Some New Evidence that Generation Y May Prefer Accurate Sound Reproduction

[tonmeister]

I posted a new article that summarizes some recent experimental evidence where I tested a group of teenagers' preferences in loudspeakers and MP3 (128 kbps) versus CD music formats. This is just the beginning of a more thorough longer study, so the results are very preliminary. Still I thought it would be interesting to get some feedback.

I could find no evidence that these high school students preferred the "sizzling sounds of MP3" over higher quality lossless formats, as reported by Jonathan Berger. I also found they preferred the most accurate, neutral loudspeakers when given the opportunity to hear and compare them with something less accurate and neutral.

These results are not too surprising to me, but the media seems to have been reporting a different story over the past year.

 

[AJinFLA]

I also found they preferred the most accurate, neutral loudspeakers when given the opportunity to hear and compare them with something less accurate and neutral.

Yes, but how often will this happen in real life?

Is standard listening on the shuffler always done in stereo?
Can't seem to open the MCH Listening Lab link in your article. Are the front and side walls "treated"?
TIA

cheers,

AJ

p.s. any word on the listener training software release?

 

[GO-NAD!]

I posted a new article that summarizes some recent experimental evidence where I tested a group of teenagers' preferences in loudspeakers and MP3 (128 kbps) versus CD music formats. This is just the beginning of a more thorough longer study, so the results are very preliminary. Still I thought it would be interesting to get some feedback.

I could find no evidence that these high school students preferred the "sizzling sounds of MP3" over higher quality lossless formats, as reported by Jonathan Berger. I also found they preferred the most accurate, neutral loudspeakers when given the opportunity to hear and compare them with something less accurate and neutral.

These results are not too surprising to me, but the media seems to have been reporting a different story over the past year.

Why didn't I get to go on field trips like that?! Oh, wait a minute, I remember why. It's because I grew up in a Newfoundland outport - thousands of miles from any audio equipment manufacturer.

Great read though. Did you chat with the kids after and were they surprised by their format preferences?

In such a setting, I can understand the results. However, when they go back to the their regular lives, are they going to be just as discriminating about audio quality? Will they be willing to reduce the number of songs they have in their media players by 50%-75%, in order to listen to lossless recordings?

Am I correct to assume that this study did not directly address the issue of dynamic compression in new and re-mastered recordings?

 

[lsiberian]

A one of the old members of Generation Y I know there are plenty of folks that enjoy better audio in my generation.

 

[tonmeister]

Why didn't I get to go on field trips like that?! Oh, wait a minute, I remember why. It's because I grew up in a Newfoundland outport - thousands of miles from any audio equipment manufacturer.

Great read though. Did you chat with the kids after and were they surprised by their format preferences?

In such a setting, I can understand the results. However, when they go back to the their regular lives, are they going to be just as discriminating about audio quality? Will they be willing to reduce the number of songs they have in their media players by 50%-75%, in order to listen to lossless recordings?

Am I correct to assume that this study did not directly address the issue of dynamic compression in new and re-mastered recordings?

Our Chief HR Officer is from Newfoundland. I'm still waiting for the memo that we can drink beer on the job and get extended unemployment benefits when fishing season is over

The kids got to see what speakers they preferred immediately after the test, but the analysis of the MP3/CD results were not done before they left. I have sent them the results since then.

We still need to do some correlation analysis between their listening results and the survey results. Part of that survey information included what formats they currently listen to, etc.

I don't know if they will change their listening habits but at least they are more aware of the sound quality differences among loudspeakers (we need to repeat this with headphones) and music formats. Knowledge can change people's behavior.

We didn't address the issue of dynamic range in recordings in this study. When I was screening recordings for this study, I found that many of the modern day dynamically compressed recordings (e.g. Coldplay) were poor at revealing MP3 artifacts - plus they gave me a headache. So the compressed dynamics of today's music may in fact be encouraging kids to use low quality MP3 settings ?

 

[tonmeister]

Yes, but how often will this happen in real life?

