Audyssey killing my HT sound?

[Matthew J Poes]

The key word is PREFERENCE. You don't train people on PREFERENCE.

If all they did was give a hearing test to make sure the subjects were not deaf, I can appreciate that.

But to TRAIN anyone in ANYWAY on how to listen and what to listen for and what to PREFER is to induce BIAS.

Even if they deny that they were training the subjects on what to prefer, that's what they were doing - training the subjects on what to prefer.

The world "Inception" comes to mind.

Why bother doing a DOUBLE-BLIND test when you already induced bias before the study?

People don't need to be told whether they LOVE the sound of something or love the taste of some foods.

I think you are taking my comments out of line here. I never said Harman trained on the preference study. That was a very specific study, small, and I don’t believe they did any training. I only said that sometimes they train people in their studies.

Your views on training aren’t in keeping with the science. When you want to test the audibility of a phenomena, good or bad, you have to break it down into small testable components in order to isolate extraneous variance (alternative explanations which are actual bias).

Let’s take one of my own current studies. I believe that group delay at low frequencies, of extreme enough, is audible, and that modern subwoofers introduce unacceptable levels of group delay. That is my hypothesis. It’s informed by measurements of 100’s of systems that showed this issue. How do I prove this is true?

Some might argue I should stick a group of people in a room, blinded to the conditions, and AB a system with and without this delay in the subs. The problem is the huge number of extraneous variables. For example, fixing the delay also changes the response since it optimized the phase. A sub can be perfectly in phase at any period interval and I’m arguing it needs to be aligned to less than one period, so my study has to separate the frequency specific time delay from the phase alignment.

First test then is to create test tracks with bandwidth limited (bell shaped) all pass filters. Then do an ABX over headphones. This prevents a response shift and only compares the audibility of the group delay. Because I can vary the all pass filter, I can establish a threshold.

That of course isn’t enough though. Next I need to establish not just audibility but what it sounds like. What if people prefer the group delay? Or what if the sound difference isn’t objectionable. So the next test is a MUSHRA test. I would stick with the same test tracks for this myself and see what happens. Here is where training comes in. Group delay distortion is probably subtle, especially at the lower limit. So while I want to know what it sounds like and how perceptible it is, I also want to account for the fact that some people will be more sensitive to it. Professionals who listen to it all the time. So I may train a group of listeners to more easily detect group delay. This helps me better detect the effect within my small sample since the number of people sensitive to this is probably naturally too small to randomly capture in my study. Does that bias things? Yes but not as you described. Even if I made people more sensitive to group delay an they more easily pick it up, that doesn’t cause them to prefer my product. There was not preference training. I was teaching them what group delay sounds like so they know what to look for. On top of that, the bias it does introduce (my estimate of the population effect is to large) is correctable by the non-trained group I also have.

After I do these tests I would switch to binaural versions of the test tracks. Why? Because group delay might be detectable on headphones where room effects don’t exist, but once a room is added back in, it’s possible that it is fully masked. If that’s true, then I want my study to be able to say, ok it is audible, here is the threshold, but there is a masking effect in rooms. By using binaural encoring, I can use a room transfer function with the modified group delay at low frequencies to simulate the effect of a room with minimal extraneous variables.

Only after all of that would I move to testing in a standard room. The standard room introduced bias but it’s also the most real world. The external validity of the test is highest, but internal validity is lowest. Because I already did the high internally valid test on headphones, as long as the results are consistent, we can ignore concern over bias. If the results aren’t consistent, then I need to find out the bias that contributed to that finding and design a new study.

My understanding is Harman follows a similar method. They don’t do studies to prove their products are best. They do pure science to understand phenomena. That is used to inform product development.

The preference curve that was tested came about quite differently from what is often described. A speaker that measures flat in an anechoic chamber with a specific but common directivity has a rise in the response. It roughly equals 1db per octave rise from 20khz to 20hz. That was the natural shape the speaker created. They then tested it that was preferable and designed the curve around that. The science that lead to that curve wasn’t about a preference curve and dates back to the 30’s. There is good reason that Sonarworks, B&K, Harman, and others all have very similar room curves. It’s not because people prefer or were trained to prefer boosted bass and rolled off treble. It’s because a “flat” speaker measures that way in room.

