Baffle Diffraction

[bentoronto]

So, have you read the ones that we did?

Apologies for my ignorance. I don't know about those publications. Appreciate links to the journal(s). Thanks.

B.

 

[Juhazi]

It took me some 15 minutes to shoot these for ya. The room is 3x4x2,7m.5ms gating here. Green lines with grille off.

I can see the grille's effect, do you?

The speaker is Amphion Helium II

 

[Juhazi]

ps. the response is not "natural", I forgot to disable eq set for the regular speaker used in this room. Minidsp 2x4HD for eq and subwoofer xo.

 

[Rokytheman]

What makes you think grill edge diffraction only affects the tweeters frequency range on a 39cm wide 59cm tall baffle ? On that size baffle it will have effects far below the tweeter's operating range.

Even with the speaker up on a stand in the middle of a room the longest reflection free time window I can get for measurements is about 4.2ms which only provides a resolution of about 300Hz.

Where the speakers normally sit in the living room near other objects the reflection free time is well under 2ms, perhaps 1.5ms making any such measurements pretty useless, especially when the whole point of the exercise is to find diffraction effects belonging to the speaker itself and its grill, not nearby objects right beside the speakers!

So sorry, I don't agree with your comment - there is no point taking measurements if their validity is completely questionable. I see enough poorly taken measurements on this forum already by people that don't understand how to properly perform a windowed measurement that is free from room effects that I am not going to add to them with obviously dubious measurements that prove nothing.

Bogus measurements are worse than no measurements at all. When I get the chance to take some suitable measurements I will. Until then you'll have to take your own!

I did not write about you posting some measurements

Not trying to be offensive or anything. I dont understand the issue. You can measure down to 300 Hz in your room. More than needed to see the diffraction at stake. Not saying that it is easy to get a valid full in room response. But the diffraction issues is not the most difficult part.

 

[bentoronto]

It took me some 15 minutes to shoot these...I can see the grille's effect, do you?

Thanks very much. Picture of the grill?

Looks like there is a 1 to 2dB influence over a small fraction of an octave. Even with instantaneous switching between states (if that were possible playing with a grill), likely you would not consistently hear a difference in tone using white noise.

But with random blind testing with white noise unlikely anybody could guess well. And on music, not a chance.

For sure, after a person has done all the important enhancements to their speakers, subs, DSP, and tri-amps, they should consider diffraction.

Thanks.


B.

 

[gedlee]

The papers are on my website.

I would put diffraction much higher than you do, and I have studied it a lot. "Tri-amps"!? I wouldn't even claim bi-amp as a significant advantage, nor "DSP".

 

[Juhazi]

Amphion's grille is rather nice actually. The waveguide helps too to minimize diffractions.

 

[cowanaudio]

Not quite what happens, and actually wool lining on a waveguide will also do almost nothing. The sound is actually reflected in a sense. The wave motion is always parallel to the baffle with no velocity into the baffle, so, again, in theory, there cannot be any absorption. Absorption requires velocity INTO the absorber. As I said there will be some and the thicker the greater.

Won't wool lining in a waveguide primarily absorb HOMs, not the main wave front? Surely this is beneficial?

 

[DBMandrake]

Of course, the grill plot DBManddrake shows is the worse grill ever designed by humankind since approximately the year Homer was born. Have a close look at the picture.

As I said, those speakers are a case study in how not to design a grill or front baffle, and hence were a good example of just how bad grill diffraction problems can get, (even I was shocked at how badly they measured) sadly they are very representative of a lot of cheap speakers on the market that are built with no regard whatsoever for diffraction!

I'm not suggesting that a well designed DIY speaker/grill is nearly that bad - in fact I sincerely hope my grills aren't anything like that bad! On the other hand, I specifically made them removable so I can leave them off when I want improved quality, and while the speakers still sound pretty good with them on they definitely sound better with them off. The crossovers were optimised based on grill off measurements so I guess that's not surprising.

And using 1/24 smoothing might show differences in the treble even a fly landed on the grill before the run, if I may exaggerate a bit.

Sorry, I don't agree with that. I always use as little smoothing as possible in measurements - I want to see the truth, not some "airbrushed" response that makes everything look better than it really is. In ARTA 1/24th octave is the least smoothing possible in some measurement modes, so that's what I normally use.