Is standard listening on the shuffler always done in stereo?
Can't seem to open the MCH Listening Lab link in your article. Are the front and side walls "treated"?
TIA

cheers,

AJ

p.s. any word on the listener training software release?

No, the standard listening on the MLL shuffler is mono because listeners are more discriminating of loudspeaker differences in mono versus stereo or surround. We recently wrote a paper on this topic which I hope to cover in an upcoming blog.

There is very little treatment of the walls in the MLL,. There is carpet, lots of chairs, and diffraction from shelves, furniture,etc. The side walls are too far away for my tastes, and we are planning to change this in the next year.

We recently lost 2 software programers in our group so the listening test software release got delayed by 1-2 months.

 

[AJinFLA]

No, the standard listening on the MLL shuffler is mono because listeners are more discriminating of loudspeaker differences in mono versus stereo or surround. We recently wrote a paper on this topic which I hope to cover in an upcoming blog.

I misread this late last night-

The relative distances in ratings between pairs of loudspeakers indicated the magnitude of preference

Thanks for clarifying that all the loudspeaker tests are in mono. I look forward to the explanation in your blog. Seems impossible to me, unless you are talking about very specific areas, such as on axis linearity, diffraction, enclosure/baffle resonances, driver linearity/compression effects, etc.
I would think placing a single speaker directly in front of the head/ears, well away from side walls, etc. would create a very different direct/reverberant field and perception, than a typically positioned stereo pair.

There is very little treatment of the walls in the MLL,. There is carpet, lots of chairs, and diffraction from shelves, furniture,etc. The side walls are too far away for my tastes, and we are planning to change this in the next year.

It would be interesting to see how the 362 (best) compared perceptually to the ML (worst), in stereo, when positioned in a typically sized/furnished living room/seated listener.

We recently lost 2 software programers in our group so the listening test software release got delayed by 1-2 months.

Sorry to hear that. Will stay tuned.
Thanks again for sharing your work with us.

cheers,

AJ

 

[DukeL]

Sean, apparently one of the speakers used in the audition was a hybrid electrostat that cost a lot more than the others, and it came off poorly in the comparisons.

Two characteristics of these hybrid electrostats could potentially work against them if not taken into account in the initial setup:

The first and most obvious idiosyncracy is their unusually narrow radiation pattern, which calls for aiming the panels at the center of listening area, and even then the tonal "sweet spot" is quite small (the "imaging sweet spot" is probably a lot less applicable in the context of the evaluation given that a monophonic signal was used). Thus listeners outside the unusually small "sweet spot" would not hear the correct tonal balance.

The second, more subtle idiosyncracy is that the tonal balance will change with distance, this because the panel approximates a line source and the woofer operates as a point source. Now I know the panel's approximation of a line source is imperfect, but the SPL will still fall off at different rates with distance. If the listening distance is beyond the normal distances the speaker is "voiced" for, the tonal balance will be off because the panels will be louder than the woofers (this based on in-room measurements I've made of a similar system).

One other possible issue: If the wall behind the speakers is absorptive, it would pretty much eliminate the contribution of the backwave energy from the panels, and the backwave contribution is usually beneficial for a dipole speaker. I don't know whether or not that wall is absorptive in your evaluation room.

So my question is, based on your first-hand knowledge of the room and how the speakers and listeners were positioned, do you think the hybrid electrostats could have been disadvantaged by the setup conditions?

Thanks,

Duke LeJeune
AudioKinesis

 

[tonmeister]

Sean, apparently one of the speakers used in the audition was a hybrid electrostat that cost a lot more than the others, and it came off poorly in the comparisons.

Two characteristics of these hybrid electrostats could potentially work against them if not taken into account in the initial setup:

The first and most obvious idiosyncracy is their unusually narrow radiation pattern, which calls for aiming the panels at the center of listening area, and even then the tonal "sweet spot" is quite small (the "imaging sweet spot" is probably a lot less applicable in the context of the evaluation given that a monophonic signal was used). Thus listeners outside the unusually small "sweet spot" would not hear the correct tonal balance.

