Hi, y'all. i'm starting to get the hang of pd now (about a year in to it), my main interest in using pd would be to create effects for processing my instruments (primarily electric bass and acoustic drums) and accessing them via an arduino footswitch and/or an arduinome i plan on making. My problem is that i can't achieve a low enough latency to have a useful patch in a live performance setting. so i have a few questions, if i can get as many of you to answer them, maybe we can have a standard setup for this type of processing (i apologize if this is common knowledge to many pders):
- in terms of processor capacity, hardware, and sound card configuration,
what would be the minimum requirements to achieve unnoticeable latency (not hear the affected signal as a slap-back type of effect)? i've read that 11ms can be achieved and is unnoticeable.
- what is your setup for live processing? what do you think is the ideal
setup for live processing? (i'd like you read Miller's thoughts on this)
- what do you think will be the next big step in making sound art, be it
musical, online group performances, new interfaces, whatever? (kind of a personal question, i know, but bear with me)
My thoughts on the last question: i'm intrigued by photonicshttp://en.wikipedia.org/wiki/Photonics, from what i understand, the amount of information handled is gargantuan when compared to electronics, and that would mean that electronic musicians (or would we then be photonic musicians?, sounds catchy) would no longer have to consider the limitations our equipment when composing/performing and our imagination would truly be the only limitation. sorry if i seem romantic about it, but it seems like such an interesting possibility. herehttp://emusician.com/futuretech/photonics_technology_electronic_musicians/'s an article by EM
here http://emusician.com/futuretech/tech-page-let-light/'s another from EM
On Sat, Jan 30, 2010 at 10:28 PM, Jeffrey Concepcion < jeffreyconcepcion@gmail.com> wrote:
Hi, y'all. i'm starting to get the hang of pd now (about a year in to it), my main interest in using pd would be to create effects for processing my instruments (primarily electric bass and acoustic drums) and accessing them via an arduino footswitch and/or an arduinome i plan on making. My problem is that i can't achieve a low enough latency to have a useful patch in a live performance setting. so i have a few questions, if i can get as many of you to answer them, maybe we can have a standard setup for this type of processing (i apologize if this is common knowledge to many pders):
- in terms of processor capacity, hardware, and sound card
configuration, what would be the minimum requirements to achieve unnoticeable latency (not hear the affected signal as a slap-back type of effect)? i've read that 11ms can be achieved and is unnoticeable.
- what is your setup for live processing? what do you think is the
ideal setup for live processing? (i'd like you read Miller's thoughts on this)
- what do you think will be the next big step in making sound art, be
it musical, online group performances, new interfaces, whatever? (kind of a personal question, i know, but bear with me)
My thoughts on the last question: i'm intrigued by photonicshttp://en.wikipedia.org/wiki/Photonics, from what i understand, the amount of information handled is gargantuan when compared to electronics, and that would mean that electronic musicians (or would we then be photonic musicians?, sounds catchy) would no longer have to consider the limitations our equipment when composing/performing and our imagination would truly be the only limitation. sorry if i seem romantic about it, but it seems like such an interesting possibility. herehttp://emusician.com/futuretech/photonics_technology_electronic_musicians/'s an article by EM
-- www.avmachinists.org Puerto Rico based Art Collective/ Non-Profit Org
"Unnoticeable" latency usually refers to the musician not noticing the difference in time between when they press the key and when the sound comes out. Any time you add a delayed signal to the original signal, you will notice it. The "slap-back" happens at longer latencies, but at shorter latencies you will hear *very* noticeable comb-filtering. And since no computer-based solution is latency-free, I think you need to re-examine what you are expecting Pd to do. Either that, or go with a dedicated DSP board (and learn the accompanying programming!) which would give you a more guitar-pedal-like "zero-latency" system.
Maybe Marco Donnarumma could give a few words here on processing instruments live. His set uses an electric bass through Pd. My guess is that even the un-processed signal goes through Pd to avoid echos or comb filtering due to latency.
Best, Derek
Jeffrey Concepcion wrote:
* in terms of processor capacity, hardware, and sound card configuration, what would be the minimum requirements to achieve unnoticeable latency (not hear the affected signal as a slap-back type of effect)? i've read that 11ms can be achieved and is unnoticeable.
