I view baffle step as an acoustic load issue. When the load changes, the actual output/efficiency of the driver changes. Many refer to the issue as sound energy spreading out around the baffle, an thus lower energy in front. This may be also true, but there is a loss of power as well.
Its half and half. The directivity index increases by 3db above the baffle step frequency because the baffle forms a 180 degree horn. The remaining 3dB of on axis increase comes from a 3dB increase in power response due to the change in loading, as you suggest.
I have not looked into this in any detail, but I would not think that the loading change would be sufficient to increase the output by 3 dB. Wouldn't that mean the loading doubles? At any rate arguing the details is not useful, the fact is that the change is from two sources that can vary in amount depending on the shape of the enclosure. At long wavelength the entire enclosure is the diffraction source, then it becomes just the front baffle and this depends on the nature of the baffle edges.
The bottom line is that the baffle step is a diffraction phenomena and does differ in its effect depending on polar location and hence it cannot be globally "fixed", whatever "fixed" means.
In my polar maps I always see problems that cannot be globally corrected right around the point where a "baffle step" should be occurring (according to theory). But as I have said since I used data that contains all of the enclosure diffractions, I am optimizing the total polar response for these diffractions. The baffle step is automatically accounted for.
The bottom line is that the baffle step is a diffraction phenomena and does differ in its effect depending on polar location and hence it cannot be globally "fixed", whatever "fixed" means.
In my polar maps I always see problems that cannot be globally corrected right around the point where a "baffle step" should be occurring (according to theory). But as I have said since I used data that contains all of the enclosure diffractions, I am optimizing the total polar response for these diffractions. The baffle step is automatically accounted for.
Agreed. This is why you don't equalize back the 6db on axis response due to baffle loss the full 6db. It will bump up the power response 3db. Equalizing back 3db of FR drop due to baffle loss will keep the power response flat.
Not only that, but as you go lower in frequency, the acoustic load changes again due to floor coupling or other adjacent boundary or mutual coupling between drivers, giving you "room gain", etc. This is why I approach the baffle loss phenomenon from an acoustic load perspective, and why simple FR equalization isn't the simple cure.
Last edited:
I'm not suggesting trying to fix diffraction with EQ, clearly that doesn't work at high frequencies where the interference pattern is dense enough to cause lobing within the listening window.
As to whether its useful knowing whether diffraction is minimum phase or not - I suggest it is useful, to help better understand the audibility of diffraction, and why it sounds the way it does.
Your study into the audibility of diffraction suggest that not only is it audible and unpleasant sounding (I would describe it as harsh and fatiguing) but that its audibility increases with SPL are quite interesting but don't address a key point - WHAT exactly is it about the sound of diffraction that we don't like. We can hear it, we don't like it, but why ?
Although its a spatial phenomenon with a different possible response at each listening point, I think that everything that sounds bad about diffraction can be perceived at any single point in space. This is also suggested by your study which if I remember right relied on electronically simulated diffraction by mixing in a delayed (and inverted ?) signal with the original signal, with differing mix ratios.
If diffraction is minimum phase (which I believe it is until proven otherwise) then it tells us something useful - everything we need to know about why diffraction sounds bad is contained in the amplitude response alone, and that if we arrive at a similar minimum phase amplitude response through some means other than diffraction, it too will sound bad and have the same perceptual sensitivity to SPL.
I'll cut to the chase, I suspect that diffraction sounds bad due to the narrow band peaks and dips in the amplitude response that result from the comb filtering of delayed signal(s) being mixed in. It's not the delayed signal that's the problem per se, its the resulting sawtooth like frequency response that results that irritates us.
Those narrow peaks and dips may be narrower than the critical bands of our hearing and thus average out as far as tonal balance on broad spectrum noise is concerned, but that doesn't make them inaudible as they are easily shown up by discrete signals or a sine sweep.
There is some comparison here to multiple closely spaced high Q resonances, as sometimes occurs with cone breakup which can result in a similar zig zag up down amplitude response that is perhaps not identical to but somewhat similar to comb filtering.