The second, more subtle idiosyncracy is that the tonal balance will change with distance, this because the panel approximates a line source and the woofer operates as a point source. Now I know the panel's approximation of a line source is imperfect, but the SPL will still fall off at different rates with distance. If the listening distance is beyond the normal distances the speaker is "voiced" for, the tonal balance will be off because the panels will be louder than the woofers (this based on in-room measurements I've made of a similar system).

One other possible issue: If the wall behind the speakers is absorptive, it would pretty much eliminate the contribution of the backwave energy from the panels, and the backwave contribution is usually beneficial for a dipole speaker. I don't know whether or not that wall is absorptive in your evaluation room.

So my question is, based on your first-hand knowledge of the room and how the speakers and listeners were positioned, do you think the hybrid electrostats could have been disadvantaged by the setup conditions?

Thanks,

Duke LeJeune
AudioKinesis

Hi Duke,

Yes, we were aware that these electrostatic panels are directional and have a small listening sweet spot, as revealed in our anechoic measurements. You can clearly hear the timbre of the speaker change (it gets duller) when moving your head a few inches, or walking past it. In my opinion, that an undesirable characteristic of a loudspeaker: An ideal loudspeaker should sound good in any room, in any position, and in any seat. Of course, not many speakers do all these things, but this one is particularly poor in that regard.

The problem is that this speaker will never sound good, no matter where you put it. No loudspeaker, listening room or listening position can change the fact that the speaker has an extremely poor octave-octave balance and multiple resonances visible in the on and off-axis measurements (look at the measurements of speaker C in slide 28 ).

To answer your specific questions:

1. The speakers were located about 3 feet from the back wall (which is reflective), near the center of the listening room, pointed at the primary listening seat, far away from the side walls. This is the optimal position according to the owners' manuall on page 10.

2. The speakers were evaluated using both high school and trained listeners. The high school students were tested as two groups of 9 each sitting in two rows around the primary seat. More of these students were sitting in non-ideal off-axis positions than the trained listeners who all sat in the primary seat or "sweet spot" in separate sessions. Yet the trained listeners were more harsh about the colorations and imbalances of this speaker than were the high school students - even though they were supposedly hearing the optimal sound from the speaker.


So to conclude, I don't think the electrostatic speaker was disadvantaged in any way. According to the owners manual we set it up according to instructions (which are frankly not every well written, vague and contradictory. It's almost like they are saying put it wherever you think it sounds best, which is not very helpful).

The electrostatic loudspeaker didn't do well because it is technically and sonically inferior to the other loudspeakers in the test. The subjective and objective measurements support this statement, as does 30+ years of loudspeaker research at Canada's National Research Council, and more recently at Harman International.

Sorry, you can't get blood from a stone.

 

[PENG]

A couple of times I A/B/C/D/E the $5,000 (Future Shop listed price in CAD) ML Vista Xstat, the $2,500 Energy RC-70, Klipsch RF82, Polk Audio RTiA9 and found the RC-70 sound much better than the Martin Logan. I thought may be the ML was at disadvantage due to placements and the not powerful enough Yamaha HTR-6295 that Future used to do everything. Now I guess may be not if even the $450 a pair P362 can do better than the ML Vista. Very hard to understand how this can be true though considering the huge price difference.

 

[DukeL]

Sean, thank you for taking the time to respond.

Regarding the frequency response curve of the hybid electrostat, it looks to me like the panel is a good 5 dB louder than the woofer at the measurement distance. Given their respective radiation characteristics, at a more typical listening distance the discrepancy would likely increase. If the panels had a level control, or even an adjustable bias, I think the speaker's in-room tonal balance could be significantly improved.

For the record, I believe that the radiation pattern matters a lot. Too bad the electrostat's frequency response was so far off; it might have been interesting to see a comparison where the on-axis responses were pretty similar, such that the main differences were in what's happening off-axis. But, I digress, as that wasn't even remotely the point of your study!

Back on topic, your findings certainly are encouraging for those of us who hope to one day persuade the MP-3 generation that music well-reproduced works its magic all the better.

Duke

 

[mtrycrafts]

...

