Can you elaborate a bit what you want to do??

I think you better process the table rather than go from one domain to another.

inverse the phase by multiplying by -1, you can "reverse" the table by reading it from back to front.

On Thu, Aug 4, 2016 at 8:53 AM, Simon Iten itensimon@gmail.com wrote:

sorry to post again.

i think sigmund~ can do this.

but i would be thankful for a hint how to calculate it. if it’s fft, then some insight how to use fft on a single wavetable would be very nice.

...

On 04 Aug 2016, at 09:34, Simon Iten itensimon@gmail.com wrote:

dear list,

i can create nice wavetables with sinesum.

what would be the easiest process to do the inverse?

reading a wavetable of known size and converting it to it’s sine

harmonics gain structure.

...

fft? i am a little bit lost on how to do this with a wavetable.

or is there a simpler way if i know the number of points and the range

of the waveform?

...

cheers

---

Pd-list@lists.iem.at mailing list UNSUBSCRIBE and account-management -> https://lists.puredata.info/listinfo/pd-list

Hi Simon,

On 04/08/16 09:54, Simon Iten wrote:

ah sorry, yes

i want to read a wavetable, say 128 steps and calculate the gain structure to aproximate this wavetable with sine waves. the idea behind this is to get a different sound from wavetables, use low resolution wavetables to get nice sounds (waldorf microwave xt)

so for a saw wave i would want the following numbers.

1 0.5 0.3333 0.25 0.2 and so on…

how to do this for an arbitrary input wavetable?

i looked at the fft examples but it is not clear to me how i would do this with a single wavetable (of known size)

You could do something like this with [rfft~], but you lose phase information which might be important depending on what you are doing (use fixed-width font to see diagram):

"set appropriate block size, turn off dsp, bang to execute 1 block" | [switch~]

[tabplay~ wavetable] | [rfft~ ] ^ ^ [*~] [*~] \ / [+~] | [sqrt~] | [tabsend~ spectrum]

or if there is a “simpler” (without fft) possibility that would be great.

I don't think you'll get simpler than FFT here. As a bonus you can also get phase information (sinesum has all phases 0, cosinesum all phases pi/2, general wavetable can have arbitrary phases).

You could use [rifft~] instead of sinesum to generate your wavetable, too. Note there may be some issues with normalization (fft->ifft has an amplitude gain equal to the blocksize, iirc).

For a more advanced use of oneshot FFT and IFFT for wave tables, see my bandlimited project: https://mathr.co.uk/blog/2015-02-12_bandlimited_wavetables.html

Claude

https://mathr.co.uk

Hi Simon, list,

On 04/08/16 22:51, Simon Iten wrote:

hi claude,

your bandlimited project works/sounds great! very nice sounds from handdrawn waves.

thanks!

however i am having trouble getting your example to work. probably just stupidity from my side.

i suppose i don’t understand your:

...

"set appropriate block size, turn off dsp, bang to execute 1 block”

message. if you find the time, can you have a look at it?

sure, just a [loadbang]--"0"--[switch~] - this turns off dsp for the canvas (so you might want to put it all in a subpatch unless you're using it as an abstraction). see attached, works for me

do i not need to run a bang into tabplay in this special one block send scenario?

with my patch i get “kind" of the right thing.

1. the table “spectrum” only updates once i click into the table (after i hit

the bang).

mm, seems to be a Pd bug or other infelicity. the attached includes a workaround using [tabwrite] which updates the GUI

2. i get only 64 values, so i guess my blocksize is still 64?

no, it's that for real signals the second half of the rfft~ output is symmetric to the first, so Pd sets it to 0 and ignores it in rifft~ to avoid confusion.

3. if i send a bang to switch~ when dsp is off, i get an error on the console:

bang to block~ or on-state switch~ has no effect.

you have to switch it off with "0" first, and have global dsp on (afaik).

cheers

hth,

Claude

https://mathr.co.uk

...

