Just put any speaker a distance of 1/4λ from the wall behind it, make a sweep and measure it. You'll see a big notch at the frequency where the distance is 1/4λ. So, for example, if the speaker is 2 feet from the wall, the notch is around 140Hz.
This is a simple case, one that's easy to test. There are also many other things that cause ripple in the 100Hz to 200Hz region, like floor bounce and room modes. The axial modes in the vertical are particularly prone to creating ripple in this frequency range because of the distances involved.
Worst thing is they someomes are staggered in that octave, such that one thing causes a dip at one note, and another thing causes a dip a little higher. They can combine to cause a large wide dip, and that's probably the worst thing that can happen.
Here's a little more on the subject:
I was playing some 78's this afternoon (mono source with consistant high frequency content, i.e. surface noise) and tried the left right shift. Its interesting how the effect varies with distance. Close to the speakers, inside the equilateral triangle where the speaker spread is more than 60 degrees, the shift is very smooth. When I am centered the image is centered and it shifts smoothly from side to side in the direction I am shifting. There is little or no phasiness. At the mid point the central image is solid and well centered, even for high frequencies.
When I back up from the speakers, especially when I get to a distance roughly twice their spread, the effect becomes dominated by the swish-swish of comb filter interference. There is still a broad shift with position but it is confused by the high frequency back and forth.
A thought on time-intensity trading: I don't see how it can work at the farther distances. The comb filtering will always be there and the proscribed level adjustment (to offset the time shift) is hard to define when the effective level at each ear is considerably altered by comb filter interference.
By the way (Earl) the numbers came from the previously suggested paper
http://www9.dw-world.de/rtc/infotheque/sound_perception/sound_perception.pdf
and it appears he took them in an anechoic chamber with 2 speakers set up in a 60 degree triangle. The author suggests that others have had different results from similar tests.
David S.
Well said Dan, but I would add another lightning:
1) All the (respectable) studies that we refer here are based on the subjective appreciations (translated in words or gesture) of human material, already this is not very "black and white".
2) In a further step, this has to be statistically averaged to see if there is a repeatable trend. I've been in this business (about cognitive perception), believe me, there are sometimes very strange and constant deviations, but a study is about searching a general law.
3) Making by ourselves some basic tests is not ridiculous if all the context is Ok.
4) What is believed true today can be demonstrated wrong to morrow, progress shows no pity for iconic values.
1) All the (respectable) studies that we refer here are based on the subjective appreciations (translated in words or gesture) of human material, already this is not very "black and white".
2) In a further step, this has to be statistically averaged to see if there is a repeatable trend. I've been in this business (about cognitive perception), believe me, there are sometimes very strange and constant deviations, but a study is about searching a general law.
3) Making by ourselves some basic tests is not ridiculous if all the context is Ok.
4) What is believed true today can be demonstrated wrong to morrow, progress shows no pity for iconic values.
What I was trying to say is that those who had a weak phantom center image, may have that due to having a crossover point right in that 2-6kHZ region. That could cause lobing, which could cause differential variations in frequency response at the listeners ears, especially when the room acoustics are brought into the picture. Imaging in the upper midrange depends on both speakers delivering a very similar frequency response to the listener. Many conventional speakers have crossover points in this frequency range. Just a guess.Are you ready for the surprise !?! There was no cross over at all !(of course not a real sursprise since I explained it already carefully, it was FE126En)
- Elias
It sounds like you are saying that you don't have a crossover point in that region? Did you have a pretty solid center phantom image at all but the 6kHZ highpass pink noise? What was your result?
Yes, "window to there" guy, for sure. Unless I crank up the good old Yamaha DSP1.
As for shifting center image, mine shifts to the same side that I do. But I have to lean a very long way to get it to shift far. By the time pink noise is high passed at 9.5K I can get it to collapse into one side or the other by not moving far. Below that I just about have to fall over to get it into only one side. I was surprised at that.
My speakers centers are 10 feet apart, I'm 12 feet back.
