Pitch bend quantization/correction
| Date | 2016-04-09 13:59 |
| From | Gerard Rodríguez |
| Subject | Pitch bend quantization/correction |
Hi everyone, I want to correct a pitch bend message so that the note on always sounds in tune but allows for vibratos and pitch slides (and ignores slow movements).For a bit of context, this is the controller that I'm using. http://www.rogerlinndesign.com/linnstrument.html When I hit a square, the pitch bend message depends on the x position of the finger, so if I were to hit the square on the right side, and use the pitch bend message as is it would sound a quarter step out of tune. The controller does have a pitch correction feature, but I want to use the x position to control other things so I need the original pitch bend message (and sometimes their correction is a bit buggy) Thanks. |
| Date | 2016-04-09 18:41 |
| From | Rory Walsh |
| Subject | Re: Pitch bend quantization/correction |
When you hit a square, you get a note on message, AND a pitch bend message? As in you get two separate MIDI messages or? On 9 April 2016 at 13:59, Gerard Rodríguez <binbiniqegabenik@gmail.com> wrote:
|
| Date | 2016-04-09 19:10 |
| From | Gerard Rodríguez |
| Subject | Re: Pitch bend quantization/correction |
That's right you get a note on, and then a continuous stream of pitch bend messages. 2016-04-09 19:41 GMT+02:00 Rory Walsh <rorywalsh@ear.ie>:
|
| Date | 2016-04-09 19:33 |
| From | Iain McCurdy |
| Subject | Re: Pitch bend quantization/correction |
|
If you are using Csound as your synthesiser then it is easy just to ignore the pitch bend data if you don't want it to affect tuning. Are you using Csound as your synthesiser? I don't think that you have stated that explicitly yet. Iain Date: Sat, 9 Apr 2016 20:10:40 +0200 From: binbiniqegabenik@GMAIL.COM Subject: Re: [Csnd] Pitch bend quantization/correction To: CSOUND@LISTSERV.HEANET.IE That's right you get a note on, and then a continuous stream of pitch bend messages. 2016-04-09 19:41 GMT+02:00 Rory Walsh <rorywalsh@ear.ie>:
|
| Date | 2016-04-09 20:31 |
| From | Gerard Rodríguez |
| Subject | Re: Pitch bend quantization/correction |
Yes I am using Csound, although sometimes I also use supercollider that's why I asked more generally for an algorithm. You didn't understand me right, I want the pitch bend data, I don't want to ignore it, I want it to be corrected for the initial touch, and the following movements, for example if I do a slow slide I don't want it to affect pitch but I want to be able to do vibratos. If it's easier to understand think a theremin, that was slightly quantized so your initial pitch lands on the correct note but you have a bit of a wiggle room to move without affecting pitch, however if you move faster to do a vibrato or a slide it works as expected (the theremini does this).2016-04-09 20:33 GMT+02:00 Iain McCurdy <i_mccurdy@hotmail.com>:
|
| Date | 2016-04-09 20:40 |
| From | Rory Walsh |
| Subject | Re: Pitch bend quantization/correction |
Use your note-on message to set the pitch, then you react to a pitch bend message only if it has changed from the initial pad press. In pseudo code it might look like this (I'm assuming that each pad press triggers an instrument to start) iNote notnum kPtichBend cpsmidib if changed:k(kPitchBend)==1 then //code here will only be reached if you change the pitch bend value after the initial press endif This might be be a little work depending on how sensitive the pads are to movement. But you get the idea? On 9 April 2016 at 20:31, Gerard Rodríguez <binbiniqegabenik@gmail.com> wrote:
|
| Date | 2016-04-09 20:54 |
| From | Gerard Rodríguez |
| Subject | Re: Pitch bend quantization/correction |
That doesn't prevent a slow slide from changing the pitch, so you can press the pad and slide slightly and it's going to sound out of tune. Also it always changes, it's continuous, and it's quite sensitive so it has very small variations. What I was thinking is filtering the bend message with a low pass filter so those very fast small variations are ignored, and with a high pass filter so only the wanted movements like a vibrato with a fast enough frequency are used.2016-04-09 21:40 GMT+02:00 Rory Walsh <rorywalsh@ear.ie>:
|
| Date | 2016-04-09 21:06 |
| From | Iain McCurdy |
| Subject | Re: Pitch bend quantization/correction |
|
Maybe you want small area around the centre of a note wherein pitch bend remains zero, but upon straying beyond this safe area, pitch bending begins. This is something that hardware pitch bend controls do to account for mechanical wear and drift that may mean that the control doesn't return precisely to its equlibrium position. You could just map pchbend, normalised 0 to 1, through something like: f 1 0 1024 -7 -1 450 0 124 0 450 1 The output then would be a line from -1 to 1, but with a small flat area in the middle around zero. Date: Sat, 9 Apr 2016 20:40:14 +0100 From: rorywalsh@EAR.IE Subject: Re: [Csnd] Pitch bend quantization/correction To: CSOUND@LISTSERV.HEANET.IE Use your note-on message to set the pitch, then you react to a pitch bend message only if it has changed from the initial pad press. In pseudo code it might look like this (I'm assuming that each pad press triggers an instrument to start) iNote notnum kPtichBend cpsmidib if changed:k(kPitchBend)==1 then //code here will only be reached if you change the pitch bend value after the initial press endif This might be be a little work depending on how sensitive the pads are to movement. But you get the idea? On 9 April 2016 at 20:31, Gerard Rodríguez <binbiniqegabenik@gmail.com> wrote:
