Oh nice!
@a.harker is the one that suggested this thesis to me a while back, which I did my best to try to implement some of the ideas from.
This sort of postprocessing is exactly what I am working on right now for our forthcoming onsets object …
2 Likes
Now that @groma hinted at it, I can say we have amplitude and spectral onset objects coming indeed. I like the amplitude only version and @groma and I have a meeting on this very soon indeed… stay tuned!
1 Like
Awesome, can’t wait!
I imagine it’s already a design consideration, but one of the main places I find most algorithms lacking is when doing things like rolls and buzzes with drums.
don’t hold your breath on these - they are incredibly difficult. But by giving you access to a few different segmentation, with parameters like fft size and that jazz, you will be able to experiment and find a creative solution!
Totally.
There would be little I could do at a parametric level to account for algorithmic/design decisions, so just putting it out there as a use case you could test with.
1 Like
I also have an idea of feedback (in the cybernetic sense) to kind-of-auto-adjust those in Max, I’ll tell more about it here when I post it. More very soon!
1 Like
Just to throw my 2c in here. I don’t hear a roll as a series of onsets but as a sustained noisy sound. Perhaps an onset detector isn’t the right thing to be looking for, but some sort of zoomed out classifier 
and all of a sudden we’re back to the other thread…
it depends on the quality of the drummer 
(although @a.harker has agreed with me that the poor quality of Quest Love press roll in the intro of Booty (Badu, 2000) does not make him a lesser groove machine after in that amazing track) a press roll is definitely a single perceptual object, but there are other types of rolls that are iterative by nature, and would be ambiguous in Schaeffer’s taxonomy!
1 Like
I agree with all of that, but I guess as roll I just meant many fast hits in succession.
So after spending some quality time with @weefuzzy on this, it seems that the thesis is doing spectral frame processing and offers some other exciting ideas… and that triggered a lot more ideas too I’ll need to check bonk~ and aubio performance in pd… stay tuned!
1 Like
Ok, based on a cool hang and chat with @tremblap this week, I’ve incorporated some of his sample-accurate funny business into the differential-based onset detection patch posted above.
The only thing that’s different now is the velocity side of things. It has a latency of 2.5ms now (vs 15ms in the previous one) thanks to the sample accurate-ness of the onset detection and envelope follower.