Diffraction requires a high measurement resolution to measure properly, why would I want to deliberately throw away all the measurement detail so that I can no longer see the issue at hand ? It would be like deliberately wearing a pair of glasses that makes everything seem blurry just so you don't see defects in the paint job of your house!

And somebody might wonder why the two curves don't superimpose perfectly below say 750 Hz like they seem to when I repeat runs quickly together.

My measurements are 100% repeatable - I always take multiple measurements with the same test conditions to make sure of this, and the measurements overlay perfectly. The changes shown in the plots as low as 600Hz are actual, real changes caused by the grill and baffle diffraction.

You seem to be under the misapprehension that diffraction effects are only a treble problem, and that the change from adding a grill should only affect the treble, but they can extend all the way down to the baffle step frequency of a speaker.

As it happens, these Technics speakers are 20cm wide and thus have a baffle step transition frequency of about 600Hz. The changes at 600Hz and 1.2Khz are a direct result of the intrusive grill interfering with the baffle step transition of the speaker as it radically alters the diffraction effects at the baffle edge.

Anybody who has done real-world FR traces knows that some plots, sims, and phenomena (like comb filtering or other phase stuff) that look profoundly horrible are not audibly detectable except when special test tones are used, if then.

Well that's a subjective opinion. All I can say to that is the difference between taking the grill on and off with those speakers was night and day. They are decidedly harsh and sibilant with the grills on, and significantly smoother and easier to listen to with them off. When you look at the frequency response, it's easy to see why. No golden ears required for that one.

The change with the grills on/off on my speakers is definitely a lot more subtle. The main effect I notice is of the rolling off of the treble which is a function of the material of the cloth not diffraction effects. However I deliberately use relatively directional drivers so that will go some way towards minimising diffraction off the grill frame or the baffle edge for that matter.

DRMandrake here and in other places has made astonishing claims for his hearing perception. So I'd still like to see human tests of diffraction detection. And if there is some sense of shortcoming when diffraction is present, does it take more than a bit of felt or proper asymmetry to make it undetectable.

I don't claim to be able to hear "diffraction" per se, but you can hear the change in frequency response that it causes at a given listening point. The real issue with diffraction is that it makes a different change to frequency response at every listening angle.

If it just caused a constant change to the frequency response you could just EQ it out. But any attempt to EQ it on one axis is wrong on another axis and very likely worse than no EQ on that other axis.

Hence why minimising diffraction happening in the first place is so important.

 

[DBMandrake]

It took me some 15 minutes to shoot these for ya. The room is 3x4x2,7m.5ms gating here. Green lines with grille off.

I can see the grille's effect, do you?

The periodic wiggle most of the way through your measurement is a clear sign that you have your gate time set too long and that a reflection has entered the measurement window.

Did you check the impulse for the first visible reflection when choosing the window time ? Another possibility is that you have a reflection from your microphone stand, which would have a time delay too short to be windowed out.

I'd try to fix that rather than applying smoothing as your measurement is misleading when it includes reflections, and applying smoothing won't fix this. I never find it necessary to apply smoothing to high frequency measurements when they are properly windowed.

 

[DBMandrake]

Not trying to be offensive or anything. I dont understand the issue. You can measure down to 300 Hz in your room. More than needed to see the diffraction at stake. Not saying that it is easy to get a valid full in room response. But the diffraction issues is not the most difficult part.

I can't measure down to 300Hz in-situ as I've explained a couple of times now. It's not possible to get sufficient reflection free time.

With the speaker up on a tall stand (about 1.2 metres off the ground) near the middle of an empty room with the microphone at 1 metre I can achieve about 4.2ms before the first reflection.

That is sufficient, but measuring in the actual living room where the speakers are and especially where they normally sit down to those frequencies is not possible.

Keep in mind also that the window length doesn't just affect the minimum frequency you can go down to at all, it also affects the resolution of the entire measurement.

So even if you can "go down to" 300Hz, the entire measurement only has a resolution of 300Hz, meaning for example if there was a sharp peak at 600Hz it would appear as a broad curve. People often forget this resolution effect and assume that everything down to the cutoff frequency set by the window time is detailed - it is not. Resolution is very low towards the bottom end of the usable range.

