Re: [Jack-Devel] Ethernet-based audio interface using (net)jack idea

In my original post, I proposed slave hardware that would synchronize its
local clock to sync packets from a regular Jack server (digital PLL),
advertise a fixed amount of (floating point) and MIDI channels to the
server, get packet/period/... data from the server and start sending
receiving data over UDP continuously. The software needs to emulate a
fixed-configuration NetJack2 client, which can eventually be done without an
IP stack, since the initial communication is done using broadcasts.
Therefore: small software footprint ==> no EEPROM or DRAM on the board
==> low pin-count FPGA ==> low cost. Since we're trying to digitize
precision analog signals, reduction of power (no high-current
switching voltage regulators) and omission of memory (no noisy digital
memory lines) is also a good thing for circuit design.

However, in the meantime we have evolved to implementing the Jack software
on an embedded CPU. Is this really feasible ? Embedded hardware is usually
seriously less powerful than a general-purpose desktop CPU. Jack is also a
fairly large piece of software, relying heavily on all sorts of libraries.

Perhaps we should think in a different direction: why not try and run Jack
on a compact, cheap but still powerful embedded PC ? An E350 board can be
had these days for under $200. Some of the industrial embedded boards even
have a 12V dc power input. Maybe the problem may then be redefined as how to
get 8 analog audio I/O in and out of such a board (through PCIe). For
slaves, we can keep disks etc. out of the box if we boot them through the
network (there IS a Linux server anyway, isn't there, and more than one
slave may use the same boot image).

1312622260.3614_0.ltw:2,a <CAPRPZsAxZGKVNb28vGLejrn6AaeqMKxUxL_OuvDJ7cFoF38jsw at mail dot gmail dot com>