As for shifting center image, mine shifts to the same side that I do. But I have to lean a very long way to get it to shift far. By the time pink noise is high passed at 9.5K I can get it to collapse into one side or the other by not moving far. Below that I just about have to fall over to get it into only one side. I was surprised at that.
My speakers centers are 10 feet apart, I'm 12 feet back.
It sounds likely that the comb filtering effects you heard further back are more because of room acoustics. Perhaps the ideal dispersion pattern would have limited dispersion in the 2-6kHZ range where imaging is all about the match of both sides, and have wider dispersion above 6kHZ where imaging isn't as much of an issue (?).
No, as explained before comb filtering has more to do with your directional hearing and whether one ear hears primarily the speaker of its side, or both speakers.
With the head related transfer function we don't have great directivty but at high frequencies and angles of approaching 60 degrees (120 degrees between the speakers) we have decent seperation and the left ear hears the left speaker only, etc. At such angles the two speakers are independent and effect each ear with simple level and time shifts.
At narrow angles between the speakers (28 degrees when you are twice the speaker spread away) the left ear hears both the left and the right speaker. Any shift in position sets up the comb filter effect, the swish-swish I've mentioned. Room acoustics and speaker directivity will determine the direct to reverberent ratio at that distance, but if you still hear the comb filtering, then the direct component is still sufficient.
David S.
On the curve you show the relevant part is the curved rising edge on the left side. This is the locus of points where the time shift and level offset can balance each other out.
It looks like 10dB offsets .75 msec. on this curve. The other author has 8dB offsetting 1 msec. Quite different?
David S.
I've tried the same thing before myself, although minus the 78
This is exactly what I suggested a few pages back - once you get past around 60-70 degrees separation, the comb filtering effect at high frequencies (at least for the direct path signals from the speakers) more or less stops being an issue, due to sufficient left/right crosstalk reduction afforded by the head.
Above 1Khz or so the brain doesn't care what the relative phase is in left and right ears, it's just doing an amplitude comparison so as you move left and right the apparent source pans smoothly to the same side as you due to the smooth amplitude change, and the wildly varying relative phase received by left and right ears means nothing.
Yes exactly. Now there is a very high level of crosstalk between each ear with both ears hearing both speakers at high frequencies, so each ear perceives the amplitude fluctuations caused by adding and cancelling at the point of the ear. Both ears don't experience cancellation at the same location so the high frequency source alternates left and right with lateral movement.
Being twice the distance away from the speakers as their separation is a very harsh test mind you - that's a separation of approximately only 30 degrees, and I wouldn't expect to get any kind of stable image there - precisely because of comb filtering.
If your angle of separation is too narrow, comb filtering will destroy it, yes. Even at 45 degrees I notice a significant "wobble" in the centre position as I move sideways, despite the overall tendency of the image to cling to the middle compared to lesser degrees of toe in.
On the other hand a wide angle of say 60-70 degrees will largely eliminate the "wobble" from comb filtering, but now causes more time delay differential per lateral movement, making the requirements for time/intensity trading even more stringent.
From this, there must be an optimum angular separation for a given speaker design for widest sweet spot, which can only be found empirically. For each angular separation a range of toe in values would need to be tried and evaluated for width and stability of the sweet spot, then repeated at the next angular separation.
I suspect wider than typical separation will probably give the best results though, due to minimizing the effects of comb filtering.
I think you're spot on there regarding a notch causing an apparent (psychoacoustic) emphasis of the frequencies around it.
Notches or dips in a frequency response are often said to be a lot more benign than a peak of equivalent Q and amplitude, but I'm really starting to question that belief these days.
I had already noticed what you describe in relation to floor bounce notches causing apparent emphasis of frequencies either side of the notch, but recent experiments of mine correcting surround dip in the 1Khz region have given much the same results - although a notch doesn't immediately draw attention to itself in the same obnoxious way as a peak, a deep notch does indeed seem to cause a perceptual increase (and therefore imbalance) in the surrounding frequencies.
So a 3dB 1/3 octave notch at 1Khz has a tendency to make the speaker sound too "hot" in the 1/3 octave (approximately) bands on either side of the notch, in this case around 700-800Hz and 1.2-1.4Khz, even if it is in fact perfectly flat through these surrounding frequencies.
