Using fluid.ampslice~ to extract amplitude information

[rodrigo.constanzo]

Although I still haven’t made sense of fluid.ampslice~, it occurred to me that using the gate output nature of it might make it useful for detecting what a peak was within that time.

So using a vanilla 1->0 crossing would be no good as that could potentially happen 100+ ms after the onset, since the 1->0 crossing happens only when the threshold criteria becomes false, and has nothing to do with when a peak might have happened.

I then thought I could use the code from the previous onset detection algorithm to trigger the amplitude measurement aspect of the previous abstraction:

----------begin_max5_patcher----------
4382.3oc68z0bipbkO64WQW5gT1UoQa+c2rISpI6to17PRU4g7VlsbgEHaxU
BzBHOiu255e66o6FPfDfPRf8jMiqw1xcCz84zmu+f4W9vMydH4agYyP+6n+N
5la9kObyM1gLCbSweeyrM9ea4Z+L6kMKN7qIO7OlM2MUd32xsCuEkrZUVXd4
Dw61jrKecXt8tHEi5FJ+ksgtkb1Lz+SwTa8yW9TT7i2mFtL2MKQ4s.OGI0rE
JsTHwXgV3wTh4Ht1LCkt.W8.hBraDXy8QuZ6hn3xMAt9JElV.iE.4MyVEsN7
4vzrnj3Za4al4ucasguo1sXvL+iD6CROuZnnX2P3pgRCeNp79kUi5mBvaN.r
6RsXiYeSxms+wjDDlFuKxtUbC9qenbKYOMh82Dls0eo6lMGZkSuGCJrHPBla
9Eu.oww6wZvo8iqSV9SgA02xyR1FFGEuMMLKLN2OuXuWMcP3J+cqyueURbdV
zOa2AD3rns4WUrCacRCLX29+gzH+0U.vioQAIwlMQiSByvkKGPcHrvlnNvXu
hX+ssby.cAfV5XxL.H2k8fep4f5g01UfVNYdRx5lSUceqCWkWL81n33Cvh4I
a6dxznGepm68gDXxM88rsyjc+tX2r2CzD42m4+bSrct+50ErrMe7eyONZied
Xdj6HfhqlLL1G.zmxVllrdcC30MyysLS.PiuL7qQA4OYWn5DCvkGssjHZV0o
bPzigY4MGK2+wrlijk+hCoWancOTvCeed3lsqAnn4EzPnVcF15B2ZL9AB4Vl
rYCP4WwNVWR2eLNC3XyP4O4miLzSYOkj+JJ2+mrCFhLGDnLeXiEBBEsC8LP5
kjhxSP9OmDEf.jNHpCEkksKLaQ8kYcTb3xjcw10hUahlBTw0loEImThkUmwr
rHZqT.tpFmxABLoyZtR6EZRJG+W+vGJ+v7KDq1P0wApORCAdxflJQ5VQxITl