These results are not too surprising to me, but the media seems to have been reporting a different story over the past year.

You are a researcher looking for evidence based facts. The media is not interested in this.

 

[AJinFLA]

Yes, we were aware that these electrostatic panels are directional and have a small listening sweet spot, as revealed in our anechoic measurements. You can clearly hear the timbre of the speaker change (it gets duller) when moving your head a few inches, or walking past it. In my opinion, that an undesirable characteristic of a loudspeaker: An ideal loudspeaker should sound good in any room, in any position, and in any seat. Of course, not many speakers do all these things, but this one is particularly poor in that regard.

Hi Sean, I'm not trying to defend the ML, but I think the scenario you are describing is entirely unrealistic as to how they would be used in the real world. If they were listened to in stereo, it would be at a position between the 2 speakers, most likely by one person. If it is an HT scenario, there may be a similar to your test larger listener area, but their would also be a center channel within the multi-speaker frontal soundfield and a different subtended angle, etc.

The problem is that this speaker will never sound good, no matter where you put it. No loudspeaker, listening room or listening position can change the fact that the speaker has an extremely poor octave-octave balance and multiple resonances visible in the on and off-axis measurements (look at the measurements of speaker C in slide 28 ).

I don't see how you can conclude that with the way you tested it. First of all, the FR shown is anechoic, not what you measured in your listening room at the listener positions (spatial averaging the seating area would seem more reasonable to me). The speaker is a dipole that transitions to a monopole (450hz xo ??), with a cardioid like pattern between. When you place it in a real room, even with semi-lossy walls and openings, there will still be significant gain at LF as it radiates into 1/4-1/8 space. No excuses of course for the resonances, but I thought you should add some caveats to the measured spectral imbalance. I would reiterate that listening to them vs the 362's, in stereo, in a typical sized living room space/decor, seems fairer before making such statements as they will never sound good.

1. The speakers were located about 3 feet from the back wall (which is reflective), near the center of the listening room, pointed at the primary listening seat, far away from the side walls. This is the optimal position according to the owners' manuall on page 10.
2. The speakers were evaluated using both high school and trained listeners. The high school students were tested as two groups of 9 each sitting in two rows around the primary seat. More of these students were sitting in non-ideal off-axis positions than the trained listeners who all sat in the primary seat or "sweet spot" in separate sessions. Yet the trained listeners were more harsh about the colorations and imbalances of this speaker than were the high school students - even though they were supposedly hearing the optimal sound from the speaker.
So to conclude, I don't think the electrostatic speaker was disadvantaged in any way. According to the owners manual we set it up according to instructions (which are frankly not every well written, vague and contradictory. It's almost like they are saying put it wherever you think it sounds best, which is not very helpful).
The electrostatic loudspeaker didn't do well because it is technically and sonically inferior to the other loudspeakers in the test. The subjective and objective measurements support this statement, as does 30+ years of loudspeaker research at Canada's National Research Council, and more recently at Harman International.
Sorry, you can't get blood from a stone.

I would have to disagree. I would personally never own such a speaker (for very different reasons), but placed in stereo, somewhere around the front corners of a living room, where the bass loading, modal coupling and dipole nulls directed at the sidewalls (unlike your 362), etc, etc. would generate an entirely different soundfield at the listener position than was experienced in this test, hardly seems fair to me at all.

Regarding the frequency response curve of the hybid electrostat, it looks to me like the panel is a good 5 dB louder than the woofer at the measurement distance. Given their respective radiation characteristics, at a more typical listening distance the discrepancy would likely increase.

Or depending on placement in a real room and listener seating distance, could increase.
As noted previously, I would much rather have seen some spatially averaged FR's at the seating area....with the speakers 3' (minimum) from the front wall....and not 10' away from the side wall.

cheers,

AJ

 

[tonmeister]

Sean, thank you for taking the time to respond.

Regarding the frequency response curve of the hybid electrostat, it looks to me like the panel is a good 5 dB louder than the woofer at the measurement distance. Given their respective radiation characteristics, at a more typical listening distance the discrepancy would likely increase. If the panels had a level control, or even an adjustable bias, I think the speaker's in-room tonal balance could be significantly improved.

