That could be another interesting thing - if the tilt has something to do with the value of the DI itself (high x low).
The higher the DI the stronger the direct field in general. Don't we want that to be flat?
- What it could also mean, BTW, is that the higher the average DI the more should be the DI curve actually tilted (upward in this case) to still have a flat LW and the "preferred" downward tilted PIR (if that's really the target). Otherwise you would be forced to tilt the LW. The more dominant the direct sound (LW) is the more weight it has in the PIR response as well.
The higher the DI the stronger the direct field in general. Don't we want that to be flat?
- What it could also mean, BTW, is that the higher the average DI the more should be the DI curve actually tilted (upward in this case) to still have a flat LW and the "preferred" downward tilted PIR (if that's really the target). Otherwise you would be forced to tilt the LW. The more dominant the direct sound (LW) is the more weight it has in the PIR response as well.
Last edited:
BTW, I've managed to run my own ABEC/VACS GUI automation via AutoIt so I will also be able to run automated optimizations soon. That's pretty exciting. I'm curious how many iterations I will be able to get over a night
Nice tip! Excited to see any progress in this area.
@ Maiky: as I interpret the curves, the KEF PIR has a sightly more tilted curve, leading to 2-3 dB lower level @10 Khz then the Revell. (Both are bluddy close anyway).
Question: would it not be easy to re-design the KEF x/o in order to make the PIR equal to the Revell? Or would the Uni-Q cone profile, acting as waveguide, be the deal-breaker here?
Question: would it not be easy to re-design the KEF x/o in order to make the PIR equal to the Revell? Or would the Uni-Q cone profile, acting as waveguide, be the deal-breaker here?
Meaningful comments that I fully endorse.
It begs the question: "how smooth is smooth enough?"
BTW, I've managed to run my own ABEC/VACS GUI automation via AutoIt so I will also be able to run automated optimizations soon. That's pretty exciting. I'm curious how many iterations I will be able to get over a night
This is really cool, mabat!
Last edited:
Toole and Olive did research the audibility of resonances years ago and I think it's still valid.
Why don't you just tell him?
In the meantime, my take on this: Strictly speaking, you could incorporate just any EQ into the crossover, shaping any chosen curve (PIR, LW, SP, etc). But that's not the point. By that you would also automatically shape all the other curves in exactly the same way and that's usually not what you want. The curves are all tied together by directivity and that's an inherent three-dimensional property of the acoustic source itself you can't do anything about. In crossover design you have only a limited means to affect directivity - only in the overlap regions and then you can only try so it's not worse than it has to be, than anything else.
That's the reason why we talk such much about DI, etc. - once this is given, you can't do much about it.
- Oh, and did I mention that you can't shape the DI curve by EQ? This curve (basically the difference of LW and SP) will stay exactly the same no matter what you do.
Last edited:
I can't say if it would be easy to redesign the KEF's XO, but you could just connect the driver to your own active DSP setup to adjust / correct factory settings (the presence boost about 2-5k ao).
What's more, you may ask yourself what's the purpose of modifying a loudspeaker that incorporates an enormous amount of R&D and the DI of which largely depends on the design of the waveguide, the Tangerine phase plug, the dome, cone etc.
Since commercial loudspeakers are being compared here, I would suggest to buy a loudspeaker with a PIR closer to your preference.
Last edited:
BTW, I've managed to run my own ABEC/VACS GUI automation via AutoIt so I will also be able to run automated optimizations soon. That's pretty exciting. I'm curious how many iterations I will be able to get over a night
Interesting! Could you elaborate on this a bit please?
A high DI means that there will be a greater "direct sound" contribution than early reflections. This will improve imaging. So I design for a high DI. In my last two designs, the NS-15 and the NA-12 the responses were nearly identical, but the DI was higher for the NS-15 because the woofer had a higher DI being 15" versus 12". Customers who have had and heard both say that the perceived differences are small, but the higher DI always wins out. These are not scientific studies though, and none have ever been done so that's all the data that I have.
That the DI should be smooth is a given, but that it should be flat results from wanting the direct sound and the reverberant sound to have the same tembre. So high and flat is my goal.
I'll cover the fall in listening axis response below.
I have tried to explain the "why" many times before, but here it is again.