zoBErCg.5MTDtjv0RMQxwR5bjVbnFkCQRJu2arzGQjKD0DYHnFF1QoVnEJFV
wYbrlhYD.s3sPzOxQi6D4PeaPN9ANo3nrnGeNqa7DsWRn4CCWwD64uzpSQ4n
IuIHGGzL6REZTxcHUKDJhGvfnnZAQX3NXXK7h6FFYcybTahRAq0EWaePwAge
qKNo852MRG6PvtEfLy2NhJKYW5xxS4xyDDoAz.ZpxihqL24uumxFAf8fNnN6
cQwC+j6BivmIaWT9vO4tf04lnXvRaTmYLDI3dmQW26mmmF8vtb2YWcitOKiC
.KUevecgp+JSGaw1gOreyY+oCGcEd27wW62ulNsk8XILfzoXX5976g5QsJrY
7ELOFUiwvHdRswwGBg2mmOBbqt9PqsOqav93ggxVGED9JR.6Lr3c.Ywz7EBi
NckGwyCr+wfrz39vUbUq3J1aEthyIX6OdKQWLm1pVQWDLsW7k9cBe4mmD7vq
mYjEFNoCEvEJvhGJu.Ovk8hFjshFHiG7BWSzlcadE8QEdxfZFHcQxDdZIHgg
PMbK8FSENqcIKiGXuE8TzVy0LUfLwYiuf0KbR5UB5+BE7HOtSUizg0DVtBlP
7inG8inG8inGcgQOBDTjGlFFf72EDkf.oYcF8G5k5HmmMdOTO7BlB6o.G5JL
dTRK8ZsKu3daBwQeHHGdIJd7QKzB2aMgxfSNElP8tiIbZCQK2kmrZ03iNJxt
RA9PcJ7g3cGevyeBkjBZ1Pfqj.SzWSRgQdBjDVypgQDAwzVhDJgYvPDp9Tnn
28nnlid.s5Bi70C9wOZh90p0I9CHVgNajAGrDZMSKELgjyElXE5Hq5KbXL92
KgadUZ3+6tv3kuLUQblhs3IBP1vwdXNkIzLOMmoMtacZL06NIkSeEnEFrxF8
YG6GG8Yv1NvRnrvJ1OzmA6URVm73K04NuPzZPzRiBe+zWFJgHQKWPodRkXen
RXR1ovuD06M9coe1R+fvWuPL0AN+zEVxYBOkADbBLix3685ifwmDKIdSBes8
oNULhEXfpvXehfY2MDi6HZ1sY2RuQ096RzTA6zfwS5yFO0oUMUHquiRIRAUi
GYgtokzmF0vwmcBQ522f29TjXNcGVZA3GcgKAcAotKAuvSKobIl6wwbBmO23
fGVQnXkVCnUllZGhZR4HQq0JpPIAQSKlpDdX3uGDjQDm.xLeIIZAC18LKjYF
g4g0bLLnCV4DLiwvBJUoDpIEv3CLqVFKK9mqiL9POxT8CY3409wDteoiz986
tShV1v8v7P9mIHanhE33+EMQn+7F+snOJwKPt+QVLlE7oygakFb3F7PRyAwk
LNa.olh0dp7DiGf+ned3qHJhzN7ROQ1IlOr7THrPoB7VTJAsEz8FbH75C9o5
oN4LgAOF95YB6kgTv96dKzWWn4ZsPe68bWwl5bwUUfemIcdyvnzCwdKm0Dhp
WflN0G1KeBNwBmfLtVdPC.sGHk2S4wKS9rPzKLSl5C5Ik.ujwtEIa8Sfq0SN
At+SSvAsi3Vn4KL0tRYPuLvqp2ruJUSMsM7TgKqCPlOXw3UQJc9P30EZfIG9
R4AF0XJz.poEF3KnJhGmRXZopnhU5uzClbN+fv09u7JHABa9dBD.vJRAOeAQ