I think most people try to draw a clear distinction between resonances (energy storage) and comb filtering from two summed but differently delayed signals (no energy storage) but KSTR's post about the duality of comb filtering and EQ was a real eye opener.
He was able to generate a comb filter by adding a delayed and attenuated copy of a signal, measure that it was indeed minimum phase, then devise a second filter consisting of many standard minimum phase PEQs to simulate the same amplitude response, then confirm that the resulting impulse response of the two filters was in fact the same.
On a practical level this suggests to me that there is nothing fundamentally different about the harshness that results from a series of closely spaced high Q resonances and the harshness that results from comb filtering from diffraction. (Or other source of comb filtering for that matter)
The cause of the two is very different, as is the cure, but fundamentally they're perceived and disliked by our brains in a very similar way, and both are therefore perceptually SPL dependant.
I've often seen it stated as fact that cone breakup "gets worse" at higher SPL, implying that somehow the damping reduces at higher SPL...which would imply that the frequency response changes significantly around those resonances at higher SPL, but is this really the case ?
Could it not be that the exact same frequency response (including resonances) becomes perceptually more noticeable and obnoxious at higher SPL in the same way that your study shows diffraction becomes perceptually more obnoxious at higher SPL ?
My belief is its these steep narrow band frequency response changes that are responsible for the listener fatigue and characteristic harsh sound of both resonances/breakup and diffraction.
In other words a smooth (low slope) response is more important than a flat response, as Toole has suggested.
For one, a mountain range shaped frequency response will cause a lot of amplitude modulation of frequency modulated signals - the classic example is strong vibrato on a female vocalist - if it happens to fall on the edge of a resonance or other steep amplitude slope the resulting amplitude modulation is obvious and very unpleasant and fatiguing. The same vibrato on a system that is smooth and flat is much easier on the ear even at higher SPL.
Now imagine massed strings sweeping across a whole series of resonances and or other narrow band amplitude response disturbances and its no wonder it can sound harsh and fatiguing.
In my experimenting I've experienced big improvements in sound quality (smoother, cleaner, lack of harshness) both from minimising tweeter diffraction and carefully dealing with narrowband response anomalies from the driver like resonances, both through damping changes to the driver, and specific corrective EQ of specific driver resonances.
The thing that's special about diffraction is that the narrow band response anomalies that it causes can only be treated in the physical domain unlike driver resonances...
A couple of comments:
The load on the woofer does vary from below the transition frequency to above. It is "a piston on the end of an infinite tube" vs. "a piston on an infinite plane", the 2 classic cases for which acoustic impedance are different but well known.
We seem to be making up new rules. "you can't equalize if the system isn't minimum phase" and " you can't equalize if diffraction is involved". Again, we are letting the perfect be the enemy of the good. The only rule should be "if the effect is audible, equalize it".
Baffle step is a combination of transition between loadings and time related diffraction effects. Certainly the fine structure of diffraction related reflections will be specific to each point of observation, but that doesn't prevent us from giving a general correction that works for the range of positions we expect to occupy. We also know that we can reduce diffraction effects greatly with careful choice of dimensions, asymmetry, radiused corners, etc. Once that is done the residual response effects are simply the loading transitions and can clearly be equalized.
These effects are important in the direct field and vary somewhat slowly with position. EQ can have a positive effect over a reasonable window, so it should be considered as an improvement.
Regards,
David
The load on the woofer does vary from below the transition frequency to above. It is "a piston on the end of an infinite tube" vs. "a piston on an infinite plane", the 2 classic cases for which acoustic impedance are different but well known.
We seem to be making up new rules. "you can't equalize if the system isn't minimum phase" and " you can't equalize if diffraction is involved". Again, we are letting the perfect be the enemy of the good. The only rule should be "if the effect is audible, equalize it".
