Questions About Diffraction and Ovenight Sensations

[movinginstereo]

Noob here. For my first project I want to do a version of Paul Carmody's Overnight Sensations based on Dieter Rams' SK4 and L1 designs. My concern is that the tight square edges will produce significant enough diffraction that by honoring Rams' design principle of aesthetic I will defeat his principle of thoroughness.

I am not planning to use a grill cover as Rams did, but instead leave the monitor and twitter exposed as in Carmody's design.

So, my questions are: a) will the diffraction be enough that I should worry about it? and if so, b) is there any other way to compensate for diffraction other than rounding the edges?

Thank you much in advance. I have spent the past week studying in here and you people are remarkable. I am in awe of the craftsmanship, skill, and knowledge you all possess and share.

[fastbike1]

Re: Questions About Diffraction and Ovenight Sensations

Well, since you asked, I think the referenced SK4 and L1 definitely violate his own principle of "Good design is a little design as possible". Also the SK4 violates his principles of useful, aesthetic, understandable, unobtrusive, and honest. I actually find the two examples to apparently have little concern for actual function, and I'm a guy that likes innovative design. A total fail.

The design of the Overnight Sensations, by contrast, seems to meet the "principles" much better.

In my opinion, both SK4 and L1 designs are overdone, non-functional, and ugly.

Noob here. For my first project I want to do a version of Paul Carmody's Overnight Sensations based on Dieter Rams' SK4 and L1 designs. My concern is that the tight square edges will produce significant enough diffraction that by honoring Rams' design principle of aesthetic I will defeat his principle of thoroughness.

I am not planning to use a grill cover as Rams did, but instead leave the monitor and twitter exposed as in Carmody's design.

So, my questions are: a) will the diffraction be enough that I should worry about it? and if so, b) is there any other way to compensate for diffraction other than rounding the edges?

Thank you much in advance. I have spent the past week studying in here and you people are remarkable. I am in awe of the craftsmanship, skill, and knowledge you all possess and share.

[Paul Carmody]

Re: Questions About Diffraction and Ovenight Sensations

So, my questions are: a) will the diffraction be enough that I should worry about it?

No, and here's the science behind it. Let's say one does a 3/4" roundover on a box. That corresponds to the frequency 18,000 Hz. That's out of the range of most people's hearing, except maybe teenagers. If a roundover is really intended to have a sonic impact on diffraction, then it needs to be quite large, like 2", which corresponds to about 6700, which is an audible range, though still pretty high.

*Note: if I'm wrong, someone please correct me. But I took this information out of Joe D'Appolito's book.

[greywarden]

Re: Questions About Diffraction and Ovenight Sensations

I thought it was more directly related to the baffle shape...

[Jeff B.]

Re: Questions About Diffraction and Ovenight Sensations

No, and here's the science behind it. Let's say one does a 3/4" roundover on a box. That corresponds to the frequency 18,000 Hz. That's out of the range of most people's hearing, except maybe teenagers. If a roundover is really intended to have a sonic impact on diffraction, then it needs to be quite large, like 2", which corresponds to about 6700, which is an audible range, though still pretty high.

*Note: if I'm wrong, someone please correct me. But I took this information out of Joe D'Appolito's book.

Although those radius distances correspond to those frequencies, their effect on diffraction will extend much lower, to about 1/4 those frequencies. For example, consider a 9" baffle. It corresponds to frequency of 1506 Hz, but the baffle's effect on diffracting the wavefront will still be measureable at 375Hz (and even much lower to a lesser extent). I would expect measureable differences in frequency response into the 3000 Hz range with a 3/4" radius.

However, that said, we have all heard many small speakers that sound very good. We need to remember that diffraction changes depending on the listening angle, so spread over a +/- 30 degree window a lot of diffraction effects fill in and smooth out naturally. Sometimes, like with my Continuum, I included the effects into the design itself as part of its frequency response. I wouldn't worry too much about it.

