Audyssey quotes (collected over time)...

Hi to All,

I've collected a sizable number of quotes from Audyssey (aka Chris Kyriakakis) that I thought by this time would be interesting to be launched and shared via a separate thread for those interested in reading such Q&A's of this well known room correction system. Please regard this thread as a platform for quotes only, while kindly request everyone not to post comments. The sole purpose is to share as many quote as possible, meantime if you have your own collection, please do not only feel free to share them, but you are explicitely asked to do the same. As you will see, in-depth explanations will follow throughout the quotes that I hope will be both entertaining and educational at the same time. Should you have a desire to comment you are kindly requested to open a separate new thread for further discussions on any of the subjects quoted here.

By no way does this thread want to be competing with the Audyssey FAQ or the Audyssey setup guide, it's a light weighted thread, kinda casual one, yet hopefully it will be taken with your anticipation.

Remember, quotes only!!

Here's the kick off for today to be follow by many more:


Audyssey test signal
________________________________________
This question has come up in some emails I received recently and I thought I would post it here as well.

Q: How can Audyssey measure anything with these silly blips? Shouldn't they use sweeps like everyone else?

A: The silly blip you hear is actually a fast sweep. It starts at 10 Hz and runs out to 24 kHz, but it weighs the frequency sweep logarithmically. In other words, the lower octaves get more energy than the upper ones. Sound familiar? In fact, if you take the time domain test signal (it's called a log chirp) and transform it to the frequency domain you will get the exact same spectrum as full range pink noise.

During measurement, the initial chirp is approx. 75 dB SPL for a nominal listening distance and speaker sensitivity. The chirp repeats several times per speaker and this has the benefit of increasing the signal to noise ratio in the measurement.

Also, Audyssey listens to the background noise in between chirps. If it's above the required minimum then it repeats the sequence of chirps at a higher level to make sure it gets meaningful measurements.
__________________
Chris

(to be continued...)
Edited by mogorf - 10/1/12 at 2:51pm

One more for today:

Regarding DymanicEQ implementation:

Q: So now I'm confused - what is Audyssey actually trying to achieve? Does it try to make a violin played at reference level sound like a violin played at the level corresponding to the master volume setting? Isn't the goal of equal-loudness compensation "just" to maintain the spectral characteristics of the whole recording, so a forte played violin still sounds forte even when the recording is played at -30 dB from reference?

A: One goal is to maintain the same perceived spectral balance when listening at levels lower than those used during the mix. A violin, for example, is playing a wide range of notes and it is mixed so that every passage is at a given balance with the other instruments. When you turn the volume down the lower notes of the violin will start to be perceived softer in level than the higher notes for that same passage. The overall perception of the violin relative to other instruments playing along with it will also be perceived differently. So, with the static part of Dynamic EQ we are trying to make spectral adjustments that follow a set of static curves. The dynamic part of Dynamic EQ adds one more level of detail: it looks at the moment-by-moment variations in the content loudness and compares them to the perceptual model. Based on that information it determines how "loud" that passage will be perceived and then, by knowing how loud it was perceived during the mix, it makes a secondary adjustment. This secondary adjustment is, well... secondary and changes continuously as the signal is playing.
__________________
Chris

Q & A with Chris K. on how MultEQ sets channel trims:


Q: When we talk about channel trim settings during auto setup are we talking about a standardized 75 dB SPL measured at the MLP with a -30 dBFS band limited (500 Hz to 2 kHz) pink noise signal? When Audyssey sets trims is this standard used? What puzzles me is how the Audyssey chirps fulfill this standardized method. Is it done by some kinda gating of the chirps that basically cover a range of 10 Hz to 24 Khz, while during trimming the speaker levels only a 500 Hz to 2 kHz range is utilized?

A: Audyssey measures the entire frequency response of each speaker. The chirps are "full range" even though it's hard to hear the low frequencies in the beginning. After that the energy under the 500-2k range is analyzed to produce an SPL estimate. The trim is the difference between that estimate and 75 dB SPL.

Q: How does that work with regard to subwoofer level...does Audyssey set the bass level using the lowest measured frequency and then make cuts to equalize the other frequencies? or does it use one particular frequency like the crossover as the reference?

A: Same as above, but the range it looks over is 30-80 Hz.
Edited by mogorf - 10/3/12 at 12:07pm

Someting about verifying trims set by Audyssey:

Q: While playing test tones (-30 dBfs, 500 Hz-2kHz band limited pink noise) off a test CD and Audyssey is turned ON and OFF will the meter (C-weighted) show different SPL levels at the MLP?

A: Yes it's possible when using a CD. It will depend on how much work the MultEQ filter is doing in that region.

Q: How about when using the AVR's internal test tones? AFAIK, the AVR turns off Audyssey filters during test tone rendering, but leaves the channel trims and distances intact.

A: That's right. The internal test tones don't see the filters so you will get the same answer with Audyssey on and off.

Q: Ok, thanks. So, for absolute SPL (75 dB) at the MLP which one is the valid test, an external CD with test tones and Audyssey ON, or the internal test tones without Audyssey filters?

A: If you are absolutely certain that the external CD was properly recorded (there are some that are not--particularly DVDs that have messed up the dialnorm setting), then that's the way to go. Because of this uncertainty, however, we always recommend to go with the internal test tone if you want to check.