On 04 Aug 2016, at 18:01, Claude Heiland-Allen <claude@mathr.co.uk mailto:claude@mathr.co.uk> wrote:

Hi Simon,

On 04/08/16 09:54, Simon Iten wrote:

...

ah sorry, yes

i want to read a wavetable, say 128 steps and calculate the gain structure to aproximate this wavetable with sine waves. the idea behind this is to get a different sound from wavetables, use low resolution wavetables to get nice sounds (waldorf microwave xt)

so for a saw wave i would want the following numbers.

1 0.5 0.3333 0.25 0.2 and so on…

how to do this for an arbitrary input wavetable?

i looked at the fft examples but it is not clear to me how i would do this with a single wavetable (of known size)

You could do something like this with [rfft~], but you lose phase information which might be important depending on what you are doing (use fixed-width font to see diagram):

"set appropriate block size, turn off dsp, bang to execute 1 block" | [switch~]

[tabplay~ wavetable] | [rfft~ ] ^ ^ [*~] [*~] \ / [+~] | [sqrt~] | [tabsend~ spectrum]

...

or if there is a “simpler” (without fft) possibility that would be great.

I don't think you'll get simpler than FFT here. As a bonus you can also get phase information (sinesum has all phases 0, cosinesum all phases pi/2, general wavetable can have arbitrary phases).

You could use [rifft~] instead of sinesum to generate your wavetable, too. Note there may be some issues with normalization (fft->ifft has an amplitude gain equal to the blocksize, iirc).

For a more advanced use of oneshot FFT and IFFT for wave tables, see my bandlimited project: https://mathr.co.uk/blog/2015-02-12_bandlimited_wavetables.html

Claude

https://mathr.co.uk https://mathr.co.uk/

---

Pd-list@lists.iem.at mailto:Pd-list@lists.iem.atmailing list UNSUBSCRIBE and account-management ->https://lists.puredata.info/listinfo/pd-list

hello,

i'm also very interested in artefact free and band limited table reading.

(my work is in the nusmuk-audio svn folder, you will find there tabosci~ that implement a naive anti aliasing filter using table as oscillator).

The best algorythm i found so far to get an alias free, band limited table reading is this one : https://ccrma.stanford.edu/~jos/resample/resample.pdf but i did not had time to implement it in pd.

I would be glad if someone want to do it before me.

cheers cyrille

Le 05/08/2016 à 00:04, Claude Heiland-Allen a écrit :

Hi Simon, list,

On 04/08/16 22:51, Simon Iten wrote:

...

hi claude,

your bandlimited project works/sounds great! very nice sounds from handdrawn waves.

thanks!

...

however i am having trouble getting your example to work. probably just stupidity from my side.

i suppose i don’t understand your:

...

"set appropriate block size, turn off dsp, bang to execute 1 block”

message. if you find the time, can you have a look at it?

sure, just a [loadbang]--"0"--[switch~] - this turns off dsp for the canvas (so you might want to put it all in a subpatch unless you're using it as an abstraction). see attached, works for me

...

do i not need to run a bang into tabplay in this special one block send scenario?

with my patch i get “kind" of the right thing.

1. the table “spectrum” only updates once i click into the table (after i hit

the bang).

mm, seems to be a Pd bug or other infelicity. the attached includes a workaround using [tabwrite] which updates the GUI

...

2. i get only 64 values, so i guess my blocksize is still 64?

no, it's that for real signals the second half of the rfft~ output is symmetric to the first, so Pd sets it to 0 and ignores it in rifft~ to avoid confusion.

...

3. if i send a bang to switch~ when dsp is off, i get an error on the console:

bang to block~ or on-state switch~ has no effect.

you have to switch it off with "0" first, and have global dsp on (afaik).

...

cheers

hth,

Claude

---

Pd-list@lists.iem.at mailing list UNSUBSCRIBE and account-management -> https://lists.puredata.info/listinfo/pd-list