For the record, I believe that the radiation pattern matters a lot. Too bad the electrostat's frequency response was so far off; it might have been interesting to see a comparison where the on-axis responses were pretty similar, such that the main differences were in what's happening off-axis. But, I digress, as that wasn't even remotely the point of your study!

Back on topic, your findings certainly are encouraging for those of us who hope to one day persuade the MP-3 generation that music well-reproduced works its magic all the better.

Duke

Duke,
The measurement distance in our chamber is 2 meters. The listening distance in the room is maybe 3 to 3,5 meters, so you are probably right.

I think a uniform radiation pattern is certainly important. In the past, for stereo playback the speakers with the widest dispersion had a tendency to be favored in our tests, probably because they created stronger lateral reflections which produces greater apparent source width (ASW) and spaciousness. With multichannel surround, the side and rear speakers effectively do the same thing, if the recording producer/engineer or up-mixer are doing their job correctly.

Most kids listen to Ipods with headphones or stereo docking stations, They probably won't experience ASW, Envelopment and spaciousness for another decade unless they go to concert halls. For now, it's good that they experience uncompressed audio with a full range speaker. Baby steps.

 

[tonmeister]

Sean, thank you for taking the time to respond.

Regarding the frequency response curve of the hybid electrostat, it looks to me like the panel is a good 5 dB louder than the woofer at the measurement distance. Given their respective radiation characteristics, at a more typical listening distance the discrepancy would likely increase. If the panels had a level control, or even an adjustable bias, I think the speaker's in-room tonal balance could be significantly improved.

For the record, I believe that the radiation pattern matters a lot. Too bad the electrostat's frequency response was so far off; it might have been interesting to see a comparison where the on-axis responses were pretty similar, such that the main differences were in what's happening off-axis. But, I digress, as that wasn't even remotely the point of your study!

Back on topic, your findings certainly are encouraging for those of us who hope to one day persuade the MP-3 generation that music well-reproduced works its magic all the better.

Duke

Hi Duke,

Here is another study we did that involved over 300 listeners (including some audio journalists, audio retailers, trained listeners and college students) and included the $11k electrostatic model from the same manufacturer (speaker M); again, note the correlations between the anechoic measurements and the listener preference ratings. Listeners like accurate, neutral loudspeakers, which can be easily determined via comprehensive anechoic measurements

So the listening results in the current test are not accidental, but very repeatable. When you have a speaker (M or D) with that measured response, it will simply not do well against more accurate models.

 

[Loren42]

I posted a new article that summarizes some recent experimental evidence where I tested a group of teenagers' preferences in loudspeakers and MP3 (128 kbps) versus CD music formats. This is just the beginning of a more thorough longer study, so the results are very preliminary. Still I thought it would be interesting to get some feedback.

I could find no evidence that these high school students preferred the "sizzling sounds of MP3" over higher quality lossless formats, as reported by Jonathan Berger. I also found they preferred the most accurate, neutral loudspeakers when given the opportunity to hear and compare them with something less accurate and neutral.

These results are not too surprising to me, but the media seems to have been reporting a different story over the past year.

I agree with the analysis... However, the proof is in what they buy.

They may agree that lossless quality is better, but if their actual buying preference says otherwise, then as a manufacture I would follow where the market goes (or the profit).

For a meaningful study (from a market perspective) you have to examine what they buy, not what they tell you.

I also can agree with the conclusion in that better education can steer the buying decision.

 

[GO-NAD!]

Our Chief HR Officer is from Newfoundland. I'm still waiting for the memo that we can drink beer on the job and get extended unemployment benefits when fishing season is over

The kids got to see what speakers they preferred immediately after the test, but the analysis of the MP3/CD results were not done before they left. I have sent them the results since then.

We still need to do some correlation analysis between their listening results and the survey results. Part of that survey information included what formats they currently listen to, etc.