Q: I see, and understand the worry about external "stuff", but in case of the internal test tones what is the rationale for turning off Audyssey if the MultEQ filters have a job there. Am I far from reality with my conclusion that in this narrow 500 Hz-2 kHz band a typical room will only show subtle differences with MultEQ filters ON or OFF? At least it will be somewhere within the +/- 2dB tolerance range of an average SPL meter due to a region free of room modes not like the bass department.

A: It's not that MultEQ is turned off. It's because the AVR has the test noise in a block that comes after MultEQ and so it doesn't see the filters. We have asked for this many times, but it's apparently too complicated for them to make the change in the architecture.

A bit more on DynamicEQ. Having it engaged hides some "secrets" that may become apperant only when taken away. Please read on, its an interesting discussion on particulars of this feature:

Chris: The Audyssey room correction (any flavor) operates on each speaker independently. It doesn't care if you have 1 or 11 speakers on at a time. The correction created for the front L and R channels is the same whether you are using them for 2ch. or multichannel listening.

Q: Apparently it is for Audyssey's particular brand of loudness compensation. My point is that I don't see why it has to be that way for loudness compensation in general. It's not like front channel levels track differently than surround channel levels as you lower the volume. The part of the ear that judges level doesn't know which direction the sound is coming from, so loudness compensation can be per-channel (based on content and offset from reference), without concern of whether it's a surround channel or front channel.

A: It actually does matter. Perception of loudness is spatially dependent. And, there is also evidence that it is cross modal. In other words the presence or absence of picture affects our estimation of loudness.
__________________
Chris


Q: The dynamic part comes in when they're messing with surround levels.

A. No, that' not true. The level increase applied to the surrounds does not look at the content. It only looks at the volume control setting and therefore is not dynamic. The frequency response changes applied to the loudness compensation look at content and master volume setting and are dynamic (i.e. changing with time).
__________________
Chris

Q: It's calibrated properly, but at no point does Audyssey determine whether my surrounds are to the sides or the rears. Side surrounds are very common IRL even if they're not technically correct - DEQ is doing more harm than good, compensating for an issue that doesn't exist.

A: Are your surrounds pointing to the first mic position? If not, then the natural off-axis level drop off that happens because of high frequency directivity will influence the level setting that MultEQ does. If that's the case then the assumed "reference" level would be too high and could cause Dynamic EQ to overcompensate.
__________________
Chris

Q: What I'm talking about is not the sensitivity, but how it tracks (whether it changes) as we lower the volume. So here's a hypothetical for you:
I sit you down in the middle of a 4-speaker set-up: the fronts and surrounds are placed per your preference. Surrounds are playing back at the same perceived level (as opposed to the same measured level) as the fronts: you hear the fronts and the surrounds as equally loud. If I start lowering the volume, will the surrounds fade faster than the fronts?

A: This was the exact experiment we did: set levels on all speakers with narrow-band pink. They all measure the same. At reference master volume they were also perceived to be at the same level. Now, bring down the master fader and adjust the surround fader so that you maintain a constant impression of surround level. To do that required a raising of the surround volume as the master fader came down. Repeating this over a number of participants gave us the level adjustment curve needed.
__________________
Chris

Q: The same measured level was perceived the same all the way around, regardless of direction? I suppose that's possible, just seems counter to what you were saying earlier. You did this with pink noise, real program material (surround mixes), or both? I'm assuming the latter.

A: At high SPL the front and rear levels were perceived as equal. As the master volume started to decrease, the rear level was perceived to be dropping faster than the front. Yes, this was done with both content and noise with very similar results.
__________________
Chris

Q: So how would the boost applied to the surrounds not result in unnecessary emphasis when the surrounds get loud?

A: The boost in the surround level only happens when the master volume is turned down. So, at a given master volume (say –20 dB) there will be a few dB of increase in the surround level. Any content that is designed to be louder intentionally will be louder but not unnecessarily so. Also, the frequency response adjustments to that louder content will be different compared to the other softer content because of the dynamic frequency response adjustment portion that looks at the content in real time.
__________________
Chris

Here's another interesting Q & A quote on treatment of in-room reflections:

Q: The simplistic version: add a counter-reflection to the direct signal that is capable of cancelling the reflection from the room.
The problem is that this counter-reflection does create a reflection in the room too, just like the original sound did. Now you need to add another counter-reflection to the direct signal in order to eliminate the counter-reflection's reflection from the room and so on.
What complicates things further is the fact that reflections are not the same at every location within the room.

A: The time domain graphs that you see on the Audysey site are shown to be aligned in time after correction. When you first measure them, the big initial spike arrives first at the microphone. But because the speakers are at different distances that spike will arrive at different times. This is why it is so important to set the time delays properly so that the speakers are aligned to each other in time.

The MultEQ algorithm does not try to isolate the direct sound from the later reflections that you see. Instead, it looks at that information from all the measurements and uses a set of fuzzy logic rules (fancy words for probability theory applied to signals) to determine what problems are most common across the listening area and how to best apply the filter power to minimize the most important ones first. So, yes, there is an opposite signal that is generated, but it's the opposite of the combined responses so that it can address the totality of the problems rather than a single reflection found by a single measurement (that may not be there in a nearby seat).

Chris