides the left and right sides of the listener's head, stretching up and down and backward and forward through the room so that all points within it are equidistant from the main Left and Right speakers.
For the
second calibration, giving us the next MMM plot, the first measurement position was straight above LPC, one foot away from the nearest point of the chair back. All of the other measurements were in the center plane, at least one foot away from the chair back, all at or above the ear position, all different distances from the chair back, and all within 1.5 feet of LPC. The resulting imaging and soundstage were good, but FR was hollow in the midbass and uneven through the midrange.
Realizing that it was important to include LPC as the first point in any calibration, even with its imperfections, the
third calibration started at LPC and after that was similar to the second calibration, but the FR (no MMM was done) ended up very light in the bass frequencies, and still rough through the midrange. Imaging and soundstage again were quite good.
At this point it was clear that the strategy of focusing all the measurements positions in the center plane was good for imaging and soundstage, but finding the right combination of points to give good FR was not quite so simple.
The fourth and final calibration finally got the right combination of factors involved. All measurement positions were in the center plane, the first was at LPC. Now imagine a right triangle with the right angle at LPC, one side pointing straight up, and the other side straight forward through the bulb of the nose. The second measurement position was straight above LPC even with the top of the back of the chair. The final measurement was taken straight forward from LPC, 12 inches away. The line from the 2nd to the final measurement positions is the hypotenuse of the triangle, and the remaining measurement positions were all along that line, equally spaced.
The convenience of this method was that the small mic stand in use, sitting on the seat of the chair, was set up so that the 2nd through 8th positions simply involved sliding the telescoping boom of the stand without moving any other angles or settings. That made it very easy to finish the rest of the measurements without a tape measure or any precise physical measurements of microphone position. The MMM average FR result is shown in the final diagram, which is very flat and sounded very good. Soundstage and imaging (SS&I) were both very good and the FR was smooth and enjoyable. The high bass amount could easily have been tailored by changing subwoofer level and the small dip just above 100 Hz could probably have been alleviated with a subwoofer positioning or delay setting change.
NOTE: At this point the best SS&I performance of Dirac Live was compared to that of Audyssey XT. While the results with Audyssey XT were very good, the Dirac Live results were sharper and more precise. Audyssey XT32
might have done a better job at this than Audyssey XT, although this can not be automatically assumed.
Some might argue that left and right data should be included in the measurements. There is no harm in doing that, however it complicates the process in that precise physical measurements are required for the symmetry of that measurement pattern, and one of the goals I was hoping to achieve was the elimination of that tedious necessity. I believe that has been achieved with the fourth pattern described above. And if one wishes to argue that the left and right data is important because of variations that need to be included, the counter argument is that if there are significant FR variations a few inches left and right of LPC, then there are serious problems at the listening position which only will be alleviated acoustically, and Audyssey or Dirac Live or any other room correction program will have no hope of making it any better.
The priorities for data collection by way of the mic setup pattern at the LP are:
- front to back variations
- up and down variations
- left and right variations
Measurements and actual listening tests verify that using the center plane for all measurement positions gets the information necessary to give excellent FR and SS&I results at the same time. And since those measurements do not involve any left/right symmetry requirements, the tedium of careful physical spacing is unnecessary. What a relief!
Beyond that, the main requirements for getting good results are that:
- the first measurement is at LPC, even though we know there are disturbances there and often no line-of-sight to surrounds, as there is important data needed at that position for the room correction program to get best performance from the front main speakers
- the remaining measurement positions, all in the center plane, should be different distances from the chair back
- all within one foot of LPC, if possible, although line-of-sight to surrounds is important for these measurements. With Audyssey, measurement points farther than 12 inches from LPC give compromised FR and softened, imprecise SS&I, so are only useful if smoothed FR across several seats is the goal.
The method described above, sliding the telescoping boom along the hypotenuse of the right triangle described, makes it very easy to accomplish this without any further physical measurements.
In conclusion, with Audyssey, where the height of a chair or sofa back extends above shoulder level, multiple measurement positions are necessary when good FR and good SS&I performance are all desired. The single mic position calibration (3 calibration steps all at LPC) is appropriate where a low chair back is in use. With Dirac Live, multiple measurement points, with the first at LPC, are always recommended, as discussed elsewhere.
