David, are you sure you're not hearing standing waves? Thiele and others ran the tests in the chambers and the results are what they are. The only discrepency with your personal test results I can see are standing waves. Can you think of any other possible reasons?
Edit: it may also be source signal. Perhaps you were using sines?
Dave
Last edited:
No, this is definitely with pink noise. I frequently use this in a new set up to confirm the speaker distances are both equal (or that I am sitting in the center of the couch). Pink noise (mono) to both speakers and shift back and forth. If you turn sideways and block your far ear then it is very evident where the exact center is, from the treble coming and going as the comb filter moves up and out of range. At the same time I distinctly remember that when facing the speakers you will hear the individual comb filtering in both ears. There is no spot that gets rid of it because it will be in one ear or the other. I can even hear it on program if there is a centered element with strong high end.
Try it and you will hear, assuming your room isn't too lively and speakers too distant.
David S.
Hello,
I agree. Furthermore I believe it depends on the individual ability to process the spatial cues when some of them are conflicting, as they are in +/-30deg stereo pair.
Mostly what I hear in stereo, if there is wideband high freq sounds, is two separated sources at the sides indicating the speaker locations. Pinna power 🙂 High directivity loudspeakers at the high freqs are the worst.
It does not have to be like this, one may even enjoy stable phantom images at high freqs, especially if one has a weak pinna 🙂
I also agree. The auditory organs can be identical, even the nerve cells can fire identically, but there can be great differences in perception at the higher level of abstraction.
- Elias
I agree. Furthermore I believe it depends on the individual ability to process the spatial cues when some of them are conflicting, as they are in +/-30deg stereo pair.
Mostly what I hear in stereo, if there is wideband high freq sounds, is two separated sources at the sides indicating the speaker locations. Pinna power 🙂 High directivity loudspeakers at the high freqs are the worst.
It does not have to be like this, one may even enjoy stable phantom images at high freqs, especially if one has a weak pinna 🙂
I also agree. The auditory organs can be identical, even the nerve cells can fire identically, but there can be great differences in perception at the higher level of abstraction.
- Elias
Are we back to a superior hearing argument or rather "differences in perception at the higher level of abstraction"? 😡
Oye,
Dan
Oye,
Dan
So your saying the correlation I found was pure coincidence? I don't think so.
Originally Posted by bentoronto
An interesting post, thank you.
But you've made an elementary error of logic: being able to tell things about (small variance) has ZERO to do with validity in ranking them.
.
So your saying the correlation I found was pure coincidence? I don't think so.
----------------------
The conversation is at cross-purposes.
In measurement (in this case psychophysical measurement), the term "validity" has a technical meaning (actually there are various forms of validity). The consequence is that the concepts of correlation and validity may have little to do with one another.
I beleive BenT was using the term "validity" in the technical sense.
An interesting post, thank you.
But you've made an elementary error of logic: being able to tell things about (small variance) has ZERO to do with validity in ranking them.
.
So your saying the correlation I found was pure coincidence? I don't think so.
----------------------
The conversation is at cross-purposes.
In measurement (in this case psychophysical measurement), the term "validity" has a technical meaning (actually there are various forms of validity). The consequence is that the concepts of correlation and validity may have little to do with one another.
I beleive BenT was using the term "validity" in the technical sense.
His gang looked into this: The Variability of Loudspeaker Sound Quality Among Four Domestic-Sized Rooms By Olive, Schuck, Sally and Bonneville. They found that rank ordering of speakers didn't change across 4 different rooms.
I'm not sure how early reflections could "fix" axial response errors. The reflections can't add in a power way and would always add their own comb filtered profile, due to delay. If you found a combination that added up flatter than either individually, wouldn't it only work at a particular spot in space?
David S.
Hello,
Yes and it's mainly because of human ability of adaptation of room acoustics. In a relatively short time human adapts himself to the local acoustic environment, it allows him to extract the most relevant information from the soundfield even the space is previously unknown for him.
More interesting is that apparently human cannot adapt to the loudspeaker sound to the same extent as to the room!