 

[AcuDefTechGuy]

I think you are taking my comments out of line here. I never said Harman trained on the preference study. That was a very specific study, small, and I don’t believe they did any training. I only said that sometimes they train people in their studies.

Your views on training aren’t in keeping with the science. When you want to test the audibility of a phenomena, good or bad, you have to break it down into small testable components in order to isolate extraneous variance (alternative explanations which are actual bias).

Let’s take one of my own current studies. I believe that group delay at low frequencies, of extreme enough, is audible, and that modern subwoofers introduce unacceptable levels of group delay. That is my hypothesis. It’s informed by measurements of 100’s of systems that showed this issue. How do I prove this is true?

Some might argue I should stick a group of people in a room, blinded to the conditions, and AB a system with and without this delay in the subs. The problem is the huge number of extraneous variables. For example, fixing the delay also changes the response since it optimized the phase. A sub can be perfectly in phase at any period interval and I’m arguing it needs to be aligned to less than one period, so my study has to separate the frequency specific time delay from the phase alignment.

First test then is to create test tracks with bandwidth limited (bell shaped) all pass filters. Then do an ABX over headphones. This prevents a response shift and only compares the audibility of the group delay. Because I can vary the all pass filter, I can establish a threshold.

That of course isn’t enough though. Next I need to establish not just audibility but what it sounds like. What if people prefer the group delay? Or what if the sound difference isn’t objectionable. So the next test is a MUSHRA test. I would stick with the same test tracks for this myself and see what happens. Here is where training comes in. Group delay distortion is probably subtle, especially at the lower limit. So while I want to know what it sounds like and how perceptible it is, I also want to account for the fact that some people will be more sensitive to it. Professionals who listen to it all the time. So I may train a group of listeners to more easily detect group delay. This helps me better detect the effect within my small sample since the number of people sensitive to this is probably naturally too small to randomly capture in my study. Does that bias things? Yes but not as you described. Even if I made people more sensitive to group delay an they more easily pick it up, that doesn’t cause them to prefer my product. There was not preference training. I was teaching them what group delay sounds like so they know what to look for. On top of that, the bias it does introduce (my estimate of the population effect is to large) is correctable by the non-trained group I also have.

After I do these tests I would switch to binaural versions of the test tracks. Why? Because group delay might be detectable on headphones where room effects don’t exist, but once a room is added back in, it’s possible that it is fully masked. If that’s true, then I want my study to be able to say, ok it is audible, here is the threshold, but there is a masking effect in rooms. By using binaural encoring, I can use a room transfer function with the modified group delay at low frequencies to simulate the effect of a room with minimal extraneous variables.

Only after all of that would I move to testing in a standard room. The standard room introduced bias but it’s also the most real world. The external validity of the test is highest, but internal validity is lowest. Because I already did the high internally valid test on headphones, as long as the results are consistent, we can ignore concern over bias. If the results aren’t consistent, then I need to find out the bias that contributed to that finding and design a new study.

My understanding is Harman follows a similar method. They don’t do studies to prove their products are best. They do pure science to understand phenomena. That is used to inform product development.

The preference curve that was tested came about quite differently from what is often described. A speaker that measures flat in an anechoic chamber with a specific but common directivity has a rise in the response. It roughly equals 1db per octave rise from 20khz to 20hz. That was the natural shape the speaker created. They then tested it that was preferable and designed the curve around that. The science that lead to that curve wasn’t about a preference curve and dates back to the 30’s. There is good reason that Sonarworks, B&K, Harman, and others all have very similar room curves. It’s not because people prefer or were trained to prefer boosted bass and rolled off treble. It’s because a “flat” speaker measures that way in room.

Personally I wouldn't call it "Pure Science".

We are talking about speakers and subwoofers, not anything truly important.

I do appreciate the "methodologies" and "studies".

And if any of the "preference" studies were done using randomized and untrained/unbiased people, that's great.

I just get turned-off reading about how Harman would like to "train" people on how to "critically" listen, which makes me wonder if they were doing that for all their "studies".

We can only speculate. But it does cast a few doubts about their "studies" and their "science".

BTW, I do prefer speakers with good measurements. Just don't tell me how I should be critically listening, watching, eating or drinking.

