Audyssey FAQ Part 2
Chris's post starting here:
-------------------------------------------------------------------------------
Here is a list of tips based on several discussion in this and other forums about MultEQ. I will add to it from time to time as new questions come up. Some additional information can also be found on the Audyssey website.
Best regards,
Chris
CTO, Audyssey Laboratories
Microphone Placement
• The microphone has been calibrated for grazing incidence and so it must point to the ceiling during calibration. Any other orientation will produce incorrect results.
• The microphone response has been calibrated to match (on average) the response of an industry-standard ¼” instrumentation microphone. It is critical to use the microphone that came with the receiver and not one from another model that may have a different calibration curve.
• It is also important to place the microphone on a tripod or other stand so that it is at ear height. We strongly recommend against holding the microphone in your hand because this can give rise to low frequency handling noise that will cause the MultEQ filters to compensate by cutting those frequencies. Furthermore, it is not recommended to place the microphone on the back of the couch or recliner. If a tripod is used, care must be taken to ensure that the microphone is placed at a height just above the seat back so that reflections from the seat do not cause problems at higher frequencies.
• The first microphone position is used to calculate the distances to each loudspeaker and subwoofer and set the delays. It is also used to measure and set the trims. So, it is important to place the microphone in the main listening seat for the first measurement.
• MultEQ measures the background noise level in the room before playing the test signal from each speaker. For the measurements to be valid, the signal to noise ratio must be above a certain threshold. If it is not, the test signal from that speaker will repeat at a higher level. If the noise in the room happens to be higher during some of the speaker measurements, then the test signals from those speakers will sound louder than the test signals from the other speakers. This does not affect the calculation of trim levels. If the room noise is too high even after the test signals increase in level, then an error message will be displayed warning the user that measurements can not be completed.
• After the first position is measured, MultEQ measures other positions in the room around the listening area. These do not necessarily have to be in each individual seat. The idea is to capture as many points around the listening area as possible so that the acoustical problems that affect the quality of sound within that area are minimized.
• For example, we recommend taking 3 positions on the couch facing the TV and then 3 more positions about 3 feet in front of the couch and parallel to the first three. Measurements up against the back or side walls should be avoided.
• Some loudspeakers have rather problematic responses when measured off-axis (i.e. more than 15° away from the imaginary straight line that points to the listening position). In these systems, measurements taken too far away from the center line will show a reduced high-frequency response that may result in overcorrection and thus overly bright sound. Although it is difficult to predict which type of loudspeaker will have these off-axis problems we have most often observed them in poorly-designed multiple-driver arrays that exhibit very high off-axis lobing. In these situations we recommend a tighter calibration pattern centered around the main listening position and making sure that the mic is not placed in extreme locations and certainly not outside the plane of the front main speakers.
Checking the Results
• Once MultEQ calibration is complete the results are stored in the receiver memory.
• It is important to activate MultEQ by selecting one of the target curves. This is not performed by default after the calibration is finished and must be selected by the user. In a THX system we recommend using the “Flat” setting that allows the re-equalization to work as intended. In other systems, we recommend “Audyssey” for movie playback and “Flat” for music playback. Unfortunately, the music industry does not have any mixing standards like the movie industry so some music program material may sound better with the “Audyssey” setting. “Front Align” also uses the Audyssey process, but it does not apply the filters to the two front loudspeakers. “Manual” is not an Audyssey setting and does not use MultEQ filters. It is a simple parametric equalizer and will be subject to all the limitations that come with parametric EQ.
• Small vs. Large speakers. This is the most commonly discussed topic by MultEQ users. The first thing to understand is that it is not a personal insult to your system if your speakers were detected as “Small”. It simply means, that in the room they were measured the – 3 dB point was detected at 80 Hz or above. This may happen even if the manufacturer’s spec shows that the speaker is capable of playing lower. In fact, there are several benefits at crossing the speakers over at 80 Hz that have to do with power handling and headroom in the bass region that will be handled by the subwoofer amplifier.
• The second most common question also relates to Small vs. Large. In the Denon receivers, MultEQ will designate as “Large” any speaker that has a –3 dB point below 80 Hz. For non-THX speaker systems this is an arbitrary definition that often causes confusion. All it means is that the speaker will not be bass managed unless the user tells it to be. Because Audyssey is not in charge of bass management, we have to abide by the manufacturers’ rules and simply report the information found by the measurements to the bass management system.
• In situations where the speakers do not play significantly below 80 Hz, an additional step must be taken to make sure that there is no loss of bass information. The user must set the speaker to Small manually so that bass management is performed properly.
• Polarity: MultEQ checks the absolute polarity of each loudspeaker and reports it to the user. This is simply a report and does not affect the subsequent calculations in any way. It just asks you to check the wiring to make sure it is connected properly to each speaker. Sometimes we get false alarms. This is usually because the speaker has a driver (usually the mid-range driver) wired out-of-phase intentionally to make up for some problems at the crossover region. If a phase warning is shown, it is not a cause of alarm. Simply check the cables and hit “Skip” if everything is fine. Again, this does not have any effect on the EQ results.
• Subwoofer distance: in many active subwoofers it is not possible to defeat the low-pass filtering. That means that the pre-pro bass management filters will be on top of the low-pass filters inside the subwoofer. The built-in low-filters introduce a delay to the signal coming in (because they have poles). This delay is seen by MultEQ as acoustical delay and is reported in the results. That is why sometimes the subwoofer distance is reported to be longer than the physical measured distance. The setting should not be changed because the blend between the sub and the satellites has been designed based on this time delay.
• The design constraints for MultEQ were that it (1) must fit within a small portion of the DSP so that other processes can also run and (2) it must use FIR filters because of the well-known artifacts that IIR filters cause particularly in the time domain response. As it turns out, these two requirements are contradicting. In order for FIR filters to be effective and capable of correcting to low frequencies, they must consist of several thousand coefficients (taps). The problem is that the CPU power required increases with the number of taps, hence the dilemma. What we did at Audyssey was to come up with a different way to partition the frequency axis so that we can use fewer taps and yet not completely give up on low frequency resolution (and therefore low frequency correction). This allows us to take a 512 tap filter that would normally have a resolution of 94 Hz (meaning that any peak or dip narrower than 94 Hz would be missed) and significantly improve its resolving power. The resolution of the filter actually varies continuously with frequency and starts at around 10 Hz. Does this mean that MultEQ can correct an arbitrarily narrow peak or dip at 30 Hz? Of course not. The reality is that in the MultEQ XT version found in receivers, we can correct broader features below 100 Hz better than narrow ones. For example, a lump that is half an octave wide at 50 Hz can be fixed. A narrow dip or peak that is 1/3 or 1/6 octaves wide and centered at 30 Hz will be improved, but not eliminated.
• The on-screen display in the receiver has very limited graphics. Therefore it is not possible to really show what the MultEQ correction filter is doing at all frequencies. It appears to only be operating on 9 bands like a parametric equalizer, but this is not the case. What is shown is a very crude approximation to the MultEQ correction and it should not be used to read exact values of cut or boost at the 9 frequencies shown.
• Furthermore, there is no display for the subwoofer filter. This doesn’t mean that there is no subwoofer correction. It was not added to the display because of interface and memory considerations.