A wider coverage angloe at HFs, which is what a flat DI will result in, yields more HF energy into the room. This means that a high DI will sound bright if the listening axis is flat when compared to a rising DI (as most speakers have.) (The link between DI and directivity was alluded to in some previous posts.) Does this judgement stem from "habit"? (Everything previously had a rising DI so that's what we are "used" to.) I think that highly unlikely because a falling axial response seems to be judged more "neutral" almost universally. It could stem from the fact that air absorption grows (dramatically) with frequency thus yielding our world to be one with a falling HF response. Some of our "distance" judgement comes from this fact. That's why too much HF sounds like the music is "in your face".
I'd like to warn that one cannot consider the greater direct field response from a high DI to be commensurate with the PIR response. That's because in the context that I am using "direct" here it means before significant early reflections, hence "direct". But acoustically this word means something different. It is the region of the sound field that is dominated by the sound directly from the source and not the reverberation field. This later definition applies ONLY to steady state responses, not transient ones. We should be careful not to mix up what we mean when we talk of "direct". The reverberant field will not rise at HF as the DI grows (depending on other parameters such as power response and axial response,) although the early sound to reverberant sound will increase with DI.
One of the reasons that I have always had concerns about the Toole-Olive approach is that they do not consider "imaging" in any of their work (not even Toole's book does.) To me this is a failing since it is so important to me - at least.
What would you like to know? I'm still working on that. Basically it allows you to automate all the mouse clicking and keyboard typing you would have to do manually if you wanted to do such task with ABEC. That's one part. The second part is the logic behind an assesment of the results. I've talked about that already - my idea is to use DI as an optimization target for the moment.
So it allows you to completely automate the repeated/iterative process of generating a profile, running a BEM analysis, evaluating the results and saving them for presentation. My vision is to set a target response in the evening, go to bed and at the morning see what's the best (or necessary) shape and size for that
Hopefully (it may be slightly more complicated than that, but not much)...Axisymmetric freestanding waveguides considred at the moment as these are really good and extremely easy to work with.
Last edited:
Maybe the downtilt preference is a result of all the close miking i.e. the sad realisation that the recording industry and stereo system is a big fail re *hifi*. Putting the mic to close also change the character of instruments and ruin the natural mix of the individual contributions in an orchestra. And as gedlee indicate - we miss out the "air absorption" in the recordings due to close miking.
It's all a mess - and here are we, being picky on quarter of a dB...
Sp why not put the mic where people sit - 10th row... no-no, you cant do that - that sounds like sh*t... eee - maybe it shouldn't and its due to the reproduction side?
//
It's all a mess
- and here are we, being picky on quarter of a dB...Sp why not put the mic where people sit - 10th row... no-no, you cant do that - that sounds like sh*t... eee - maybe it shouldn't and its due to the reproduction side?
//
Last edited:
Well that's all up to a guy who makes the recording and who mixes the spectral balance that it sounds "just right". But he uses some loudspeakers for that. Do they have a neutral spectral balance, and what is that? How these loudspeakers should be designed? That's what the circle of confusion is all about. It's not just a habit.
They might have a neutral spectral balance. But how the Mic is placed is probably a big part of it and in turn, have impacted the control speakers....
So here we see how the *whole* chain interacts....
There need to be a closed calibration loop - a recorded standard "clap" (from stage) at recording venue golden seat (studio productions need to improvise and make a virtual one) included hidden in every song/album to be used at replay side for instant calibration of replay system so to recreate the mechanical reference "clap" perfectly as original.
//
So here we see how the *whole* chain interacts....
There need to be a closed calibration loop - a recorded standard "clap" (from stage) at recording venue golden seat (studio productions need to improvise and make a virtual one) included hidden in every song/album to be used at replay side for instant calibration of replay system so to recreate the mechanical reference "clap" perfectly as original.
//
Last edited:
Is it impossible to do something for:
21/2” SB65WBAC25-4 / Aluminum – Sbacoustics
It is a very nice driver at low cost.
40 cm is OK
//
21/2” SB65WBAC25-4 / Aluminum – Sbacoustics
It is a very nice driver at low cost.
40 cm is OK
//
Hey, rock'n roll for example is about having fun, live fast die young, there is no point tweak the mic position when the loud message is delivered The enjoyment comes through and the enjoyment follows, even with crappy recordings and even with crappy speakers. I'm sure there are plenty of very good recordings (never mind the mic placement) as there are the poor ones across all genre. Besides, lots of modern rock is done one performer at a time, very polished stuff, I don't subscribe to that but many do! People making art for the fellow people. Sometimes some explore more ground and sounds evolve. Good speakers are really much fun for trying to get part of that and see how much good music I can find, they give some extra impact.