 

[Phase 2]

Thing I never liked about ARC is you can run it 10 times and get 10 different settings. Now before I get jumped, I only used the Mic the unit came with. I did use a tripod. Actually my Son and I wanted to see just how many different settings it would create. The speaker distances were always off never the same. The levels were off never the same. Most always ARC had the speaker levels trim down -4.5 on the surrounds or the Sub level -9, couple of times it said we had a speaker out of phase.

 

[lovinthehd]

Always good to see for yourself.

I've listened to a couple of systems using DIRAC and ARC. I didn't like any of them.

Bottom line again: some people like RC, some people don't.

Yet you like artificially high levels of bass...which is a form of RC isn't it?

 

[Matthew J Poes]

Personally I wouldn't call it "Pure Science".

We are talking about speakers and subwoofers, not anything truly important.

I do appreciate the "methodologies" and "studies".

And if any of the "preference" studies were done using randomized and untrained/unbiased people, that's great.

I just get turned-off reading about how Harman would like to "train" people on how to "critically" listen, which makes me wonder if they were doing that for all their "studies".

We can only speculate. But it does cast a few doubts about their "studies" and their "science".

BTW, I do prefer speakers with good measurements. Just don't tell me how I should be critically listening, watching, eating or drinking.

I think the training is a little different.

For example, I have been doing some work with spaciousness. It’s a function derived from the -L and -R components of a recording, predominantly at low frequencies. The effect of turning spaciousness on and off by canceling those components is really subtle and easily disrupted.

I recently had a conversation with Todd Welti, @gene and @shadyJ and Todd and I had each played around with David Griesinger’s work to setup listening tests. We both ran into problems. One of the trick I would use it I was to do this over is to train listeners to hear the spatial effects. The training increases their sensitivity to pick it up. It would t bias them toward a product or teach them how to listen to music at home. Purely teaching them what to listen for in this experiment.

 

[Hetfield]

Not to get involved in the "trained" people for listening tests but I have heard of other shady things in listening tests. Like hand picked people for the listening test or planting someone or some people in listening tests to steer people into thinking one way or another.


Sent from my Pixel using Tapatalk

 

[AcuDefTechGuy]

Yet you like artificially high levels of bass...which is a form of RC isn't it?

I should say "ARC" - Automatic Room Correction, since we are talking about Audyssey ARC.

I'm sure the room acoustic influences the bass levels. But I don't know if I would consider manually adjusting the subwoofer 20-80Hz region using the subwoofer Parametric EQ as "Room Correction".

 

[AcuDefTechGuy]

Not to get involved in the "trained" people for listening tests but I have heard of other shady things in listening tests. Like hand picked people for the listening test or planting someone or some people in listening tests to steer people into thinking one way or another.


Sent from my Pixel using Tapatalk

Good point. It's not like these "speaker studies" are being regulated by the FDA or other professional medical committees.

 

[Hetfield]

Good point. It's not like these "speaker studies" are being regulated by the FDA or other professional medical committees.

All I know is yesterday I loved the sound of my system and my speakers. I'm not gonna I'm out of the woods and everything sounds perfect because I need more time to evaluate but I was impressed yesterday. I'll do more listening this weekend. Dam the 5.1 music I listened to yesterday sounded incredible. Full range, wide open, no cut off or roll off. I love my speakers.

Sent from my Pixel using Tapatalk

 

[lovinthehd]

I should say "ARC" - Automatic Room Correction, since we are talking about Audyssey ARC.

I'm sure the room acoustic influences the bass levels. But I don't know if I would consider manually adjusting the subwoofer 20-80Hz region using the subwoofer Parametric EQ as "Room Correction".

Semantics? Schoeder frequency effects sure. Parametric EQ tho is a form of eq which is bad, right?

 

[Phase 2]

Not unless you make the adjustment manually than its not ARC. But you gotta start somewhere right?

 

[PENG]

I should say "ARC" - Automatic Room Correction, since we are talking about Audyssey ARC.

I'm sure the room acoustic influences the bass levels. But I don't know if I would consider manually adjusting the subwoofer 20-80Hz region using the subwoofer Parametric EQ as "Room Correction".

Let's agree for the time being that ARC, or REQ if some prefer, in fact crapped things up for you, and/or would hurt more than help in general. Now, recall that some of reasons cited so far for their bad behavior/effects might have been due to the "flattened" low end and high end bands, rendering the sound "thinner", or "harsher", or both.

