John Roberts {JR} wrote on Thu, 12 June 2008 08:40 |
The missing piece is effective appropriate bandwidth communication between the parts. Control input and indicator feedback is reasonably low bandwidth. Engine to I/O A/D/A will require gobs of bandwidth.
Logically the I/O and engine should be collocated to simplify this communication. The processing engine will be modest in size compared to other parts so it likely will be rolled into the mic preamps and converter box.
|
As someone who has for over a decade co-authored international standards for data storage and transmission as a means of feeding my family, this is an area where I have some specialized knowledge. I am not convinced about any need for unrealizable amounts of bandwitdh between the AD/DA IO and the processing.
Let's break this down. Currently, the vanguard of digital audio is running at 24 bit word depth, and 192Ksamples/sec. Accordingly, each channel requires 4.608 Mb/s of bandwidth.
A fairly sizable (though certainly not huge) system might have 64 inputs and 32 outputs - or a total of 96. That is a total *aggregate* bandwidth of less than 442 Mb/s.
There are numerous current technologies that can provide this bandwidth today, with acceptable latency.
Even lowly Gigabit Ethernet does 1.25 Gb/s in each direction simultaneously (minus a couple % for framing). Strip off the clumsy TCP/IP layers, and it is pretty low latency besides. This could comfortably yield 200 channels upstream and 200 channels downstream, simultaneously. This should accommodate any but the largest live sound scenarious.
Not only that, 10 GbE is going mainstream in IT as we speak. While I have not done any studies of audio on 10 GbE, it should be capable of supporting perhaps over 2000 channels of 24/192K in each direction simultaneously.
Not that I necessarily believe Ethernet is the logical best coice for such transmission. However, as an unbiquitous technology, it provides a useful baseline.
I agree that it is logical to keep the processing and the IO co-located. However, I think it has more to do with the cost of sheet metal than the cost of data transmission links.