I don't know if they will change their listening habits but at least they are more aware of the sound quality differences among loudspeakers (we need to repeat this with headphones) and music formats. Knowledge can change people's behavior.

We didn't address the issue of dynamic range in recordings in this study. When I was screening recordings for this study, I found that many of the modern day dynamically compressed recordings (e.g. Coldplay) were poor at revealing MP3 artifacts - plus they gave me a headache. So the compressed dynamics of today's music may in fact be encouraging kids to use low quality MP3 settings ?

You're out of touch with the times buddy - Newfoundland is hard at work in the offshore oilfields so that Ontario can be kept in the lifestyle to which they have become accustomed!

I was wondering how the combination of dynamic compression and 128 kbps MP3s compounded the degradation of a recording. I only listen to CDs, so I don't have any means to investigate that myself.

Thanks for the info!

 

[PENG]

Here is another study we did that involved over 300 listeners (including some audio journalists, audio retailers, trained listeners and college students) and included the $11k electrostatic model from the same manufacturer (speaker M); again, note the correlations between the anechoic measurements and the listener preference ratings. Listeners like accurate, neutral loudspeakers, which can be easily determined via comprehensive anechoic measurements

So the listening results in the current test are not accidental, but very repeatable. When you have a speaker (M or D) with that measured response, it will simply not do well against more accurate models.

Dr. Olive, I would like to ask some off topic questions since people here often talked about how their speakers came alive with more power. In all those studies you linked, how did you take into account of the way different amplifiers would interact with different speakers? Did you just make sure you have more than enough power headroom for the speakers that were the hardest to drive, or you did something else as well? In terms of amplifier power, how much headroom would you have allowed for?

I figure that if the tests are to be conducted at an average SPL of 78 dB from 3 meters, then the sensivity of a speaker with say 87.5dB/W at 1m should produce just about the targeted 78 dB of SPL assuming no room reinforement effect. So if you allow for 30 dB of headroom power (or 15 dB for current), you would need to have 1024W (MONO) of peak power capability on hand, or less depending on how much room reinforcement you get in the non anechoic chamber where the listening tests are conducted. If you only allow for 20 dB of peak power headroom then any 200W mono block should be fine right?

 

[tonmeister]

You're out of touch with the times buddy - Newfoundland is hard at work in the offshore oilfields so that Ontario can be kept in the lifestyle to which they have become accustomed!

I was wondering how the combination of dynamic compression and 128 kbps MP3s compounded the degradation of a recording. I only listen to CDs, so I don't have any means to investigate that myself.

Thanks for the info!

I guess I am out of touch Since I moved from Canada to California I don't hear much about Newfoundland anymore.

I haven't done any systematic investigation of the interaction between dynamic compression and how that affects the audibility of MP3 artifacts. My sense is that material that is dynamically compressed makes MP3 artifacts more difficult to hear, but I have no proof to support that yet.

An other area that is affected by MP3 is the apparent sound quality of up-mixers. At 128 kbps you can hear more spatial steering artifacts in up-mixers because the inter-channel phase differences on which the up-mixers operate are corrupted somewhat by the MP3 encoding process. So the audibility of MP3 may be more audible and offensive when listened through 7 speakers (via up-mixer) versus 2 channels. This is a research area that I'd like to explore.

 

[AJinFLA]

Hi Duke,

Here is another study we did that involved over 300 listeners (including some audio journalists, audio retailers, trained listeners and college students) and included the $11k electrostatic model from the same manufacturer (speaker M); again, note the correlations between the anechoic measurements and the listener preference ratings. Listeners like accurate, neutral loudspeakers, which can be easily determined via comprehensive anechoic measurements

So the listening results in the current test are not accidental, but very repeatable. When you have a speaker (M or D) with that measured response, it will simply not do well against more accurate models.

The (repeated) test is valid for detecting non-linearities in a single loudspeaker as tested.
It is not valid for concluding that speaker A will be preferred speaker B, listened to in stereo (or multi-ch), positioned in a real room, even using the exact same panel of listeners.
And that is how they will be used in the real world. The same way you had listeners testing the codec.