Hi!

This is more of a request for tips rather than a technical issue, but i figured it was most appropriately posted in this section of the forum.

Basically i'd like to know your opinion about something. I am making a looper, which is intended to be fairly big and not so linear. That means that i don't seek to implement to much sync-effectivity to regular rythms (4/4, 3/4, 214/0,3 and so on...).
As of now i'm writing the recorded audio to disk via [writesf~] and opening it in a dedicated array. However, tonight it struck me. I could just write it into an array without creating .wav file for every loop recorded. It also seems like an ideal thing to do if overdubbing is wanted (and it is), while writing to disk would require to recreate the .wav file and reload it to the array everytime an overdub is recorded (quite a nasty thing to do if looping in a live situation i'd guess).

What are your thoughts? What are the pros and cons of using either method? Maybe there is another more effective?

I see that I may have been quite ineffective at postulating the question, and if more information is required i will cooperate However, mine is more of a speculation than a specific technical issue, so it might be difficult to be any clearer. Any thought is appreciated, even if it at first seems impertinent to the question.

If you are going to use the looper in any sort of live setting, arrays are certainly the way to go. A .wav file that is 74 minutes in length takes up about 700mb, and if you keep it in RAM, you can access it without any hang. But writing it to disk with writesf~, and then reading it back with readsf~ can be a source of system hang. If you want to save and recall, soundfiler is my preference.

I am making a looper, which is intended to be fairly big

How big is big? Note that Pd's data type (single precision float) lack the resolution to index big tables. Beyond 2^24 samples it goes wrong and aliasing will be the audible result. For fractional speed reading in big tables the limited resolution is even a bigger problem.

See patch B15.tabread4~-onset.pd in Pd's audio help patches for a description of this problem and a possible workaround.

Katja

@sunji said:

If you are going to use the looper in any sort of live setting, arrays are certainly the way to go. A .wav file that is 74 minutes in length takes up about 700mb, and if you keep it in RAM, you can access it without any hang. But writing it to disk with writesf~, and then reading it back with readsf~ can be a source of system hang. If you want to save and recall, soundfiler is my preference.

You have a very good point. I do think i'll settle with arrays, which will require quite an overhaul, but it seems to be worth.

@katjav said:

How big is big? Note that Pd's data type (single precision float) lack the resolution to index big tables. Beyond 2^24 samples it goes wrong and aliasing will be the audible result. For fractional speed reading in big tables the limited resolution is even a bigger problem.

See patch B15.tabread4~-onset.pd in Pd's audio help patches for a description of this problem and a possible workaround.

Katja

Thanks a lot for the warning. I will study the help patch thoroughly when I get some time over. This is great information as I do regard myself as a novice (which is why I may have misunderstood your question, as my upcoming answer may give away ).

The "bigness" refers more to the desired features than the size of the table(s) I'd like to store data inside. However, I want to have more than one table looping at the same time (ideally as many as my computer/Pure Data is able to handle). Which begs the question: do you think that a lot of small tables will cause the same problem you described if their collective size goes beyond the aforementioned limit of 2^24?

As I already stated, I'm a novice, so if I did misunderstand what you meant, I ask for forgiveness and would be more than happy to learn something new.

do you think that a lot of small tables will cause the same problem you described if their collective size goes beyond the aforementioned limit of 2^24?

If you have many loopers each with a few seconds length there will be no precision problem because you don't need large numbers for indexing.

Katja