PURIFI finally did a fully purified passive speaker design! The SPK 16 prototype is here - with a PTT tweeter

[Matias]

The cabinet is almost identical to Sointuva AWG, so roundeovers the same.

As for wider x narrower treble dispersion, IMO it is secondary to the end result of having better linearity in predicted in-room response. I would (and did) prefer a narrow dispersion with very linear tilted response (Sointuva AWG), over a wide dispersion which possibly/probably will have a treble boost in response (SPK 16). Let's wait and see to be sure, but I fear the SPK 16 is going to be like that.

 

[boXem]

You still need to terminate the edges of waveguided speakers with round overs, have plenty of my own data showing the benefits. Joseph crowe is a speaker designer who does waveguides/horns and his take it a step further and round all the way to back of the horn mouth.

My understanding is that the waveguide shape optimisation is different for rounded/sharp corners. The diffraction can anyhow easily be seen in measurements.

I fear the SPK 16 is going to be like that

I do not remember that the PIR would indicate such a thing, but my memory could be wrong. Treble boost was clearly not something I noticed when I listened to the SPK16.

 

[Matias]

Amém, fingers crossed here. If it is only wide dispersion without treble boost = best of both worlds.

 

[Ktacos]

My understanding is that the waveguide shape optimisation is different for rounded/sharp corners. The diffraction can anyhow easily be seen in measurements.

All waveguides need round overs for optimal performance, they all produce waves that will reach the edge of the box.

 

[Ktacos]

The cabinet is almost identical to Sointuva AWG, so roundeovers the same.

As for wider x narrower treble dispersion, IMO it is secondary to the end result of having better linearity in predicted in-room response. I would (and did) prefer a narrow dispersion with very linear tilted response (Sointuva AWG), over a wide dispersion which possibly/probably will have a treble boost in response (SPK 16). Let's wait and see to be sure, but I fear the SPK 16 is going to be like that.

The round overs are not the same, the ones on the purifi are larger.

I see no reason to assume the purifi will have a treble boost.

 

[jaakkopetteri]

All waveguides need round overs for optimal performance, they all produce waves that will reach the edge of the box.

Define "optimal performance"

 

[kharan]

Dispersion charts show diffraction issues on those speakers from their edges.

Can you explain what too look for in dispersion charts? Preferably by using actual charts from speakers and circling the specific thing to be noticed.

 

[Ktacos]

Define "optimal performance"

Simple to define, you want the waveguide to be doing its thing without the baffle impacting it's performance.

Can you explain what too look for in dispersion charts? Preferably by using actual charts from speakers and circling the specific thing to be noticed.

Sure, but I'll have to edit a few images when I get back to a desktop, trying to share the info over my phone is gonna be impossible.

 

[jaakkopetteri]

Simple to define, you want the waveguide to be doing its thing without the baffle impacting it's performance.

Sorry, but that's a 100% circular definition

 

[Emlin]

Sorry, but that's a 100% circular definition

No it's not.

 

[Da cynics]

This is interesting

Edge Diffraction Testing - Shapes

That will make a difference but it is not likely to be a positive one. Creases are not good and making something like a waveguide start at the edge of the tweeters faceplate is also a way to introduce negative diffraction effects. This image shows where the edge rounding starts and as close to...

www.diyaudio.com

 

[fluid]

Whenever there is a finite baffle there is diffraction at the edge of that baffle. What the shape or size of an edge does is control how that diffraction happens.

A sharp edge creates a secondary sound source that interferences with the outgoing wave causing peaks and dips in the response. The size of the baffle will determine where in the frequency range the effects will be seen. It is most easily seen as a ripple in the on axis response, there will be peaks and dips from the interference. The thing that is often missed is that it will create an imbalance between the near axis and further off axis responses. There can be a peak in the off axis response that is higher than the on axis in one or several places. This makes the speaker difficult to equalize in that range, what works well in practice will be harder to determine from measurements alone. It will create a compromise and this will also make the speaker less universal to all music, there will be some tracks where the compromise makes it sound bad.

Speaker design is balancing compromises. The more compromises that are introduced the harder it is to balance them successfully.

The shape of the edge and the shape and size of the cabinet behind, if there is one, is an important variable in the design process of a waveguide. The wider the directivity of the waveguide the more importance it will have, the wider the frequency range the high frequency driver covers the more important it will be. Lars has spoken about it and I have seen the same thing in simulation, ignoring the rest of the cabinet will lead to a waveguide design that is not quite what was intended or sometimes one that is really quite awful.

It is expensive to make large roundovers on cabinet edges in production speakers. In a cost concsious segment they are rare to see. Most manufacturers use diffraction control techniques in their flagship products, because of cost and also to allow them to measure and perform better than the cheaper speakers. 25mm edge rounding is where it starts to pay off and 36 to 40mm is where going bigger desn't tend to make much difference on conventional box speakers. That can be quite a lot of cabinet material and a lot of weight. For full shape control like in the blade moulds and plastic make a lot of sense.

