My main objection to Audyssey is, ironically, the part that seems to most strongly influence people's decision to purchase Audyssey products. The room correction software. To explain why I'll first have to explain (review for many) some fundamentals about sound reproduction in a room.
Speakers differ from electronics in that they radiate in three dimensions. Because we are listening in a room we will hear both the off-axis sound reflected through the room and the on-axis direct sound. Not only does attention have to be paid to the spectrum of the on axis sound but also to the spectrum of the off-axis sound.
Speakers (omnidirectional speakers notwithstanding) become more directional with increasing frequency. This is because because frequency is inversely proportional to wavelength; as the frequency increases the wavelength shrinks. Eventually the size of the driver is substantial compared to the wavelength. At this point the driver begins to become highly directional or "beam". Most speakers are at least two way designs to achieve bandwidth close to human hearing. Because the different drivers are different sizes they will become highly directional at different frequencies. This can create a directivity mismatch at the crossover frequency that looks like this:
The far off axis sound of a well designed speaker will smoothly and gradually decrease in amplitude without evidence of a directivity mismatch between drivers at the crossover frequency. The following is a good example:
Check out Andrew Jones's interview on Home Theater Geeks for a good explanation.
My first objection to Audyssey is that in consumer versions it begins with the assumption that your speakers are designed like in the first image. It is referred to as "Midrange compensation" and is discussed by Dr. Kyriakakis here.
The reflected sound field is often a bigger determinant of the measured frequency response in a room than is the direct sound. If the reflected sound is a large contributor to the in-room frequency response then speakers with a directivity mismatch will measure with a hole in a frequency response at the crossover region
But our binaural hearing system is not entirely represented by microphone measurements. We seem to be able to "listen through" the reflections and hear the spectrum of the direct sound (see Sound Reproduction by Floyd Toole for more info). So instead of compensating for the hole in the in-room response, which may make the on-axis response sound colored (assuming it was balanced in the first place), Audyssey instead starts with the assumption that your speakers were designed for with a relatively flat on-axis frequency response and a directivity mismatch. This is probably a very reasonable assumption, but there are more speakers than ever that are designed without the directivity mismatch.
If one uses a speaker with well controlled directivity then "Midrange Compensation" simply introduces an arbitrary dip centered at 2kHz to the on-axis sound. As we are particularly sensitive in this region, this is probably undesirable.
My second objection is the choice of the target curve. As I mentioned earlier, measured in-room response is substantially influenced by reflections. Because of the unavoidable increasing directivity with frequency, if the on axis sound is balanced to be flat, this will result in in-room measurements showing a frequency response that follows downward slope from the low frequencies to the high frequencies. If the in-room response is equalized to be flat, the direct sound will be pushed to have an upward slope.
Here is an idealized representation of the relationship:
Normal frequency response for a "perfect speaker"
"Perfect speaker" equalized to flat in-room response.
Audyssey's house curve does appear to take this into consideration somewhat, but as far as I'm aware this is only with high frequencies. If the rest of the in-room response is equalized flat then there will be a decrease in amplitude of direct sound as frequency decreases, making the speaker sound thin or even bright.
In the only controlled, double blind study I've seen Audyssey performed poorly. This can be found here. The names of the room correction products are not revealed, but it's not hard to figure out from it's characteristic "Midrange Compensation" dip that RC6 is Audyssey . It's worth noting that listeners found RC6 to be thin and colored which is consistent with the points that I've made.
I think the Audyssey Room Equalizer hardware used in the comparison has MultEQ rather than MultEQ XT32, which has less FIRs and no dual sub EQ. Dynamic EQ also probably would have made things sound less thin, so there are some limitations to drawing comparative conclusions from the study. Nevertheless, the conclusion that the a flat target curve is not ideal is very useful and contributes to the discussion
My final point is that sound behaves very differently below a certain transition frequency that varies depending on the room, but it is often around 300Hz. Below that transition frequency the measured frequency response is dominated by the position of the listener and loudspeaker. Above that frequency the measured response is largely determined by the sound source and reflections. Equalizing above the transition frequency is mostly equalizing the loudspeaker. This is best done with access to the direct sound rather than with in-room measurements.
Most of the limitations of Audyssey don't exist for the pro version. This version has an adjustable target curve and defeatable Midrange Compensation. It requires an expensive license and an even more expensive AV receiver equipped with it. I hope in the future Audyssey will make a version that is some kind of compromise between the consumer and pro versions with at least a toggle for the Midrange Compensation.
The compromises that Audyssey makes make sense for an average consumer, but this is the AV Science forums, and we are most likely not average consumers. We often recommend well designed speakers, so we should be cautious about recommending room correction software that nullifies the speaker designers' hard work.
- Midrange compensation may be robbing your speakers of the natural timbre that your speaker designers worked so hard for.
- The target curve that Audyssey uses is probably making your speakers sound thin.
- Above the transition frequency you are equalizing the loudspeaker. If this is done based on in-room measurements it could do more harm than good.
Edited by SyntheticShrimp - 12/30/13 at 3:56pm