+3db below 200Hz and -3db above 3kHz... Generally speaking that looks like approximately 6db end to end. It would seem that we almost have some measure of agreement with the notion put forward in the subject of this thread.
SL said, "We have increasing sensitivity to high frequencies when the real source is at a 30 degree angle, than when it is in front of us. The exact relationship is complicated but there is a broad trend to the sensitivity increase with a slope in the order of 2 dB/decade to 3 dB/decade."
ITU control room: http://www.aes.org/technical/documents/AESTD1001.pdf
Here's the EBU's: http://tech.ebu.ch/docs/techreview/trev_274-hoeg.pdf
These are not dead rooms. In fact my living room and BR fall w/i these guidelines, but they are treated.
Now look at what they recommend to evaluate classical music:
http://www.wseas.us/e-library/confe...?dfa51120?3e245730?dfa51120?3e245730?dfa51120
Sounds familiar.
It would make sense to me that being even across the spectrum might be more important than an absolute number and that could explain why the Orion works at all.
Dan
Here's the EBU's: http://tech.ebu.ch/docs/techreview/trev_274-hoeg.pdf
These are not dead rooms. In fact my living room and BR fall w/i these guidelines, but they are treated.
Now look at what they recommend to evaluate classical music:
http://www.wseas.us/e-library/confe...?dfa51120?3e245730?dfa51120?3e245730?dfa51120
Sounds familiar.
It would make sense to me that being even across the spectrum might be more important than an absolute number and that could explain why the Orion works at all.
Dan
There does seem to be general agreement supporting a 1dB/octave rolloff starting at about 2kHz *in overall (room) response at the listening position*. What is not so clear is how that is to be accomplished, and the significance, if any, of the frequency response of the *direct* sound from the loudspeaker. One can accomplish that rolloff by applying a global rolloff to a constant directivity loudspeaker, by maintaining a flat-on-axis response with a changing (narrowing) directivity index as frequency rises, or with a flat and constant directivity loudspeaker in a room with rising absorption at high frequencies. Presumably those options, although providing the same measured pink-noise result at the listening position, would sound different to a human listener on a recording of music as opposed to random noise, at least as regards "imaging" and "spaciousness".
With regard to the ORION example that's the question I just asked . . . assuming a relatively flat room is it better to implement the rolloff globally (as it now is), or would it be preferable to keep the front tweeter "flat" and apply the (necessarily greater) rolloff to the rear tweeter only (to bring the desired rolloff in "room response")? Or in other words, do we keep a constant DI and change the frequency response, or keep a constant (on axis) frequency response and change the DI (an option almost uniquely available with ORION)?
Front and rear should do the very same so that the reflections match the direct sound as best as possible. So in my view this is not a viable option.
I guess that answer pretty much depends on the application of a speaker. For recreational listening it depends mostly on the room. And then there remains preference...
I guess you meant Pluto 😉 But Pluto is not flat either. The Aura tweeter has an inherent rolloff starting around 15KHz. But the 30° off axis repsonse follows nicely so I assume it has a uniform DI.
And it sounds as if SL equalizes it also to a monotonically falling response since the intro of the Seas woofer.
+3db below 200Hz and -3db above 3kHz... Generally speaking that looks like approximately 6db end to end. It would seem that we almost have some measure of agreement with the notion put forward in the subject of this thread.
SL said, "We have increasing sensitivity to high frequencies when the real source is at a 30 degree angle, than when it is in front of us. The exact relationship is complicated but there is a broad trend to the sensitivity increase with a slope in the order of 2 dB/decade to 3 dB/decade."
From Linkwitz a few years back: (I believe this feature was incorporated into the Audio Artistry products as well.)
"Finally, for a speaker like the PHOENIX that is flat on-axis and similar in its power response, some recordings may sound bright because the microphone pickup was too close to the instruments. To correct for those situations I have found a frequency response downward tilt of 3 dB/decade to be a subtle, yet effective remedy. The tilt is also beneficial when listening very close to the speaker in a small space, or when listening at lower volume levels.
A tilt control can be added to the crossover/eq by adding circuitry to the buffer stage that drives the three filter channels."
Crossover/EQ
http://www.linkwitzlab.com/images/graphics/3db_buff.gif
Dave.
