On Wed, Jan 23, 2013 at 8:14 PM, Charles Z Henry <czhenry@gmail.com> wrote:


On Jan 23, 2013 5:13 PM, "Paul Davis" <paul@linuxaudiosystems.com> wrote:
>
>
>
>
> On Wed, Jan 23, 2013 at 5:10 PM, Len Ovens <len@ovenwerks.net> wrote:
>>
>>
>> I have been reading what I can on audio interfaces... I am thinking there
>> are very few good ones. I am thinking that some parts of the system are
>> vastly over priced like ADAT and MADI interfaces, both of which seem to
>> cost about 10X (at least) what they should cost... like how come a Gb NIC
>> is well under $100 and MADI PCI(e) card (100Mb) is $800?
>
>
> Gb NIC sales: in the millions
> MADI PCI(e) card sales: low thousands if you are lucky, hundreds if not
>
>
> > For audio quality, it seems FW, USB and ethernet _could_ give better results if designed right.
>
> none of them live "on the bus" the way a PCI card can. all of them are subject to a separate clock, in addition to the audio sample clock and the PCI bus clock (which is so high a frequency as to be mostly irrelevant).

Anyway you got to have a breakout box with any decent pcie card.  The noise in the power supply has to be bad on the pcie bus.  It's better to use a good external power supply.  So all you really get out of the pcie is the bus transport to some other processor in a breakout box that runs the audio clocks adcs and dacs. 

My ideal audio interface would run on something like infiniband or 10Gb cable.  Insane bandwidth. 


"but all you really get out of the of network interface is the bus transport to some other processor in a breakout box that runs the audio clocks adcs and dacs"
 
if you know what you are doing and build the machine entirely from the ground up (e.g. RADAR from iZ), you can deal with the in-box issues very successfully. it involves clever stuff to do with "power rails" that i don't claim to understand. it also isn't an option with a regular computer.