 

[scottjoplin]

Rather than (or perhaps in addition to) smoothing does anyone use two microphones to measure to more accurately approximate how we hear, wouldn't some effects, like comb filtering, reduce?

 

[DBMandrake]

Rather than (or perhaps in addition to) smoothing does anyone use two microphones to measure to more accurately approximate how we hear, wouldn't some effects, like comb filtering, reduce?

Depends what the source of the comb filtering is.

If comb filtering is coming from a single location - such as from one speaker due to baffle diffraction, we hear it.

However comb filtering at high frequencies (~>2Khz) that is a result of the interference pattern between a left and right speaker is hardly audible, provided the speakers have sufficient angular separation, due to the occlusion of the face.

The left ear hears the treble of the right speaker greatly attenuated and vica versa. This means that the comb filtering perceived by each ear while still present is minimal.

And it's a good thing too, because if our ears worked like an omni-directional microphone the traditional stereo speaker configuration and stereo imaging would not work. Anyone who has tried using a narrow band RTA at an equidistant listening point and then moved it just a few inches to either side will have noticed the severe comb filtering that occurs in the treble, but we don't perceive that, not because we can't hear comb filtering (we can) but because there isn't actually significant comb filtering occurring at the ear due to the occlusion.

A fun test is to put mono (correlated) pink noise on both speakers, sit at your normal equidistant listening position then move your head sideways to see if you can hear the "swishing" effect of comb filtering on the noise. You may hear a little bit but if your speakers are separated enough you won't hear much.

Then just turn your head 90 degrees to the side and repeat the same test by moving your head forwards and backwards (still sideways relative to the speakers) and you will absolutely hear a large change in the noise and the phasing effect of varying comb filtering - it should be extremely obvious. This is because the ear nearest to both speakers hears the treble of both speakers relatively equally aside from slight HRTF differences, so comb filtering forms at that ear.

If your speakers are too close together in angular separation the cross-talk between opposite ear and speaker increases and comb filtering gets worse when the head is pointing forwards. This is one reason why speakers that aren't spaced apart enough have a very narrow sweet spot - if you're off that sweet spot you get comb filtering.

None of this ramble is applicable to measuring a single speaker though. What you might be thinking of is that some designers will measure a defined "listening window" and take an average of that instead of a single on axis measurement. For example measure 0, 5, 10 degrees to either side and then take an average. This tends to average out any direct on axis diffraction effects.

As most people listen slightly off axis it may be more representative of what you'll hear in typical conditions, however this kind of measurement window is only really necessary if the speaker has significant diffraction problems. If it doesn't, it shouldn't vary much at all across +/- 10 degrees, and that should be one of the design goals.

In some regards I consider this averaging of a few measurements around the on axis point to get a spatial average of the on axis response to be "cheating" in the same way that applying 1/3rd octave smoothing is cheating. Both are hiding true response anomalies of the speaker.

 

[Rokytheman]

DBMandrake: Cool !

 

[scottjoplin]

And it's a good thing too, because if our ears worked like an omni-directional microphone the traditional stereo speaker configuration and stereo imaging would not work.

Thanks Simon. Yes, it's a wonder the imaging works at all considering the things that conspire against it, and lucky for us. It goes to show just how adaptable our perception is.
Sometimes I will listen to my mono recordings through a single central speaker....how it should be

 

[bentoronto]

Thanks Simon. Yes, it's a wonder the imaging works at all considering the things that conspire against it, and lucky for us. It goes to show just how adaptable our perception is.
Sometimes I will listen to my mono recordings through a single central speaker....how it should be

If it sounds different, as you say, then stereo is a hoax.*

B.
*a virtual sound image should - in theory and allowing for some slight room acoustical artefacts - be indistinguishable from a mono speaker

 

[scottjoplin]

No denying that, but our brains have to position things in space I suppose to try to make sense of them

 

[gedlee]

Won't wool lining in a waveguide primarily absorb HOMs, not the main wave front? Surely this is beneficial?

It is true that HOMs will see more absorption than the normal mode, but it would still be very small unless the material is thick. Going with the thickest possible, i.e. filling the waveguide then yields the most HOM/Normal mode reduction possible and this will also dampen any normal mode standing waves in the device. This is why I fill the waveguide with foam.