Combined those suspects with your own experience in liking the effects of Audyssey but only with DEQ "On" (since you wanted DEQ at the time you had no choice but to leave Audyssey "on") and L/R "bypassed", do you then at least wonder if in fact you "prefer" the bumped up bass by DEQ? If yes, then I would further surmise that you are not necessarily a super bassoholic as such, but that you don't listen close enough to reference level, therefore missing the proper balance between the low, mid and high. In other words, if you were fine with listening at reference level at the time, you most likely could have been fine without DEQ.

If that's all or even partially agreeable, then the remaining puzzle is, was it Audyssey that turn music to crap, or it was simply a case of tonal balance, or imbalance, due to your preferred lower listening level? It could help clear things up a little more if you can confirm/remember which of the following sounded better to you:

1. Audyssey completely off.
2. Audyssey on, Left/Right bypassed, DEQ on.

If the answer is 1. then I would say your speaker's in-room response might have just happened to be able to compensate the tonal imbalance, in this case aka weak bass, by some bass bumps (aka room reinforcement) in the right place and with the right magnitude. I know I might have made things complicated or confusing to understand, and I hope @lovinthehd might understand the point I am trying to make too, that you probably do prefer accuracy/neutral than super exaggerated bass. Afterall, you are guy who own/owned at least two pairs (KEF Reference 201/2, Revel Salon2 of speakers that have the flattest looking FR curves I have ever seen published online.

 

[AcuDefTechGuy]

Semantics? Schoeder frequency effects sure. Parametric EQ tho is a form of eq which is bad, right?

No, since it’s only EQ 20-80Hz, it is 100% awesome.

It’s bad when you EQ above 200Hz.

 

[AcuDefTechGuy]

Let's agree for the time being that ARC, or REQ if some prefer, in fact crapped things up for you, and/or would hurt more than help in general. Now, recall that some of reasons cited so far for their bad behavior/effects might have been due to the "flattened" low end and high end bands, rendering the sound "thinner", or "harsher", or both.

Combined those suspects with your own experience in liking the effects of Audyssey but only with DEQ "On" (since you wanted DEQ at the time you had no choice but to leave Audyssey "on") and L/R "bypassed", do you then at least wonder if in fact you "prefer" the bumped up bass by DEQ? If yes, then I would further surmise that you are not necessarily a super bassoholic as such, but that you don't listen close enough to reference level, therefore missing the proper balance between the low, mid and high. In other words, if you were fine with listening at reference level at the time, you most likely could have been fine without DEQ.

If that's all or even partially agreeable, then the remaining puzzle is, was it Audyssey that turn music to crap, or it was simply a case of tonal balance, or imbalance, due to your preferred lower listening level? It could help clear things up a little more if you can confirm/remember which of the following sounded better to you:

1. Audyssey completely off.
2. Audyssey on, Left/Right bypassed, DEQ on.

If the answer is 1. then I would say your speaker's in-room response might have just happened to be able to compensate the tonal imbalance, in this case aka weak bass, by some bass bumps (aka room reinforcement) in the right place and with the right magnitude. I know I might have made things complicated or confusing to understand, and I hope @lovinthehd might understand the point I am trying to make too, that you probably do prefer accuracy/neutral than super exaggerated bass. Afterall, you are guy who own/owned at least two pairs (KEF Reference 201/2, Revel Salon2 of speakers that have the flattest looking FR curves I have ever seen published online.

Back when I was using the AVP-A1HDCI, my preferred setting was Audyssey Bypass + DEQ, which bypassed Audyssey for the Front Left and Right.

When listening to music in Stereo Mode, that means Audyssey is completely off.

What I liked about DEQ was the boosted Subwoofer Bass from around 20-80Hz.

The Subwoofer was boosted, not the Front Left and Right, which were set to Small and XO at 120Hz.

By manually boosting all the trim levels (except the subwoofer) by about 5dB, the master volume was usually set to -30dB and this produced very loud volume. Using the Galaxy digital SPL meter, the measured total volume was up to 110dB (C-weighted) many times when listening to 2.1 music.

Because the volume was at -30dB, DEQ boosted the subwoofer bass.

In my current setting (Yamaha) as of today, I use the PEQ for only the subwoofers and boost the ~ 20, 40, 60, and 80 Hz by 2dB. This produces about the same as what I was used to hearing from DEQ before.