For some reason the first onset it reports is always wrong (it reports a dB of 0.0 (0.909091 after scaling)) no matter what the input is. I don’t know if it’s some weird signal-rate order of operations stuff, but I don’t see why/where this would be happening, and why it only happens for the first attack.
But other than that, it works pretty well.
Give it a spin and let me know what you think (and if anyone has thoughts on why the first onset is wrong, I’m all ears too):
----------begin_max5_patcher----------
3589.3oc0b08aqaaE+4b+qfvOTjL3qm32RqMEsaCX8kAzG1a8VbgrEsiZkk7
jjStoEM+suCIkjsjksksXRyBPbbHoD4uy2mCozu+galLO6KphIn+F5mP2byu
+gatwzjtgap9+alrN7KKRBKLCaRp5or4+xjo1tJUeozz7usNbC5iBuYHwLD7
IdV8P1DVt3g3zUeNWsnzNSRxLuoHoOdFmQ88YDojxnR9TD1yzEA9D8yU2fzs
qiSSTkl4muqwrsk0shqZMNxrXfE3Gox5EfcbkOuQYm8ISP+rtm+3CeP+wzwA
7UgkpWPDD9D3kvnF.Sjy39BLwS3wYBeMf4AmFujdwK4P7R7OJdKhWkFlLYZy
2bI9UQqTubBri8XZDJjxYBhOkKBnTdfgYyNCuFOTrKoGE6yCSWoQt4utD2Eo
gaJdHq7kyKmSwZnyvx.KOWvtFddOx3RxQw8xjrvRmB3EO.jvAwpAwbgzCXxD
AVhsh47qgU2GjwmSL+sT3tVwtGKYNS312+OAg6vGNubMGPMVRHLLA6Y0nkTG
IYKjukrYXEBC6THlELS.+HwRNIf.5w9.GFLoICjTJQJnhJAcew0PAXGRAXjg
aOeZs99zWCE+HUR3yufvXO8umR+mX.OGz+kbN7eAbJlnIJANRrf7lp8uIWsQ
kFgh1lGVFmkdBnSk9y38a5KHvQl9HduMQzrtnLqHb8lhSoPPIdZ.y05.XhOI
f6G3as7AM4FKe3fAnB3d48UpzWPeWMSGgkB1obA3SOV7MXByQrd14E62eEpx
qPcEruYxx3D0ip7BsT7to3lIga1rWy2r2knoU+Rl4F4Osoo3TaSdMMkqdLt9
5oMsFlCzoRfHsM2rRm7kFhn91jEoxS2FaVJ1FAtV0RxveRCWqJ1DtvdwQEal
Abk5avj87DA9fzNfM5XAXC8ly71QvAYfUIYK9UUz9K6IYfhcbJnfWnRKsJ26
2cjZY31jxOuLKsrH92LqBr1oWO8urZU1amZbXfv2mGqkWqFxp73nrT8hnE2P
2b8zAlSMwLo+bGXLi.h7rmKFjS.xxQ5r..41h4g4Zl07D0dZbfjeVVR6tZtt
D0xxpt2Dml1gJVls43clGu5gSbsyyfNWep6somhOuM016mA4hxOWD9XapcYX
RRkhb6a+WBSiWCImUFaYADulNUog.PenXQdVRRK7Z64wd5IBjyWndJNp7AyD
suv.L73M0BQSZ3xQwqTEksaqLbUQ6VJJe1Rz2qosyqzi+boZ8lD.EsGPqr12
WoceSdsZuiouEfFndjS20eKw0u4eCBQFEvnuc+A0mWeowqOFazAEBaHfz8Ea
qu8cUUNlAwiaT7f3D6hfdzVOlGy85VSNL2yeHFb9k+LxpubKvku6q+T9mR+w
v7v0MABbKig8zR.HXseO1ND8uwoKxUqASJn6QfgIzes4R1Mjknas2cz2bOxC
8UeExlY0swo36P2CW3c+tYvkUCCZwd0k+g9OpjBEZu6x2h7tC08Jp9xGQ21r
jta+axmJ0eBTELL356jYt+5I0jEi2zJayWoXVKOrs8xFm1TxjyIYYx1f500W
5ghOdWp3SqUvwhp50mV.y7PIFLrugZHOO033JSd8SMX8w5246IIN8XFcLqWc
+8SGJx1lunlpVo1hZuxAinkwoMdh+ol3dziaPbgKcMfG3ZPbz0vNRiJJtTyt
luZQVhMBoeBbQDLcuOz1Cr2E6s3BCVcQ1ZsV7gopmj8DRk9nJAb5W2qlLrHa
aZ49QV2WZs1n3If3DCrb64WE4JOP2LkdYAt50Sfq7INDqKgH.tZrR.sEMVkj
WKrxbIVuU6SBEFEtoTEgVlmsF8ieO5+.1xmCoie2opNi0ErutBTcKutu0twd
AH2xwLYTD.rSI.EKBSTnxLDjxw5vDvudDJW6n7RY7BClgTTOjb3YhYwA7dN0
oPWk.qdjJtDBADoqtGJKGoE6WDW9LJxTVhXcqoHctT8RQnmH4c.2lL00IWzY