Plus, if someome wants to take things to another level there's always felt. Not as pretty as a nice radius, but more effective in taming diffraction.

Jeff

[movinginstereo]

Re: Questions About Diffraction and Ovenight Sensations

Well, since you asked, I think the referenced SK4 and L1 definitely violate his own principle of "Good design is a little design as possible". Also the SK4 violates his principles of useful, aesthetic, understandable, unobtrusive, and honest. I actually find the two examples to apparently have little concern for actual function, and I'm a guy that likes innovative design. A total fail.

The design of the Overnight Sensations, by contrast, seems to meet the "principles" much better.

In my opinion, both SK4 and L1 designs are overdone, non-functional, and ugly.

I agree somewhat and you hit on exactly what I am in conflict about. From a less is more standpoint the OS design is spot on, but the aesthetics of the Rams design has always appealed to me. On the other hand, I'm a designer and the idea of adding decoration that degrades performance is abhorrent. On the other hand, I like the idea of representing something iconic.

This is why designers seem to be in angst so often -- we struggle with this conflict of interests in everything we do.

Thanks, you got me thinking.

[movinginstereo]

Re: Questions About Diffraction and Ovenight Sensations

Although those radius distances correspond to those frequencies, their effect on diffraction will extend much lower, to about 1/4 those frequencies. For example, consider a 9" baffle. It corresponds to frequency of 1506 Hz, but the baffle's effect on diffracting the wavefront will still be measureable at 375Hz (and even much lower to a lesser extent). I would expect measureable differences in frequency response into the 3000 Hz range with a 3/4" radius.

However, that said, we have all heard many small speakers that sound very good. We need to remember that diffraction changes depending on the listening angle, so spread over a +/- 30 degree window a lot of diffraction effects fill in and smooth out naturally. Sometimes, like with my Continuum, I included the effects into the design itself as part of its frequency response. I wouldn't worry too much about it.

Plus, if someome wants to take things to another level there's always felt. Not as pretty as a nice radius, but more effective in taming diffraction.

Jeff

Thanks, this is very helpful. Do you know where I can find out more about using felt?

[Sydney]

Re: Questions About Diffraction and Ovenight Sensations

...Do you know where I can find out more about using felt?

In SpeakerBuilder and AudioXpress have been numerous articles addressing diffraction with the application of felt.
http://www.speakerdesign.net/felt_am...ltssenter.html
http://www.speakerdesign.net/felt/fe...vs_blocks.html
http://www.speakerdesign.net/audioXp...ffraction.html

[dlr]

Re: Questions About Diffraction and Ovenight Sensations

Although those radius distances correspond to those frequencies, their effect on diffraction will extend much lower, to about 1/4 those frequencies. For example, consider a 9" baffle. It corresponds to frequency of 1506 Hz, but the baffle's effect on diffracting the wavefront will still be measureable at 375Hz (and even much lower to a lesser extent). I would expect measureable differences in frequency response into the 3000 Hz range with a 3/4" radius.

It is odd how some references only cover the radius in terms of a single cross-sectional cut, given that for any straight baffle edge with a circular-cut roundover there is a single ray from the center of the driver to the edge that is circular, while all other intersections of rays from driver to edge follow a path that results in an elliptical pattern.

For those interested in visualizing this, go here. You'll see how Jeff's point comes about.

dlr

[neildavis]

Re: Questions About Diffraction and Ovenight Sensations

It is odd how some references only cover the radius in terms of a single cross-sectional cut, given that for any straight baffle edge with a circular-cut roundover there is a single ray from the center of the driver to the edge that is circular, while all other intersections of rays from driver to edge follow a path that results in an elliptical pattern.

For those interested in visualizing this, go here. You'll see how Jeff's point comes about.

dlr

Interesting. The baffle diffraction program I wrote uses ray-tracing, but it uses the rounded-edge methodology that Paul Verdone describes in the BDS user manual. Paul's method doesn't account for the elliptical elongation. It should be fairly easy to model the effect by calculating the intersection angle. Maybe I'll try that.

