thanks for the reply. i am still working out the feasibility of the whole concept at the moment, but if it looks promising i'm sure i will be asking more questions around here :P thanks porl 2008/7/23 Emanuel Rumpf <xbran(a)web.de>de>:

Correction: the Indigo IO is 2 in / 2 out; the Indigo DJ is 0 in, 4 out Unless I am mistaken. I'm guessing that the F/OSS options are basically: Custom board design with PCI envy24 chipset => ALSA : JACK / ~2ms IEEE1394 devices made by ECHO (AudioFire2/4) => FFADO : JACK / ~4ms Cardbus devices made by ECHO (Indigo IO/DJ) => ALSA : JACK / ~4ms USB devices (48kHz native rate) => ALSA : JACK / ~8ms The cardbus is going to be easiest for a one-off project. ALSA drivers for the Indigo products are relatively mature. Some smaller ITX computer boards feature cardbus interfaces built-in. I seem to remember seeing a 64studio-related hardware device that integrated an AudioFire2 though... On Wed, Jul 23, 2008 at 8:10 AM, Phil Frost <indigo(a)bitglue.com> wrote:

On Wed, 2008-07-30 at 21:52 +0100, Dan Mills wrote: excuse me, but are you crazy? does latency actually matter? :) i think you just want something (chipset) that can function as a busmaster. then that can move data to/from host RAM. the host picks it up/delivers on the interrupt. forget IDE, PATA etc. etc.

On Wed, 2008-07-30 at 20:21 -0400, Lee Revell wrote: I was not planning on using the VFS layer, but thinking about it, mmap /dev/sdc does have amusing possibilities.... playing a .raw file with a call to sendto..... It was the physical layer interface that interested me, fairly quick, well documented, parallel, and unlike the pci bus, the state machine is fairly simple. It actually feels a lot simpler then some of the approaches using conventional soundcard chipsets.

On Thu, Jul 31, 2008 at 2:06 AM, porl sheean <porl42(a)gmail.com> wrote: One thing that comes to mind is the small gumstix modules. It may be a more expensive solution than you have in mind, but the advantage is that you get a *tiny* little computer pre-loaded with Linux. Add the audiostix attachment and you get stereo audio in/out. Or the roboaudio, which has an on-board Atmega, will give you some 10-bit analog inputs for sensors. But yeah, it's about $100 to $150 for the motherboard and $40 for the audiostix expander, so this might be out of your price range. Also add a few more $$ for the ethernet module. Anyways, worth looking into. I've played with them a bit for a sensor interface, and it was lots of fun, and way easier than building it myself. Never got around to trying the audio side of things. Steve

i am aware that proprietary solutions exist, but they are a) really expensive and b) proprietary. i was thinking if these things worked with netjack etc, it would essentially be possible to use them directly as audio inputs to a standard linux pc over ethernet as well as being networkable audio distributors. as for the gumstix etc, i looked at them a while ago and thought they looked pretty cool, but i would be surprised if the audio interface was up to the quality required. i noticed a couple of mentions that echo audio do custom jobs. does anyone have any idea how much would that be costing though? obviously i suppose i would use their normal interfaces for prototypes and developing the interface, but i wonder if the custom builds would be more appropriate? thanks porl 2008/8/1 Eliot Blennerhassett <linux(a)audioscience.com>om>:

On Fri, Aug 01, 2008 at 10:27:29AM +1200, Eliot Blennerhassett wrote: Since you seem to be into this sort of thing: can cobranet be implemented in software, using standard network interfaces (assuming you have a realtime kernel, and the interfaces are dedicated to this task) ? I would be nice to have an open source standard for this sort of thing. Found it quite funny to read on Audinate's site that the problem with 'the others' is that they are closed and proprietary - Audinate isn't any better AFAICS. Ciao, -- FA Laboratorio di Acustica ed Elettroacustica Parma, Italia O tu, che porte, correndo si ? E guerra e morte !

if there was an open standard for this sort of thing that we could implement in software and interface directly with jack etc then it would make my idea redundant.... i would love to see such a thing though :) what does netjack (or netjackmp) lack compared to these implementations? i'm still not entirely sure of netjack's synchronising abilities for instance. porl 2008/8/1 Paul Davis <paul(a)linuxaudiosystems.com>om>:

any ideas as to how long before we can hear any info on this thing you are working on? will it be compatible with jack? or is it possible to expand netjackmp to allow communication between arbitrary nodes without sending the audio signals to the master node and back out again? i thought it was possible to do this with the netjackmp implementation but as i said i don't know much on the actual technical abilities of it. porl 2008/8/2 Fons Adriaensen <fons(a)kokkinizita.net>et>:

On Sat, Aug 02, 2008 at 09:50:51AM +0200, Stéphane Letz wrote: There's an added complication to the system I'm developing these ideas for. In this case all (identical) computers *do* have their own sound card. The cards are word-clock synced, but of course the jack periods are not. A second requirement is that two or more computers must be able to output (on their soundcard) a signal received from the network with exact sample sync. Ciao, -- FA Laboratorio di Acustica ed Elettroacustica Parma, Italia O tu, che porte, correndo si ? E guerra e morte !

Fons Adriaensen wrote: Livewire does have a software implementation (as a closed source driver). AFAIK There is no cobranet software implementation, But I don't see any technical reason why it couldn't be done. The packet format is not published, but anyone with the publically available information http://www.cobranet.info/en/support/cobranet/developer/tech_data_sheet.html and a packet sniffer and some cobranet nodes could figure it out. Cobranet nodes all have software/hardware to allow them to sync to the system master clock (the 750Hz 'beat' packet). Having a PC play/record audio (e.g. from HDD) into a cobranet network might not be difficult but if you wanted to also use the PC soundcard, this is much more difficult unless you have a way to sync its sampleclock. This may be of interest http://www.ieee802.org/1/pages/avbridges.html The open standards they use are (at least) zeroconf, ieee1588 and RTP. Its how they are put together... As with all the others, a big part of the "secret sauce" is how the node clocks are synchronised. The other part is how the audio routing is set up, and the audio format negotiated between nodes. http://www.proavmagazine.com/industry-news-print.asp?sectionID=1618&art… Cobranet uses the beat packet and VCXO audinate/Dante uses IEEE1588 Livewire and Ethersound I'm not sure of the exact mechanism. To achieve low latency, some form of hardware assistance is probably needed to regenerate the isochronous clock (in order to avoid samplerate conversion/buffer adaptation schemes) Finally, another overview paper http://nas.cim.mcgill.ca/wp.pdf -- Eliot

On Fri, Aug 01, 2008 at 05:12:00PM -0400, Phil Frost wrote: Running audio over cat5 is possible, provided you use shielded cable and perfect balancing at both ends. It's even available commercially, see http://www.etslan.com/pdf%20docs/12.%20PA200%20Series%20InstaSnake.pdf Ciao, -- FA Laboratorio di Acustica ed Elettroacustica Parma, Italia O tu, che porte, correndo si ? E guerra e morte !