The No.1 Website for Pro Audio
DIY Binary Amplitude Diffuser anyone?
Old 6th March 2014 | Show parent
  #31
Lives for gear
 
boggy's Avatar
Quote:
Originally Posted by robertopisa View Post
Hello Boggy,

I have a large 150x200x30 porous absorber (Caruso IsoBond with an internal metal plate resonator) on the front wall between the two speakers.

The sound is a bit dead, but I read here on GS that your binary MLS pattern, which I like, works better when the entire front wall is covered with porous absorber, which is not my case.
Other room treatment may be a reason too, because response to sound "a bit dead"
Also the binary MLS pattern is not "mine", I use it extensively, but I did not invent it.
Quote:
Originally Posted by robertopisa View Post
I thus decided to use slats with the 534 sequence pattern. My question: between the speakers is it better to have vertical or horizontal slats on the front wall?
534 "pattern" is not real binary pseudo random sequence. Ok, it is better than fully periodic sequence.
Between speakers it is better to have vertical slats.
Quote:
Originally Posted by robertopisa View Post
I saw pictures for both choices, and I know how their 1D diffusion operates but it is unclear to me which are the advantages of each choice.
They can soften loudspeaker diffraction a bit when they are parallel to the (loudspeaker) edges.
Quote:
Originally Posted by robertopisa View Post

Thanks
-Roberto
You're welcome
Quote:
Originally Posted by robertopisa View Post

Edit: The topic of vertical vs horizontal slats is addressed here but mainly for side walls Horizontal versus vertical slats. While most agree that vertical slats are better for side walls, there is no explicit agreement for the front walls. I would opt for horizontal ones but I have no clear argument for that.
As I said... slats parallel to the loudspeaker edges... are better arrangement for the absorbers between speakers.



Old 6th March 2014 | Show parent
  #32
Gear Maniac
 
robertopisa's Avatar
Thanks Boggy, very helpful!

Can I cover the slats with the panel cloth or does this diminish their effects?

-R
Old 6th March 2014 | Show parent
  #33
Lives for gear
 
boggy's Avatar
Quote:
Originally Posted by robertopisa View Post
Thanks Boggy, very helpful!
My pleasure!
Quote:
Originally Posted by robertopisa View Post
Can I cover the slats with the panel cloth or does this diminish their effects?

-R
I never recommend something as that.


Old 7th March 2014 | Show parent
  #34
Gear Maniac
 
robertopisa's Avatar
Thanks Boggy, I hope this Python code is useful to generate a specular/symmetrical MLS for the vertical slats to be placed on the front wall between the two speakers.