[dlr]

Re: Questions About Diffraction and Ovenight Sensations

Interesting. The baffle diffraction program I wrote uses ray-tracing, but it uses the rounded-edge methodology that Paul Verdone describes in the BDS user manual. Paul's method doesn't account for the elliptical elongation. It should be fairly easy to model the effect by calculating the intersection angle. Maybe I'll try that.

Are you sure about Paul's method? It may be part of the difference between simple vs. complex calculation options. I would think that this would make a significant difference.

If points along an edge are assigned linearly for ray traces, then the results will be skewed towards the closest point on the edge. This will in turn focus more in the area where the path is closer to a circular pattern.

If the rays are uniform with fixed angles between adjacent rays, then a more representative result should be obtained (as I see it anyway) since this will distribute the points along the edge with equal weighting from point to point, with increasing point density as the angle of incidence of rays with the edge changes.

Of course this way of viewing it may not prove to be the best. It just seems to me that equal angular spacing ought to work best.

dlr

[neildavis]

Re: Questions About Diffraction and Ovenight Sensations

Are you sure about Paul's method? It may be part of the difference between simple vs. complex calculation options. I would think that this would make a significant difference.

If points along an edge are assigned linearly for ray traces, then the results will be skewed towards the closest point on the edge. This will in turn focus more in the area where the path is closer to a circular pattern.

If the rays are uniform with fixed angles between adjacent rays, then a more representative result should be obtained (as I see it anyway) since this will distribute the points along the edge with equal weighting from point to point, with increasing point density as the angle of incidence of rays with the edge changes.

Of course this way of viewing it may not prove to be the best. It just seems to me that equal angular spacing ought to work best.

dlr

I used equal angles between the rays rather than fixed distances around the perimeter. Paul's code did this, also. However, I didn't adjust the length of the roundover based on the angle. I just assumed that the roundover distance was always equal to the radius, and I used Paul's weighting scheme to account for the attenuation at 4 points along the radius. My understanding of the "Complex Method" is simply that it uses more bounce points along the edge. Hmmm...on page 33 of the user manual he says he calculates the distance to the outer edge and inner edge, so maybe that is where he accounts for the projection at an angle. I guess I need to add this.

[dlr]

Re: Questions About Diffraction and Ovenight Sensations

I used equal angles between the rays rather than fixed distances around the perimeter. Paul's code did this, also. However, I didn't adjust the length of the roundover based on the angle. I just assumed that the roundover distance was always equal to the radius, and I used Paul's weighting scheme to account for the attenuation at 4 points along the radius. My understanding of the "Complex Method" is simply that it uses more bounce points along the edge. Hmmm...on page 33 of the user manual he says he calculates the distance to the outer edge and inner edge, so maybe that is where he accounts for the projection at an angle. I guess I need to add this.

One other consideration that may be an issue. Maybe Jeff could answer it. When using multiple points along the edge radius, the ellipse in this case, is there an equal weighting for each point? I would expect two influences, these being the distance (decreasing waveform pressure at farther points) and the angle at these points. That is, is there a relationship with regard to each point given that a line that is normal to the curve at that point will change direction with increasing change with distance and that this change is not a linear relationship with distance?

Sounds a bit convoluted, I'm not sure how to say it more concisely at the moment.

dlr

[neildavis]

Re: Questions About Diffraction and Ovenight Sensations

One other consideration that may be an issue. Maybe Jeff could answer it. When using multiple points along the edge radius, the ellipse in this case, is there an equal weighting for each point? I would expect two influences, these being the distance (decreasing waveform pressure at farther points) and the angle at these points. That is, is there a relationship with regard to each point given that a line that is normal to the curve at that point will change direction with increasing change with distance and that this change is not a linear relationship with distance?