If you had no idea what the frequency response looked like and were to attempt to EQ the apparent imbalance purely by ear there is a surprising tendency to try to cut the response on either side of the notch rather than increase the response at the notch. Obviously this leads to a slippery slope of greater and greater errors in the response, and one reason why EQ'ing by ear only is a very dubious concept no matter how experienced you are.
I was initially very surprised to find that increasing the response at the notch by compensating it actually made recordings that had sounded too forward in the 500-800Hz range sound more balanced and neutral, rather than just sounding even more forward.
I think what's going on here is our perceptual masking patterns are being fooled by a deep notch which is roughly the same width as the critical bandwidth of our ears.
Lossy codecs like mp3 make extensive use of the masking effect, where a loud sound at one frequency masks quiet sounds in nearby bands, if you essentially notch out an entire 1/3 octave band in broadband program material, with nothing to "suppress" the bands adjacent to the notch they then seem emphasised, as they are not receiving their usual amount of perceptual masking.
It may be similar to what we do with contrast ranges visually - the visual system has a very non-linear response which takes more notice of "local contrast" within individual parts of a picture, so how bright something looks to us depends not on its absolute brightness within the scene, but how bright it is relative to the details immediately around it. (Something which a "tone mapped" HDR photo attempts to simulate)
Long story short - notches aren't as immediately obnoxious or obvious as peaks but they do matter just as much, and it's often not until they're eliminated that we realise what was so wrong with them being there in the first place, and this definitely applies to notches in the upper bass as much as the midrange.
Last edited:
I'm reading the article now.
Deutsche Welle Radio Training Centre said:
I've of course noticed the effect that people's voices sound brighter when you're in the open and at some distance from each other, but I don't understand the above explanation. Once you're in the far-field, as far as I know the ratio of low to high frequencies shouldn't change. What am I missing here?
I always attributed said effect to what is explained in the quote below.
Deutsche Welle Radio Training Centre said:
---
Last weekend I did some of those imaging tests with Markus' pink noise samples.
I currently have the luxury of a dedicated listening room a little over 3.4m high, about 4.5m wide and 6.5m long. The speakers are quite directional (see signature) and they're setup symmetrically across the length of the room, a little under one meter from the sides and about 1.80 from the rear. The sofa is between 2.5 an 3 meters from the back, the speakers are crossed a little in front of the listening position (probably about 0.5m) and the ipsilateral sidewall reflection accurs at close to 90 degrees off-axis. The room is fairly live. I measured RT30 with ARTA and it shows that from 500 hz to 8 khz RT30 decreases smoothly from about 0.5s to 0.4s. Below 500 hz it increases to about 0,65, to decrease again below about 200 hz, to about 0.3 below in the bass.
* Correlated Pink Noise Full-spectrum sounds centred but also somewhat broad. Moving my head left or right causes the timbre of the noise to sound lower in frequency. Moving forward also causes the noise to sound lower (maybe because I'm getting closer to the speaker's axis), but to a much lesser extent. Moving only a little sideways causes the image to shift to the nearest speaker. When I get to a little over a foot, it's almost entirely located in the nearest speaker.
* Correlated Pink Noise 1500 hz HP sounds just as centered, but a bit more compact. Moving my head sideways causes it to sound a bit phasey, but once I kept my head still again, the phaseyness went away. The shifting of the image to the nearer speaker was similar to what happened with the full-range noise.
* Correlated Pink Noise 3000 hz HP also sounds well-centered. Moving my head didn't lead to as much phaseyness as the 1500 hz HP sample. Also, the image seemed to stay more centred upon moving my head sideways.
* Correlated Pink Noise 6000 hz HP sounded nicely centred, but moving just a little to the sides made it sound really funny and a bit phasey again. Probably I was hearing the many closely spaced interference peaks and dips.
It seems that with my setup and my ears the image falls apart faster than with most others'. I know Earl Geddes and some others believe speakers with high directivity that are crossed in front of the listener the image should be very consistent across the width of the couch, but in my situation this clearly isn't the case. However, when sitting in the centre between the speakers the image is very precise.
