Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[fiasco ( P.M.DuMont )]

Hello.

I'm posting to show you folks a D.I.Y. project I did, and pose a question.

Materials @ 20 USD (although I didn't have to pay... work scrap), @ 5 hrs  labor.
1/2 inch birch plywood, polyurethane, no stain. Sort of a trial run.









My question: Referring to the depth of the pockets, how much of a part would math play in a piece like this?
The manufacturer of a similar piece states, "quasi random".
Huh?
Isn't random, by definition quasi?

I am in the beginning stages of treating my space and look forward to you fine folk's help.

[franman]

That's a fine 'knock-off' of a 2 dimnesional omnifussors by a well known manufacturer, (who probably wouldn't appreciate your post)...

anyway, math would play a large role in making this scattering device into a true 'diffusor'. The math of diffusion can be deep, but RPG uses a series of QRD patterns to make their Omnifussor (tm). The math for QRD is readily available (master handbook or acoustics and other places)....

It seems you did a very nice job.. FYI, the 'original' unit is four inches deep, and if I recall correctly there are five unique depths to the wells. The pattern and the actual depths are what the QRD math is for.

Other observations: the thickness of the dividers should be a thin as possible (for future revisions). This is a basic concept of phase gradient diffusors.

Careful about posts regarding patented devices... obviously you aren't intending to sell these, but still, take care.

[Greg Reierson]

From the pictures it looks like the effective frequency range of that diffusor would be pretty narrow. What are the dimensions of the wells?


GR

[fiasco ( P.M.DuMont )]

franman wrote on Tue, 17 June 2008 21:19

Careful about posts regarding patented devices... obviously you aren't intending to sell these, but still, take care.

Thanks, I actually hadn't thought of that. It's almost like throwing a patent on a shadow box.
They are just for me.

Thanks for the math info, I will look into it.

[fiasco ( P.M.DuMont )]

Greg Reierson wrote on Tue, 17 June 2008 23:19

From the pictures it looks like the effective frequency range of that diffusor would be pretty narrow. What are the dimensions of the wells? GR

The squares are 2 5/8 inches and the depth was totally random.
The overall depth of the piece is 3 inches.
Any suggestions?

[fiasco ( P.M.DuMont )]

Francis, upon reflection, I will have no problem if this thread were removed.

Your call.

[bruno putzeys]

I hope manufs will take a healthy attitude towards DIY'ers. Personally I've never tried discouraging DIY folk from "recreating" my amps for personal education & pleasure as long as they do so based on publicly available info (such as patents, papers) and stuff they find out for themselves. Reverse engineering is a no-no of course.

Just my

[fiasco ( P.M.DuMont )]

Bruno Putzeys wrote on Thu, 19 June 2008 08:15

But to the topic, I've always wondered if the regularity in the other 2 dimensions doesn't invite some sort of cumulative effect at a wavelength of 4 5/4"?

Meaning while the depth varies, the height and width of the pockets remain constant?
Seems that could be self defeating or even problematic.

[AndreasN]

Fiasco wrote on Wed, 18 June 2008 12:42

The squares are 2 5/8 inches and the depth was totally random.

The problem is that 'totally random' using the dice method requires a really long number of throws to be random. A short string of numbers is actually very hard to make random! The beauty of the correct math for diffusors is that the string ends up having the property that it is noise. A fourier analysis on such a string gives a flat spectra, just like white noise. The base sequence number is the number of fourier bins. More numbers in the base sequence gives better spread to the spectra.

Bruno Putzeys wrote on Thu, 19 June 2008 14:15

I've always wondered if the regularity in the other 2 dimensions doesn't invite some sort of cumulative effect at a wavelength of 4 5/4"?

Perhaps the well sizes needs to be the same to get a correct blend from the delay array? Like wavefront syntesis?


With the disclaimer that I'm just an amateur that read Cox's and D'Antonios book skimming the math. Hope someone corrects me if there's anything wrong!


Regards,

Andreas Nordenstam

[J-Texas]

Bruno Putzeys wrote on Thu, 19 June 2008 07:15

... I've always wondered if the regularity in the other 2 dimensions doesn't invite some sort of cumulative effect at a wavelength of 4 5/4"?

franman wrote on Tue, 17 June 2008 20:19

Other observations: the thickness of the dividers should be a thin as possible...

I've always understood these to relate to one another? Am I stoopid?

[bruno putzeys]

AndreasN wrote on Fri, 20 June 2008 01:54

Perhaps the well sizes needs to be the same to get a correct blend from the delay array? Like wavefront syntesis?

I think an irregular array would require a more complicated method for computing depth. So the regularity would be a carryover from the method used to compute the depths. The method you've explained abstracts the array to a string of numbers which only works with uniform spacing (otherwise you'd need more than one number to represent one box). So what I'm worrying is that at higher frequencies you'd get some very obvious spectral aliasing. Perhaps someone could get rich figuring out a way to calculate a non-uniform array of holes.

[fiasco ( P.M.DuMont )]

AndreasN wrote on Thu, 19 June 2008 19:54

Fiasco wrote on Wed, 18 June 2008 12:42 The squares are 2 5/8 inches and the depth was totally random. The problem is that 'totally random' using the dice method requires a really long number of throws to be random. A short string of numbers is actually very hard to make random! The beauty of the correct math for diffusors is that the string ends up having the property that it is noise. A fourier analysis on such a string gives a flat spectra, just like white noise. The base sequence number is the number of fourier bins. More numbers in the base sequence gives better spread to the spectra.

Which led me to question the term "quasi random".

[jimmyjazz]

If there is a relatively easy way to numerically characterize a given geometry -- say, a spatial Fourier analysis -- then it should be possible to iterate through a sequence of designs iteratively, using some sort of an optimization routine to hunt for the "best" design within a given set of constraints.  I think it would require some sort of finite element or boundary element method.

How you manufacture the resulting design is another question entirely!  I suppose one could mill the cavities into a thick slab of refletive (or slightly absorptive) material.

[franman]

As this is a DIY project, and it's stimulated some discussion, I think we will leave this thread up.... nobody here is trying to sell anything for a profit. It's all good IMHO.

[fiasco ( P.M.DuMont )]

franman wrote on Tue, 17 June 2008 21:19

Other observations: the thickness of the dividers should be a thin as possible (for future revisions). This is a basic concept of phase gradient diffusors.

This makes sense to me. I could certainly go to 1/4 inch stock, however 1/8 inch would seem to compromise the rigidity.
I guess I could possibly do 1/8 inch in a hardwood, ash perhaps.

So, in anyone's opinion, would the direct opposite of this design...
hmm, I don't know how to word this
...(?)risers set at different heights on a plane(?), be more or less effective.

Staggered combination of the two?

Look in a different direction?