To generate a MLS of length 2^4-1 = 15 and its specular MLS, where each MLS is rotated by 3 positions, it suffices to open a terminal shell and launch the following command, where the long run of symbols '###################################################################' visually represents slats, so allowing us to visually check if the resulting sequence is OK.
Code:
python MLS2.py 4 3
Output:
Code:
position: 0014 ***** 1  ###################################################################
position: 0013 ***** 1  ###################################################################
position: 0012 ***** 1  ###################################################################
position: 0011 ***** 1  ###################################################################
position: 0010 ***** 0  
position: 0009 ***** 0  
position: 0008 ***** 0  
position: 0007 ***** 1  ###################################################################
position: 0006 ***** 0  
position: 0005 ***** 0  
position: 0004 ***** 1  ###################################################################
position: 0003 ***** 1  ###################################################################
position: 0002 ***** 0  
position: 0001 ***** 1  ###################################################################
position: 0000 ***** 0  
position: 0000 ***** 0  
position: 0001 ***** 1  ###################################################################
position: 0002 ***** 0  
position: 0003 ***** 1  ###################################################################
position: 0004 ***** 1  ###################################################################
position: 0005 ***** 0  
position: 0006 ***** 0  
position: 0007 ***** 1  ###################################################################
position: 0008 ***** 0  
position: 0009 ***** 0  
position: 0010 ***** 0  
position: 0011 ***** 1  ###################################################################
position: 0012 ***** 1  ###################################################################
position: 0013 ***** 1  ###################################################################
position: 0014 ***** 1  ###################################################################
In my case, I am using slats of 6cm width and offset=1:
Code:
python MLS2.py 4 1
I obtain a nice pattern:
Code:
position: 0014 ***** 1  ###################################################################
position: 0013 ***** 1  ###################################################################
position: 0012 ***** 0  
position: 0011 ***** 0  
position: 0010 ***** 0  
position: 0009 ***** 1  ###################################################################
position: 0008 ***** 0  
position: 0007 ***** 0  
position: 0006 ***** 1  ###################################################################
position: 0005 ***** 1  ###################################################################
position: 0004 ***** 0  
position: 0003 ***** 1  ###################################################################
position: 0002 ***** 0  
position: 0001 ***** 1  ###################################################################
position: 0000 ***** 1  ###################################################################
position: 0000 ***** 1  ###################################################################
position: 0001 ***** 1  ###################################################################
position: 0002 ***** 0  
position: 0003 ***** 1  ###################################################################
position: 0004 ***** 0  
position: 0005 ***** 1  ###################################################################
position: 0006 ***** 1  ###################################################################
position: 0007 ***** 0  
position: 0008 ***** 0  
position: 0009 ***** 1  ###################################################################
position: 0010 ***** 0  
position: 0011 ***** 0  
position: 0012 ***** 0  
position: 0013 ***** 1  ###################################################################
position: 0014 ***** 1  ###################################################################
The code is attached below.
Attached Files
File Type: zip MLS2.zip (1.3 KB, 345 views)
Old 7th March 2014 | Show parent
  #35
Gear Maniac
 

Quote:
Originally Posted by robertopisa View Post
Thanks Boggy, I hope this Python code is useful to generate a specular/symmetrical MLS for the vertical slats to be placed on the front wall between the two speakers.

To generate a MLS of length 2^4-1 = 15 and its specular MLS, where each MLS is rotated by 3 positions, it suffices to open a terminal shell and launch the following command, where the long run of symbols '###################################################################' visually represents slats, so allowing us to visually check if the resulting sequence is OK.

[...]
Thanks for this program Roberto. And thanks Boggy for sharing your great work. I will use slats too and this is really helpful.

When running the MLS program, it always outputs x slats ("1") and x-1 holes ("0"), or the contrary. For 127 values, there are 64 slats and 63 holes. If there are no gaps between slats it means 50.4% of slats. For 15 values it's 53.3%.
If necessary it can be compensated by spacing the slats or the holes by a few millimeters (Boggy uses this technique to get a specific balance between slats and holes, like 45% and 55%, or any other balance).

If I understand the use of slats in front of bass traps correctly:
* it should be as aperiodic as possible (randomness is a good solution for this)
* the balance between slats and holes should be controlled
Is it correct?


Something completely different, I found an online true-random number generator (based on atmospheric noise), here is the version for integer numbers:
RANDOM.ORG - Integer Generator
Before running it many times read this:
RANDOM.ORG - Quota Top-Up

Is it possible (or recommended) to use true-random for slats? A true random system is likely to generate a 15-number sequence like this:
0 0 1 1 1 1 1 1 1 1 1 1 1 1 1
It can also happen when flipping coins or similar random experiments. What do you do in this case?

True-random is not something controlled. But maybe it's possible to generate a long true-random sequence, then extract a segment that you like, then compensate the balance of 0 and 1 with spacing the slats by a few mm. As a room treatment I am not sure it sounds different to MLS though.

Julien
Old 8th March 2014 | Show parent
  #36
Lives for gear
 
boggy's Avatar
Thanks Roberto!!!

Quote:
Originally Posted by julien79 View Post
Thanks for this program Roberto. And thanks Boggy for sharing your great work. I will use slats too and this is really helpful.

When running the MLS program, it always outputs x slats ("1") and x-1 holes ("0"), or the contrary. For 127 values, there are 64 slats and 63 holes. If there are no gaps between slats it means 50.4% of slats. For 15 values it's 53.3%.
If necessary it can be compensated by spacing the slats or the holes by a few millimeters (Boggy uses this technique to get a specific balance between slats and holes, like 45% and 55%, or any other balance).