That is one theory, that may be correct, at least up to some frequency. But it is not obvious that it should be so . . . in the concert hall the highs roll of in both the (direct) distant and the reverberant field, why should reproduction be different?
No . . . I meant ORION. One cannot change the DI of PLUTO, it is fixed by the directivity of the tweeter (as frequency rises). One *can* change the DI of ORION by the simple expedient of (electronically) changing the frequency response of the rear tweeter. There are only a few other speakers that offer that (variable DI) option.
Hi Pano,
Sorry about that. I think the graph is worth explaining as it concisely illustrates this whole concept. BTW, I won't refer to it as "SL's" concept as I arrived at it many moons ago (and no doubt have others).
Here is the graph attached. What it shows is the sound pressure incident at the "drum reference point" in the ear canal caused by a flat broadband signal in an anechoic condition, under various conditions which I'll explain. Stated another way, the curve shows the frequency response of your head, outer pinna and ear canal as measured at the ear drum. For DRP definition see section 4.4 in
http://eu.sabotage.org/www/ITU/P/P0057e.pdf
Stereo speakers in the "classic" stereo triangle are 30 degrees to the right, or to the left, in front of the listener. The magenta curve is the frequency response of your near ear if the sound is playing from one of these locations. To illustrate, lets say the sound is playing from the right loudspeaker. The magenta curve is the head/ear frequency response measured at the DRP of your right ear (Near Ear). The yellow curve would be the frequency response at your left ear (Far Ear).
Now, imagine the same signal is played in both the right and left speakers (mono). The light blue signal is the frequency response in either the right or the left ear. Its the combination of the "near ear" response and the "far ear" response, using a power summing rule. Caveat: the power summing rule isn't always the most accurate rule, it depends on circumstances. The perception of loudness depends on the source signal frequency content, dynamic characteristics and duration i.e. sometimes "double" loudness is +6 dB as I showed here, sometimes its closer to +10 dB, or in between. Zwicker and others have shown this in DBT tests (Ben, this no doubt plays a part in your question: how do you measure loudness?).
So, the light blue curve is the frequency response at both DRPs for the "phantom" center image.
The dark blue curve is the frequency response at both DRPs when a sound is actually physically directly right in front of you.
Comparing the light blue "phantom image" curve to the dark blue "physicaly present" curve, you can see that the phantom image, generated by speakers at +/- 30 deg, will present 2 dB more sound pressure to your ear drum, from ~ 5 to 8 kHz, than the same signal if it was actually physically located at the phantom position, in front of you.
This difference is the purple curve.
This is one inherent error in stereo itself. Caveat again: head frequency response vary with a high sigma over a population of individuals. This calculation uses industry agreed standard curves. I think the trend is universal though, only the magnitude will vary from person to person. The reason is that a source at 30 deg sees less acoustic shadowing from the head/pinna than a source which is directly frontal incident.
You could take this approach and calculate the error for any phantom image within the +-/30 deg window. Now, some are arguing here that the recording engineers will correct for this. My response is, some might, some might not. Who knows what they individually do or think? If they do , its by ear. From what I can tell, standard pan pots do not use azimuthaly dependent frequency shaping (though they could post eq). The side effect of the "error" magnitude being user dependent also means that the "correction" required is user dependent. What may sound "right" to a recording engineer might not sound "right" to you, for very objective, concrete, physical reasons.
Many have found that a gentle tailing of the high frequency response in playback sounds more natural. I posit this phantom image error phenomenon may be one reason why (aside: I think another reason is that too many recordings sound like cr@p due to over heavy use of compressors and effects, and toning down the top end makes them more bearable: humans tend to avoid discomfort as a priority, and tuning for bad recordings over good recordings, given bad recordings far outweigh good recordings in number, may be tempting).
Earl uses 3 dB shelf to reduce the top end. Lynn Olson used 5 dB in the Ariels. I use 2 dB as a starting point, but tweak by ear depending upon dispersion and playback room intended for use (livelier and smaller it is, more shelf I use).
All 3 of us find flat speakers with broad dispersion to be too bright. So do other experienced designers I know. I have long campaigned against the "tyranny of flat".