CIvDSFUL430GBbIMwTEJfVT.hA.raYJ.sDhVO6IfVEoVD97kRNvbiBCmi6T2
p.tiHETgKIE57Y0gL.5FgOlECVGRBgQFgBKKCW7qEWrW.rwKf.rLh6TBi.gi
TPntUAYwCqiKAJ.jj4JU9k6JPNSyeY.hIcqfK2yQPlHm3thRUhHmxjOjtk1I
Ffq.BySRHBtPiPmUAJ7wKGcbpaq+V4I2KUguM.3WOnJeyPZNLCJvZVgprw59
4gN3+ZVv9QvfwmA5zdgNsG2XGuBySq98pvu81c9MRymEbvFoYS7i5MZ8w5ZE
zXUQ2g0jI51Zmrgq2jDWtEB8BR9WagEO6tlfshSiTFX54PaYWzJnWKbGsPi8
WuZOKk1+vvgrUbzAl9ZTf6VRFGR2fAEud658nWOAwKEmhVrMOWWBlS4bfdtv
E09D8oA7f.Q0daF36pfAbZbQgyKxRzlQRyhKfHgRxxtX+gbrmEa83NT3pH.H
NMUvvsQwYn3SsmbUULQWIRu.hf3AVWzLSevGO8fzftNPI8bIndHdCLDsNOjl
W1iJz+7eborx5P6w.ujy8j9l.Ys6KuKXhANLllsoazV9VhVN4ZrjQY3Y1SZg
znpJB7BZNiM58kj6jcgjHNlwrSe1adU1axhj3H0KPhq.x73StNG.zdih3baR
OcvgJ5coDrWg8tuhPoKTu4iqSBSdj.tN2oTb3zJ76.Z0h7rhhWPD90QkHTC0
.qIV6r334ny0XvURfdcNsi+kWFiNW0IiwSv2YnBG3pyLEg7NPV5aeAocJdMD
IF2F4UeGqJF2UDI16.hT4C4phGLlwwixHduzJLULi6HK4AukTKyJ77FZzoB1
Mftykjn2fO13m7PV0WBjcBD8UIGw+kJUkGBA4WUR3vlhBGtqrvOFlrERZzTn
TUDZ9yHHDv0E2sqnowovDnqarYQO6SoeJ86rxhn6qtbKbVuUWc0urPoKpXZD
BlzTHlzxs4on4J8NNMWsLKWA+Qe6pxzRO3Ugwo2BywTD.r6f6ecwJuGoOHF0
WWE.VDpuecud8lVifENb05paq2NaUZTQGveqotI20px32t6ayUkOofUMjCsF
A5jng6mI8H3F1WVSUH6oGhW7.zdc0TgzJSRZuJ0qun+t99UEw98.ocC.xp+c
Yt5+tUkt3YHWzxrkKQaxhgDR0W1O7a8mKu7prXZKbaeJDbp8zC4M9bbvBGlM
+IqDVkrXsHikqTQS2eeNtHhD2Z0vX7.hfiwBHV6jjyU1LQuTHQOGQ6K0JYS0
zlN1Jqcritp83CElup.Uqha71LCUoC.doQMJmL36UzYHJPjUsyneMFRd6HBY
zCSvXucK2rqSdMefGcdaT7UwUtJtvwra+HXjAL5EEuboRWSJfrgbXo77raiA
Ex0m5i883LbPszszYaiwqu1+GGUT0THHH+tv+r0DevQLFveGDwXXYVz7WFSR
rGtgILeGsgIAh++lBUKG8JRgjuGnPyKFE8gP763AkJcE8wI4YXVDsOlZVxU2
iFVEQ6fiDFo0YB6XmGLfWhvGgwL3Yx2dSN4L0Y4rNNxDGYE3HAlyyjTHN7gJ
0X6638vAiicepMv1sP9vtZl+r7H6CVf2nftlWeVhLdHPmHBlw7kLhzy9iwYn
u7H88mCP7ZMnloFOtoVNfoV63puo160ep0O7Zmk8QEPXKcqaOOXFt6CZWOCz
4DTwPIn3+TlZB6TDzJYdpfNyGD54r.6O9TvGlPbj9btjAeP5CA6NOsW+TM.q
mAtXhzO7sm2iPqwz5D.eXNMia4.yD8b3tEsYXYX0QdRv8s1P4yvATPqqxFJN
PuM3zy0Gl6MiaKHQ0Oxo0J88zmCnIryPS3GiCgeM3PmUYVbIKm9YQPtNZxIC
q8w6C49ToVKk5c7FaOdaG1oCPGYHlpFq.H1bnrY5ylMX8usAMqfTO84.vcNi
J3fiqqG3KHLgGKf4wvLlAkdPrTdRfEDHo9DSSDIixv.aQR3RwoYx91GxVPWi
J7D5qfWuGjTaIu6ouQRFzuLMNqHrjbBadBN34Nfx.DynlG+AtG32FnB9bISn
EQ0jFgOkP7v.+CXotP2qyhp+UNU9dbkOjDOpYLialFh8I+waeRDXBIQpSmPJ
v9zJsTsdAgSNReiDbB4kK55nn3F1TieMhHeHgcIeyb.xGhUTxa1xwjIBdHoz
O5YZnZVis1AT4HiIcDV45WY22DKhOVdvQJsRU+vtFKM.O.1pCLiEHLPif4yf
7Qa8Xiw5umwBLugjH5IAFdH.iXW9XtPGNtutx41eHSOQetnpWmCb0ob85Wes
N1FFd80HAyn5e39RASv70N0F53z4ie3Q1fHNrrG8LRW60hHFbJ6ithmhg3Xv
LUiVLaPlhIt.TCIhSlKjm8F9D45bSF0BGOjR0zAcuSV4j2nPkk1C21aanx3g
jwQfvAhtMU767piiepFBnNLXO6FG04cgjdR57NPpy6+nCe2Gc726QcemGYde
GYe2uzYypZ1Vt0geQz7NZo9sJTaJP3hEpzxV5K9ZANPNgfE5uIAG3dr1xesd
KErKjgJI1cNYz6bnYjeNNUSYTmqplSq3V5i5P2WqMSVCq935rc1m61Bv6cF.