 

[Phase 2]

All I know is yesterday I loved the sound of my system and my speakers. I'm not gonna I'm out of the woods and everything sounds perfect because I need more time to evaluate but I was impressed yesterday. I'll do more listening this weekend. Dam the 5.1 music I listened to yesterday sounded incredible. Full range, wide open, no cut off or roll off. I love my speakers.

Sent from my Pixel using Tapatalk

Man, just when I thought I had ARC figured out. After reading this Thread on ARC, now I'm even less assured and on top of that, Samsung just released Vertical HDTVs! Wtf!! It took me a long time to get use to the black bars on my horizontal 16.9. lol just messing around guys. But yeah don't know if any of you guys read that Samsung did just release vertical TVs..

 

[lovinthehd]

Man, just when I thought I had ARC figured out. After reading this Thread on ARC, now I'm even less assured and on top of that, Samsung just released Vertical HDTVs! Wtf!! It took me a long time to get use to the black bars on my horizontal 16.9. lol just messing around guys. But yeah don't know if any of you guys read that Samsung did just release vertical TVs..

ARC isn't a very good abbreviation for automatic room correction, as there is a specific automatic room correction algorithm called ARC (in Anthem gear, stands for Anthem Room Correction)

 

[TheWarrior]

But to TRAIN anyone in ANYWAY on how to listen and what to listen for and what to PREFER is to induce BIAS.

That's not true. The Harman How to Listen software trains you identify peaks and dips in an ever narrowing band of frequencies and can be used on any computer, with any speakers.

I don't mean to condemn, but I think more people could learn to listen, better, if they try to train their ears. And I highly recommend this program to anyone!

But also to give perspective, while some are still debating the effectiveness of Audyssey, (LOL!) Harman has created an AI system to address room/cabin induced problems as well as a sound cancelling system that effectively creates electronic sound proofing for any room/car. The car version seems to be more well published, as it uses the car's sensors to also anticipate bumps and microphones to fine tune the cancellation it emits through speakers - anti nodes combine with the anticipated outside noise and cancel each other out.

I am not sure which corrected response @Matthew J Poes is referring to with, "The Harman work did, in fact, show that the corrected response was preferred."

One important point is that the Harman systems also work with the available anechoic data for any of their loudspeakers - you cannot take full advantage of room curve data, without knowing what the speaker sends out, free of reflection.

But at the same time, an SPL meter is one of the simplest tools to balance the output of each speaker at each listening location. I think the consumer could get receiver's etc. a lot cheaper if we weren't stuck paying for that Audyssey sticker...

 

[AcuDefTechGuy]

The Harman How to Listen software trains you identify peaks and dips... I think more people could learn to listen, better...

I think some people take this HOBBY way too seriously.

It is a hobby and the only purpose is entertainment.

Again, it's not really THAT important. It's not engineering a bridge or vehicle and sure as heck isn't surgery or medicine.

I'm sure some people think you should learn how to eat, drink, smell, look, and do all kinds of things better too.

I don't need anyone telling me how to enjoy my music, movies, foods, or drinks.

I'm sure Harman would love for everyone to say, "Oh, I shouldn't love the sound of my B&W, Paradigm, GoldenEar speakers because I can hear better and can identify the peaks around 10kHz".


BTW, I personally wouldn't buy a $20K-$30K speaker that has a 5-6dB peak around 10kHz either. But I'm playing devil's advocate.

 

[AcuDefTechGuy]

All I know is yesterday I loved the sound of my system and my speakers. I'm not gonna I'm out of the woods and everything sounds perfect because I need more time to evaluate but I was impressed yesterday. I'll do more listening this weekend. Dam the 5.1 music I listened to yesterday sounded incredible. Full range, wide open, no cut off or roll off. I love my speakers.

Sent from my Pixel using Tapatalk

The important thing is that you know EXACTLY what sounds awesome to you and you don't need anyone telling you what is the definition of "amazing great sound".

What? You didn't hear those dips and peaks around 3kHz and 10kHz?

 

[AcuDefTechGuy]

ARC isn't a very good abbreviation for automatic room correction, as there is a specific automatic room correction algorithm called ARC (in Anthem gear, stands for Anthem Room Correction)

Well, maybe @Phase 2 means that both ARC's aren't his cup of tea.