Hi Jeffrey! I ve been trying to minimize latency in Pd for a year now, experimenting with various OS and hardware. I m using Pd for the same purpose, that is live processing of electric instruments (mainly a guitar). I would recommend using a Linux distro, because they have "realtime" kernels, and the JACK server, plus you can get the hid object in Pd (which does not exist on windows). If you want to play live you want to go for latencies below 7 or 6 ms. I get a 5 ms latency on an old Dell laptop (1 Ghz) with the latest Fedora with the CCRMA realtime kernel. I can also give you a couple of hints about the interface (I personnaly have hacked a cheap gamepad and it works great). You can reasonably expect to get a low-latency live set at a very low cost, provided that you have a quite recent laptop to work with.
Pierre
2010/1/31 Derek Holzer derek@umatic.nl
"Unnoticeable" latency usually refers to the musician not noticing the difference in time between when they press the key and when the sound comes out. Any time you add a delayed signal to the original signal, you will notice it. The "slap-back" happens at longer latencies, but at shorter latencies you will hear *very* noticeable comb-filtering. And since no computer-based solution is latency-free, I think you need to re-examine what you are expecting Pd to do. Either that, or go with a dedicated DSP board (and learn the accompanying programming!) which would give you a more guitar-pedal-like "zero-latency" system.
Maybe Marco Donnarumma could give a few words here on processing instruments live. His set uses an electric bass through Pd. My guess is that even the un-processed signal goes through Pd to avoid echos or comb filtering due to latency.
Best, Derek
Jeffrey Concepcion wrote:
* in terms of processor capacity, hardware, and sound cardconfiguration, what would be the minimum requirements to achieve unnoticeable latency (not hear the affected signal as a slap-back type of effect)? i've read that 11ms can be achieved and is unnoticeable.-- ::: derek holzer ::: http://macumbista.net ::: ---Oblique Strategy # 18: "Balance the consistency principle with the inconsistency principle"
Pd-list@iem.at mailing list UNSUBSCRIBE and account-management -> http://lists.puredata.info/listinfo/pd-list
Hello, latency delay is noticeable at ~25ms, below there are artefacts grain caused by phase decay if the source and the processed signal are played together at the same place with almost same amplitude.
11ms is only the buffer size, and other elements in sound processing need to be taken in consideration like the distance expressed by the sound speed in the air (about 340m/s at sea level 15°C), so you can add about 3ms per meter, also almost all effects needs a processing window so the more processing power you have, the lower the achievable latency.
Any other digital processor added to this chain would add latency, like digital guitar pedals and amplifiers.
The dsp needs realtime access, so the Operating System have to be configured for giving high priority to PureData, and a low latency audio driver needs to be used directly from pd or through jack, anyhow there are much possibilities with jack server.
Selon Pierre Massat pimassat@gmail.com:
Hi Jeffrey! I ve been trying to minimize latency in Pd for a year now, experimenting with various OS and hardware. I m using Pd for the same purpose, that is live processing of electric instruments (mainly a guitar).
2010/1/31 Derek Holzer derek@umatic.nl
"Unnoticeable" latency usually refers to the musician not noticing the difference in time between when they press the key and when the sound comes
even the un-processed signal goes through Pd to avoid echos or comb filtering due to latency.
Jeffrey Concepcion wrote:
type of effect)? i've read that 11ms can be achieved and is unnoticeable.
Thanks for the suggestions, i'll deffinately be looking into your suggestions (i'm not familiar with some of the terms ). For now i must deal with what i have at home, which is an acer netbook w/ windows Xp @ 1.6 GHz, 1GB RAM btw! No external soundcard as of yet. i should get (in theory) a similar result to Pierre's so i'll be trying that approach for now. i also have access to new macpros at my university and a motu 8 channel interface so i'm hoping for much better results there.
As far as upgrading goes, i'm considering building a pc dedicated to my audio needs, probably running a linux distro, any recommendations would be much appreciated. Perhaps i should include the dedicated DSP board that Derek mentioned. Can those be used to process pd patches?