I don't claim to be able to hear "diffraction" per se, but you can hear the change in frequency response that it causes at a given listening point. The real issue with diffraction is that it makes a different change to frequency response at every listening angle.

Simon, there is also an effect of diffraction beyond just frequency response and that is due to the delay time. Delay also showed as "significant" in our studies. I think that you probably knew this, but your comment may have been misleading to others.

 

[DBMandrake]

Simon, there is also an effect of diffraction beyond just frequency response and that is due to the delay time. Delay also showed as "significant" in our studies. I think that you probably knew this, but your comment may have been misleading to others.

I wasn't trying to be misleading, I just have a slightly different view of why diffraction is audible and undesirable than you do, and it's a difference that we've discussed in other threads not so long ago.

While I agree with your research findings that the effects of diffraction are audible, detrimental, and also that the perceived effect gets worse with SPL, (and is therefore often confused with non-linear distortion as it sounds worse at higher SPL as we would expect of non-linear distortion) I don't agree that there is some "intrinsic" property of diffraction that causes this to be so that is independent of other properties such as the frequency response.

My view is that diffraction sounds bad not through any intrinsic property of it's own (such as it being a "delayed signal") but simply due to the undesirable effect it has on the resulting frequency response at any given listening point in space.

It's tempting to think of diffraction as a delayed impulse but on any real speaker the time delay to each diffracting part of a speaker (like the perimeter of a rectangular baffle) is a constantly varying value meaning that the delayed "impulse" is extremely smeared, hence the very irregular frequency response that results instead of a classic comb filter that a delayed but not smeared impulse would result in.

When you talk about delay caused by diffraction you also have to keep in mind that baffle edge diffraction is always going to be minimum phase at any given listening point in space (something else we don't seem to agree on based on previous discussions) because this must be so when the delayed signal is always lower in amplitude than the direct signal.

The result could only be non-minimum phase if the delayed diffracted signal were to be greater in amplitude than the direct signal from the drivers - and short of putting an object deliberately in the way of the drivers but not the baffle edge, I can't see how that could happen with any normal speaker, so I discount it as a realistic possibility.

As well as thinking about it from a theoretical perspective I've also taken quite a few measurements of a number of speakers looking for any signs of any non-minimum phase behaviour due to diffraction and have not found any whatsoever. The only source of non-minimum phase behaviour in all normal speakers is the excess phase introduced by a crossover and/or driver acoustic centre offset errors, not from diffraction itself.

If the total response of driver plus baffle edge diffraction is always minimum phase at any given listening position that means the frequency response irregularities caused by diffraction are enough to explain the phenomena of diffraction sounding bad, and at a given point in space the frequency response irregularities caused by diffraction would sound equally bad (and getting worse with SPL) if they were generated using EQ and played through a non-diffracting speaker.

It also follows that a non-diffracting speaker that suffers from resonances that happen to give the same frequency response irregularities as the diffracting speaker would also sound bad in the same way at that same point in space.

If they're both minimum phase and have the same frequency response, the final impulse response must be the same, and there isn't anything left to make one sound worse than the other.

So I see avoiding frequency response aberrations arising from resonances as equally important to avoiding frequency response aberrations arising from diffraction, although the method of dealing with them is totally different.

The key difference between them is that peaks and dips resulting from resonances can be corrected on all listening axes with EQ using a single transfer function and thus can be "fixed" electronically, (and any crossover with notch filters in it is attempting to do just that) while peaks and dips resulting from diffraction cannot be fixed electronically with EQ as a different transfer function would be required for every off axis angle - clearly impossible.

Therefore diffraction can only be fixed mechanically/acoustically, using directivity, damping, and the baffle geometry design, and as a consequence is a lot more challenging than compensating for resonances.

So my original comment that I can hear the frequency response effects of diffraction "not diffraction itself" is precisely because that's what I believe we are hearing when we hear the effects of diffraction, not some other intrinsic delayed signal phenomena.

 

[gedlee]

Simon

I also don't share your propensity for being verbose, so in a nutshell, how do you explain our result that delay was a "significant effect" meaning that greater delay's are more audible and that this was shown to be "significant"? Both will have equal mean effects on frequency response, albeit different in detail, but the greater delay is more audible. This contradicts your position.