If I understand the use of slats in front of bass traps correctly:
* it should be as aperiodic as possible (randomness is a good solution for this)
* the balance between slats and holes should be controlled
Is it correct?


Something completely different, I found an online true-random number generator (based on atmospheric noise), here is the version for integer numbers:
RANDOM.ORG - Integer Generator
Before running it many times read this:
RANDOM.ORG - Quota Top-Up

Is it possible (or recommended) to use true-random for slats? A true random system is likely to generate a 15-number sequence like this:
0 0 1 1 1 1 1 1 1 1 1 1 1 1 1
It can also happen when flipping coins or similar random experiments. What do you do in this case?

True-random is not something controlled. But maybe it's possible to generate a long true-random sequence, then extract a segment that you like, then compensate the balance of 0 and 1 with spacing the slats by a few mm. As a room treatment I am not sure it sounds different to MLS though.

Julien
I think you complicate too much. It is not needed to be that precise with size of reflective and absorptive surface... It is even possible to go up to 75% reflective covering with "two dimensional MLS" as shown at image below:


Also generate a much larger sequence and use whatever part you like from it...

again 2D:


Old 8th March 2014
  #37
Gear Maniac
 
robertopisa's Avatar
@boggy: Interesting the 2D MLS

@julien79: The random.org stuff is good, the owner was once generating random bits using some radioactive isotopes in his isolated basement !!

However, the MLS sequence is better for two reasons:

- random sequences make sense when they are sufficiently long (e.g. 1000 bits), but typically we use few slats (15 or 30 in my case), so randomness on 15 or 30 bits is not very meaningful from a statistically point of view

- MLS of order k (k=4 in my example) guarantees that if you slide a "window" of k or more bits on the MLS, you obtain 2^k - k distinct bit configurations (they repeat only once), just shifting the window from one position to the next (e.g. De Bruijn sequence). It is balanced in this sense and this cannot be guaranteed by short random sequences (while this is true with high probability if you take a sufficiently long random sequence, but it is too long for our application).


Technically speaking, MLS cannot be said to be random as it exists an algorithm to generate it, but it behaves better than truly random short sequences for our purposes. If we had to build 1000 consecutive slats, then random would be ok.

Edit: OT nerd stuff Truly random bits exist but we cannot decide if any sequence is truly random.

Last edited by robertopisa; 8th March 2014 at 09:27 AM.. Reason: typos
Old 16th March 2014
  #38
Boggy thats a really inspiring pic! Have you ever experimented with ternary diffusers? (1,0,-1) or someone else? If you did is the better scatering really noted?
Old 18th March 2014 | Show parent
  #39
Lives for gear
 
boggy's Avatar
Quote:
Originally Posted by Sebastian Vera View Post
Boggy thats a really inspiring pic!
Thanks!
Quote:
Originally Posted by Sebastian Vera View Post
Have you ever experimented with ternary diffusers? (1,0,-1) or someone else? If you did is the better scatering really noted?
I did not try ternary, but senidenary (hexadecimal) diffusers (phase grating!)... And yes, they had much better (audible!) performance than binary type (amplitude grating basically)


Old 18th March 2014 | Show parent
  #40
Quote:
Originally Posted by boggy;9952861

I did not try [I
ternary[/I], but senidenary (hexadecimal) diffusers (phase grating!)... And yes, they had much better (audible!) performance than binary type (amplitude grating basically)


Wow never heard of that!. But then if it is phase diffuser i guess it would perform better.

Are this type of diffuser (senidenary) easyer to construct or design than the qrds?
Old 19th March 2014 | Show parent
  #41
Lives for gear
 
boggy's Avatar
Quote:
Originally Posted by Sebastian Vera View Post
Wow never heard of that!. But then if it is phase diffuser i guess it would perform better.