I found it interesting that Earl qualified it that only "broad dispersion" speakers required the shelf. This mirrors my findings. The D26NC is a very low distortion and extraordinarily flat tweeter, on axis. When I designed with it, I found that no shelf was required. Any shelf made it sound dull. The tweeter has a larger high frequency directivity index than most soft domes.
BTW, I also agree with Earl's findings regarding the need for a low frequency rise in typical rooms. Without it, speakers in my experience sound thin and lack foundation or impact. This is very different than an alignment Q or port tuning issue, its achieved through careful cabinet shape, driver placement and crossover tuning. This one I have no theory for as the speaker is already launching far more low frequency power into the room than at mid and high frequencies. I "think" this sounds "right" because it better mimics the sound power real instruments would present in the room, and because it mimics the mastering environment, but why a rise? Is it representative of the environment used by the recording/mastering engineers: speakers with a rising low end?
Dave Dal Farra
Attachments
Yes . . .
Except . . .
that's not how it's implemented in ORION 3.2.1, where the rolloff is fixed, and applied to *all* recordings. In effect that puts *every* performance on a concert hall stage, with the listener "out in the audience", and it does seem to work for most "classical" recordings (it does quite satisfyingly for me). But that is neither the artistic intent nor the listener's desire for many non-clasical recordings. There is no "control" offered to correct for that . . . only a one-size-fits-all frequency response correction. And it leaves unanswered the question of whether there should be (electronic) correction for room response irregularities (by varying DI or by other means), or whether those variations should be dealt with entirely by room "treatment".
The problem with "correcting" for that rising hf response in the "phantom center" by rolling off at the loudspeakers is that one then (incorrectly) rolls off the real "side image", and we end up seeking the "best compromise" (perhaps necessarily, but it's always contentious).
That becomes even more an issue for the growing number of "multi-use" systems that (for some recordings) actually *have* a physical center channel, and are playing recordings mixed for it. The "rolled-off" Left and Right speakers are, for those recordings, clearly wrong. This *could* be corrected for "in the mix" . . . but I see no evidence in the literature that it is, or that any "standard" exists for so doing.
Of course it could also be argued that these effects, while real, are all but inconsequential using most reproduction systems in most listening environments, being swamped by a multitude of other response "errors" in recording, equipment and room, and "sonic misrepresentations" commonly present in the recordings as well.
Deward, I think you may want to read the posted documents again. It sounds like you are missing the point. Your last post is better however. What DDF is saying makes sense if there was some idealized recording process and no one ever listened to and tweaked the recordings prior to sale and the phantom center was the only goal.
Dan
Dan
Last edited:
Yes, this is absolutely correct. I pointed this out in my earlier post.
Agreed, as I pointed out earlier.
This echoes what I pointed out, that many recordings are awful and users may tailor for this over other considerations. Either way, IME, a 1 dB broadband (covering an octave or more) change is very audible and this effect will be user dependent so YMMV. Your consistent points about classical recordings needing diferring requirements than multi-miced pop also apply.
Last edited:
Here's a thought. If there is a ~2dB HRTF boost in the mid-treble at 30 degrees from the direct on-axis sound from the speaker, how much HRTF boost in mid-treble is there for angles of ~60 degrees, where the first side wall reflection will come from ?
I'm guessing a lot more, as at 90 degrees - side on to the ear - there is a LOT of apparent boost in the treble compared to sound sources directly in front of you, so much so that it's clearly observable as a big treble increase rather than just a change in perception of direction.
Maybe the reason flat on-axis speakers with extremely broad dispersion in the treble sound too bright in smaller/live rooms is not so much that there is too much overall power response in the treble, but more to do with the fact that a great deal of the reflected sound including the primary early side-wall reflections are entering the ear at 60-90 degrees, and thus being greatly boosted by the HRTF, even over and above the "stereo error" of 2dB at 30 degrees ?
Even though a wide dispersion tweeter will still generally have some power response roll off and the room reflections will have increased attenuation at higher frequencies, (resulting in a measured "room response" tapering off in the treble) it's probably far outweighed by the increased high frequency sensitivity of the ear as you go towards 90 degrees. So the reverberant field appears to have far too much treble content relative to the direct field mostly due to the direction from which it enters the ear boosting the sensitivity to treble.
I've noticed the same, I've tried conventional face plate dome tweeters, wave-guide loaded ribbon tweeters, and large full range drivers which all have very different directional characteristics in the treble, in very dead and live rooms.
