How hard is IEEE1394 (firewire) ?

[Colin Howarth]

Hi,

I'm designing an external computer audio interface (192 kHz, 24 bit,
stereo using the Cirrus CS5381 ADC) and, I'd like to connect it using
firewire.

I'd thought of using an Oxford Semi OXFW... chip, but they got taken
over by some outfit called PLX last year and all I can find is NAS/DAS
storage chips (SATA) and PCI(e) bridges...

What I was looking for was the simplest sort of serial to FW bridge.

I suppose a TI TSB41AB1 transceiver and TSB12LV01B link layer controller
(plus serdes used the wrong way round?) would do it, but (as an amateur)
I'm wondering whether I'll be able to get this going without the IEEE
standard since the datasheet for the TSB12LV01B does fairly warn

"This document is not intended to serve as a tutorial on 1394; users are
referred to the IEEE 1394-1995 serial bus standard for detailed
information regarding the 1394 high-speed serial bus."


I guess I'll end up using the FT2232H USB Hi-Speed UART ... :-(


Thanks,

colin

 

[Ecnerwal]

I guess I'll end up using the FT2232H USB Hi-Speed UART ... :-(

Given that 1394 is becoming an interface that's not on new computers,
avoid the agony of designing your product to use it. USB2.0 will get 'er
done and work on non-vintage computers. I don't consider the loss of it
a good thing, but the fact is it's not there...

 

[Colin Howarth]

But:

Why?!?!?

Why go with FireWire when USB* will do what you want, and far easier?
FireWire is a peer-peer bus, which really, really, really made a lot of
sense for the product I was working on (the other end of the digital
link was a board that turned the video into RS-170 analog, with no PC in
sight). But the fact that it's peer-peer means that -- with a few
exceptions -- every node has to have all the capabilities of any other
node, which makes the peripheral design much more complex than a
master/slave bus like USB.

You, on the other hand, are talking about something that's well within
the capabilities of even fairly slow USB hardware -- and you can hardly
walk by a computer these days without tripping over a USB cable, while
at the same time it's hard to find a computer equipped with FireWire.

For a lot less effort than equipping a peripheral with FireWire you can
get a USB-capable microprocessor and do whatever you want.

I'd recommend that you find a chip that'll handle USB isochronous mode
(to get you guaranteed bandwidth for that audio that you don't want to
hiccup), and get to it. It'll be a _lot_ easier than messing with
FireWire, and when you're done you'll be able to plug it into any new
computer on earth.

* Several of my colleagues from that period of time would be rolling on
the floor, pointing at me and laughing, just for making that statement.

Thanks (both of you). I suppose that backs up the difficulty in finding
suitable chips.

The why is partly due to what I'd read about FW and USB. And because I'd
like to get around to doing my own (DSO) oscilloscope interface (i.e.
like the electric guitar interface, but running at 12bit / 500 Msps

Maybe I'll get the book, and then do it with USB anyway...

 

[JW]

And because I'd
like to get around to doing my own (DSO) oscilloscope interface (i.e.
like the electric guitar interface, but running at 12bit / 500 Msps

That's pretty high sample rate for a 12 bit A/D converter - those should
cost a few bucks. I just repaired a rather unusual scope that had 12 bit
A/D converters - a Nicolet Integra 20. It had 4 channels of differential
input. However it's sample rate was 1MS/s and it's bandwidth was rated at
500KHz.

 

[Colin Howarth]

That's pretty high sample rate for a 12 bit A/D converter - those should
cost a few bucks.

True enough. Looks like around 180 USD. I actually just picked the
number out of the air, rather than going for the 12bit 1 GSPS chip from
TI -- that one costs around 800 USD Sensibly I might consider going
down to 250 MSPS Even more sensibly, I want to do the audio
interface first.

I just repaired a rather unusual scope that had 12 bit
A/D converters - a Nicolet Integra 20. It had 4 channels of differential
input. However it's sample rate was 1MS/s and it's bandwidth was rated at
500KHz.

Specially designed for those hard to catch slow events? Or for high
accuracy,more likely.

 

[Colin Howarth]

I was on a team that implemented an infra red imager that talked on IEEE
1394, back before it was apparent that FireWire was going to go down in
flames in the marketplace

Ah, the marketplace always knows best...

So, the chips cost a dollar or two more and the USB guys can write the
number 480 (instead of 400) on the box.

That USB loads the processor more and has "substantially" lower
throughput is irrelevant.

*sigh*

I've ordered the book by the way.

 

[JosephKK]

I was on a team that implemented an infra red imager that talked on IEEE
1394, back before it was apparent that FireWire was going to go down in
flames in the marketplace (of course, that was back when Microsoft was
still pretending to support it -- I actually went to a FireWire user's
conference in Redmond).

We got it working, and working quite well. It met every single dang one
of our needs except for "works on a popular consumer-oriented bus".

If you want to go the FireWire route, get a copy of "FireWire System
Architecture" by Don Anderson of MindShare. It is a near-perfect book
for this -- we got copies for every developer, and each one of us was
practically able to prop the book up next to our monitor to figure out
what needed to get done.

But:

Why?!?!?

Why go with FireWire when USB* will do what you want, and far easier?
FireWire is a peer-peer bus, which really, really, really made a lot of
sense for the product I was working on (the other end of the digital
link was a board that turned the video into RS-170 analog, with no PC in
sight). But the fact that it's peer-peer means that -- with a few
exceptions -- every node has to have all the capabilities of any other
node, which makes the peripheral design much more complex than a
master/slave bus like USB.

You, on the other hand, are talking about something that's well within
the capabilities of even fairly slow USB hardware -- and you can hardly
walk by a computer these days without tripping over a USB cable, while
at the same time it's hard to find a computer equipped with FireWire.

For a lot less effort than equipping a peripheral with FireWire you can
get a USB-capable microprocessor and do whatever you want.

I'd recommend that you find a chip that'll handle USB isochronous mode
(to get you guaranteed bandwidth for that audio that you don't want to
hiccup), and get to it. It'll be a _lot_ easier than messing with
FireWire, and when you're done you'll be able to plug it into any new
computer on earth.

* Several of my colleagues from that period of time would be rolling on
the floor, pointing at me and laughing, just for making that statement.

And i think i may join them. USB 3.0 is out now and the connectors are
different and a lot of other things. See also eSATA. Must also mention
PCI-E(xpress) 3.0. Move over InfiniBand, there are new pigs at the trough.

 

[John Nagle]

FireWire is more or less obsolete at this point.

I once wrote a FireWire camera driver for QNX, to support
stereo cameras. But that was 7 years ago.

FireWire is really a local area network, not a "bus".
Linux supports IP over FireWire, and Windows XP did, but
but Microsoft took that capability out of Vista.

John Nagle