fEJlB7QxI0S2uFNxjW9AEwDNXWpybuQzvFSU2HZWtmW+fMdxK1jr7DSwpm85
EaYyoo+YtBHtU4A8pqi3M0v+igwuh9b4gNhnpJjhwu.TZy.GBs2B5i2N7++m
qFkfrsKfSkVpHEUgMSVVFOWkPJ9Q6L8iBR4GEjxIxpdPXsZ7p+jFzfP928W.
xKKqYvue10kYAkKQcDWOcVjaNl3fHn2FO1g1HNn.q2AyrKHTANn6OEAJ+ReA
43WtENku629kzuD+W8S82TYHvslZz0FsPXo9DwcIluihWlFZhdO5S1fR8uUc
K6ujUnaKZKke2mPXzu42fbNWcaTL4Nzmfa7tewdw4EWFLh6ty+UyuBWmEhp8
T98H7cnCuihO7QzsUao6p+P9Rt4mvYFAt3xmjcs+sydaRyYT7E25RClxhtOm
U8lMytSY0aCt.fv9PF3AAs7hN76j8dz4UqESRdjjCMppSXKtPF3dP1SfcqTa
DDkaNTd3wlgyddserOH0WT3WaVGT0acfjuhBieNbMnzub11J2o1IpZifh6wM
1lyLVUcXGAHrlsWulmZPZIFMiy6BdWY5k1wEdA+uuT3kO4v6sFcSH+.+s4gl
xSJYC5u9GP+MPl9Cfi42M6bAVkxJUTaJk6CCodQeKVy74FZeos3M1zi.xV5u
NzznzfqGa7WCJvCPoFEliyguS3olQNFc3pFqtN6EroGzCWC6STXT9SlZbD.Z
TRJxP5uLJ+ETfM7Dl1FGbS03SUqXD14hQXfhDqe3lfNqkDJVhEEgjgPIG143
M3G7lbbhMdT.tHCHC.vtg3.zJvp8juB3pfvk9uLNnChvxUHTjEhChTgmnOjA
SN4HCimsFiGpdMBrZsObkAH+7b+k+T1HoOfXEBJU7EDplIjdLVQrJ7j8whvd
CXQV9zlnb.C.ta9XX5HAvfBASkvInb32BBU5IX1ZiSf6CdopoN1T4H5YFKp5
8tdqx+3Nk8TuEdTNVQoRgz.p8GEJBc5AUxHCpROWMself5jenlBPRHxT7r4U
B2aGzYcFq44EeO.3mB143AtaiqB33I5f31axyK.9aT6cmg8YtbkoEjixU1wU
cWawEn8Pk1nVMsxKQ2Vpi07tZIJeGX40s3EFwqjE2UYqUUc2gGQAYm0N3LDu
0cyB2Fl1kxEsQL2gVCQD8I5izKpdD5D3tvaEsFaM70WAykhhQK2klZhDSeZF
t.iAbu3r.iATGloNOcuVCL8lF4+PVxZinj3jnLvXn0IIijBQSygakbxVvobB
lpzLy6AFivSYul.PmdOBOnNqOG+9jkQLA6QkTrl6IwdlNwnnif6PthBO4.UQ
QQC78fmdIOGh9u9OGoiRf214rG1T5+JWM+6xAemGhSdN21Eu0H8aUUyJ0QNG
OiW8AmQx3KwIsl.dkxNWm4iUN0R95x9nmRCyd5UjIw7uGueMZKi8DlbQO3Jv
i5o9EFQ8dgXTRTe4KGA5wBI523CFt224j8zoHSiAF+2gwgo9f1ghvI3WEPA+
8gT3Y+06.KNrNYCtQ9vKH5Bwlr616vcTLr.lXNX2zK9R7Wh+riTD8oha2ANa
1Y7L+aKCMtiFGXdq2ECByx2kFidHzD0xGBWkjFB+x73JTSat3G8ihuEVi4H.
vtCd9kt49IyqIuvx6q..V5addGd+lDeffMNb2lHiXRIRXbP1A.+sVituqQTU
