Hi, I've read miller's book and I've always wanted to know a bit more about reverbs. I also want to better explain these examples to my students. One thing I wonder is about the choice of delay lengths for the early reflections and also for the recirculating delay lines. They seem quite arbitrary, right?
All I could make of it is that these numbers are just broken down so they don't become multiples of each other, resulting in a rather complex mesh of reflection times.
But also, it seems that a "room size" parameter can be controlled by the size of the delay lines, right?
Anyway, I'd appreciate if I could learn more about these abstractions and also some cool references for study (such as "digital reverb modelling for dummies").
cheers
Hi, Alexandre.
On Thu, 7 Feb 2019 at 02:26, Alexandre Torres Porres porres@gmail.com wrote:
Hi, I've read miller's book and I've always wanted to know a bit more about reverbs. I also want to better explain these examples to my students. One thing I wonder is about the choice of delay lengths for the early reflections and also for the recirculating delay lines. They seem quite arbitrary, right?
All I could make of it is that these numbers are just broken down so they don't become multiples of each other, resulting in a rather complex mesh of reflection times.
Yes, as far as I understand, the idea is to have an FDN, with whatever matrix you've chosen, whose impulse response is as close as possible to white noise. The filters that you put inside the network can then be used to model different environments.
But also, it seems that a "room size" parameter can be controlled by the size of the delay lines, right?
There is an algorithm on the website of J. O. Smith to calculate the best approximation for a desired size while keeping the coprime relationship between the lengths: https://ccrma.stanford.edu/~jos/pasp/Prime_Power_Delay_Line_Lengths.html.
My reverb in PD is a 16th order FDN with Hadamard matrix and 1pole lowpass filters to model the absorption of high frequencies when bouncing on the walls. For simplicity, I pilot the delay lengths through a signal to change the exponent of 16 prime bases. Computationally acceptable and I like how it sounds.
Anyway, I'd appreciate if I could learn more about these abstractions and also some cool references for study (such as "digital reverb modelling for dummies").
I found these good reads:
Rocchesso, D. 1997. “Maximally diffusive yet efficient feedback delay networks for artificial reverberation. ” In Signal Processing Letters, IEEE 4 (9), 252255.
Rocchesso, D., and J. O. Smith. 1997. “Circulant and Elliptic Feedback Delay Networks for Artificial Reverberation.” I EEE Transactions on Speech and Audio Processing 5 (1):51–63.
Ciao, Dario
cheers _______________________________________________ Pd-list@lists.iem.at mailing list UNSUBSCRIBE and account-management -> https://lists.puredata.info/listinfo/pd-list
thanks everyone for all the "reverberation"
Em qui, 7 de fev de 2019 às 09:23, Dario Sanfilippo < sanfilippo.dario@gmail.com> escreveu:
as far as I understand, the idea is to have an FDN, with whatever matrix you've chosen, whose impulse response is as close as possible to white noise. The filters that you put inside the network can then be used to model different environments.
cool, thanks
There is an algorithm on the website of J. O. Smith to calculate the best
approximation for a desired size while keeping the coprime relationship between the lengths: https://ccrma.stanford.edu/~jos/pasp/Prime_Power_Delay_Line_Lengths.html.
awesome
My reverb in PD is a 16th order FDN with Hadamard matrix and 1pole lowpass filters to model the absorption of high frequencies
yeah, I saw that on the facebook group, and also that you mentioned there it was based on a model by "Rocchesso".
I found these good reads:
Rocchesso, D. 1997. “Maximally diffusive yet efficient feedback delay networks for artificial reverberation.” In Signal Processing Letters, IEEE 4 (9), 252255.
Rocchesso, D., and J. O. Smith. 1997. “Circulant and Elliptic Feedback Delay Networks for Artificial Reverberation.” I EEE Transactions on Speech and Audio Processing 5 (1):51–63.
yup, that's the source then :) I was able to find them both online!
Now, a question. You say it uses a Hadamard matrix, but I've realized it's basically the same matrix as used in [rev3~], it's only flipped vertically (I hope you know what I mean) and I ask if it is the same matrix or if this difference grants it to be another matrix that goes by a different name?
Ciao, Dario
Tchau, obrigado
Look up Sean Costello’s (valhalla reverbs)postings on forums and his blog. He shares lots of things, but you have to dig for it all.