On Sun, Jan 31, 2010 at 9:20 AM, Pierre Massat pimassat@gmail.com wrote:
Hi Jeffrey! I ve been trying to minimize latency in Pd for a year now, experimenting with various OS and hardware. I m using Pd for the same purpose, that is live processing of electric instruments (mainly a guitar). I would recommend using a Linux distro, because they have "realtime" kernels, and the JACK server, plus you can get the hid object in Pd (which does not exist on windows). If you want to play live you want to go for latencies below 7 or 6 ms. I get a 5 ms latency on an old Dell laptop (1 Ghz) with the latest Fedora with the CCRMA realtime kernel. I can also give you a couple of hints about the interface (I personnaly have hacked a cheap gamepad and it works great). You can reasonably expect to get a low-latency live set at a very low cost, provided that you have a quite recent laptop to work with.
Pierre
2010/1/31 Derek Holzer derek@umatic.nl
"Unnoticeable" latency usually refers to the musician not noticing the difference in time between when they press the key and when the sound comes out. Any time you add a delayed signal to the original signal, you will notice it. The "slap-back" happens at longer latencies, but at shorter latencies you will hear *very* noticeable comb-filtering. And since no computer-based solution is latency-free, I think you need to re-examine what you are expecting Pd to do. Either that, or go with a dedicated DSP board (and learn the accompanying programming!) which would give you a more guitar-pedal-like "zero-latency" system.
Maybe Marco Donnarumma could give a few words here on processing instruments live. His set uses an electric bass through Pd. My guess is that even the un-processed signal goes through Pd to avoid echos or comb filtering due to latency.
Best, Derek
Jeffrey Concepcion wrote:
* in terms of processor capacity, hardware, and sound cardconfiguration, what would be the minimum requirements to achieve unnoticeable latency (not hear the affected signal as a slap-back type of effect)? i've read that 11ms can be achieved and is unnoticeable.-- ::: derek holzer ::: http://macumbista.net ::: ---Oblique Strategy # 18: "Balance the consistency principle with the inconsistency principle"
Pd-list@iem.at mailing list UNSUBSCRIBE and account-management -> http://lists.puredata.info/listinfo/pd-list
On Wed, Feb 3, 2010 at 9:22 PM, Jeffrey Concepcion jeffreyconcepcion@gmail.com wrote:
Thanks for the suggestions, i'll deffinately be looking into your suggestions (i'm not familiar with some of the terms ). For now i must deal with what i have at home, which is an acer netbook w/ windows Xp @ 1.6 GHz, 1GB RAM btw! No external soundcard as of yet. i should get (in theory) a similar result to Pierre's so i'll be trying that approach for now
Don't be fooled by the processor Hz count. First of all, different operating systems deal with system resources in different ways. Whether you can get reliable, low latency performance from XP on a netbook is beyond my experience but AFAIK, Windows and MacOS are not as transparent and easily tuned as *NIX are. I still use very lean and small window manager when I want to be sure to have clean audio without dropouts in low latency.
Beyond that, the hardware performance has a lot to do with latency. The cheap audio interfaces that are bundled with today's computers, especially netbooks, are not necessarily useful for for real-time, low latency work. They will usually require a relatively large buffers to convert analog<->digital without hick-ups and that will add up to the buffers needed by the audio driver, pd, CPU's workload etc.
That said, you may want to look at this thread (and/or google similar): http://linuxaudio.org/mailarchive/lau/2009/5/6/154392
There are some people using linux on netbooks for realtime processing and therefore it is possible. YYMV, depending on your needs, your perception of "real time" and the hardware/software configuration. Note, that if you want to give linux a try you can grab pure:dyne or ubuntu and install it on a USB stick or HD and take it from there until you decide which OS works best for your particular needs. pure:dyne will give you immediate access to media applications (including pd) and realtime kernel.
I am actually working these days with an Acer netbook (aspire one) running ubuntu 9.10. Using a rt kernel improves audio performance somewhat. I just used a stock rt build, I could probably roll my own, tuned to my liking but I think I am getting too old for that :). However, the hardware cannot really keep up with very low latencies. My focus is not on getting lowest possible latency right now so I do not have any better advice at the moment.
half a cent from yours truly.
./MiS
hi, I totally agree with Michal about the OS. I haven't tried MacOS but i know for sure that XP can't beat a good rt Linux kernel. That being said, I personnally spent about 6 months trying various Linux distros before i found a good match with my hardware. Although I have had very little luck with Ubuntu, you should probably start with this OS, and then try other options (Pure:Dyne, Fedora with Planet CCRMA, 64 Studio...). Most of these come with a live version which you can try without installing anything. Here's what I'd advise you to do everytime you install a new distro:
- Install it or run the live cd (of course)
- Run QJAckCtl or Jack Control to try the audio server first (google it to
find the proper settings for realtime use). If you don't get a decent latency without any Xruns (shouldn't be higher than 6 ms on your hardware), then you can either try another distro OR install the rt-kernel for your OS (in case the standard install did not include it). 3) Once you are sure that JACK is working fine with a low latency and no Xruns, you can try your Pd patches.
