SmokeDoepferEveryday

i-am-sitting-in-a-room-with-vanilla-objects

I'm not aware of the Elsea/Lyon methodologies, but wouldn't it just be [read-buffer]-> [delay/filter network]->[write buffer], then copy contents of write-buffer into read and begin again.

tapping buffer: but you wouldn't want to tap every iteration? that would be messy. Doesn't it need a variable record length then automated record wrapping of the process thereafter [and the source needs to playback out the dac and only recording in from the adc] with some sort of feedback gain control too.

The Elsea and Lyon got me wondering why they turned to their externals [for finesse/accuracy I guess?] adding the patches for discussion:

Elsea Lucier.maxpat

Lyon Lucier.maxpat

let me rephrase. "tapping buffer" was somewhat wrong:

- a simple delay in the feedbackloop which is as long as the material is

since this will run endless, you have to click the stop button yourself.

 - a reverb
 - a feedback loop around the reverb
 - a simple delay in the feedbackloop which is as long as the material is

since this will run endless, you have to click the stop button yourself.



yeh, there's a tonne of ways to implement that. I just wondered what would be most economic. I get the process by the way, I'm just keen to see how people would literally approach it. There doesn't seem to be a definitive version or method, yet it's become a common thing to program in Max rather than revert to the hardware performance.

You create two buffers. Loading a source audio to the first buffer. Playing this buffer (I mean, the sound should be audible in the room). In the same time you are recording the sound (mic->adc->[second buffer]). Now you are repeating the procedure, but playing second buffer and overwriting the content of the first buffer. You repeat the procedure again (each time playing the last recorded buffer). With each repetition, the acoustic parameters of the room affect some frequencies.

Staying "inside" computer and using filters... it seems to me to act in the opposite to Lucier's assumptions..

You create two buffers. Loading a source audio to the first buffer. Playing this buffer (I mean, the sound should be audible in the room). In the same time you are recording the sound (mic->adc->[second buffer]).  Now you are repeating the procedure, but playing second buffer and overwriting the content of the first buffer. You repeat the procedure again (each time playing the last recorded buffer). With each repetition, the acoustic parameters of the room affect some frequencies.

Staying "inside" computer and using filters... it seems to me to act in the opposite to Lucier's assumptions.. 

I see what you're saying @yaniki but my understanding of the filtering in the DSP is because the feedback will be far more acute and onset than the analogue, so to prolong the convolution through iterations some filtering is useful.

https://willparry.files.wordpress.com/2014/09/screen-shot-2014-09-05-at-14-46-19.png

this seemed pretty economical to me: https://willparry.files.wordpress.com/2014/09/screen-shot-2014-09-05-at-14-46-19.png 

yes, leaving the reverb part out is very economic. just not very "room".

are you talking about making an artificial space? the reverberance comes from whatever space the microphone and speaker is in otherwise.

yes of course, i thought you are going to emulate all parts in software. :)

aaaah! I thought our tracks were differing, sorry! I was just thinking for the replication of the original using Max instead of the tapes. I'm considering making an article about the considerations. But of course a simulated version is interesting too - I found an interesting rendition in Python not too long ago: 

https://github.com/zo7/sitting-in-a-room/blob/master/sitting-in-a-room.py

aaaah! I thought our tracks were differing, sorry! I was just thinking for the replication of the original using Max instead of the tapes. I'm considering making an article about the considerations. But of course a simulated version is interesting too - I found an interesting rendition in Python not too long ago: https://github.com/zo7/sitting-in-a-room/blob/master/sitting-in-a-room.py 

uses buffir -- best to use a large signal vector size to alleviate cpu

here's the soundfile to use (put it in your search path):

lucier_excerpt.wav.zip

uses buffir -- best to use a large  signal vector size to alleviate cpu

here's the soundfile to use (put it in your search path):


solved