Are this type of diffuser (senidenary) easyer to construct or design than the qrds?
It depends on what you want to achieve,... I plan to design one studio following this concept, and to publish results and achievements later (AES Convention or my local TELFOR conference), I believe that this will be much better than telling stories before any real experience...


Old 19th March 2014 | Show parent
  #42
Gear Maniac
 

Quote:
Originally Posted by boggy View Post

[...]

I think you complicate too much. It is not needed to be that precise with size of reflective and absorptive surface... It is even possible to go up to 75% reflective covering with "two dimensional MLS" as shown at image below:

[image]

Also generate a much larger sequence and use whatever part you like from it...

again 2D:

[image]

Thanks for your answer boggy, and congratulations for these great studios

I made a quick program that shows the slat positions depending on sequence length and starting point (at the moment it works only for fixed room and slat dimensions , making it universal takes time but I can try if someone if interested). The four walls are displayed around the floor (white rectangle in the middle). Slats are white and absorbent is black. The ceiling is not shown.

After bass traps my room will be 330 (L) x 240 (W) x 205 (H) cm. The slats I found are 22 x 48 mm (construction wood). Each side of the room needs a sequence of 100 slats or holes (4 mm gap). Here are two examples:

DIY Binary Amplitude Diffuser anyone?-mls_127_9.png DIY Binary Amplitude Diffuser anyone?-mls_511_15_inverted.png

[EDIT: manual calculus error! It's more than 100 slats per side even with 4 mm gap, but this does not change anything to the discussion]

After trying configurations I came to these questions:
- When using only a segment of an MLS sequence, does it sound similar no matter where you start?
- Is it better to have global similarity or as much disorder as possible? For example see the difference between the images above
- Is it possible that slats produce flutter echo when continuous groups of 5 slats (or more) are facing each other on side walls?
- Is it possible to paint slats?

Thanks again for your help
Julien
Old 19th March 2014 | Show parent
  #43
Gear Maniac
 

Quote:
Originally Posted by robertopisa View Post
[...]
The random.org stuff is good, the owner was once generating random bits using some radioactive isotopes in his isolated basement !!

However, the MLS sequence is better for two reasons:

- random sequences make sense when they are sufficiently long (e.g. 1000 bits), but typically we use few slats (15 or 30 in my case), so randomness on 15 or 30 bits is not very meaningful from a statistically point of view

- MLS of order k (k=4 in my example) guarantees that if you slide a "window" of k or more bits on the MLS, you obtain 2^k - k distinct bit configurations (they repeat only once), just shifting the window from one position to the next (e.g. De Bruijn sequence). It is balanced in this sense and this cannot be guaranteed by short random sequences (while this is true with high probability if you take a sufficiently long random sequence, but it is too long for our application).


Technically speaking, MLS cannot be said to be random as it exists an algorithm to generate it, but it behaves better than truly random short sequences for our purposes. If we had to build 1000 consecutive slats, then random would be ok.
Thanks for your answer Roberto. I agree, true random is not adapted to a (relatively) small number of slats. I am using MLS now.

Quote:
Edit: OT nerd stuff Truly random bits exist but we cannot decide if any sequence is truly random.
+1
http://www.random.org/analysis/dilbert.jpg

Julien
Old 19th March 2014 | Show parent
  #44
Quote:
Originally Posted by boggy View Post
It depends on what you want to achieve,...

What I mean is, a n53 qrd has lots of diferent depth so cutig may become tedious... A binary or ternary may be more easy, less diferent cuts, and may be cheaper also!

Quote:
Originally Posted by boggy View Post
I plan to design one studio following this concept, and to publish results and achievements later (AES Convention or my local TELFOR conference), I believe that this will be much better than telling stories before any real experience...