But some loudspeaker adaptation certainly takes place, if we think about 'breaking in' of speakers they allways improve but never get worse 😀
There are a lot of things to the room reflections in terms of perception (not so much of measurements).
For example the loudness perception is formed by integration, as presented in this simplistic graph (but it shows the idea):
Also, comb filtering may not occur that much in real life (despite measurements would indicate so), because most music is no steady state and consists of bursts of short durations.
- Elias
Yes and it's mainly because of human ability of adaptation of room acoustics. In a relatively short time human adapts himself to the local acoustic environment, it allows him to extract the most relevant information from the soundfield even the space is previously unknown for him.
More interesting is that apparently human cannot adapt to the loudspeaker sound to the same extent as to the room!
But some loudspeaker adaptation certainly takes place, if we think about 'breaking in' of speakers they allways improve but never get worse 😀
There are a lot of things to the room reflections in terms of perception (not so much of measurements).
For example the loudness perception is formed by integration, as presented in this simplistic graph (but it shows the idea):
An externally hosted image should be here but it was not working when we last tested it.
Also, comb filtering may not occur that much in real life (despite measurements would indicate so), because most music is no steady state and consists of bursts of short durations.
- Elias
Do you happen to have this paper? It'd be a good read. The main point I take from it from what I have gleaned off the web is that with the speakers and test conditions here, none of the speaker's tonal errors was "fixed" by a room to the point that it was preffered over another that bettered it in a different room.
However, Toole even acknowledges in his book that the side wall reflection response balance will impact tonality. Also, it seems at odds with much practical experience that some speakers just do better in some rooms than others. Were the speakers in the study relatively well controlled and similar in dispersion? I would expect that the more different the radiation patterns and the greater the difference in change in score from room to room.
Toole, in his white paper for Harmon "Audio, Art Science" supports this common sense perspective:
"The complex interactions between room boundaries and speaker directivity at middle and high frequencies, and speaker and listener position at low frequencies are powerful influences in what we hear. With careful acoustical design and electronic signal manipulations, we are finding ways to make speakers more “friendly” to rooms, both in recording studios and in homes."
If the early reflections didn't affect tonality, why bother worrying about them?
Again from Toole: "These will be convincing only if the loudspeakers can deliver the appropriate sounds to the listeners’ ears. If this “spatial dynamic range”, as I call it, is to be achieved in normal – i.e. not acoustically treated – listening rooms, we will very likely need loudspeakers of differing directivities in different locations and, if we are truly fussy, we will need more than five channels."
I've been harassed on other boards for stating that differing room treatment affects the off axis targets, but there it is from the patron saint of objectivists directly.
David, not saying this applies to you at all, but I see Toole's name thrown out with abandon by people who misrepresent his positions, and those of the company that paid for much of his research, in their pursuit of a simple cookie cutter objective theory they can wrap their head around. It just isn't that simple.
Dave
They used an interesting technique of binaural recordings of 3 speakers in 4 rooms. They found that a listener would rank order the rooms and also rank order the speakers, but neither variable changeed the preferences of the other. In other words a good speaker was good whether the room was good or bad. They were open to the possibility that a bad room perfectly complemented the LF response of a bad speaker, but didn't find it in the limited number of variables.
I've noticed this as a change in Toole's position over time. I think the data in his early paper leaned heavily towards axial response being key and power response having little effect. The data of the 85 paper clear shows that, even if he waffles a little in his conclusions regarding power response. His strongest statement about power response is that it is useful for revealing resonances, but he tends to fall back into: "If a speaker measures well in every regard then it will be well ranked by listeners." Hardly a surprising conclusion.
Later he seemed to get into the in-room simulations that seperately added the direct energy, the early reflection energy and the later reverberent energy. Using Harman computer power he was able to create good simulations of in-room response, but he never makes a convincing arguement that the end result is strongly indicative of performance. I asked him about this once and he said "read the book". (I have.)