Hope this helps.
Pierre
2010/2/7 Michal Seta mis@artengine.ca
On Wed, Feb 3, 2010 at 9:22 PM, Jeffrey Concepcion jeffreyconcepcion@gmail.com wrote:
Thanks for the suggestions, i'll deffinately be looking into your suggestions (i'm not familiar with some of the terms ). For now i must
deal
with what i have at home, which is an acer netbook w/ windows Xp @ 1.6
GHz,
1GB RAM btw! No external soundcard as of yet. i should get (in theory) a similar result to Pierre's so i'll be trying that approach for now
Don't be fooled by the processor Hz count. First of all, different operating systems deal with system resources in different ways. Whether you can get reliable, low latency performance from XP on a netbook is beyond my experience but AFAIK, Windows and MacOS are not as transparent and easily tuned as *NIX are. I still use very lean and small window manager when I want to be sure to have clean audio without dropouts in low latency.
Beyond that, the hardware performance has a lot to do with latency. The cheap audio interfaces that are bundled with today's computers, especially netbooks, are not necessarily useful for for real-time, low latency work. They will usually require a relatively large buffers to convert analog<->digital without hick-ups and that will add up to the buffers needed by the audio driver, pd, CPU's workload etc.
That said, you may want to look at this thread (and/or google similar): http://linuxaudio.org/mailarchive/lau/2009/5/6/154392
There are some people using linux on netbooks for realtime processing and therefore it is possible. YYMV, depending on your needs, your perception of "real time" and the hardware/software configuration. Note, that if you want to give linux a try you can grab pure:dyne or ubuntu and install it on a USB stick or HD and take it from there until you decide which OS works best for your particular needs. pure:dyne will give you immediate access to media applications (including pd) and realtime kernel.
I am actually working these days with an Acer netbook (aspire one) running ubuntu 9.10. Using a rt kernel improves audio performance somewhat. I just used a stock rt build, I could probably roll my own, tuned to my liking but I think I am getting too old for that :). However, the hardware cannot really keep up with very low latencies. My focus is not on getting lowest possible latency right now so I do not have any better advice at the moment.
half a cent from yours truly.
./MiS
2010/1/31 Derek Holzer derek@umatic.nl
"Unnoticeable" latency usually refers to the musician not noticing the difference in time between when they press the key and when the sound comes out. Any time you add a delayed signal to the original signal, you will notice it. The "slap-back" happens at longer latencies, but at shorter latencies you will hear *very* noticeable comb-filtering. And since no computer-based solution is latency-free, I think you need to re-examine what you are expecting Pd to do.
I think it is a reasonable expectation that you don't have to play a note one 16th early to have it sound 'on the beat', which is what one has to do when dealing with 23 ms latency in a 163 bpm piece. Attached is a little patch that shows some rhythmical implications of any given soundcard latency in ms. gr, Tim
Either that, or go with a dedicated DSP board (and learn the accompanying programming!) which would give you a more guitar-pedal-like "zero-latency" system.
Maybe Marco Donnarumma could give a few words here on processing instruments live. His set uses an electric bass through Pd. My guess is that even the un-processed signal goes through Pd to avoid echos or comb filtering due to latency.
Best, Derek
Jeffrey Concepcion wrote:
* in terms of processor capacity, hardware, and sound cardconfiguration, what would be the minimum requirements to achieve unnoticeable latency (not hear the affected signal as a slap-back type of effect)? i've read that 11ms can be achieved and is unnoticeable.-- ::: derek holzer ::: http://macumbista.net ::: ---Oblique Strategy # 18: "Balance the consistency principle with the inconsistency principle"
Pd-list@iem.at mailing list UNSUBSCRIBE and account-management -> http://lists.puredata.info/listinfo/pd-list
-
colet.patrice＠free.fr -
Derek Holzer -
Jeffrey Concepcion -
Michal Seta -
Pierre Massat -
tim vets