 

[Ktacos]

Finally at a computer. Time to make way too long of a post.

I gathered up a few speakers mentioned earlier that had spins and the typical combo of a ~6.5" with a roughly 6" circular waveguide, with the exception of the Kali lp6 v2 which are oblate spheroid but still similar in width and exhibits the issues I mentioned previously. Some are worse than others, most people probably wouldn't care about these problems but after hearing speakers with 3" radius round overs I'm kind of a convert and the problem just really bugs me now. Subjectively I find the issue smears the sound of music when the content gets is a little "busy" or dense if that makes sense. Ever listen to a song with a clear vocal that gets kind of drowned out when the music picks up? That very well could be diffraction getting in the way. A distcint aspect I noticed in speakers with 2-3" radius round overs is that you don't lose track of song elements at any given time, reverb is rendered much better and the tails seem to stick around longer and in a more convincing way. This makes sense to me as a dense track and reverb are generally going a more sustained aspect of music so they will be more prone to exhibiting audible interference from edge diffraction. More staccato/pizzacato/percussive stuff doesn't seem to be affected much. One user on this forum whom I shared my experiences with was on the same page as me and described diffraction as sounding "restless" and I agree. It's a sound that makes you want to get up and try to fix something with the speaker.

First up, a Buchardt speaker. These have no round over, just a hard edge. Based on my own testing with a similar waveguide, I expect to see some diffraction artifacts at around 5khz, and that's exactly what we see. There's some lower stuff but probably not as annoying as a secondary source at 5khz.

Next up Kali LP6 v2. These have a taper to the baffle on both sides and a very small round over. You can see that the lack of a hard edge helps in the region the buchardt has trouble, but you still get issues around 3khz as you generally need around a 1" radius to avoid that. This speaker is also a good example of how diffraction works it's way down to the woofer region as well, it's not just about helping the tweeter.

Here's the Kef LS50 meta. It has largely no roundover but does have a vertical taper. The woofer does a good job of preventing the tweeter from having much trouble, but the woofer around 1-2k is a diffraction factory.

Here's probably one of the worst examples I've ever seen, the criton 1td. It has no round over and no waveguide for the tweeter. The wilson tunetot has similar problems but I really do not care to share that speaker as it's kind of just a joke of a product. I am not identifying the diffraction in the graph because it's so obvious.

Now for my own measurements, which are not going to be as accurate but display some things I wanted to share.

Here's a 6" waveguide I had placed on a baffle with hard edges, then on a one with a 3/4" round over. Not the best data but you can clearly see how the response has been smoothed out from the round over. Top is hard edges, bottom is rounded. In the top graph we can see a few issues, such as the depression off axis at 3khz, an on axis dip at 5khz, and a big dip axis at 10k. We can see on the bottom graph most of the issues are alleviated with the roundover. These may appear to be small differences I can tell you that these speakers sounded completely different from each other, the speaker with the hard edges was just annoying to listen to.

For fun I wanted to share my most beautiful speaker I've ever made, some cnote drivers and box with some PVC around it and active filters. Not really much going on with the tweeter filter here, the round overs make the response very smooth and easy to work with. There is a 2nd order high pass and a tilt EQ from 4k-20khz on the tweeter, and I think a low pass at the top end at 15khz (stuff above that kind of annoys me). My current goal is trying to find a way to manufacture this in a visually pleasing manner.

Anyway, those are my thoughts on the importance of proper edge termination, there's definitely a bit more to it all, but so far in my experience the larger the round over the better the speaker sounds. There's plenty of better info out there on the topic and as a DIY'er, I do find a lot of DIY is people finding an aspect they don't like about speakers and obsessively pursueing solutions to that one problem, so I'd say edge diffraction is that for me.

 

[thewas]

Here is also a nice overview of the impact of edges and waveguides:

Basic article | Baffle geometry

About baffle design, edge diffraction, secondary sound sources and their influence on the reproduction quality of loudspeakers.

heissmann-acoustics.de

 

[AudioJester]

The cabinet is almost identical to Sointuva AWG, so roundeovers the same.

As for wider x narrower treble dispersion, IMO it is secondary to the end result of having better linearity in predicted in-room response. I would (and did) prefer a narrow dispersion with very linear tilted response (Sointuva AWG), over a wide dispersion which possibly/probably will have a treble boost in response (SPK 16). Let's wait and see to be sure, but I fear the SPK 16 is going to be like that.

Did you audition the Sointuva AWG before buying?

 

[Matias]

Did you audition the Sointuva AWG before buying?

No, I had to trust data, reports and videos of it playing.

 

[AudioJester]

No, I had to trust data, reports and videos of it playing.

Having heard them, you made a great choice!

 

[boXem]

Finally at a computer. Time to make way too long of a post.