|
| Date | 2016-04-09 21:06 |
| From | Rory Walsh |
| Subject | Re: Pitch bend quantization/correction |
Ok, lowpass filtering with something like tonek should smooth the signal. It's still sending values between 0 and 127 yeah? Why not do something like this: kPitchBend cpsmidib This will only print something if the signal has changed by 10% of the full range. Is this any better? On 9 April 2016 at 20:54, Gerard Rodríguez <binbiniqegabenik@gmail.com> wrote:
|
| Date | 2016-04-09 21:54 |
| From | Gerard Rodríguez |
| Subject | Re: Pitch bend quantization/correction |
I tried my idea of high filtering the signal and it kinda works, the only disadvantage is that I can't slide outside the pad, but I fix it by just filtering inside the pad, that means that if I slide outside and land on a note then I can still get that note out of tune but It's OK. I guess I shouldn't have been afraid to try, Csound scares me a bit :P
if (kPchBnd < 0.5) then kPchBnd atonek kPchBnd, 4 endif you get a float between 0 and the pich bend range with pchbend (well you can get whatever you want but you want that to be able to sum it to the midi note number) I tried your idea but I don't quite get it Rory. Ideally I still would like to get the note after a long slide to be filtered in the same way but it doesn't bother me that much. Thanks for your ideas. 2016-04-09 22:06 GMT+02:00 Rory Walsh <rorywalsh@ear.ie>:
|
| Date | 2016-04-09 22:59 |
| From | Rory Walsh |
| Subject | Re: Pitch bend quantization/correction |
My idea assumes that the pitch bend varies between 0 and 127 but you can change that. It basically quantises the signal so a call to changed doesn't return 1 until you've moved your figure 1/10 of the width of the pad. I'm not sure how the high pass filter helps here but if it's working for you then plough ahead. On 9 Apr 2016 9:54 pm, "Gerard Rodríguez" <binbiniqegabenik@gmail.com> wrote:
Csound mailing list
Csound@listserv.heanet.ie
https://listserv.heanet.ie/cgi-bin/wa?A0=CSOUND
Send bugs reports to
https://github.com/csound/csound/issues
Discussions of bugs and features can be posted here
|
| Date | 2016-04-09 23:21 |
| From | Gerard Rodríguez |
| Subject | Re: Pitch bend quantization/correction |
It works because when the signal doesn't change faster than the high pass filter frequency then it stays at 0, so no pitch bend unless you do a vibrato faster than the high pass frequency. At least that's how I understand it. What you propose still would allow for a slow slide on the pad to change pitch right?2016-04-09 23:59 GMT+02:00 Rory Walsh <rorywalsh@ear.ie>:
|
| Date | 2016-04-09 23:28 |
| From | Gerard Rodríguez |
| Subject | Re: Pitch bend quantization/correction |
How would I check if a signal hasn't changed in x seconds? I think I know how to fix the issue with the high filter method. If I quantize the signal to get the pad my finger is on during a slide, I can check if it stays longer than x seconds there and then filter the signal like I do in the initial touch. 2016-04-10 0:21 GMT+02:00 Gerard Rodríguez <binbiniqegabenik@gmail.com>:
|
| Date | 2016-04-10 00:07 |
| From | Gerard Rodríguez |
| Subject | Re: Pitch bend quantization/correction |
Ok this does the trick, not the pretiest code ever though :P kChangePadTime init 0 kLastPad init 0 ktime timeinsts kPortTime linseg 0,0.001,0.01 iRange = 12 ; pitch bend range in semitones iMin = 0 ; equilibrium position kPchBnd pchbend iMin, 2*iRange kPchBnd tonek kPchBnd, 10 kPchBnd portk kPchBnd,kPortTime kCurrentPad = round(kPchBnd) if (kCurrentPad > kLastPad) then kLastPad = kCurrentPad kChangePadTime = ktime endif if (ktime - kChangePadTime) > 0.4 || kCurrentPad == 0 then if (kPchBnd < 0.5 && kPchBnd > -0.5) then kPchBnd atonek kPchBnd, 3 endif endif 2016-04-10 0:28 GMT+02:00 Gerard Rodríguez <binbiniqegabenik@gmail.com>:
|
| Date | 2016-04-10 00:37 |
| From | Gerard Rodríguez |
| Subject | Re: Pitch bend quantization/correction |
Not quite, it has to be a bit uglier if (ktime - kChangePadTime) > 0.4 || kCurrentPad == 0 then if (kPchBnd < 0.5+kCurrentPad && kPchBnd > -0.5-kCurrentPad) then kPchBndT atonek kPchBnd, 3 if (kPchBndT < 0.5 && kPchBndT > -0.5) then kPchBnd = kPchBndT+kCurrentPad endif endif endif Well, with this I'd consider my problem solved, thanks for reading and helping 2016-04-10 1:07 GMT+02:00 Gerard Rodríguez <binbiniqegabenik@gmail.com>:
|
| Date | 2016-04-10 09:51 |
| From | Rory Walsh |
| Subject | Re: Pitch bend quantization/correction |
Nice one. I wouldn't worry if it's ugly code. It works. Happy days. On 10 April 2016 at 00:37, Gerard Rodríguez <binbiniqegabenik@gmail.com> wrote:
|
| Date | 2016-04-10 17:01 |
| From | Oeyvind Brandtsegg |
| Subject | Re: Pitch bend quantization/correction |
Aren't you the one with the obsessive coding syndrome? :-) 10. apr. 2016 10.52 skrev "Rory Walsh" <rorywalsh@ear.ie>:
Csound mailing list
Csound@listserv.heanet.ie
https://listserv.heanet.ie/cgi-bin/wa?A0=CSOUND
Send bugs reports to
https://github.com/csound/csound/issues
Discussions of bugs and features can be posted here
|