I'll concede that fairly early off axis reflections will change a speakers sound balance. What I'd debate is how likely it is that a poor on-axis response would be fixed by a particular off axis response. What I take away from relevant papers is that, in evaluating speakers, we perceive the direct sound at high frequencies, a fairly short time window at mid frequencies (enough to catch a floor bounce or close sidewall), and the steady state response at LF. If you measured a system response in a room with 3 corresponding time windows, it is unlikely (not impossible) that early reflections would take a poor direct response and make it good.
I see the same thing. when I was at PSB, Paul Barton frequently quoted Floyd. I ended up re-reading the 85 paper at least twice a year to keep in mind what he really said.
David S.
But they found that too. When they used binaural recordings they could switch the speaker listened too, or choose one speaker and instantly switch the room it was played (recorded) in. Then they found that the room was a strong variable (you could rank order good to bad) and that your preferences would hold whether it was a prefered speaker or less prefered speaker being played.
Definitely you adapted to whatever variable was being held stationary.
Indeed!
I haven't found this to be the case. Any audible effects I've heard on pink noise (steady state), I've eventually heard on music. As discussed previously, it helps if the music is fairly "busy" with a full spectrum.
David S.
Hi,
To emphasize the importance of 'power response' we can take a normal small bookshelf speaker and turn it around facing backwards to the listener. Now if we measure the on axis response it is extremely poor.
But then place the speaker (still facing backwards to the listener) in a room with high reflective boundaries, and if we listen at relatively long distance it can sound very good! It's a simple test one can do.
This is to say that for a speaker with little or no directivity the power response dominates the sound in a room.
On the other hand in my experience the speakers with very high directivity, dipole line arrays for example, the direct sound can mostly define the tonal balance. If we integrate few tens of ms and include first room reflections, the energy of the direct sound can play a major role of the total energy within the integration window.
I would even claim that for a speakers with very high directivity the off axis response is not that important. But the directivity must be very high, just a normal dipole may not be high enough.
- Elias
To emphasize the importance of 'power response' we can take a normal small bookshelf speaker and turn it around facing backwards to the listener. Now if we measure the on axis response it is extremely poor.
But then place the speaker (still facing backwards to the listener) in a room with high reflective boundaries, and if we listen at relatively long distance it can sound very good! It's a simple test one can do.
This is to say that for a speaker with little or no directivity the power response dominates the sound in a room.
On the other hand in my experience the speakers with very high directivity, dipole line arrays for example, the direct sound can mostly define the tonal balance. If we integrate few tens of ms and include first room reflections, the energy of the direct sound can play a major role of the total energy within the integration window.
I would even claim that for a speakers with very high directivity the off axis response is not that important. But the directivity must be very high, just a normal dipole may not be high enough.
- Elias
David, were they able to use individualized HRTFs? Back when I was at Nortel, we worked on auralization using HRTFs, as a method to increase intelligibility in audio conferencing over speakers. Basically the different talkers incoming to a conference bridge were given differing HRTF processing to physically place them in a different location in space, when played back over the spaced computer speakers. The intent was to use spatial discrimination to try and allow greater intelligibility during double talk scenarios. IMO, it didn't work well since the generalized HRTF is such an artificial construct for individual users.
I'd have to read the paper, however using generalized HRTFs IME won't allow generalized conclusions to be reached the experiment (especially with such a small sample group of loudspeakers), just provides a useful data point within the significant limitations of the study.
I know for myself if I've taken two of my own designs, I would at times prefer one speaker in one room and another in the the other. They have to be near similar quality though, to start with. But I actually voice my designs for their specific rooms. I start the design off as flat, with as well controlled directivity is possible, and through long listening and re-optimization, arrive at an on-axis that sounds most accurate and natural, to me, in that room. I spend a lot of time at this over the years, and its why I'm of the opinion that you can adjust the on-axis to help mitigate room impacts to tonality.
I haven't seen a controlled study that investigated this aspect. So, taking two speakers designed for "IEC rooms" and then finding they rank the same in differing rooms is understandable. I think this is why companies design them for IEC rooms, which tend to be some sort of room average. This allows the company maximized chance of success in differing rooms.