6Gxs.KiPA+k5EBUpqDj9afk.N10Kq0phhv5G879WXyWsLNIoY4s+aeq58sbx
p7vn3cGTjJzfa3cl2wfS68a38eMPYtLR8kAVBA5vzd+1gWV8UQB7vAFYDZkL
h9aPS3NWUX5ppWXTx8dqGsIOaSVdyaNqYPhD0ieaYVCPqyts4E7kqjpZqcLL
t39OFhCP5xrVXfXkmwGCCnRzfqVxxZjom2tYZT7g+3C+OZKcmv.
-----------end_max5_patcher-----------
Sexy!
A few ideas for improvements:
- we used 110Hz instead of 250 for the high pass - trying different values there will change the nervousness / sensitivity to the bass - I use a 4th order highpass now.
- I now do the dB conversion before the slices (as discovered in a compressor paper)
- I now make a minumum of dB for the silence instead of an absolute. It does the same thing and is cleverer
- the steepness of the slide~ for the fast envelop will give you access to how nervous it is
Ugly patch attached - it is more nervous, and yet another variation. It seems that none are perfect, and all work differently. I use the brush drum to test.
----------begin_max5_patcher----------
2176.3ocwaszjaiaD9rleEnTkZqYpRZBdwWYiS4bKG1C495TSQQBog6RQpPB
YOSbY+aOM.H4PJQpgTBZjWuTz3Ui9qefF.M+9cyluJ+EQ4bzeC86nYy99cyl
oKRUvrp+8r4aCeIJMrT2r4qSyy1uc9BSUkxWSE5xqKIeuLUHkutSXFUnBz7U
gYali9OUMYWnL54jrMOUHhjlVQCBdDu.43qehUOozGwMcYcdw1PcacqJAlDI
Y.kzyJRUgIw54R9p+XIkMugbEgaERQwShrvUloK9sQwLg0CCUU3Ot6N0iEiD
O1JJKC2HpIlT7hzfSEh+6dQVzqn+BYARC.S.yNEVwvbE9DnQIBgcDX0AZn8.
Mj48x9jyf8yDeCFxi49jT.viEkIaxPeNGdq.wQetPTtKOqTfdNYyy6fw.8YY
9t7z7MuhVsWBc4a4Exmm.REmDISxyBKdcLXFg3o9g49NfVO5SDuqNnklGFqT
mfoI1xJKTsgkOexrMy8py1QgkQgwheNAVVoVEldRFm.1GTGPjy0xZWuIal3v
rGqWFku6TbXOL.GzU8.Ffwb0ptLbyOilEZDdEf+mJvaoudfVxOXfZwgX6Ibg
1jrc+1ehVR3NWGILkpYnfoKg41S.a59jDvTrQ.6V4TBW+b7Fm98y.0qtEkuc
qHS1YhjjEKdwxVvaDY+D8438EgJewJAhkE00Pkimi1Hf5NYGYACHqaSTQQEu
Ww7yTqiI9pnnDXqVsd17vc6ZU7rVcQgX+Qtdf7VzTTRloHVSQEhulT2emlRC
K.VWB789BCp8hasRpZXxg0Qy1mzH9zxt6pGvZzxymnsvIXpIRAh9Gm1V7ylK
hSjJTd0lHXA3BSWwOFrn0CR6NrZSZdzeJhaogARychrjrcvB6fhlV72o5Xw5
v8oxmVmmIKS9eZdh.hs9pecXjXvNmAgvoAj+YQBnrT2jMEIw4YpIQGwip3Zx
ANJLJMNsYFcKxB20SmAsE.JGnxRfI2WtJrPI8pBljVWoLOOsaUM8KUrVVU8t
jrrCPQHHngqr.hU5D8cUNT41SM15ZJeZelo1m.yY4SkgesKZKCSSqru6N7uD
lk.AeKjIFQ.E2ToIf5mKiJxSS6vulZ9ZO0DCJ9QhukDCw4oHTakAn4I6pUhl
2HkiS1HJkcKSFtoraIG4pAJZ+pJC6mjhs6RAtnaC5r8m1Vws8D1o7C7HFAlj
pVt3s5Oddb5v1p1fywF.CstmmIf.hdKStUwxxZqdWOh8XxMjuxgBcnYr52Bb
POpUK.Ea5y+JoTlW7JxX9bOHze3W+RwWx92p8lgpW43dNmf0KCByuOQLMQ8+
IYQEB0RYnOo8I8Wa5xaMYM5dyni96eBgQ+xufhdVE0y8IYjGPeB53CeW2XYU
yfRL8V9C0OhTXeIsFk+AB+.5vdT8xRz8MSoGZOHeQpdBnAAZb8Hoo8uNuFVz
q4V469L055rNb20hSxPjKVa7DZcz2hUo85uGqWg6WuhbBMmOFDBn3Xfngw.N
QuaJl26CBCXbwGFDv8.BusrUZR1P9qzyWU88iNk46Khpk1UF3ntybv+KDQPy
h3+dybU0tQIal5bfLx4f6fygpBMkXmnYg0ICi9SzZ0+MgvXWCaeWpNrqCd4j
6NmUsEFyQd4FL4vZo8GVK6ZbpDTv..szydmMQM2a1.DgL8n5s3IYUuYoC4+B
QJR9LDf6ynos8NfcTbGiGXB+z3yD24TM6aI52moaD5CttrE1P+P3Q3pxZ739
vUJmoKcvOLsi1nVva1nhEgFmaIxrzGOIXvgS06pk6z5L5rBJbSUPVxOOXfUs
qd6AC2VsAmyDFLWBhEgA2aJL3dlvf49MrHL3eSgAuyz2.kYYXH3VBCSz0.28
QsiAC2ysEFvuopBrIhANZLvIvxXf2sDCnSFCTa3JfXYL3l5YjbNX.uh6sGFb
SWjbhNE0G9.2wbdtAVCB3eDPvUJwIp2kgGoUpSvGBT3MfxzynBxHfoqatVbk
fPOtIhiO.Hz4VCgmwEQ6xd7BuF527125dn0iyGykPa1t5OQDL7W7U5pIY9l6
ZqmbW588Jw5QWAO4H0r3U5VllDK.D6JbYtM2au4lK7o1Bw7tk.1xqT9qvvZT
hwNKTpGiw2Njha.JEJyiWcsxzm.yoy6eVHUOdq8o2dCP0MhoeXWLCVxqkIHA
aMaPe1sT6ZU40R2xC2+U9bt5VL9GANccBXh4243aulALw7t0ALomdVVohEnw
Oei1jS8y9imB2SnAjSmLSiNuuzTr6kJZvjCuHuJj4nKvi04VDG316Tohc802
cD5OVJQFCgr.cTAQ7tDRsF5ESI9HHzAyFchTOryfwRY5X.SUBOe4RMuQPIUh
1BQZdYTpVI6zJHjOH4la+hM7kAl3wnaFXAVjMFwF2BRMmwH0XbanIRFql3kR
owHjTYt6ESHe1XcUcohIe5H0GtXVBOFqX70vo3nH8glfVxxdLtTNPZ2PY5kY
oOJRaiEVUiA88LK7rDgHumbzwRD5cUXrhuKmQqZRuTRwGMo55TwDJ4A4EuhJ
GjO7GjK7GmG7CmC7Gl+65be2DS9Agu1rQAHTd2lzysNgx6BApcOpy.9mRxTy
81GO4aOZkI7yCihfnq6j079pl4yBnDW0ad3.eLuau5jm8jpwr5A9QmZvbQk3
p.Hwga2X9VX1mT6tos3sCC2ZKK8yxcl5bJygpS3elG0gqei5wcHUSeqMs57k
RZKYwpMqSRSaXn1e6F0aJa9lhv3j2t8oJ9mzH6bg+3sn2257APn6Fstad99.
xsn22Nta08hFfIAZcDVkNh5MnHxA8B1Dd0WWfWqTjeWQ9t7hlOyhGYAMseuL
ugQaxiz57hrewWWqiwIEa+gNMBsKsLiCpUXp9MfmYA1TyRFt6TSl1Ga.3852
Bk4neS8AUL+bnOzmeb2+GAw02lA
-----------end_max5_patcher-----------
1 Like
Ok, here’s those ideas incorporated, and tidied.
I have to say I find the slide~ 1 2205 waay too nervous. I get those almost ‘late’ onsets all the time (with brushes). 5samps seems like a good middle ground while still being fast.
I still get the same “first velocity value is wrong” problem, but the normal onset side of stuff is fast/updated.
edit:
It looks like I’m always one onset late, hence the first one being 0.0, so I’ll troubleshoot that and post the fixed version.
(I removed the code for now)
1 Like
Ok, I’ve played with this more and I can’t seem to fix the problem. It seems like an order of operation thing, but nothing I’ve tried seems to fix it.
It’s this specific part of the patch that’s a problem:
It’s super inconsistent too as sometimes I get the correct value, sometimes I get the previous value, sometimes nothing at all.
I’ve tried delay 0 0 (and up to delay 20 20) before the change~ and before the snapshot~. I’ve tried deferlow and delay 0 after the edge. And nothing seems to make this particular section of the patch work consistently.
Here is the isolated section of the patch showing what I mean.
----------begin_max5_patcher----------
1578.3oc6YszaiaCD9r8uBCc1IP7sTu01C6ksm5ghhEEAxxLJZirngdrIoK1
82dGRJIS4HaK6HGrsnAHBzTjbl4advOR804y7VodVV5s3mV7oEyl804ylY5R
2wrleOyaSzywYQklg4sptpRk6sz9p75Mo4YxJy6PMctMpJ9gz7j6JjwU10lP
X2xVtfPI2JB3LtuOKfERDPeX5s91mK9qcqpptZ+k01U0Kak10zaUTdhW2jRW
azO0pOeivqSSJh1HqjE2IyiVkYlnu9Uea9b8ikiznAEZkr36CZ03lNuWkWkC
RyLgetHMJy6v3AFItkC+EH7QHAJPCDgAZffwMvA9fvQq7.UHcS8FcW2v8gI0
o2scuqSstceTr09ODbVlljqU5Ed2mohp1graTqktNBGjl54Hgxz+1NNbmfg0
znfLPWtDbW9DHkVYTIe1.edwYowO98yNFDa.3.s1sfDdBXFcBXZn3Nh2EXiW
XBk0XH31jmINEBO0oPPLEnSmqYh4LsoQstKcTznxMtWUrIxrD7CY4537CZ8H
+o172TmUkVlktVVb1X.02ACvlefH9mDDJkUf..svtL1hDKbm2ffxvQ1ASMfT
Fq1JOdE0AvBjua7fAENNXzVHp.b0xd.A2YNtt9.ueD1f.1nzItmO539onNOd
fxA3wUneRKyKWmHO6p73l56CvzfhGILNbYxiUvPbQk9OfcWlGss7AU0YmczT
ofDR.xELLFi3hPtgbAmdo62sWPhqQimPiN9AcF5k5twfIG5KBBEgzPrkOEic
E1hWflRGczCWpOFSAGJNDSXAjv.pI9F4KtBVLWLgVrcCoy2nGftLBP.r.ERw
HR.uoyfSUojd5ph6Web4wxAnSYNvZYVzKeeAB4q++bAoFJgnP1sHlOSPDDep
M6O3Zv1EeQoBZisOkWvLrleaGqhheLoPUmut2oNRxUfwpo9618fEJpKkqRT4
wpLUwIJfngqWcRLGlV3Cysfru+C4sS95NUeQV7JcvXXRiowG9DZHm00wH.c0
3ecebBpbv+FvscT8.kOA6VyXK68zcI0hu2fIzPe7xAZfN.UJL4nrFcMSWjRK
Kyx29vc86w+fzgeqR1Gm7g8+Cf5i5VTAvBHn+BkIShheomQh9OP7LgY1omvu
rPX7+GB2KDl+u2HXy37xRy2+B9LQM596GVWppKhaw9lqQawNosVVVklGUkBQ
5NiAaGzfYNiUP7wHI3fYtCRUnOJsdGwqunw3gEM4MIZAdLVs+vh1+5KZxA.b
zaRzzQHY5TDTQFiiEcMPWxXMQzTCtixsNAX6dZ+vBpM05sYQnQXRFIgdOJBI
vWibBwXLRA48pRq3pjUHHW11I1swh1tE1fsrYzFg.7v9rcWufkysePD6OM6n
5UH+RZ63Mzb7hJ.xRU.So5B6kv8L298KL27UQdcZiqDLOPjkUujs+Vm6t+t5
0opeuJppt7teSlWaoI.V08Q0YU8ghUI2mlk0sG8rt2.LGZX63Ye6xlt6SjHj
hPAFtJHh.yMsfFFxNs3YybPsShxZI3P4XA0xgPSlfz6JfalFdmr7QgVI3GF3
SssftP8nd.XYdhkCDVzw.DhEJTaUEs9TXpgciutRkTDsNUZIGahkrtXKbaoX
Au0.6N7msWcd6bbi75bFOZuX8+P6cOli3D7PmDcANb.WehginFcL1tKMWGTK
aQe60y27vkoWTbLnM8hJLADAjPLx3tDMtK2Y8ZFk6d.bkmPC1bhkQh7PbHDF
hCMgVPTL0zRGihlZ+vFYYYThbZ8EiIatKRe3jy.8ENHVNXKzgSN2cZfW250S
6GxT59YGiyKZOW3nitL9LmxefMC57DFYYu7teIqVdC5XZ0.IsDjfSMtadmBF
p6BcETvB0S4msFtSuXc4ljqjF9quDc9JnONjQsk7DDB1l6fQHDY5UvOTHkWf
FZPOh9OcKDOfiESux8mxrL0Sms1EfYHgsxAJ.Ypuw.Dkc36n3SluyGKz2OXp
L.yWK8Cpr0ms5qCE8MteFr0GyXIH2pYShBVXITn0vQW2gS4bRXS3nPXCLAfM
jbjckeuLnpnsGyPb+zz.S7OFUoV7wzjGz2614KeKicyUjpWzxsMWJn4lTm+s
4+CfB4zaz
-----------end_max5_patcher-----------
that’s the bit you changed from my original patch. My version was oversampling but consistent. This version might have the problem of Max lag as I was getting similar issues…
Yeah, I compared it to the version in your patch, but I think the delay~ is in the signal path in your version and avoids the “order of operations” problem as a result.
I couldn’t get that code to freak out at all, but you’re saying you ran into some issues with it?
I get that there would be Max slop since it’s going back to the Max domain, but if anything, the bang from edge~ would come late, and there would be +2/3ms delay on the reported onset, instead of sometimes not reporting at all.
Is there a better way to go from MSP to Max on command (that doesn’t use a snapshot~ 1)?
I ran into issue with similar code to yours (banging a snapshot~) so I kept it in audio.
technically banging like you did (and I did before) is the best. in practice, it is quite fragile I found, so I use snapshot~ 1 → change and it is the best I’ve had. It is in all my descriptor-based processes.
It’s just so expensive…
So is this change~ -> `edge~’ thing an issue of Max slop or is it a bug-y thing?
not really - every millisecond, one number is sent to one comparator then it stops there…
@a.harker explained to me one day why I was getting constant inconsistency but I cannot remember if it had to do with max slop, timings of message processing, bad coding from them, or wrong expectations from me. @weefuzzy might have a clever answer too.