Youp i believe it too! Looking forward to hearing from that! Good luck!
Old 19th March 2014 | Show parent
  #45
Lives for gear
 
boggy's Avatar
Quote:
Originally Posted by julien79 View Post
Thanks for your answer boggy, and congratulations for these great studios

Quote:
Originally Posted by julien79 View Post
After trying configurations I came to these questions:
- When using only a segment of an MLS sequence, does it sound similar no matter where you start?
I didn't find any (audible) difference between them...
Quote:
Originally Posted by julien79 View Post
- Is it better to have global similarity or as much disorder as possible? For example see the difference between the images above
It is irrelevant from my perspective.
Quote:
Originally Posted by julien79 View Post
- Is it possible that slats produce flutter echo when continuous groups of 5 slats (or more) are facing each other on side walls?
No.
Quote:
Originally Posted by julien79 View Post
- Is it possible to paint slats?
Yes. Water based coating with hard surface after drying, three times with brush or roller, with light sanding between.
Quote:
Originally Posted by julien79 View Post
Thanks again for your help
Julien
You're welcome


Old 19th March 2014 | Show parent
  #46
Lives for gear
 
boggy's Avatar
Quote:
Originally Posted by Sebastian Vera View Post
What I mean is, a n53 qrd has lots of diferent depth so cutig may become tedious... A binary or ternary may be more easy, less diferent cuts, and may be cheaper also!
True.
Quote:
Originally Posted by Sebastian Vera View Post
Youp i believe it too! Looking forward to hearing from that! Good luck!
Thanks



Old 19th March 2014 | Show parent
  #47
Quote:
Originally Posted by julien79 View Post
After trying configurations I came to these questions:
- When using only a segment of an MLS sequence, does it sound similar no matter where you start?
- Is it better to have global similarity or as much disorder as possible? For example see the difference between the images above
- Is it possible that slats produce flutter echo when continuous groups of 5 slats (or more) are facing each other on side walls?
- Is it possible to paint slats?
Julien
1, yes it sounds the same.
2. It doesnt matter much, either is good. Becouse both are rigid by the same principle, this secuence has the propiety of avoiding the effects of repetition so lobing (what i think you are talking about) is unlikely to be important to be notice by ear

3. No. The canges in impedance will prevent the flutter echo.

4. Paint, like normal paint wont be an issue. It may help if you use a latex paint or somthing that helps the reflection.
Old 19th March 2014 | Show parent
  #48
Gear Maniac
 

Many thanks to Boggy, Sebastian and Roberto for your help!
Old 21st March 2014 | Show parent
  #49
Gear Maniac
 

Hello

Finally I found 2 types of slats:
- 48 x 22 mm
- 27 x 13 mm

I read people use 20~80 x 10~30 mm slats so I guess both are working. But as I have the choice, is there a reason to choose one particular type, apart from aesthetics and price? This is for a very small control room.

* The big ones (48 x 22 mm) can be used as structural elements (eg for carrying ceiling clouds), making the building very simple. But the maximum working frequency for amplitude diffusion is only ~3600 Hz.
Also, is it possible that thick slats make low frequency absorption (behind slats) less good?

* The small ones (27 x 13 mm) require an additional structure to mount them, but the maximum working frequency is ~6400 Hz.

Cheers
Julien
Old 7th April 2014 | Show parent
  #50
Lives for gear
 
jhbrandt's Avatar
MLS sequence (optimized)

I thought I would share this for anyone who is brave enough to attempt.
The zip file is an Excel Spreadsheet.
This sequence gives a uniform/equal amount of 0s and 1s.
The pdf layout is in a grid of 61 by 83 positions without repeats.
Fun time.

Cheers,
John

-- I love math!
Attached Files
File Type: pdf Binary (MLS) panel 5063 sequence.pdf (527.7 KB, 1509 views) File Type: zip MLS sequence to 5099.zip (154.0 KB, 664 views)
Old 7th April 2014 | Show parent
  #51
Gear Maniac
 
robertopisa's Avatar
Thanks for sharing your project
-R
Old 13th April 2014
  #52
Gear Head
 

Thanks for sharing!
Old 13th April 2014 | Show parent
  #53
Lives for gear
 
jhbrandt's Avatar
Thanks & you are welcome. BTW, this number sequence was calculated using MatLab.

We are using this large sequence on curved metal panels in a Home Theater build in Saudi Arabia.

