Nice graph. Maybe it shows why that pesky 1Khz test tone seems to come from everywhere and nowhere. FWIW, I have read that about 7Khz, more or less, is the center of the height effect. 8Khz seems right in there. It's caused by HRTF, I think.
I did not get any height differences on the various versions of the test.
I did not get any height differences on the various versions of the test.
The in-room response is flat (+/- 2 dB or so) with only a ~10 dB moderate Q null at 60 Hz.Check your tweeters on axis response for a peak in the ~8Khz region if you consistently notice imaging well above the speaker.
See the attached image that shows what frequency ranges localize offsets in which axis - oben is german for "above", and the graph shows that a fairly narrow frequency region around 8Khz is responsible for elevation perception.
I had always noticed my ribbon tweeters imaged quite a bit higher than their physical location (sometimes nearly as much as you describe) and a careful measurement showed a 1/3rd octave 1dB bump around 8Khz was present - I EQ'ed it out and presto, the apparent image location has moved down quite a bit and is now only slightly higher than the tweeters. (Quality and tonal balance has improved significantly as well despite the fairly small error)
Note: even if your speakers are flat in this region the music itself may have some emphasis at 8Khz, and the cumulative effect of the two if the speakers also have a bump there could be excessive and cause an unusually high apparent image source location.
The graph was posted in another thread, I've lost track of which thread it was and who posted it, but thanks to whoever it was, as I've found it very accurate and enlightening.
(English translation: Green - in front, Red - behind, Blue - above)
I have played in an orchestra that did Mars and the trumpets match where they were sitting.
if you consistently notice imaging well above the speaker.
So what is the low down on vertical imaging? I believe there is a false effect due to speakers and room. Can height cues be preserved and revealed?
In room (steady state) response doesn't matter much at 8Khz, its the direct field from the speaker having a bump that would cause it. +/- 2dB doesn't really tell us much, the bump that I EQ'ed out was only 1dB above a response that was flatter than that, (and therefore well within a +/- 2dB margin) and it still lowered the apparent source height by at least a foot.
According to that graph, height cues come almost entirely from the ~8Khz region, which will be due to the HRTF.
So as far as I can see unless the recording is made with a binaural head the actual height cues of the original sound source locations relative to the microphone will not be encoded in the recording, and thus any sense of a sound originating from well above the speakers is an artefact of the recording/mixing process or the speakers.
Of course its possible that some engineers have learnt that emphasising 8Khz raises the apparent source height, and so might apply that deliberately to certain instruments to spread the vertical height of the mix as one of their tools in trade... or it may just happen by accident due to microphones with a peak in their high end response or the incidental EQ applied during mixing.
However I do note that the graph shows a small contribution to height perception from low frequencies around 300-500 Hz as well, perhaps this will be captured in the recording (since its well below HRTF frequencies) and perhaps is related to the way floor bounce interferes with the lower midrange... (in other words there may be a weak but valid height cue encoded in the low frequencies on recordings that are not close mic'ed)
While experimenting with floor bounce elimination I had sometimes noticed a shift in image height as well as depth, perhaps this graph explains why. (Getting a bit speculative now
)
Last edited:
Contain height cues in the recording ? Yes, definitely. Retain actual original raw recorded height cues ? Probably not, sans binaural head. Most likely they are manipulated in the recording/mixing/processing process.
If even a simple peak around 8Khz can raise the apparent source location, it should be relatively easy to add convincing height cues to individual instruments using HRTF convolution. I'm sure I've heard it done on a few recordings too.
I don't think un-manipulated recordings made with normal microphones would do it though, except by incidentally putting a peak at the right frequency, but then everything that microphone recorded with significant high frequency content would have its apparent source location skewed upwards...
If it helps any, the massed strings stayed between the tweeters.
How does bumping up 8 kHz raise the apparent origin of sounds not at 8 kHz? Is the opposite also true, where pulling down 8 kHz results in a lower sound origin?
I'd agree with you Simon, had I not heard it myself in the company of several dozen other people. That was the most amazing take-away from the demos at Espace Kiron. The uncanny sense of height that some recordings had. I remember one orchestral piece with choir that had people shaking their heads and looking at each other. It actually sounded like the choir was in three rows behind the orchestra - solid as could be. No idea if it was, but that's what we all heard.