The enjoyment comes through and the enjoyment follows, even with crappy recordings and even with crappy speakers. I'm sure there are plenty of very good recordings (never mind the mic placement) as there are the poor ones across all genre. Besides, lots of modern rock is done one performer at a time, very polished stuff, I don't subscribe to that but many do! People making art for the fellow people. Sometimes some explore more ground and sounds evolve. Good speakers are really much fun for trying to get part of that and see how much good music I can find, they give some extra impact.
Last edited:
I was thinking mostly about orchestra music like:
Beethoven Symphony No 5 C minor John Eliot Gardiner Orchestre Revolutionnaire et Romantique 2016 - YouTube
//
Beethoven Symphony No 5 C minor John Eliot Gardiner Orchestre Revolutionnaire et Romantique 2016 - YouTube
//
I hope I am not pestering everyone with my posts…
Toole 3rd Edit. p350/351 about the 2013 target:
[…]More data would be enlightening, but this amount is sufficient to indicate that a single target curve is not likely to satisfy all listeners all the time. Because these preferred curves were subjectively determined, they include the “circle of confusion” issues. Just as they changed with listener experience and training, they are likely to change with different programs. It is proof that we need easily accessible bass, treble and/or tilt tone controls in playback equipment. The first task for such controls would be to allow users to optimize the spectral balance of their loudspeakers in their rooms, and, on an ongoing basis, to compensate for spectral imbalances as they appear in movies and music.[…]
(The proposed EQ strategy is what we do in headphones now)
Until such time, maybe not so far in the future, as the audio reproducing apparatus (not necessarily a set of loudspeakers) can adapt itself to
the media (not necessarily music) it reproduces (think emphasis / de-emphasis on CD players), and
IF there is no escaping the circle of confusion for now, then shouldn’t we embrace that circle and design loudspeakers accordingly?
What would be the point of making the “technically perfect” speaker, whatever that mean, if a lot of the music played on it sounded horrible as opposed to
making a speaker that might not be perfect but does reproduces the music the way is was intended?
You can design state-of-the-art waveguides with any reasonable and smooth DI with reasonably flat ON/LW, that is great no question about it.
Naturally however, the question is then what would that DI should be to maximize the performance of the speaker?
The DI design becomes synonymous with PIR target, unless of course you suggest that the smoothness is the ruling factor.
Remember, any model is a simplification but as stated earlier nothing prevents us from checking its predictions agains new observations.
I started with extrapolating the 2003 Olive model for predicting the preference for speakers.
What came from that is, at the limit, the model would rate a “flat everything speaker” higher than one with a steeper slope.
At the same time however, the model would rate a speaker with a steeper PIR slope higher than a still “smooth and flattish” one
but with a less that perfect “flat everything” even if closer to the preferred PIR.
A “flat everything speaker” is, to me, a better approximation of the example (Revel) closer to the preferred PIR than the one with the steeper slope (Kef).
But that is debatable I guess.
Both “DI types” can produce high score but as new observations came in (new generations of speakers),
it seems that the model limit prediction, the “flat everything speaker”, is a better approximation of the high rated speakers (Revel).
I might be straying too far there but, IF the Harman team was blindly following the model,
THEN the speakers they built should have had a DI similar to that of the Kef speaker,
UNLESS the limit case is a “better” approximation of the preferred PIR.
The model was published in 2003 and by 2013 the general PIR target shape to be achieved was done and dusted,
especially after they found it consistent with their separate work on headphones.
Many of the Loudspeakers produced by the Harman group (Pro and Consumer) are now designed with it in mind.
The flattish character of the DI being consistent but the level and bandwidth depending on the price bracket and intended usage.
Now, the same way that we can ask: how smooth is smooth enough? We can ask: how flat is flat enough?
The mock-up Spinorama I shared (which is still rather self consistent) does not have a perfectly flat DI
but flat looks to my eyes a better approximation to the “smooth and flattish” speaker I drawn, consistent with the limit case of the model.
But that is debatable again I guess.
As Dr Geddes pointed out (thanks for your input) the imaging, the Olive model paper states that:
[…]we found that timbre-related attributes accounted for 94% of the variance in comments whereas nonlinear distortion and spatial-related attributes only contributed 3% each. It is our experience that timbre is the dominant factor related to loudspeaker preference, and speakers that accurately reproduce timbre generally have favorable spatial properties.[…]
I would expect the Headphone and Loudspeaker target curve to be consistent, maybe not identical but definitely consistent,
same sensory path same organ, same function or am I delusional?