Sounds a bit convoluted, I'm not sure how to say it more concisely at the moment.

dlr

There was a flurry of emails a couple of years ago where Jeff and others were discussing this--I've got them stashed away at home and can forward them to you if you want. The conclusion was that the radial model with equal weights was the "right answer", as it accounted for the waveform pressure correctly.

There are two additional considerations that I thought were important enough to add to my model. One is taking into account the attentuation (1/n^2 loss) due to additional travel distance from the edge to the summation point. For large distances (such as a summation point at 8') the travel distance from the driver or the edge is nearly the same, and the additional distance can be ignored. However, for closer measurements the difference is significant. That's why my model doesn't show exactly -6db for maximum baffle step. Also, I calculate the attentuation for the backwave in the dipole configuration.

The other consideration is phase. I ray-trace both sine and cosine signals so that I can look at the phase relationships of the summed signal. The phase doesn't change a whole lot, but it is enough to impact the crossover design (around 30 degrees below the baffle step region). The Edge program also calculates phase, but as far as I know, none of the other diffraction simulation programs do this. And none of the other programs use the distance attentuation. Since each ray has an attention factor associated with it, it wouldn't be too hard to add the effect of using felt or other materials, assuming you could model the felt as a simple attentuator.

Another feature is that my program is multi-threaded so that it is much faster on a quad-core CPU :D. Because if it wasn't, it would be very, very slow :o

[Jeff B.]

Re: Questions About Diffraction and Ovenight Sensations

There was a flurry of emails a couple of years ago where Jeff and others were discussing this--I've got them stashed away at home and can forward them to you if you want. The conclusion was that the radial model with equal weights was the "right answer", as it accounted for the waveform pressure correctly.

There are two additional considerations that I thought were important enough to add to my model. One is taking into account the attentuation (1/n^2 loss) due to additional travel distance from the edge to the summation point. For large distances (such as a summation point at 8') the travel distance from the driver or the edge is nearly the same, and the additional distance can be ignored. However, for closer measurements the difference is significant. That's why my model doesn't show exactly -6db for maximum baffle step. Also, I calculate the attentuation for the backwave in the dipole configuration.

The other consideration is phase. I ray-trace both sine and cosine signals so that I can look at the phase relationships of the summed signal. The phase doesn't change a whole lot, but it is enough to impact the crossover design (around 30 degrees below the baffle step region). The Edge program also calculates phase, but as far as I know, none of the other diffraction simulation programs do this. And none of the other programs use the distance attentuation. Since each ray has an attention factor associated with it, it wouldn't be too hard to add the effect of using felt or other materials, assuming you could model the felt as a simple attentuator.

Another feature is that my program is multi-threaded so that it is much faster on a quad-core CPU :D. Because if it wasn't, it would be very, very slow :o

My original model calculated phase as well. However, in my newest, since everything was designed to work by combining the calculated diffraction response with the infinted baffle response I decided to let the HB phase extraction take care of all of it at once, especially since all of these effect are minimum phase to begin with.

Regarding the weight of the ray. As you know, my model calculates the ray in 10 degree angular spacings forming circle as well. (For what it's worth, I expanded my model one day to 360 1 degree rays for a 10 fold increase in calculations. The resultant diffraction response was identical to what I was getting with 36 10 degree rays, but took up a whole lot more spreadsheet. After that I decided to forgo the exta calculations and dropped back to the 36 ray method.)

Anyway, it seems intuitive to me, but I could be wrong, that the strongest or most weighted point at each edge would be the one just off the flat section of the baffle or the point where the radius begins. The reason for this is because as the ray begins to wrap around the radius less of its diffracted energy would combine with the on axis energy and more of it would be angled off-axis. It would also seem to me that the bigger the radius the wider the spread in frequencies diffracted compared to a sharp edge, thus the effect is more diffuse because the affected frequencies are spread out. Of course, I am just picturing the mechanism in my head, so if you disagree I am open to discussion.

Jeff B.