tc+mdHL+qgvtFL.y.AFKvfmmUuJ7.aScaAj80mhV9DLdoe3fMIqW2bWZ1eA+
GlmWgn9OAn1s.PV7mUMhXQiwf1l.92Xus+zO2tgfpwyPvBW9aipW3JXGaBvZ
kI+8RfnyMpBZwRRF2oRANsdLxZhjjmujxJWtlOP2uDtVLmBxMwfuW7pB9g1u
+WsWvexIGeVjBZ+zGM8xoCsZU1r.UvC.FBgpXN4vmKvylhEyw1Y4u5WkdsLl
PZoULtvClAbPzzBHvcmdqm72MGbdFDx.B8BhVsJz3PC.5no1OPNwE.LNUuPe
3KbLU+w.6ch6GjF+d7JYyvMqUZojBF23BIQukulm2kherOslIg0gtNLwmEHs
iR3oWi2oeckw0Zo67nClAuRCXgL0r99VAZHMj0Eucz5AreZjR5oc6.qD+D6F
QW6FxTranmhnf08Y00zxVEOuMQAViaJnqkfPGCeDnnQSqhQKCD6b3ni.nyNE
nSdSAcospfoFQrMMQPw6cBhxEfFv1rx+ZDPNmhmgJOGxz1AYg.6dcyXewgnw
BO2WJk480.1oF53IGAnibBni30sHgw+nWncEDMPA.fszbGUUDLiYCKaKyckn
AEanRp6RznT.1E3w.qmnblsTWDXhxCvBZghKMmhFTiTynTLg6YJy8wPr0Aap
NZoY02i6bxPv4rQPMrZHrv5qmEVgsQ8QQjKTJIQyJ3RM7ETIoi4tRfSpNeRW
6Kqltib43tzj1WZ7UszB7.VZaX1tZpGwP.x2LyTDCQfM8Ma6vGh5Q6gE4ZWI
0PWoQmXaPKMuIR+.QCLrk8mSwG6GoTufbTKvc7EN1bub4PfpwP9KeHZWOfRp
c7mw3Wrx7JEgP4XJy91Bi4I6XtQmPXP.hZJj0xoCfwuTf70sRCQur2XHYkOT
8xWqrCl5Jc78JrQpc5YOqw9ZSKebP8ETXnvwScs3.x.HfFAif7jt+GQgnLkI
X8xqovCmil4ZArgXJvAD0i4apnKdiSGhtyxX1b8qDYHJvHmP.rBTVINnVXY5
ELyWBsRxkbswb2gdcvTKTfehLpRxJyzroUnOVi3wW4XKicPnIJY.5oZC7c80
U2yHLtFPnZpG3BbQg0RbEk3wSM1Z1nxAaO80JL1tTjAsTzqcoFhpLqRyqFnF
jPnoINJW0Fm3MDZd72i6b5aka8tZc6s0s9JeWGBE0QYG4f1Y2rHGzF6GzB6G
295c255G1151VV209tGjQlprLsw+axp1rsrwvahALIvxd7beTrYuG1cVGJPI
MZxLRwryKOX2Khze4xv37FDsZyU.GVThz7IE1SaTyWinbdwokISiG1Yxy1.6
9nRgt0Oca.v0JAm1A4FaIyK.Yp0zRFvhvseBXZDjwdasILKyurqgFuyhUQqW
WAP0ekKTlNwYOl5GDU6UXtC9IUmIVM9ya8Sj589u81nk2FHlAvbya8SGeak2
E0CS7rm8rhydymfgHGbW9wOV7VBPUqU22llrMIs50kvBvmfxqeWdREfV04rk
c5X6GeM4NF1oX8RHe.TW1yLNPVgshi4.Ly7FSJqb+s8sYpmyYP10e1OOA8mM
uXDlcIquSHWG+20xG90O7+YSnTgJ
-----------end_max5_patcher-----------