I would think that the imaging difference between headphones and Loudspeakers would be an important difference although Xfeed can go a long way.
Not even talking about binaral recordings/media that will change everything.
Stereo widening system using binaural cues for headphones
(PDF) Stereo widening system using binaural cues for headphones
Characterizing the Frequency Response of Headphones – a new Paradigm
Why would you think they ought to be different?
Toole 3rd Edit. p350/351 about the 2013 target:
[…]More data would be enlightening, but this amount is sufficient to indicate that a single target curve is not likely to satisfy all listeners all the time. Because these preferred curves were subjectively determined, they include the “circle of confusion” issues. Just as they changed with listener experience and training, they are likely to change with different programs. It is proof that we need easily accessible bass, treble and/or tilt tone controls in playback equipment. The first task for such controls would be to allow users to optimize the spectral balance of their loudspeakers in their rooms, and, on an ongoing basis, to compensate for spectral imbalances as they appear in movies and music.[…]
(The proposed EQ strategy is what we do in headphones now)
Until such time, maybe not so far in the future, as the audio reproducing apparatus (not necessarily a set of loudspeakers) can adapt itself to
the media (not necessarily music) it reproduces (think emphasis / de-emphasis on CD players), and
IF there is no escaping the circle of confusion for now, then shouldn’t we embrace that circle and design loudspeakers accordingly?
What would be the point of making the “technically perfect” speaker, whatever that mean, if a lot of the music played on it sounded horrible as opposed to
making a speaker that might not be perfect but does reproduces the music the way is was intended?
You can design state-of-the-art waveguides with any reasonable and smooth DI with reasonably flat ON/LW, that is great no question about it.
Naturally however, the question is then what would that DI should be to maximize the performance of the speaker?
The DI design becomes synonymous with PIR target, unless of course you suggest that the smoothness is the ruling factor.
Remember, any model is a simplification but as stated earlier nothing prevents us from checking its predictions agains new observations.
I started with extrapolating the 2003 Olive model for predicting the preference for speakers.
What came from that is, at the limit, the model would rate a “flat everything speaker” higher than one with a steeper slope.
At the same time however, the model would rate a speaker with a steeper PIR slope higher than a still “smooth and flattish” one
but with a less that perfect “flat everything” even if closer to the preferred PIR.
A “flat everything speaker” is, to me, a better approximation of the example (Revel) closer to the preferred PIR than the one with the steeper slope (Kef).
But that is debatable I guess.
Both “DI types” can produce high score but as new observations came in (new generations of speakers),
it seems that the model limit prediction, the “flat everything speaker”, is a better approximation of the high rated speakers (Revel).
I might be straying too far there but, IF the Harman team was blindly following the model,
THEN the speakers they built should have had a DI similar to that of the Kef speaker,
UNLESS the limit case is a “better” approximation of the preferred PIR.
The model was published in 2003 and by 2013 the general PIR target shape to be achieved was done and dusted,
especially after they found it consistent with their separate work on headphones.
Many of the Loudspeakers produced by the Harman group (Pro and Consumer) are now designed with it in mind.
The flattish character of the DI being consistent but the level and bandwidth depending on the price bracket and intended usage.
Now, the same way that we can ask: how smooth is smooth enough? We can ask: how flat is flat enough?
The mock-up Spinorama I shared (which is still rather self consistent) does not have a perfectly flat DI
but flat looks to my eyes a better approximation to the “smooth and flattish” speaker I drawn, consistent with the limit case of the model.
But that is debatable again I guess.
As Dr Geddes pointed out (thanks for your input) the imaging, the Olive model paper states that:
[…]we found that timbre-related attributes accounted for 94% of the variance in comments whereas nonlinear distortion and spatial-related attributes only contributed 3% each. It is our experience that timbre is the dominant factor related to loudspeaker preference, and speakers that accurately reproduce timbre generally have favorable spatial properties.[…]
I would expect the Headphone and Loudspeaker target curve to be consistent, maybe not identical but definitely consistent,
same sensory path same organ, same function or am I delusional?
I would think that the imaging difference between headphones and Loudspeakers would be an important difference although Xfeed can go a long way.
Not even talking about binaral recordings/media that will change everything.
Stereo widening system using binaural cues for headphones
(PDF) Stereo widening system using binaural cues for headphones
Characterizing the Frequency Response of Headphones – a new Paradigm
Why would you think they ought to be different?