But isn't this just as one would expect? High directivity leads to a strong direct sound and strong direct sound means good imaging. Strong early reflections lead to a more smeared image, that's wider and less precise. This means that reflections contribute to imaging and that with wide dispersion speakers they thus do so to a greater extent than with high directivity speakers. Moreover, the sum of all reflections is more diffuse in character than the direct sound. When you move off the centre axis between the speakers you are changing your position with respect to both speakers. With speakers that have a strong direct sound this should lead to a greater change in imaging than with speakers that rely of a multitude of reflections for imaging.
Nothing. As far as I'm concerned that reference is talking a load of twaddle in this particular quote
High frequencies fall off at less than 6dB per distance doubling just because they're more directional ? I don't think so....
Not in the far field of a point source anyway. A conical midrange horn is much more directional than a closed box woofer at bass frequencies, but both fall off with inverse square law as soon as you are out of their immediate near-field.I think the answer is much simpler than that - near-field to far-field transition, for example the baffle step effect, and that's probably what the reference meant to say, but grossly over simplified it to the point where they made it wrong.
Take a person speaking with a mic right in front of their mouth. You'll get a lot of bass here, as you're measuring the near-field response of their mouth.
Now measure at 1 metre and there will be a lot less bass relative to higher frequencies - because their head is basically a small baffle, and frequencies from 600Hz or so and below are below the baffle step frequency of the head, so there will be a 6dB loss in bass as the low frequencies wrap around their head.
This 6dB baffle step transition will be complete in well under one metre though for a persons head, beyond this all frequencies are falling off with 6dB per distance doubling, until you get far enough away for air losses at high frequencies.
With much larger sound sources you have to get further away before the near-field to far-field transition is complete of course, but I'm not sure that this near-field to far-field transition is a significant distance cue for us.
I would say this is a far more dominant factor than their claim in the first paragraph.
I've finished reading the article. It is indeed very relevant to the discussion about imaging (thanks for the link, speaker dave). They come to the same conclusion as I did on the basis of listening to those noise samples: wider dispersion (omni in their case) has a wider sweetspot but it has "the disadvantage of reduced sharpness of localization".
OK, I'll try to answer that. Might be a little difficult 20+ years on.
Yes, that probably had an important role to play. But if it were all that dominant, I think that a lot of the "virtual space" we heard would have been similar. I wasn't, it changed from recording to recording - often drastically. From recording studio to night club to concert hall to cathedral to immense cave. What was so mind boggling was the very large spaces. We all commented on it, there were lots of looks of disbelief and head shaking. How could those immense spaces fit inside that little black box theater?
So I would say that Yes, the space was important. But it was not the entire effect. I knew that acoustic space well, by that point I had spent 1000s of hours working in it. It only sounded like that during the demos. The big take away for most of us wasn't how great the system was, but how much info is hidden in recordings.
Not sure what you mean. A combination of many random sinus waves? I was using pink noise from the DEQ2496, if that matters.
That's a brunch of good posts. Manifestly, through the different opinions and descriptions, there is still a significant percentage of problems with the HF restitution, just because their damned wavelength is smaller that our head (comb, cross talk, swish -swish).
Here I recommend again to think to the alternative configs that propose a radical solution : centered source for the Hf (stereo for the ISVR and it's OSD, or even mono for Stereolith and assimilates). Having tried both, my preference goes BTW to a third pattern that radiates mono on the axis and stereo on the sides. Barbaric but efficient.
@Pano : Sorry, I mean that your description makes me think of what happens with a pure sinus, of course it's probably just interferences. You have been in France, so you must remember how much the French are good in English...
Here I recommend again to think to the alternative configs that propose a radical solution : centered source for the Hf (stereo for the ISVR and it's OSD, or even mono for Stereolith and assimilates). Having tried both, my preference goes BTW to a third pattern that radiates mono on the axis and stereo on the sides. Barbaric but efficient.
@Pano : Sorry, I mean that your description makes me think of what happens with a pure sinus, of course it's probably just interferences. You have been in France, so you must remember how much the French are good in English...
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.