Dirac > Audyssey XT32 ... Sure, Always?

[Andrej]

Any preference for a "target" curve response is intimately tied to the loudness level. I have found that I personally prefer flat response in my listening spaces, ranging from a heavily treated room, to no acoustic treatments (but both speakers and myself away from the walls and the listener relatively close to the speakers). I am a big fan of DSP, and that includes the automatic approaches like Audyssey (I have had several XT32 receivers, as well as XT) and Dirac (a friend has the PC based version). On his system we both also prefer flat target curves.
However, in order to get satisfactory perception of "realism" in the reproduced sound, I found that I needed the volume to be at reference level, or having some form of dynamic equalization engaged. In general I aim for "fidelity" rather than "personal preference", and so I inherently find preference where I find accuracy, ie. flat response (which I then modify with the DEQ)
When I build a new 2-way satellite pair of speakers my favorite first listen is with an inexpensive old Onkyo XT32 receiver with a built in digital crossover (those were manufactured for only a short amount of time). All I need is a pair of speaker boxes, bass units, tweeters, and, if so desired, a subwoofer. Everything else is done by the receiver, which can be bought used for less than $400. Cannot beat that value for money.

 

[peng]

That doesn't explain how adding a dip to the target curve helps Audyssey better correct the peak it failed to correct earlier. Unless that dip isn't fixed due to spatial averaging (that is, it doesn't consider the dip as it's only in 1 place). If we just measure in the same spot, can you still get any flatter response by setting a non-flat curve? How does that actually work?

I understand your point and I do not know the exact reason why it works, just know that it works for me and some others who tried. I do have some ideas, just educated guess but it will take too long to get into the details. I can use a quick numerical example here though, just to illustrate my logic (can be wrong obviously). Let's say in the case of just using 3 mic positions and without REQ, REW shows the following dips and bumps at 100 Hz:

Obviously again, the following are all made up, not real results, and all my assumptions could be wrong, or partially wrong but I could be partially right too. I actually did discuss my logic with Audyssey, it was a lengthy email exchange.

A) Audyssey off:
MLP: +7 Q=5
12" to the right: -1 Q=3
12" to the left: -3 Q=6

B) Audyssey On:

MLP: +5
12" to the right: + 2
12" to the left: +3

C) Now if we assume the following:

- Audyssey's algorithm is perfect, or close to perfect
- The mic is as good as Audyssey claimed per their factory calibration, and say assume +/- 1 dB (don't remember what it is supposed to be)
- The operator follow instructions to the letter, such as keeping the room quiet, mic point to ceiling at ear height, at least X ft from any walls..et..

If the assumption in C above is true the "corrected" results would be (also assumed):

MLP: + 2.5
12" to right: 0
12" to left: +1

So in this example, Audyssey missed the target for the MLP by 5-2.5 = +2.5 dB and the other positions by +2 dB
Since C) assumes the algorithm is near perfect, the error would be due to:

D) the mic's tolerance, room noise, mic positions, operator related etc..

In this case if I apply a - 2 dB, that approximate the correction needed to improve all 3 positions would be what I call "feedback" to Audyssey that in order to offset the imperfection resulted from error due to D) and the resulting "errors" would be reduced as shown by the new REW graphs.

Finally, I think is is reasonable to assume in addition to D), Audyssey's algorithm is not perfect enough to flatten the dips/bumps at 90 Hz, and in that case, applying the same - 2 dB would still bring down the bump(s) for all 3 positions.

And you are right, doing as described above would mean in order to correct the 90 Hz bump, I would have to put a dip on the target curve. In other words, in order to achieve a flatter frequency response for the corrected range, I would end up using a target curve that is not flat.

Back to my amplifier analogy, for an amplifier to have the lowest possible distortions, a feedback loop is necessary such that measured "error" measured at the output is fed back to the input, and the error signal would thereby "correct" the input signal in order to achieve a less distorted output signal. The advantage the feedback scheme an amplifier has is, obviously it is continuous, i.e. a feedback loop. Doing what I did with Audyssey (App, Ratbuddyssey etc..) is a one off, though I can repeat the process manually if I am willing to spend time on it.

Ratbuddyssey actual shows the measured results by that cheap looking mic, for each position. Below is an example of mine:

So it isn't hard to plot the corresponding "corrected" graphs for all 8 mic positions to see if any positions are "made worse". I typically would just use REW to compare the Audysey on vs off graphs for a bunch of positions by moving the mic and/or point it at different angles to the speaker.

 

[randomer]

My only real complaint with Audyssey XT32 is the default reference and flat curves. If you simply run Audyssey in a Next/Next "wizard" like function, it doesn't sound good. Audisssey's belief appears to be flat frequency response is the desired goal. This belief is flawed as people prefer a downward sloping curve. It basically sounds both bright and no base at the same time. If they would add a better downward sloping default curve (like a harmon type curve) it would make things much easier. My favorite instructions for getting Audyssey to produce good results is via this" https://docs.google.com/file/d/1YH_eNHRCxKFCwXMddi28kzXqnUwYHfrD/edit?filetype=msword

Al least I have had good results with it.

@Dj7675 the file seems to be deleted do you have another link or source? ty

 

[Dj7675]

If I recall correctly ( might not be) I believe they were from user @Jon AA . I haven’t used Audyssey in some time. If I recall correctly he used Audyssey for speaker correction. I think he may be using the pc app now.