The opening track of Andreas Vollenweider's Caverna Magica was another jaw dropper. Voices and footsteps on gravel that walk into the cavern. Suddenly the small, black box theater was a huge space 100 feet high and deep, absolutely vast. It was a twist on reality that still gives me chicken skin - so hard was it to wrap our minds around. Lots of big eyes and comments of "how is that possible?" Nothing but early Lexicon there. I wish the Lexicon guys could have heard it, they'd have been "bouche bée" as well.
The hight is in there, but it's not easy to get out.
The opening track of Andreas Vollenweider's Caverna Magica was another jaw dropper. Voices and footsteps on gravel that walk into the cavern. Suddenly the small, black box theater was a huge space 100 feet high and deep, absolutely vast. It was a twist on reality that still gives me chicken skin - so hard was it to wrap our minds around. Lots of big eyes and comments of "how is that possible?" Nothing but early Lexicon there. I wish the Lexicon guys could have heard it, they'd have been "bouche bée" as well.
The hight is in there, but it's not easy to get out.
In the CD Highlights from Miss Saigon, original London casts, in the beautiful duets, Sun and Moon, The Last Night of The World by Lea Salonga and Simon Bowman, the heights of images are always amazing and interesting on various speakers I've made along all these years.
The male voice appears obviously higher than the female. In some speakers, it seems she is at about the hieght of his shoulder; in some others, it seems she is sitting and he is standing high up on an elevated platform.
While most other sounds are more or less at the same height with her. Only he is obviously standing higher.
Maybe there's a higher proportion of HF harmonics in the voice of the male singer.
And in some other live concert records (I forget which), I've heard (seen) some cymbals floating so high that they seem appear at a very far end of the extending straight line formed by my eyes and the top of my 1.8m speakers.
No, my speakers don't have peaks around 8kHz.
The male voice appears obviously higher than the female. In some speakers, it seems she is at about the hieght of his shoulder; in some others, it seems she is sitting and he is standing high up on an elevated platform.
While most other sounds are more or less at the same height with her. Only he is obviously standing higher.
Maybe there's a higher proportion of HF harmonics in the voice of the male singer.
And in some other live concert records (I forget which), I've heard (seen) some cymbals floating so high that they seem appear at a very far end of the extending straight line formed by my eyes and the top of my 1.8m speakers.
No, my speakers don't have peaks around 8kHz.
Some comments on height of scene:
You can have large spaciousness all the time by increasing reverberation. But it only comes in all directions at once. In stereo you can't have a high cathedral without a wide cathedral IMHO. Same for caves etc.
If there are any "real" height impressions, they have to come from floor (or ceiling) bounce cues, which are in the recording. This will work all the better, if the floor bounce cues of the playback room/system are weak. I believe this would be the case in a large theater room. Didn't Toole write about it?
A big caveat: Ventriloquism works very well - at least for me -, even when I don't know about the recording situation. If you know in general from sight, where and how a choir is arranged behind the orchestra, your brain is more than willing to place it there acoustically - if the recording does not say otherwise.
Rudolf
You can have large spaciousness all the time by increasing reverberation. But it only comes in all directions at once. In stereo you can't have a high cathedral without a wide cathedral IMHO. Same for caves etc.
If there are any "real" height impressions, they have to come from floor (or ceiling) bounce cues, which are in the recording. This will work all the better, if the floor bounce cues of the playback room/system are weak. I believe this would be the case in a large theater room. Didn't Toole write about it?
A big caveat: Ventriloquism works very well - at least for me -, even when I don't know about the recording situation. If you know in general from sight, where and how a choir is arranged behind the orchestra, your brain is more than willing to place it there acoustically - if the recording does not say otherwise.
Rudolf
I'm not sure that that disagrees with what I said.
It's definitely possible to encode height into a recording such that when its played back by a neutral speaker system some sounds will be perceived to be elevated well above the speakers - I've certainly heard it enough times myself to know it can happen, and I kinda like the effect of an image having a range of height as well as width and depth. Not many recordings do, but some do.What I disagree with is the notion that the actual spatial height of the instruments relative to the microphone was captured by the microphone at recording time. This is particularly the case when we have no idea how the performance was mic'ed and mixed.
If we just place a single microphone or even a stereo microphone in front of an orchestra at one location, its not going to capture any HRTF cues at high frequencies, thus the main cue for height at ~8Khz will not be there. It's possible a small height cue in the low midrange may be captured, but its unclear how strong this cue might be in isolation, probably not strong enough to account for the very convincing impression you've noted.