On Tue, 12 Mar 2019 at 12:02, Alexandre Torres Porres porres@gmail.com wrote:
thanks everyone for all the "reverberation"
Em qui, 7 de fev de 2019 às 09:23, Dario Sanfilippo < sanfilippo.dario@gmail.com> escreveu:
as far as I understand, the idea is to have an FDN, with whatever matrix you've chosen, whose impulse response is as close as possible to white noise. The filters that you put inside the network can then be used to model different environments.
cool, thanks
There is an algorithm on the website of J. O. Smith to calculate the best
approximation for a desired size while keeping the coprime relationship between the lengths: https://ccrma.stanford.edu/~jos/pasp/Prime_Power_Delay_Line_Lengths.html.
awesome
My reverb in PD is a 16th order FDN with Hadamard matrix and 1pole lowpass filters to model the absorption of high frequencies
yeah, I saw that on the facebook group, and also that you mentioned there it was based on a model by "Rocchesso".
I found these good reads:
Rocchesso, D. 1997. “Maximally diffusive yet efficient feedback delay networks for artificial reverberation.” In Signal Processing Letters, IEEE 4 (9), 252255.
Rocchesso, D., and J. O. Smith. 1997. “Circulant and Elliptic Feedback Delay Networks for Artificial Reverberation.” I EEE Transactions on Speech and Audio Processing 5 (1):51–63.
yup, that's the source then :) I was able to find them both online!
Now, a question. You say it uses a Hadamard matrix, but I've realized it's basically the same matrix as used in [rev3~], it's only flipped vertically (I hope you know what I mean) and I ask if it is the same matrix or if this difference grants it to be another matrix that goes by a different name?
Ciao, Dario
Tchau, obrigado _______________________________________________ Pd-list@lists.iem.at mailing list UNSUBSCRIBE and account-management -> https://lists.puredata.info/listinfo/pd-list
like this: https://valhalladsp.com/category/reverb/ ? yeah, seems like a good discussion going on ;) cheers
Em ter, 12 de mar de 2019 às 03:49, Matt Davey hard.off@gmail.com escreveu:
Look up Sean Costello’s (valhalla reverbs)postings on forums and his blog. He shares lots of things, but you have to dig for it all.
On Tue, 12 Mar 2019 at 12:02, Alexandre Torres Porres porres@gmail.com wrote:
thanks everyone for all the "reverberation"
Em qui, 7 de fev de 2019 às 09:23, Dario Sanfilippo < sanfilippo.dario@gmail.com> escreveu:
as far as I understand, the idea is to have an FDN, with whatever matrix you've chosen, whose impulse response is as close as possible to white noise. The filters that you put inside the network can then be used to model different environments.
cool, thanks
There is an algorithm on the website of J. O. Smith to calculate the best
approximation for a desired size while keeping the coprime relationship between the lengths: https://ccrma.stanford.edu/~jos/pasp/Prime_Power_Delay_Line_Lengths.html .
awesome
My reverb in PD is a 16th order FDN with Hadamard matrix and 1pole lowpass filters to model the absorption of high frequencies
yeah, I saw that on the facebook group, and also that you mentioned there it was based on a model by "Rocchesso".
I found these good reads:
Rocchesso, D. 1997. “Maximally diffusive yet efficient feedback delay networks for artificial reverberation.” In Signal Processing Letters, IEEE 4 (9), 252255.
Rocchesso, D., and J. O. Smith. 1997. “Circulant and Elliptic Feedback Delay Networks for Artificial Reverberation.” I EEE Transactions on Speech and Audio Processing 5 (1):51–63.
yup, that's the source then :) I was able to find them both online!
Now, a question. You say it uses a Hadamard matrix, but I've realized it's basically the same matrix as used in [rev3~], it's only flipped vertically (I hope you know what I mean) and I ask if it is the same matrix or if this difference grants it to be another matrix that goes by a different name?
Ciao, Dario
Tchau, obrigado _______________________________________________ Pd-list@lists.iem.at mailing list UNSUBSCRIBE and account-management -> https://lists.puredata.info/listinfo/pd-list
Ther's a wonderful article here:
Blesser, B. (2001). An interdisciplinary synthesis of reverberation viewpoints. Journal of Audio Engineering Society 49(10): 867-903.
At some point I was able to find this online, but I can't anymore - perhaps you can get it on paper at an old-fashioned library.
cheers Miller On Wed, Feb 06, 2019 at 11:23:24PM -0200, Alexandre Torres Porres wrote:
Hi, I've read miller's book and I've always wanted to know a bit more about reverbs. I also want to better explain these examples to my students. One thing I wonder is about the choice of delay lengths for the early reflections and also for the recirculating delay lines. They seem quite arbitrary, right?
All I could make of it is that these numbers are just broken down so they don't become multiples of each other, resulting in a rather complex mesh of reflection times.
But also, it seems that a "room size" parameter can be controlled by the size of the delay lines, right?
