Audyssey MultEQ XT, XT32, PRO, Applet, MultEQ-X (màj)

[Hugo S]

Bjr,

Voici la copie ici de 2 intéressantes interventions qui éclairent les contours de fonctionnement de la correction acoustique MultEQ XT(32). Ceci plus particulièrement du point de vue des raisons et objectifs des résultats auditifs recherchés > 1kHz et pour une zone d'écoute.

Maintenant si on lit entre les lignes, on peut percevoir ici, tout l'intérêt potentiel du Kit Audyssey Pro (tolérance micro +/- 0 .5dB), ceci en termes de précision de l'image audio finalement possible après cette calibration Pro.

Let's put it this way, what causes "an excessively bright tonal balance"?

Hi Markus,

One reason could be taking measurements too far off axis. The tweeter listening window response may be such that off axis measurements tell MultEQ that there is not "enough" high frequency content and that can result in unnecessary boost. Hence, the recommendation to keep the measurements within the opening angle of two front speakers.

There is evidence in the literature that points to a reason to apply correction at frequencies above Schröder. It is the BBC research that showed that stereo imaging is critically dependent on having matched speaker responses (±1 dB) well above 1 kHz. That's nearly impossible to achieve just by buying two "identical" speakers since it is outside the manufacturing tolerance of most mass-produced components. So, even if you don't believe that room correction is needed you may want to consider the correction beneficial if it ends up providing a better speaker-to-speaker match in the high frequency range.
__________________
Chris

I agree that matching speakers is a very important goal. Is that all Audyssey is trying to correct for at higher frequencies?

This sounds like a trick question . In the frequency domain, it is trying to get each speaker to match the target curve. The result is that each speaker is also better matched to each other. In the time domain, it is trying to clean up the impulse response so that the contributions of the acoustical environment (room, nearby surfaces, etc.) are minimized. For example, a speaker close to a large TV screen will produce reflections from that screen that will combine with the direct sound and create time domain related problems above 3-5 kHz. Shouldn't those be "corrected"? We believe that they should.

Also, in response to an earlier comment you made about averaging. Frequency domain averaging has been tried and it can improve things greatly over single position measurements (assuming it is done correctly, i.e. in the RMS sense). But, it still has a problem: all measurements are equally weighted when, in fact, the acoustical problems are not equally distributed. To avoid this problem, we developed a different method to combine the measurements that is not based on traditional averaging. I would be happy to send you some published papers on the C-means fuzzy clustering technique that we use.
__________________
Chris

Hugo

PS: http://translate.google.fr/#en|fr|
PS': édition tolérance micro Audyssey Pro +/- 0.5dB

[alex_t]

Le big boss de AUDYSSEY confirme la calibration individuelle de chaque kit pro

Source : AVS FORUM

The Audyssey Pro-mic sensitivity is calibrated to within ±0.5 dB of reference because it is individually calibrated. The consumer mic sensitivity tolerance is ±2 dB.

[palm]

OK très bien donc les kits Audyssey qui viennent en complément des préamplis Onkyo, Marantz, Denon, NAD ne servent qu'à améliorer la prise de mesure alors que je recherchais une solution complémentaire pour améliorer la correction.

(améliorer la correction : augmenter le nombre de fréquences sur l'égaliseur)

A+
- skYfIrE.

La correction consiste en un filtre FIR - pas en un égaliseur 9 bandes qu'on trouve dans certains amplis
Certains amplis permettent de copier la correction de la réponse en fréquence mesurée vers l'égaliseur, mais en pratique on perd alors quasi tout l'interet de la correction Audyssey qui agit aussi dans le domaine temporel.
http://www.hometheaterforum.com/forum/t ... st_3366684

Avec le kit pro on peut jouer sur la courbe cible, mais effectivement ca ne changera pas la precision du filtre (mais au sens nombre de coefficients, pas au sens fréquences)

[Hugo S]

Hi Alex,

Le big boss de AUDYSSEY confirme la calibration individuelle de chaque kit pro

Source : AVS FORUM

The Audyssey Pro-mic sensitivity is calibrated to within ±0.5 dB of reference because it is individually calibrated. The consumer mic sensitivity tolerance is ±2 dB.