I don't know what the usual recording technique for orchestras is, but I don't think it uses a single microphone pickup location - from what I've read some have microphones hanging from wires above the audience to capture overall performance and ambience, plus some closer (but not "close" in the studio sense) to specific instruments to allow them to be brought forward in the mix or not. There may be some time delays applied to separate mic's as well, to bring them closer into time alignment.
The point is, none of these microphones will have a head HRTF function, and you are mixing various mic's at different locations which have different height perspectives of a given instrument together to get the final performance. If there is a strong sense of elevation in the final reproduced result for some instruments, I think it's almost certainly a product of mixing technique, as clever or realistic as it may sound. (Just like the convincing lateral localisation of instruments in many recordings is entirely created within the mixing process, and not captured by the mic)
If you have an example of a recording with convincing height cues which you know for certain the exact mic'ing and mixing method used, we might be able to go beyond speculation.
Last edited:
My comment on this is, when tuning speakers I experienced that the height increase with increase so a louder part of the HF in the music.
No difference in reflection that created the change in this effect. So finding the right balance in the response is crucial.
Last edited:
With modern convolving reverbs you can recreate the characteristics of any room you like.
If you choose a high but narrow cathedral to emulate that is what you will get.
Or an underground parking lot which would be long, wide and low.
On the topic of height, I was watching a movie tonight (I only have stereo no surround) and there was a scene, inside a house with a thunderstorm outside. The thunder was quite distant and not particularly loud, but it sounded like it was coming from outside my house and way higher than the ceiling. It was VERY convincing, It really sounded like there was a thunderstorm in the distance outside.
Tony.
Tony.
That happens to me a lot in recordings that have weather or birds, cars, doorbells, muffled conversation from other rooms, etc. I think some of it is your brain being used to hearing these sounds naturally, so if a recorded one is good enough, your brain places it there where it normally is.
For instance, in this house, I commonly hear muffled conversation from the kitchen, so when I hear well recorded muffled conversation in a movie, it always seems to come from the kitchen. Car doors closing always seem to come from the locations around the house where cars park, etc.
I do agree that a lot of it has to do with were you expect the sounds to be. That's gotta play a big role in natural sound, too.
By why do some recording and some systems do it, and others not? We just got two comments saying that the displacement happens when it's well recorded. What makes it well recorded? And why are some systems better at it than others?
Sure, I know how a choir and orchestra are laid out - so why did this one particular recording on this system sound so accurate, so real, when others do not? Just a coincidence of room, recording and speakers? If it's just a mind trick, why doesn't it happen more often? My guess: The system got out of the way of the recording and let us hear what was there. It did so on all recordings, but some recordings are more spectacular than others.
By why do some recording and some systems do it, and others not? We just got two comments saying that the displacement happens when it's well recorded. What makes it well recorded? And why are some systems better at it than others?
Sure, I know how a choir and orchestra are laid out - so why did this one particular recording on this system sound so accurate, so real, when others do not? Just a coincidence of room, recording and speakers? If it's just a mind trick, why doesn't it happen more often? My guess: The system got out of the way of the recording and let us hear what was there. It did so on all recordings, but some recordings are more spectacular than others.
There are many things, which make a recording more
"spectacular" (meant in the sense of "believable" here).
Some which come to my mind quickly:
- low bass content (there seems to be some "rumble" in many halls, even
if just the violin is playing or there is "silence" ...)
- side noises e.g. from instrumentalists, sheet of music falling to the ground, timber piles crackling before
cello plays solo e.g. (very nice effect, i do have some favourite crackles...), whisper from the audience,
"believable applause" ... etc.
- determined musical fore- and background (according to
apparent distance of singers and instruments)
Concerning the latter there are e.g. some DECCA recordings i like very much.
"spectacular" (meant in the sense of "believable" here).
Some which come to my mind quickly:
- low bass content (there seems to be some "rumble" in many halls, even
if just the violin is playing or there is "silence" ...)
- side noises e.g. from instrumentalists, sheet of music falling to the ground, timber piles crackling before
cello plays solo e.g. (very nice effect, i do have some favourite crackles...), whisper from the audience,
"believable applause" ... etc.
- determined musical fore- and background (according to
apparent distance of singers and instruments)
Concerning the latter there are e.g. some DECCA recordings i like very much.
Last edited:
Don't forget that a lot of soundtracks, in addition to AC3 5.1 tracks, will Dolby Surround encode the stereo soundtrack (for pro-logic decoding). This can add a lot of out-of-phase info for effects, that will always image well off screen.
David S.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.