Baffle step is a combination of transition between loadings and time related diffraction effects. Certainly the fine structure of diffraction related reflections will be specific to each point of observation, but that doesn't prevent us from giving a general correction that works for the range of positions we expect to occupy. We also know that we can reduce diffraction effects greatly with careful choice of dimensions, asymmetry, radiused corners, etc. Once that is done the residual response effects are simply the loading transitions and can clearly be equalized.
These effects are important in the direct field and vary somewhat slowly with position. EQ can have a positive effect over a reasonable window, so it should be considered as an improvement.
Regards,
David
Simon - I basically don't have a problem with what you are saying. I was not as interested in the "why" as I was with the nature of the effects and what they meant for audibility. That the effects act similar to other effects is fine with me. The bottom line is "what to do?" and that becomes clear. And, from experience, if you do those things the speakers sound better - what's not to like?
Dave - EQ is fine as long as you EQ the right things. That's not as easy to determine as it sounds. I EQ, most people I know do as well, but I do believe that I do it differently than most.
"you can't equalize if diffraction is involved" is, I believe, correct if you just do it at a single place, like on-axis, or a single point in a room. As to "if the effect is audible, equalize it" - as long as that doesn't make things worse, and it can if not done properly.
The point about diffraction is that it cannot be completely resolved with EQ like a resonance can, and it is far more effective to correct diffraction acoustically than it is to do it electronically.
Dave - EQ is fine as long as you EQ the right things. That's not as easy to determine as it sounds. I EQ, most people I know do as well, but I do believe that I do it differently than most.
"you can't equalize if diffraction is involved" is, I believe, correct if you just do it at a single place, like on-axis, or a single point in a room. As to "if the effect is audible, equalize it" - as long as that doesn't make things worse, and it can if not done properly.
The point about diffraction is that it cannot be completely resolved with EQ like a resonance can, and it is far more effective to correct diffraction acoustically than it is to do it electronically.
Spread in space produces unpredictable results because the room factor,is always a variable. Measuring close to listening axis, goves you understanding how direct sound wil be influenced. I think if you do enough cross relation studies, you can better understand it. Would love to see your measurements. I think studying diffraction spread is useless because of the room variation, but maybe you have more studies to show otherwise?
Last edited:
I think you forget that the driver has to have an SPL,response close to the power response to conduct EQ for near field SPL. Once you can do that with a driver, then you need to understand the way Ultimate Equalizer software does equalization, in this case using near field reduce how diffraction effect the EQing. The reason why you want to reduce EQing of diffraction is because EQing such is only perfect at one point, and we cannot position our ears precisely on that point during listening, in such case results in duplicates diffraction effects to the direct sound we hear.
As you can see, the higher fidelity one desired, the more integrated design of every part is the most important factor.
So what is your point? You still use it in rooms of uncertainty and results are still unpredictable. The best you can hope for is just to reduce it as much a you feel worth the additional cost and effort; your second choice would be spread it as randomly as possible, ENABLE 😀 like patterns or curly hair, etc., the further away from the source the less effective it is.
If you want to study the effect then you would need to look at it in isolation. That's my point.
You've missed my point though. Near-field measurement data at high frequencies is completely invalid and bears no similarity to the drivers actual far-field response.
Therefore if you EQ high frequencies based on a near-field measurement the EQ correction is also invalid. You're EQ'ing response aberrations that don't really exist - except right in front of the dust cap.
If it was valid, we'd all be using near field measurements instead of gated (or anechoic) measurements taken at a distance! 😉 As it is, near-field data is only used for bass response measurements up to about 200Hz depending on driver size.
It's invalid at high frequencies because when you get too close the time delay (and therefore phase) of the radiation from each point on the cone surface to the listening point changes (due to geometry) relative to what it is when the listening point is approaching infinity, so the summed response is different. This leads to notches etc in the response which don't actually exist in the far-field where the listener is.