But to reach a conclusion that designing a speaker for a specific room doesn't improve its tonality in that room, and that mass rankings could be room dependent, would require that approach to be taken, and blindly measured. I've done this for myself, sighted (but with equal affinity for both designs, we love all our children equally 🙂 ) and reached the conclusion that it works, for tonality. I've yet to see others try this approach and post about it. I highly encourage it as I see this to be the number one benefit of diy speaker building. Its like buying a finely tailored suit vs buying off the rack. I know this was a sighted test and therefore biased even if I don't realize this, however in the absence of experimental evidence to the contrary, I'll trust my ears on this one. I was being very objective in my approach and had no expectations for any one or the other outcome to succeed. In fact, life would be simpler if tuning for the room didn't work. 🙂
Instead some people continue to hold up the Olive and Toole studies, which looked at something else, and claim the approach doesn't work.
I used to read all his papers (and Sean's too) when they were published and I never read into them the absolutist positions that too many web participants take from them. The outcomes apply to the extent that the control factors allow them to.
I like to relate an experience we had back when vocoders were being studied by the ITU, for eventual use in wireless, during its early days. The differing member groups of the ITU would run double blind testing on the various coders up for consideration, mass rank them (DMOS often) and then report back to the Study Group. Usually mass rankings were the same, but not always. Sometimes, you got anomalous results from a botched experiment. But at other times, you received results that were scientifically sound but had different mass rankings than most of the group. The most illuminating one I recall was with Slavic talkers, where one coder favoured by the group was felt to be poor by the Slavic telecom test house. So we went back and blind retested it with slavic talkers, and lo and behold, the coder had a particular weakness with slavic users due to the nature of their language and how they perceived quality when reproducing it.
This was all very rigorous, double blind and controlled, for the ITU-T and a very important study to determine the QoS of future wireless communication.
But to me it went to show that the control factors can have a profound and unexpected outcome on double blind testing, and making absolutist positions based on one study is dangerous. Again, I know you're not doing this at all, but this conversation just offers a good opportunity to relate these experiences.
Has anyone attempted to repeat the famous '85 study, but with differing control factors? For example, bipole, dipole, line source, CD etc, and in different real rooms? That would be interesting. I think a proper study would add base control factors that the '85 study used, as an absolute anchor. Without that, it's impossible to compare results between studies.
I know we aren't solving world hunger, but I always found this topic fascinating.
PSB made some great inexpensive speakers, congratulations!
Dave
Last edited:
Hi Elias, I've reached a similar conclusion. In fact, I designed a very small (wide dispersion) high quality 2 way for use in a corner application next to my TV. For regular lower level day-day use, I now keep the tv speakers on (they're OK, if unspectacular) and play the 2-way speakers facing backwards into the corner. Naturalness and, surprisingly, intelligibility have gone up.
Dave
We used to have heated debates over at Classic Speakers about various directional patterns. There was an anti-Floyd contingent based on his lack of reverence for Allison speakers and their particular polar pattern. I think, if you want to generalize, then the particular directional pattern can't be as important as the overall directivity index and how it rises with frequency. That is, alternative radiation patterns should be viewed based on their d.i. Bipole, dipole, etc. will only determine the primary direction of the extra energy. Every room will have its unique surfaces where that energy hits, so it is hard to generalize beyond the broad directivity variations. If you can accept that, then it becomes a question of direct to reflected sound ratio and we would do well to remember that sitting near vs. far or having a lively room vs. a dead room can vary the direct to reflected ratio as much or more than typical loudspeaker d.i. variation.
Floyd does talk in his book about alternative directional pattern systems. He spends a good bit of time on the ESL63, a typical dipole. He doesn't like it primarily for a lack of spaciousness in mono (slightly improved in stereo). Certainly the bulk of his testing was with typical directivity speakers. (Forward firing systems with gradually rising directivity and usually with power response holes at each crossover point).
