Hello,

I'm trying to make a polyphonic version of this patch: http://www.lubbertdas.org/pd/guitar1.html (original monophonic version).

I followed malestrom's poly tutorial (http://puredata.hurleur.com/sujet-5883-polyphonic-voice-management-using-poly) and polyphony seems to be working fine.
However, for some reason MIDI notes 64 and 76 (both E) are much more muted than all the other notes. Some notes also ring out for longer than others but this isn't as important.

Any insight on this anybody please?

Thanks!

http://www.pdpatchrepo.info/hurleur/MMM5_Poly.pd

It's probably because of the body formant filters. The harmonics for those notes are in the band-stop regions. Maybe reducing the Q of those filters will bring them up a bit.

Also, it would probably be better to use [vd~] instead of [delread~] here. [delread~] quantizes to integer sample lengths, so notes with wavelengths that aren't exactly integer lengths will be slightly out of tune. There is some lowpass filtering as a result of the interpolation, which will make those notes seem to decay faster, but they'll at least be in tune.

Hi, thanks,

I attached a slider to the Q of all 4 filters and tested at different values but it didn't seem to have much effect. It started to make a feedback noise then when I got to almost zero.

I checked the original monophonic version of the patch and it doesn't have this problem, seems odd now that it would change.

I made this patch a couple of years ago, so I barely remember, but there's definitely going to be a little bit of variability in sustain because of the filters. You can try playing around with the center frequencies too; just make sure that no notes end up feeding back. I also read somewhere that you can combine the body response into the impulse somehow, obviating the need for body formant filters in the feedback loop entirely, but I haven't looked into that (and that would probably just make it cheaper without changing the sustain characteristics).

I kind of prefer the second one I made, although it's more computationally expensive. That one has polyphony, albeit modeled after a guitar (six strings tuned to different open-string frequencies). It could probably be improved in lots of ways by someone with a more advanced understanding of waveguides.

I'll try looking at the original one again, but could you post your version? Because there might be several different ways of making it polyphonic.

EDIT: Ignore that last part -- I forgot that the forum doesn't show attachments to posts when you're on the page to post a reply.

OK, I found your problem.

The issue is that the "pick location" slider value isn't being sent, since it's being sent to $0-pick-location, and if the stringunit patches are separate, the $0 identifiers will be different (see here and here). I need to explain a bit more for this to make sense.

What the pick-position-delay-calculator and pick-position subpatches do is create a comb filter for the impulse based on where you pluck the string, filtering out harmonics that would have a node at that point. For example, if you pluck an open string on a guitar at its midpoint, it should make a relatively hollow/fat/warm sound, because the even harmonics (the second harmonic and all its multiples) are filtered out. So since in the pick-position-delay-calculator subpatch I have the frequency of the open string set to MIDI note 40 (the loadbanged message) and the default pick position set somewhat arbitrarily to 1/4 up the string from the bridge (the [/ 4] object), and the slider isn't changing the pick position anymore in your patch, the comb filter will filter out every fourth harmonic of MIDI note 40, or every harmonic of MIDI note 64, 76, etc.

You can remedy that by changing the name $0-pick-location in the properties of the slider and in the stringunit patch to pick-location. But there's also a somewhat subtler issue that I should also explain.

Generally you need to be plucking a little way up the string for the note to sound pleasant -- it will sound somewhat harsh/tinny if you pluck too close to the bridge. But, assuming a constant pick position, you can't actually fret indefinitely high notes, as you can't fret higher than you pluck (well, you can, but the model doesn't make sense that way -- really I should have limited the range of note values, with a lower limit at the open string and an upper limit at the pick location or lower). Obviously if you fret and pluck the note at the same point, you'll get nothing, as is the case for MIDI note 64 plucked 1/4 of the string length from the bridge on a string tuned to MIDI note 40.

So that means that if you want a wider range of notes than from the open string to the point at which the string is plucked, you'll have to change the pick position dynamically based on the note you're playing, or have the different open strings tuned differently, or have different picking locations on the different strings, etc. Does that make sense?

Also, I'm guessing Maelstorm is right about using [vd~], although I can't say I noticed if it was out of tune, probably because it sounds like a banjo anyway to begin with.