For a more rigorous proof have a look at the first few pages of this article:
http://www.artalabs.hr/AppNotes/AP4_FreeField-Rev03eng.pdf
Notice in particular figure 4 which shows where the 3x diameter (6x radius in the text) minimum measurement distance requirement comes from. Also note the response graphs towards the end of the article (pages 8-10) of the near-field and gated far-field response of the same driver on the same baffle - totally and utterly different at high frequencies.
Even if you could somehow measure the high frequency response accurately in a near-field measurement (which you can't) a near-field measurement will also not show the baffle step effect, so your EQ will not include any baffle step correction. This too is clearly wrong, unless you plan to hold your ear in the near-field as well...
Last edited:
DBMandrake, I think you need to look at the data I posted and talk about the frequency response of that to talk about high frequency. I hate to just speculate without looking at some data. If you have some data that shows your point of view, I certainly would be interested in seeing it. The reason I don't discuss data from a third person is that we get stuck when we ask about certain specifics which become unknown. I call near field within 5mm of the diaphragm, for a set of complicated reasons I am not interested to elaborate here to keep typing to minimum until other people show their own data. What I can say is, you just pick the measurement method that gives you the best result both technically and in real listening. Different people are picky about different aspects.
Basically, if EQ is not working right, you can do different things to adjust the low end till you feels is adequate, and flat is not the best.
Basically, if EQ is not working right, you can do different things to adjust the low end till you feels is adequate, and flat is not the best.
Last edited:
I've looked at your data already, however as soon as I got to this sentence:
and also saw that all your measurements go from 500Hz right up past 20Khz, I knew straight away that your methodology is flawed.
I'm trying to point out that taking a high frequency near-field measurement and then using it as a basis for calculating EQ is invalid. You can show all the data you like but if the methodology is fundamentally flawed in this way the data means nothing.
Again we are creating a rule that we shouldn't EQ something if it varies in space. Who's rule is that?
Sure, correcting the causal issue is preferable, but if not, then it can be EQed to the extent that it is constant through your listening window.
For example, floor bounce is audible and typically gives a large dip in the middle hundreds. Should we EQ it? The EQ is erroneous if we put our ear to the floor? Frequently LF room EQ is counterproductive if we move well away from the initial measuring point. Is room EQ bad?
In every case it makes sense to define the extent of a likely listening window and then find the average frequency response across that window. EQ to that average and you will find that you have made a useful improvement.
Knowing the cause of response aberations is worthwhile as in distinguishing between resonances (universal across position) and reflections (effect varies with position). Still, the arguement that one should be corrected and the other mustn't doesn't make sense if both are audible.
David
So do you see wrong in both final equalized results both measured at the same distance? Even though equalization was based on different data? Let's see if we can have a reasonable explanation between EQ basis, and listening impression, versus why you say EQing based on near field measurement is wrong. Surely you have a good explanation.
David, I am not saying anything is a rule, I am just explaining what will happen if you do that. Anyone can share their experience and data on the contrary if they have done the research. Personally, I find EQing anything caused by reflection and diffraction will cause some adverse results, so I avoid it.
If diffraction is a problem, then solving it at the cause is the best way.
If diffraction is a problem, then solving it at the cause is the best way.
Last edited:
Nearfield, as in dustcap measurements, are just not representative of farfield response. Above the piston band you are sampling one position of many while the far field is the summation of all points. For low frequencies the nearfield response ignores the changing loading of the baffle.
David S
David S
Of coarse it is necessary to fully understand all these which may influence quality of reproduction. As you can see, all the measurements I have posted were taken with the driver in the enclosure. Changing load of the baffle would relate with baffle step? On a 17cm wide baffle, it seems to me this is not a problem. The far field response seems quite nice if not perfect, but there is a difference in the overall sonic reproduction as I have explained. But of course, different people will focus on different issues. Additionally, different drivers need to be measured at different locations of the surface, it is also true that you cannot use this technique on lots of drivers. So it is important to understand the vibration characteristics of the diaphragm, and decide whether or how to apply the measurement.
Last edited: