Hi all, I remember some months ago I did the suggested test using oscs from: Max/MSP, Pd, PWGL, Csound and maybe (not sure anymore), SuperCollider.
Well, they produces the same results. Anyway it was interesting to experience it!
Here attached is the audacity project file with only max and pd; max cycle~ output is shorter so one can hear pd osc~ output starting to play only when the other one is finished. I reversed the phase using Audacity, to be honest, but I don't think this makes the test irregular :-) ciao
libero
Message: 6 Date: Fri, 07 Mar 2008 10:57:49 +0000 From: Damian Stewart damian@frey.co.nz Subject: Re: [PD] Pd sounds better than Max? To: reduzierer@yahoo.de Cc: PD-List pd-list@iem.at Message-ID: 47D11FAD.1020000@frey.co.nz Content-Type: text/plain; charset=ISO-8859-1; format=flowed
Roman Haefeli wrote:
hey funny... i also heard people saying something similar the other way around.
since the same digital algorithm produces the same results on two different machines or in two different softwares, i think there are only very esoteric reasons to believe, that one sounds 'fuller' (what does it mean technically?) or 'richer' (more harmonics?) than the other. for me this goes to a similar direction as the discussion, if oxygen free, golden plated 8mm-diammeter speaker cables sound better than others (i would rather suspect a difference there than between max and pd).
well, he also said that it was because the [osc~] had a larger table size in Pd than in Max, which would make sense.
my initial assumption was that it was to do bit-depth. i used to scoff at people who claimed 24 bit was better; but then i spent some time in a studio working with 24 bit audio, and, well, you notice. (but both Pd and Max are 32 bit float, right?)
i hear you about the speaker cables; there are differences even amongst digital stuff though. for example when Ableton Live clips, to my ears it clips a lot nicer than ProTools does. (actually ProTools in general sounds very dead - its precision means that you have to work your ass off to get colour into your sound.) and back when i was composing in a multitrack sequencer environment, i remember choosing to use Cubase SX because its audio engine just sounded nicer than any of the other apps of the time (Cakewalk and Logic being the main competitors).
hm.. thinking more about that, i wonder whether this guy thinks, that pd people do just different, probably subjectively better sounding stuff. or does he really think, that [phasor~] in pd sounds nicer than the [phasor~] in max? this would be actually quite easy to test, if there is any difference at all. create a wav with same frequency and phase of a [phasor~], once in pd, once in max, and then subtract the one from the other and if you do not get a completely silent file, then............... *i shut up* ;-)
nice idea, but i'd try it with an [osc~]. anyone want to volunteer?
-- damian stewart | +351 967 797 263 | damian@frey.co.nz frey | live art with machines | http://www.frey.co.nz
hmm, sine waves are one kind of sound, but musical pieces use more than sinewave generators. if there are really technical and software related dsp differences in max and pd then you would find them more in other objects. the wavetable readers could be different (type of interpolation...) the oversampling techniques could be different, noise generators could be different, internal float precision could make a big difference. on the hardware side: da converters can make a difference. (is it overall true that pd runs on cheaper hardware, using cheaper digital-analog converters creating a richer, more distorted sound?). max could use additional filter magic that we don't know of (no source code available...). marius.
Libero Mureddu wrote:
Hi all, I remember some months ago I did the suggested test using oscs from: Max/MSP, Pd, PWGL, Csound and maybe (not sure anymore), SuperCollider.
Well, they produces the same results. Anyway it was interesting to experience it!
Here attached is the audacity project file with only max and pd; max cycle~ output is shorter so one can hear pd osc~ output starting to play only when the other one is finished. I reversed the phase using Audacity, to be honest, but I don't think this makes the test irregular :-) ciao
libero
Message: 6 Date: Fri, 07 Mar 2008 10:57:49 +0000 From: Damian Stewart damian@frey.co.nz Subject: Re: [PD] Pd sounds better than Max? To: reduzierer@yahoo.de Cc: PD-List pd-list@iem.at Message-ID: 47D11FAD.1020000@frey.co.nz Content-Type: text/plain; charset=ISO-8859-1; format=flowed
Roman Haefeli wrote:
hey funny... i also heard people saying something similar the other way around.
since the same digital algorithm produces the same results on two different machines or in two different softwares, i think there are only very esoteric reasons to believe, that one sounds 'fuller' (what does it mean technically?) or 'richer' (more harmonics?) than the other. for me this goes to a similar direction as the discussion, if oxygen free, golden plated 8mm-diammeter speaker cables sound better than others (i would rather suspect a difference there than between max and pd).
well, he also said that it was because the [osc~] had a larger table size in Pd than in Max, which would make sense.
my initial assumption was that it was to do bit-depth. i used to scoff at people who claimed 24 bit was better; but then i spent some time in a studio working with 24 bit audio, and, well, you notice. (but both Pd and Max are 32 bit float, right?)
i hear you about the speaker cables; there are differences even amongst digital stuff though. for example when Ableton Live clips, to my ears it clips a lot nicer than ProTools does. (actually ProTools in general sounds very dead - its precision means that you have to work your ass off to get colour into your sound.) and back when i was composing in a multitrack sequencer environment, i remember choosing to use Cubase SX because its audio engine just sounded nicer than any of the other apps of the time (Cakewalk and Logic being the main competitors).
hm.. thinking more about that, i wonder whether this guy thinks, that pd people do just different, probably subjectively better sounding stuff. or does he really think, that [phasor~] in pd sounds nicer than the [phasor~] in max? this would be actually quite easy to test, if there is any difference at all. create a wav with same frequency and phase of a [phasor~], once in pd, once in max, and then subtract the one from the other and if you do not get a completely silent file, then............... *i shut up* ;-)
nice idea, but i'd try it with an [osc~]. anyone want to volunteer?
-- damian stewart | +351 967 797 263 | damian@frey.co.nz frey | live art with machines | http://www.frey.co.nz
PD-list@iem.at mailing list UNSUBSCRIBE and account-management -> http://lists.puredata.info/listinfo/pd-list
Quoting marius schebella marius.schebella@gmail.com:
on the hardware side: da converters can make a difference. (is it overall true that pd runs on cheaper hardware, using cheaper digital-analog converters creating a richer, more distorted sound?).
now that is an interesting observation.
probably Pd runs on cheaper hardware (like arm-based pdas), but then i
am often using it on the HDSP-MADI card with quite good RME dacs
(though probably not audiophile)...
anyhow, i am pretty sure that pd and max use the same table-sizes and
interpolation for osc~ (ask miller to make sure), so there shouldn't
be any difference (doesn't msp's splash screen say "pure-data for max"
or something like this?)
so in the end i too believe that the only difference between the 2 in
a direct comparision would be the audio-hardware, in which case the so
called "listening test" would be rendered naught.
fgmadsr IOhannes
This message was sent using IMP, the Internet Messaging Program.
But the problem becomes appearent when you actually do some signal processing with audio data. Max has a flaw about it's floating point precision. At first I was thinking the problem was just about visualisating data(like, when you monitor a signal value, it is rounded and looks wrong but actaul data passing from cables are correct) but later I figured that it was not the issue when I tried to use signal values as control data, it was all quirky. I haven't been using max for a couple of years so I can't find my patches and demonstrate various problems but here is what I remember, might be wrong though someone has to try it: [sig~ 800] | [%~ 300]
this should give 200 but it was giving something like 199.9998 (I really can't remember).
If I were to test this value with [==~ 200] I was getting zero. So it was the calculation itself that was wrong.
There were many problems like this that I experienced and they were really appearent when I tried to use these signals as control sources(by not leaving the audio domain) because the logical processes were failing and I was having to build structures for exceptional rounding errors.
And when you process audio and give those to speakers, you can hear that float precision really matters. Max was always sounding dull to me and I always blamed float precision.
To my ears and eyes, PD and Supercollider are VERY transparent in this sense.
BB
PS: I'm not very aware of the technical issues resulting this decreased float precision. So what I believe might be folklore, and I'd be happy to be corrected on this issue. Basically what I think that max/msp has serious flaws in its floating point handling and this results to BAD sound which has that dull msp'ish character.
Libero Mureddu wrote:
Hi all, I remember some months ago I did the suggested test using oscs from: Max/MSP, Pd, PWGL, Csound and maybe (not sure anymore), SuperCollider.
Well, they produces the same results. Anyway it was interesting to experience it!
Here attached is the audacity project file with only max and pd; max cycle~ output is shorter so one can hear pd osc~ output starting to play only when the other one is finished. I reversed the phase using Audacity, to be honest, but I don't think this makes the test irregular :-) ciao
libero
Message: 6 Date: Fri, 07 Mar 2008 10:57:49 +0000 From: Damian Stewart damian@frey.co.nz Subject: Re: [PD] Pd sounds better than Max? To: reduzierer@yahoo.de Cc: PD-List pd-list@iem.at Message-ID: 47D11FAD.1020000@frey.co.nz Content-Type: text/plain; charset=ISO-8859-1; format=flowed
Roman Haefeli wrote:
hey funny... i also heard people saying something similar the
other way
around.
since the same digital algorithm produces the same results on two different machines or in two different softwares, i think there
are only
very esoteric reasons to believe, that one sounds 'fuller' (what
does it
mean technically?) or 'richer' (more harmonics?) than the other.
for me
this goes to a similar direction as the discussion, if oxygen free, golden plated 8mm-diammeter speaker cables sound better than
others (i
would rather suspect a difference there than between max and pd).
well, he also said that it was because the [osc~] had a larger table size in Pd than in Max, which would make sense.
my initial assumption was that it was to do bit-depth. i used to scoff at people who claimed 24 bit was better; but then i spent some time in a studio working with 24 bit audio, and, well, you notice. (but both Pd and Max are 32 bit float, right?)
i hear you about the speaker cables; there are differences even amongst digital stuff though. for example when Ableton Live clips, to my ears it clips a lot nicer than ProTools does. (actually ProTools in general sounds very dead - its precision means that you have to work your ass off to get colour into your sound.) and back when i was composing in a multitrack sequencer environment, i remember choosing to use Cubase SX because its audio engine just sounded nicer than any of the other apps of the time (Cakewalk and Logic being the main competitors).
hm.. thinking more about that, i wonder whether this guy thinks,
that pd
people do just different, probably subjectively better sounding
stuff.
or does he really think, that [phasor~] in pd sounds nicer than the [phasor~] in max? this would be actually quite easy to test, if
there is
any difference at all. create a wav with same frequency and phase
of a
[phasor~], once in pd, once in max, and then subtract the one from
the
other and if you do not get a completely silent file, then............... *i shut up* ;-)
nice idea, but i'd try it with an [osc~]. anyone want to volunteer?
-- damian stewart | +351 967 797 263 | damian@frey.co.nz frey | live art with machines | http://www.frey.co.nz
PD-list@iem.at mailing list UNSUBSCRIBE and account-management -> http://lists.puredata.info/listinfo/pd-list
-
Batuhan Bozkurt -
Libero Mureddu -
marius schebella -
zmoelnig＠iem.at