Merci pour cette précision. Dans mon post plus haut je voulais écrire +/- 0.5 , mais mon doigt a glissé et on lisait +/- 1.5dB (totalement incompatible avec ce qui est décrit au dessous)... aussi j'ai donc édité le post en question.

Merci à Toi.

Hugo

PS:... la neige ne me réussit pas du point de vue attention...

[alex_t]

De rien Hugo

HS : La neige a noël c'est sympa mais là elle commence sérieusement à m'ennuyer vivement une météo plus clémente ....

[Hugo S]

Edit.

Pas pertinent.

[Hugo S]

Bjr,

Une précision sur le traitement par Audyssey MultEQ, de la réponse impulsionnelle, dans un milieu réverbérant.

Hi Chris,
It's possible I misremembered or misinterpreted your original response.
In particular, by "correct time response" I was thinking that it meant what I just described.
Perhaps you meant Audyssey corrects the frequency domain response of time domain effects, i.e. reflection(s) causing perceived brightness?

A time-domain response (or impulse response) is corrected by MultEQ so that it more closely resembles a perfect impulse. In doing so, the reflections (copies of the original signal that arrive at various later times) are suppressed. This correction is done by creating an FIR filter that is applied to the impulse response. When the signal in each speaker plays through this FIR filter it is convolved with it as it plays and the corrections are applied continuously.

You can see some examples of before and after time domain responses here. Click on Seat 1, 2, etc.

The effects of the reflections (that show as blips later in time) are reduced. But, that is done by the filtering operation and not by going in with "tweezers" to kill each reflection. The amount by which each is reduced depends on the spatial distribution of the effect they have on the response.
__________________
Chris

et le corollaire sur rendu psycho-acoustique complémentaire produit par le DSX W+H:

Chris,
I may be over-simplifying things but would you say that:

1) MultEQ reduces random reflections in order to remove unpleasant "ringing"

2) DSX Width and Height channels inject back "good" reflections at the proper angles so as to enhance Apparent Source Width and Listener Envelopment.

Mark

Hi Mark,

1) I prefer to say "MultEQ reduces the effect of unwanted reflections". It's hard (if not impossible) to remove individual reflections.

2) Yes
__________________
Chris

et l'impact de tout ça sur l'égalisation du rendu timbral inter-enceintes

Hi Chris,
Thanks for the explanations regarding correcting the effects of reflections.

It is interesting to note that this discussion of reflections started from a question about the value of electronic timbre matching. (Refer to the postings quoted below.) I was hoping that we could get your thoughts on whether you share Markus' point that inorder to do timbre matching that you need to take free-field measurements. Obviously we're not really interested in removing all reflections to do timbre matching, but does filtering the impulse response hurt timbre matching if we lose frequency resolution in the bass region? Any general explanation of how Audyssey does timbre matching would be appreciated.
Thanks.
Larry

Hi Larry,

Timbre matching is a tricky thing. First of all let's start with a little experiment. Take two identical speakers and place them next to each other so that they are touching. Now play pink noise that alternates from one to the other. You may be shocked to find out that they sound completely different. That's because even a slight variation in placement changes the response. In a real setup with three front speakers that most often do not include an identical center things get even more unmatched.

Even if you have free field measurements, once the speakers are placed in the room the free field data becomes less useful. The timbre of the speaker is determined by the balance of axial response vs power response. This balance changes dramatically because of placement, listening distance, room acoustics, and most importantly: speaker directivity. In an anechoic chamber one typically measures a listening window response that involves 7 measurements on an ellipse (or circle depending on your school of thought). These measurements provide a certain standardized weighting of on- and off-axis responses that can be used to compare different speakers. It has been found to have reasonably good correlation to the perceived timbre although there are other factors as well.

In a real room we don't have the luxury of listening window measurements for each speaker. But, we can extrapolate a quasi-listening-window response if we spatially sample the listening area where the speakers are pointing. That is done by selecting the appropriate weighting for each measured response and is something that simple averaging can't do.