Cheers,
John
Old 1st July 2014 | Show parent
  #54
Lives for gear
 
almaelectronix's Avatar
 

Quote:
Originally Posted by boggy View Post
It is even possible to go up to 75% reflective covering with "two dimensional MLS" as shown at image below:

Is there any tips you can share for the slat placement?

Is there a gap between absorption and the slats? should i put any spacers or I just nail it on the absorption's frame? ..Of course after finishing with the garden felt and the fabric.

Also, should I just nail only the edges of the slat using one nail for each end or nail it really hard like putting 5-6 nails across the slat?

For the 2d slat design, is there a gap between the two layers of slats, or I just nail the horizontals over the vertical slats?

Lot of questions but I am afraid of any possible rattling noise coming from the slat movement.
Old 2nd July 2014 | Show parent
  #55
Quote:
Originally Posted by jhbrandt View Post
Thanks & you are welcome. BTW, this number sequence was calculated using MatLab.

We are using this large sequence on curved metal panels in a Home Theater build in Saudi Arabia.

Cheers,
John
Many thax for sharing the secuence!!!!

and 1 question for you.

John, Would you expect for a 1d binary diffuser using an optimized curved surface shape to show a true diffusion characteristic? or at least enchance it?
Old 8th July 2014 | Show parent
  #56
Lives for gear
 
almaelectronix's Avatar
 

My questions were answered on this thread:
Help with MLS Sequence - Small Room
Old 13th August 2014
  #57
Lives for gear
 
blizt's Avatar
There is something I don't understand about the fact that you can use any part of the sequence. I understand that the sequence is "random", but say I'm going to pick 13 positions to put the slats in front of superchunks. If I pick the first 13 positions there is a lot more "1" than picking from position 60 to 72. I want to use 40mmx20mm slats fwiw.
Old 14th August 2014
  #58
Lives for gear
 

You will need to have a 1-off sum of 0 and 1 for it to be correct. 7 and 6.
Old 14th August 2014 | Show parent
  #59
Lives for gear
 
blizt's Avatar
Quote:
Originally Posted by OpusOfTrolls View Post
You will need to have a 1-off sum of 0 and 1 for it to be correct. 7 and 6.
Then it's wrong what was said about picking any part of the sequence you desire or maybe I didn't get it right. Thanks for your help
Old 14th August 2014 | Show parent
  #60
Gear Maniac
 
robertopisa's Avatar
I do not think so at it depends on the percentage of reflection you want. If you want 50%, then take a sequence with an equal number of 1s and 0s. But if you want higher (for example on the front wall in my case), you can have more 1s than 0s. I am getting slightly less than 70% to get the lively level I prefer (it's up to your preference). On the two side walls, I used 50%. If you look at previous posts, I put the Python code to generate the symmetric sequence for the front wall.

Quote:
Originally Posted by blizt View Post
Then it's wrong what was said about picking any part of the sequence you desire or maybe I didn't get it right. Thanks for your help
📝 Reply

Similar Threads

Thread / Thread Starter Replies / Views Last Post
replies: 79 views: 31863
Avatar for rids
rids 1st July 2009
replies: 169 views: 120141
Avatar for MoyenChari
MoyenChari 17th September 2020
replies: 65 views: 8762
Avatar for RonaldDumsfeld
RonaldDumsfeld 26th March 2010
replies: 57 views: 69443
Avatar for navu28
navu28 20th April 2018
Post Reply

Welcome to the Gearslutz Pro Audio Community!

Registration benefits include:
  • The ability to reply to and create new discussions
  • Access to members-only giveaways & competitions
  • Interact with VIP industry experts in our guest Q&As
  • Access to members-only sub forum discussions
  • Access to members-only Chat Room
  • Get INSTANT ACCESS to the world's best private pro audio Classifieds for only USD $20/year
  • Promote your eBay auctions and Reverb.com listings for free
  • Remove this message!
You need an account to post a reply. Create a username and password below and an account will be created and your post entered.


 
 
Slide to join now Processing…
🖨️ Show Printable Version
✉️ Email this Page
🔍 Search thread
🎙️ View mentioned gear
Forum Jump
Forum Jump