If all are equalized to have flat on axis response, and sound roughly right on axis in a heavily damped room or outdoors, the dome tweeter placed in a reflective room becomes much brighter, needing some significant shelving down in the treble to get an acceptable balance, the full range driver becomes much duller, needing considerable variable slope boost in the treble, while the ribbon tweeter seems to change relatively little in balance, with maybe 0.5dB less treble in the highly reflective room being optimal, but still sounding acceptable even without making a change.
From that it's clear that if the power response at high frequencies is too far down it will tend to sound darker in a reverberant room, while if the power response is too close to ideal, it will sound brighter - too bright, so some intermediate directivity profile may be able to minimize the perceived shift in balance with different rooms, by roughly balancing the reduction in power response of the tweeter (and increased high frequency absorption in the room) with the "gain" from the ear averaged over the angles where most room reflections come from.
Maybe the ribbon tweeter by luck comes somewhat close to this - it's much wider than the full range driver (at least horizontally) but considerably narrower than a dome tweeter, and has more of a constant directivity profile in the horizontal plane due to the narrow ribbon driving a waveguide, than a dome tweeter that tends to roll off at the highest frequencies first with angle due solely to diaphragm diameter.
If this is the case, it argues for a constant/controlled directivity design where the directivity in the treble is deliberately restricted - but not too much, and a flat on-axis response.
It also suggests that "onmi" directional speaker designs are heading in entirely the wrong direction, at least if the aim is to design a speaker which has the least change in perceived high frequency balance with changes in direct/reflected ratio due to room size and damping...
Maybe this is not so contentious.... if we qualify the statement, "Flat is not correct for stereo" by clarifying, "this generally applies only to STEREO playback of STEREO recordings in a relatively lively playback environment of the average "living room" when comparing the playback environment to that of a relatively diffuse mastering room of a recording studio"?
I'm not so sure... the most common comment from most friends who have heard my "non-flat" speakers say they sound more realistic than any other speaker they have ever heard. (granted... none have heard the Orion and my reproduction system is certainly not like most 🙂 )
But I haven't the resources nor space in my living room to build and test a full multichannel "non-flat" system. I would love to hear some comments in this regard.
I'm not the least surprised . . . that's what I find with ORION and with all the other speakers I've tried in my listening room. It's not "on-axis" that engenders the comments . . . flat on axis with falling power response (beaming) sounds OK, and broad dispersion with falling power response sounds OK (which implies not flat on axis). They sound different with regard to "spaciousness" in particular, but similar with regard to overall tonality. What sounds more "right" depends on the recording . . .
Non flat sounds more realistic compared to what? By definition, non flat is non realistic.
Dan
Dan
Last edited:
To me (and many others) the experience has been; When listening in a average home environment and on many recordings but not all, non flat sounds more realistic when compared to the experience of a live performance. (i.e. can't pinpoint location of L & R tweeters) Therefore, "it sounds more 'Real' ". In a home environment, depending on the recording, flat can often tend towards a very "harsh" sound and sometimes even genuinely painful when listening at a realistic (i.e. live) volume level.
Non-flat doesn't define as non-realistic... just non-flat. I don't think anyone has disputed that "flat" remains the "standard" against which any modifications are compared and "flat" is obviously preferred for mastering and/or the recording studio.
I agree, it doesn't entirely make sense, henceforth the great debate.
To me (and many others) the experience has been; When listening in a average home environment and on many recordings but not all, non flat sounds more realistic when compared to the experience of a live performance. (i.e. can't pinpoint location of L & R tweeters) Therefore, "it sounds more 'Real' ". In a home environment, depending on the recording, flat can often tend towards a very "harsh" sound and sometimes even genuinely painful when listening at a realistic (i.e. live) volume level.
Non-flat doesn't define as non-realistic... just non-flat. I don't think anyone has disputed that "flat" remains the "standard" against which any modifications are compared and "flat" is obviously preferred for mastering and/or the recording studio.
I agree, it doesn't entirely make sense, henceforth the great debate.
Ah! To the live performance. Something that's not on the recording and your memory doesn't know. I suppose it will always be the great debate.
Dan
- Status
- Not open for further replies.