Unfortunately that’s not possible with fluid.ampslice~ since the output of the envelope followers is not available, and that’s what’s being used here to determine when a peak happens.

Now, I can just run the first bit of this code in parallel with fluid.ampslice~, and use a 0->1 transition to trigger the minmax~ in the same way, but it’s a bit of duplicate code, and quite possibly not the best way to go about it.

This would also be quite fast as the code in question adds a latency of 110samples (2.5ms) to tell what the amplitude has been in that window

I also thought about using fluid.loudness~, or even better fluid.bufloudness~ but with the former I don’t think it’s possible to accurately say that the onset detected by fluid.ampslice~ lines up with the current analysis frame of fluid.loudness~. With fluid.bufloudness~ it would be possible to analyze the precise frame in question, but that can then come with an overhead of up to 500ms(!).

SO

Is there a way to tell the amplitude of an onset within the fluid.verse~ in a fast and sample accurate manner? (Or is there the possibility for this coming in the future).

[tremblap]

@rodrigo.constanzo this post is on my important things to do before unplugging for 3 weeks. That day was supposed to be today, but since the Pd help files are sooooo almost done (and I found many bugs doing them so they are urgent) I will most probably unplug progressively in a few days… but not before I’ve checked this - it should be helpful to see what a user get in my Machiavelli interface design…

[tremblap]

@rodrigo.constanzo

Is there a way to tell the amplitude of an onset within the fluid.verse~ in a fast and sample accurate manner?

Now, this is an interesting question. I would use fluid.loudness for that, and get the RMS for the same price. You are in real time here, but the JIT approach is good here. Obviously that induces the latency of the window you want to analyse to ‘catch’ the peak.

as for this:

@rodrigo.constanzo

I still haven’t made sense of fluid.ampslice~ , it occurred to me that using the gate output nature of it might make it useful for detecting what a peak was within that time.

I really don’t know what you would do to do that, since it is a complicated gate, not a discrete output of amplitude

but

as it is very quick you could use its change of state 0->1 to trigger a sample and hold. if you tweak the latency right, you could catch the peak, especially if you are peak-riding with a fast attack and slow decay on the relative gate.

as for this:

@rodrigo.constanzo

(Or is there the possibility for this coming in the future).

in the long term, there is another idea in the pipeline for a way to output the curves. that has been talked about in various threads, but will happen only after the 2nd toolbox has a viable alpha, so not for a good 6 months… @groma has some dreams and great ideas there.

[rodrigo.constanzo]

@tremblap

Now, this is an interesting question. I would use fluid.loudness for that, and get the RMS for the same price. You are in real time here, but the JIT approach is good here. Obviously that induces the latency of the window you want to analyse to ‘catch’ the peak.

Totally, but that brings me back to the “analyzing the correct frame” problem. There’s no way to get the output of fluid.loudness~ to be perfectly in sync with a triggered onset. I can take a whole window and then take a peak value from that, but again, a bit sloppy and susceptible to Max slop(?).

@tremblap

as it is very quick you could use its change of state 0->1 to trigger a sample and hold. if you tweak the latency right, you could catch the peak, especially if you are peak-riding with a fast attack and slow decay on the relative gate.

Again, possible, but if the decay is set too long, or if my criteria exceeds the threshold for a longer period of time, all of that adds to additional latency. Or worse, additional (pseudo)random latency, since it’s not something you have direct musical control over.

@tremblap

in the long term, there is another idea in the pipeline for a way to output the curves. that has been talked about in various threads, but will happen only after the 2nd toolbox has a viable alpha, so not for a good 6 months… @groma has some dreams and great ideas there.

That would be cool, and it would be less…erm… of a black box object than it presently stands…

For me, however, that would just be a workaround to the larger (in my case) problem of wanting something in real-time and being able to be sample accurate about it. Where presently is in a no-man’s-land in the fluid.-verse.

[tremblap]

@rodrigo.constanzo

There’s no way to get the output of fluid.loudness~ to be perfectly in sync with a triggered onset.

I was thinking of the JIT approach,

@rodrigo.constanzo

Again, possible, but if the decay is set too long, or if my criteria exceeds the threshold for a longer period of time, all of that adds to additional latency.

now, you could also use the output to check the max of the last X samples. A JIT or a hybrid with framelib is most probably what would serve you best.

[rodrigo.constanzo]

@tremblap

I was thinking of the JIT approach,

Ah right. The JIT has a +/- of 50ms jitter with spikes up to 500ms, so I wouldn’t trust it for fast/accurate loudness measurement.

@tremblap

now, you could also use the output to check the max of the last X samples. A JIT or a hybrid with framelib is most probably what would serve you best.

Indeed, though that’s basically what the original patch in this thread is doing. So I’d be back to duplicating code with an envelope follower, or involving other frameworks (framelib).

I guess part of this post in general was just to see if “the loudness of an onset” (or change in magnitude for the spectral one) was on the radar for the fluid. stuff, as those seem like fairly common use cases that presently require work arounds to get at.