The best work on the subject of speaker directivity was the simple study that Lipshitz and Vanderkooy did. They took a KEF 104.2 and placed a sideways firing Quad ESL63 on top of it. Since the listener was in the null of the ESL63, they could vary the energy going into it and manipulate total sound power radiating into the room independently of the direct sound. Their conclusions (heavily paraphrased): Flattening sound power by messing up the direct response is always bad. Having flat sound power along with flat axial response tends to sound too bright. Holes in the power response are fairly benign. Peaks in the power response were less benign.
They did comment on special cases were the ESL63 energy bounced off a sidewall directly to the listener and that that increased brightness compared to the effect in a larger room. That tends to support your point about the reflected sound possibly enhancing the direct response if the direct response isn't just right.
I'm sure I've voiced speakers to sound right at a particular place in a particular room. That was never the goal but it is hard to avoid. Whenever we did blind tests we found that the particular position of a speaker could force a preference between one and another. My recollection was that this was for lower midrange and lower effects and, like you said, for systems that were getting pretty close anyhow. That is the driving reason behind some of Floyd's work, and also a study we did at KEF with the University of Denmark (the Soren Bech papers). we were concerned that a good speaker might only be a good speaker in a room it suited. If that were the case then it would be hard to design anything of absolute quality. It was a relief to find that good speakers tended to sound good in any room, not that the room didn't have an effect, but that you didn't have to design a unique speaker for every room. I guess that doesn't mean you couldn't optimize one for a particular room.
David S.
This is maybe more for Ben as it seems no one else really wants to discuss it but it has to do with the whole "flat" issue: Great Sound Quality at Low Volume - Audyssey | Dynamic EQ
It looks like it address some of the issues of previous "loudness" buttons that probably account for my objection to them(unknown output level relative to mastering), but maybe introduces one of its own:
Arguing its improvement over previous implementations:
Natural objects don't get more/less bass with volume level that tracks our equal loudness curve. Should we want our stereo/HT to? Would this end up sounded less realistic/natural?
Dan
It looks like it address some of the issues of previous "loudness" buttons that probably account for my objection to them(unknown output level relative to mastering), but maybe introduces one of its own:
Arguing its improvement over previous implementations:
Natural objects don't get more/less bass with volume level that tracks our equal loudness curve. Should we want our stereo/HT to? Would this end up sounded less realistic/natural?
Dan
I think this is true for tonality and for rooms that aren't too big, but perhaps not for spaciousness. As the room gets larger, the delay of the first reflections starts to make pattern more important to tonality, nothing new there that you don't know.
This certainly makes commercial sense (IEC "average" room), if less so for the DIYer.
Can you please reference the Bech papers? I've read a number of his works from the 90s and they've all been keepers, but I don't recall these.
Thanks for taking your time to share these experiences.
Dave
Perception of Reproduced Sound: Audibility of Individual reflections in a Complete Sound Field, AES preprint 3849 ('94)
and
Perception of Reproduced Sound: Audibility of Individual reflections in a Complete Sound Field II, AES preprint 4093 ('95)
Soren Bech simulated a speaker playing in a room with an anechoic setup of a primary speaker and a number of surrounding secondary speakers that all radiated simulated reflections based on a typical room. In part 1 he does the test with the appropriate direction and general level for the calculated reflections. Part 2 repeats the test with actual off-axis response curves as appropriate for each reflection.
Since the reflections are all electronically generated he could individually vary the level of each reflection. The test was to vary the level of each reflection while playing speech and also pink noise. Listeners were asked to find the level that caused timberal differences and also to detect a level of just noticable difference. He found that a typical room's floor reflection would easily cause a timbre change on noise. Later arriving reflections were generally below the detection threshold. The side wall bounce and back wall bounce might be detectable. Reflections of speech were typically 5 dB less detectable than noise.
He also found that the octaves from 500 to 2k were most critical, implying higher speaker directivity in those octaves would reduce the detectability of reflections.
David S.
Last edited:
Hi guys
A couple thoughts to muddy the water.