I gathered up a few speakers mentioned earlier that had spins and the typical combo of a ~6.5" with a roughly 6" circular waveguide, with the exception of the Kali lp6 v2 which are oblate spheroid but still similar in width and exhibits the issues I mentioned previously. Some are worse than others, most people probably wouldn't care about these problems but after hearing speakers with 3" radius round overs I'm kind of a convert and the problem just really bugs me now. Subjectively I find the issue smears the sound of music when the content gets is a little "busy" or dense if that makes sense. Ever listen to a song with a clear vocal that gets kind of drowned out when the music picks up? That very well could be diffraction getting in the way. A distcint aspect I noticed in speakers with 2-3" radius round overs is that you don't lose track of song elements at any given time, reverb is rendered much better and the tails seem to stick around longer and in a more convincing way. This makes sense to me as a dense track and reverb are generally going a more sustained aspect of music so they will be more prone to exhibiting audible interference from edge diffraction. More staccato/pizzacato/percussive stuff doesn't seem to be affected much. One user on this forum whom I shared my experiences with was on the same page as me and described diffraction as sounding "restless" and I agree. It's a sound that makes you want to get up and try to fix something with the speaker.

First up, a Buchardt speaker. These have no round over, just a hard edge. Based on my own testing with a similar waveguide, I expect to see some diffraction artifacts at around 5khz, and that's exactly what we see. There's some lower stuff but probably not as annoying as a secondary source at 5khz.

View attachment 395022

Next up Kali LP6 v2. These have a taper to the baffle on both sides and a very small round over. You can see that the lack of a hard edge helps in the region the buchardt has trouble, but you still get issues around 3khz as you generally need around a 1" radius to avoid that. This speaker is also a good example of how diffraction works it's way down to the woofer region as well, it's not just about helping the tweeter.

View attachment 395025

Here's the Kef LS50 meta. It has largely no roundover but does have a vertical taper. The woofer does a good job of preventing the tweeter from having much trouble, but the woofer around 1-2k is a diffraction factory.

View attachment 395028

Here's probably one of the worst examples I've ever seen, the criton 1td. It has no round over and no waveguide for the tweeter. The wilson tunetot has similar problems but I really do not care to share that speaker as it's kind of just a joke of a product. I am not identifying the diffraction in the graph because it's so obvious.

View attachment 395029

Now for my own measurements, which are not going to be as accurate but display some things I wanted to share.

Here's a 6" waveguide I had placed on a baffle with hard edges, then on a one with a 3/4" round over. Not the best data but you can clearly see how the response has been smoothed out from the round over. Top is hard edges, bottom is rounded. In the top graph we can see a few issues, such as the depression off axis at 3khz, an on axis dip at 5khz, and a big dip axis at 10k. We can see on the bottom graph most of the issues are alleviated with the roundover. These may appear to be small differences I can tell you that these speakers sounded completely different from each other, the speaker with the hard edges was just annoying to listen to.

View attachment 395034

For fun I wanted to share my most beautiful speaker I've ever made, some cnote drivers and box with some PVC around it and active filters. Not really much going on with the tweeter filter here, the round overs make the response very smooth and easy to work with. There is a 2nd order high pass and a tilt EQ from 4k-20khz on the tweeter, and I think a low pass at the top end at 15khz (stuff above that kind of annoys me). My current goal is trying to find a way to manufacture this in a visually pleasing manner.

View attachment 395037

View attachment 395038

Anyway, those are my thoughts on the importance of proper edge termination, there's definitely a bit more to it all, but so far in my experience the larger the round over the better the speaker sounds. There's plenty of better info out there on the topic and as a DIY'er, I do find a lot of DIY is people finding an aspect they don't like about speakers and obsessively pursueing solutions to that one problem, so I'd say edge diffraction is that for me.

I appreciate the time you took for your demonstration , thank you! Since I still had some doubts (sorry for this), I searched for other speakers with sharp edges to see if they all exhibit the behavior you highlighted. Here is the first contour plot I found and I fail to see these diffractions artefacts. Did I miss something? Or is the issue a bit more complicated?

 

[fluid]

Here is the first contour plot I found and I fail to see these diffractions artefacts.

Normalized contour plots are not a great way to see diffraction effects.

A non normalized line chart makes it easier to see. The Kef R3 does not have a lot of edge induced ripple because the cone of the woofer narrows the directivity quite a lot.

In the left ellipse there is bunching of the traces at 2.5K and a general increase in off axis pressure, in the right at 8K there is an on axis dip with peaks further off axis. In a coaxial the higher frequency wiggles come from the tweeter to cone interface and Kef does this quite well in general. In the early reflections and PIR curves you will see a corresponding rise at 2.5K. There is also a similar pattern issue in the vertical at 2.5K.

 

[kharan]

That's interesting. Indeed, edge diffraction has been a long known factor in speaker design. It's good to put a number on the amplitude of the phenomenon and the shape it takes on a chart.

My initial thought is that it's audible though I wouldn't go further and speculate on how much.