Anyway, I'd appreciate if I could learn more about these abstractions and also some cool references for study (such as "digital reverb modelling for dummies").
cheers
Pd-list@lists.iem.at mailing list UNSUBSCRIBE and account-management -> https://lists.puredata.info/listinfo/pd-list
Somewhere out there is a Fiboverb I made that tunes [fdn~] with scaled segments of fibonacci sequence. Neighbours have relative coprimality giving a wonderfully clear, uncoloured overall reverb, but with nice "character" depending on where you start in the sequence.
I recall at the time I was fascinated by Golomb rulers and Dirichlet/Riemann sequences in a mad quest to find crafty ways to make cool reverbs, alas the patches are lost to the winds and my mind is a stranger to the maths once again... but I suggest its a fertile to place to go digging as I had some nice results.
best, Andy
On Wed, Feb 06, 2019 at 11:23:24PM -0200, Alexandre Torres Porres wrote:
Hi, I've read miller's book and I've always wanted to know a bit more about reverbs. I also want to better explain these examples to my students. One thing I wonder is about the choice of delay lengths for the early reflections and also for the recirculating delay lines. They seem quite arbitrary, right?
All I could make of it is that these numbers are just broken down so they don't become multiples of each other, resulting in a rather complex mesh of reflection times.
But also, it seems that a "room size" parameter can be controlled by the size of the delay lines, right?
Anyway, I'd appreciate if I could learn more about these abstractions and also some cool references for study (such as "digital reverb modelling for dummies").
cheers
Pd-list@lists.iem.at mailing list UNSUBSCRIBE and account-management -> https://lists.puredata.info/listinfo/pd-list
There are no good references for designing reverbs because the info falls under 'Trade Secrets' in the audio industry. I develop commercial reverbs and other effects as part of my job and can give you a few tips:
- FDN is not used in any of the 'classic' reverb designs (Concert Hall,
Plate, Chamber, etc). Eventide did start to use it in the late 80s, but their most famous reverb is not FDN based at all. A number of plugins and devices do use FDN now mostly for 'Room' type algorithms.
- The few patents for old reverb boxes are not very useful because they
leave out all of the 'secret sauce'.
- Modulation is where the magic happens.
- Andy's on to something with investigating mathematical sequences and
ratios
- Interpolation. When to do it and how to do it is also key.
- Filters. what type and where to use them
- Phase. Relates to filters and interpolation and creates some surprising
results when used well.
- Move to Boston. Almost all of the major reverb developments where done
here: Blesser(EMT/Lexicon), David G(Lexicon), Chris Moore(Ursa Major), Bill Gardner and so on. I think Miller even wrote the FDN paper at MIT.
Chris
On Wed, Feb 6, 2019 at 8:26 PM Alexandre Torres Porres porres@gmail.com wrote:
Hi, I've read miller's book and I've always wanted to know a bit more about reverbs. I also want to better explain these examples to my students. One thing I wonder is about the choice of delay lengths for the early reflections and also for the recirculating delay lines. They seem quite arbitrary, right?
All I could make of it is that these numbers are just broken down so they don't become multiples of each other, resulting in a rather complex mesh of reflection times.
But also, it seems that a "room size" parameter can be controlled by the size of the delay lines, right?
Anyway, I'd appreciate if I could learn more about these abstractions and also some cool references for study (such as "digital reverb modelling for dummies").
cheers _______________________________________________ Pd-list@lists.iem.at mailing list UNSUBSCRIBE and account-management -> https://lists.puredata.info/listinfo/pd-list
You might also be interested in having a look at Tom Erbes reverb patches:
http://tre.ucsd.edu/wordpress/?p=285 http://tre.ucsd.edu/wordpress/?p=625
On 07.02.19 02:23, Alexandre Torres Porres wrote:
Hi, I've read miller's book and I've always wanted to know a bit more about reverbs. I also want to better explain these examples to my students. One thing I wonder is about the choice of delay lengths for the early reflections and also for the recirculating delay lines. They seem quite arbitrary, right?
All I could make of it is that these numbers are just broken down so they don't become multiples of each other, resulting in a rather complex mesh of reflection times.
But also, it seems that a "room size" parameter can be controlled by the size of the delay lines, right?
Anyway, I'd appreciate if I could learn more about these abstractions and also some cool references for study (such as "digital reverb modelling for dummies").
cheers
Pd-list@lists.iem.at mailing list UNSUBSCRIBE and account-management -> https://lists.puredata.info/listinfo/pd-list
-
Alexandre Torres Porres -
Andy Farnell -
Chris Clepper -
Dario Sanfilippo -
Matt Davey -
Max -
Miller Puckette