MultEQ treats the speaker and room as a combined system. When it measures impulse responses it's not trying to make the anechoic speaker responses match. It's trying to match them in the room they are in. This involves taking the effects of reflections into consideration. Once the measurements are collected a kind of target curve is imposed on the time domain impulses. Think of it as a perfect impulse: all the energy appears in the first spike and nothing after that. Of course, that's not going to happen in a real room. But, we can show that the filtering process pushes each measurement closer to a perfect impulse than before. Doing so reduces the energy in the blips that happen after the first impulse and therefore the effect of the reflections.

It's not true that this type of filtering reduces the low frequency resolution. The frequency resolution is determined by the length of the impulse response (the time that it extends over). The longer the response the higher the low frequency resolution. This is a fundamental property of the Fourier Transform. If you truncate an impulse response by cutting off its tail (and shorten it) then you will get what Markus is talking about.

So, getting back to timbre matching: If your goal is to do the first experiment I mentioned and expect to hear no difference then you will not succeed. It's not possible to equalize the directivity of a speaker. The focus has to be on a solution that takes into account the in-room balance of direct and reflected sound and produces filters that bring each speaker closer to the target response.
__________________
Chris

Meilleurs Voeux à vous tous, pour cette nouvelle année naissante 2011.

Hugo

PS: http://translate.google.fr/#en|fr|

[palm]

Je posterai les courbes mais j'ai fait des mesures avec IK ARC (Mult EQ XT) et comparé avec DRC (solutions basées sur un ordinateur donc) et DRC va beaucoup plus loin sur la correction de la réponse impulsionelle. Avec Audyssey c'est plus anecdotique en comparaison.
Mais bon c'est un résultat, je ne généralise pas, surtout qu'a l'écoute j'ai une préference pour Audyssey.
Je pense que c'est tout simplement du a la méthode de mesure vu qu'Audyssey ne va pas essayer de corriger une reflexion presente sur une seule mesure, alors que DRC ne se base que sur une seule mesure pour le sweet spot.

[Mahler]

Palm, il y a une petite écoute chez moi samedi, tu es invité si tu veux. Ca a un peu changé depuis ta visite (dac et placement)

[liolinux]

j'aurais une petite question sur audyssey quand je fais ma calibration sur mon 808 tout se passe nickel. Il me met même mon caisson à 75db et +8 niveau volume toutes mes enceintes sont en négatif vu leur rendement c'est normal. Après cette calibration, il me donne un volume de référence de 82 pour le THX .question très bête: est ce que c'est le niveau d'écoute d'un film avec le label THX ? Car franchement, il est quasiment impossible d'écouter à ce niveau sonore, JUSTE SI ON VEUT DEVENIR SOURD DE BONNE HEURE Pourrai-je avoir des détails sur ce niveau de référence? Je n'ai pas tout saisi. Merci d'avance.

[transeuropexpress]

Le volume de reference est en général TRES fort dans nos petites salles peu peuplées, je confirme, on écoute plus souvent à -8 DB ou moins par rapport à cette référence

a+

TEE

[liolinux]

Merci pour la réponse c'est bien ce que je pensais .je suis facile à -15 sur certain film et d'autre -10 .il faut combien de m2 pour pouvoir utiliser ce volume de 82. Toujours question bête je sais c'est juste pour avoir une idée de grosseur de salle.

[transeuropexpress]

Je pense sans me tromperque le niveau THX est calibré pour une salle de cinema de taille standart

a+

TEE

[JCGB]

Bonjour a vous tous , une petite question sur l'Audyssey DSX en mode "Hight" , voila avec mon 5008 j'aimerai rajouter 2 enceintes suplementaires "Hight" , mais ces enceintes doivent elles obligatoirement etre dans le meme plan que les voies avant , car vu la configuration de ma piece (salon + mezzanine et toiture en pente) je ne pourrais les positionner que devant les voies principales (environ 1.5m devant et 1.5m au dessus ) est ce que ce positionnement est possible ou bien est ce que les voies H doivent obligatoirement etre alignées avec les frontales ?.
Merci a ceux qui pourraient m'apporter plus de précisions .

[Macaille]

L'Audyssey doit pouvoir compenser plus facilement la distance que l'angle d'écoute.

Donc si tu peux garantir un angle d'écartement plus important aux "High" qu'à tes principales ça ne devrait pas poser de problème particulier.