Consider that the idea of “Flat” made an easy engineering target and logical way to show the system had no preference (within X+ - dB tolerance) over some specified bandwidth. Also, that when the switch was made to CD’s, one overall effect was much brighter recordings. It was necessary after all to have CD’s sound enough different to justify switching, a large part of that was more hf content (look how bright and clear it is). Bottom line here is what recordings “sound like” is a matter of taste, often part of the mastering is done using what is hoped to be representative of the average music buyer’s speakers. Now, there is much more emphasis on making a product that sells than is a stellar quality buts sound wimpy on a 6 inch car speaker. LP’s played back through “flat” speakers (back then) sounded great except LF feedback was a perpetual issue if you had significant LF output and extension at your disposal.
Also, while most are hung up on the one meter response, many people don’t listen at one meter. What is not appreciated is that a speakers directivity or rather changing directivity can cause it to have a different frequency response at any new distance.
A real possibility and common occurrence is that one can have a speaker that is a point source at low frequencies but have a different behavior higher up, that condition would allow a flat response at one meter but tipped bass shy response further away.
I guess one could say that the response at the listening position is what you hear, to the degree it is not generally flat, is exaggerating some frequencies, to the degree the room have made it unrecognizable is the degree the system interacts with the room producing reflections. Directivity is the only weapon yet directivity must be handled carefully or it can make the near / far response very different.
If one has constant directivity, then one has the same response shape close or far, just quieter far away.
Bottom line, I would say that “flat” remains a good engineering target and “makes a nice graph”, what is done while using the device is a different ball of wax, subject to popular whims, like cupping a microphone and grunting / howling along with music sounding like it was produced by cutting aluminum sheet on a table saw with a rip blade.
Best,
Tom
A couple thoughts to muddy the water.
Consider that the idea of “Flat” made an easy engineering target and logical way to show the system had no preference (within X+ - dB tolerance) over some specified bandwidth. Also, that when the switch was made to CD’s, one overall effect was much brighter recordings. It was necessary after all to have CD’s sound enough different to justify switching, a large part of that was more hf content (look how bright and clear it is). Bottom line here is what recordings “sound like” is a matter of taste, often part of the mastering is done using what is hoped to be representative of the average music buyer’s speakers. Now, there is much more emphasis on making a product that sells than is a stellar quality buts sound wimpy on a 6 inch car speaker. LP’s played back through “flat” speakers (back then) sounded great except LF feedback was a perpetual issue if you had significant LF output and extension at your disposal.
Also, while most are hung up on the one meter response, many people don’t listen at one meter. What is not appreciated is that a speakers directivity or rather changing directivity can cause it to have a different frequency response at any new distance.
A real possibility and common occurrence is that one can have a speaker that is a point source at low frequencies but have a different behavior higher up, that condition would allow a flat response at one meter but tipped bass shy response further away.
I guess one could say that the response at the listening position is what you hear, to the degree it is not generally flat, is exaggerating some frequencies, to the degree the room have made it unrecognizable is the degree the system interacts with the room producing reflections. Directivity is the only weapon yet directivity must be handled carefully or it can make the near / far response very different.
If one has constant directivity, then one has the same response shape close or far, just quieter far away.
Bottom line, I would say that “flat” remains a good engineering target and “makes a nice graph”, what is done while using the device is a different ball of wax, subject to popular whims, like cupping a microphone and grunting / howling along with music sounding like it was produced by cutting aluminum sheet on a table saw with a rip blade.
Best,
Tom
Thanks, I found II in my prepreprints, along with Bech's third installment, looking into the affect of reflections on localization. I'll reacquaint myself with them. Is it worth buying I from AES, given II?
I also recommend Moulton's "The significance of Early High-Frequency Reflections from Loudspeakers in Listening Rooms", along with Olives paper.
Most notably, with a 6ms lateral reflection, Moulton found that you don't hear the comb filtering and if the reflected image had the same response as on axis, timbre actually didn't change. It significantly impacted timbre if it differed in response by much. It also reduced detail and narrowed the image size.
Dave
- Status
- Not open for further replies.