Firewire and USB tranference data question.

[Robman2]

Question: Phil and Craig...this is a what if question. Not a thread to argue in, thanks so much.

 

A few folks, maintain that one of my brands outperform other studio standard brands, digitally.

 

I know what they are saying, based upon physics etc, and so we are launching DVI, VGA and HDMI in this series.

 

I've also suggested we develop Firewire and USB cables and our Webmaster asked me if that's something which is hardware throughput dependant.

 

In other words, if we developed a less resistant cable, would that be bottlenecked by the send and recieve platforms transfer rate?

 

Can the chipsets, CPU's and error correction schemes involved with Firewire and USB, work more efficiently if data transfer occurs at a faster rate, across the cable from point A to point B.

 

Thanks,

 

Rob

[Phil O'Keefe]

Interesting question Rob.

 

To be honest with you, I don't know for certain. But I just did a nice big Google search for "IEEE 1394 spec", and from what I've read, I don't think a less resistive cable is going to make a significant difference in anything, due to the serial data rate and way that the interface deals with it.

 

In addition to an architecture that scales with silicon technology, IEEE 1394 features a unique isochronous data channel interface. Isochronous data channels provide guaranteed data transport at a pre-determined rate. This is especially important for time-critical multimedia data where just-in-time delivery eliminates the need for costly buffering.

Source.

 

Again, I'm far from being an expert on the internal workings of Firewire, so I could very easily be wrong.

[where02190]

It's not the cable that's the bottleneck, it's the interface speed.

 

CAT-5 is tons faster than FW or USB, and the cable is no bigger.

[franknputer]

 

Originally posted by where02190 CAT-5 is tons faster than FW or USB, and the cable is no bigger.

 

Huh? If I understand it correctly, the throughput for CAT5 is 100 Mbps, while USB2 & FW400 are closer to 400 Mbps, with FW800 going to 800 Mbps. Now, there's gigabit ethernet - but you can't do it with CAT5, you have to use CAT5E or CAT6, which is a thicker cable.

 

Not that I think the cable's going to make a lot of difference, unless it's not up to spec of course.

[deanmass]

Cat5 cable, specifically cat 5 e, goes gigabit, or 1000 mbps, and hypothetically to full duplex, to 2 gbps. It is all the in the spec. The cable is not the limitation, but the length of the cable can hypothetically speed up the transfer. The interface/bus is the key, then the processor, etc.

 

So, that being said, ( and I am not 100% on those opinions, so no flames please), I find firewire to be more consistent, but those in my own eyeball tests,not hard data. None compare to PCI, or the new PCI spec ( I forget what it is called). What amazes me about all this stuff is that fiber is not the interface of choice. It just has not caught on. CAT5 is not a practical alternative to USB or Firewire due to the interface access/usability. You could hack transfer, but most users would not.

[boseengineer]

All in all there are about 5 or so physical contenders in shipping digital audio around. These are

USB, Firewire, AES/EBU or SPIDF (Coax or optical), CAT5/6, ADAT.

 

Some of those come in different flavors

 

CAT5 (or to be precise RJ45 jacks with twisted pair CAT5, CAT5e, or CAT6 wiring) is used in a couple of music industry applications, such as

- Gibson Magic (based on UDP Ethernet)

- Line 6 Variax Port (proprietary)

- CobraNet (based on UDP ethernet )

 

Yamaha's MLAN is basically a customized version of Firewire (or IEE1394), etc.

 

Unfortunately, it's the wild, wild west and what the world really needs is a widely accepted standard for multi-channel digital audio.

 

So what are the differences ?

A 44.1 kHz 16 bit stereo PCM stream requires a bandwidth of about 1.5MBit/s. All of the formats above can handle that without breaking a sweat. Even USB 1.0 (probably the slowest of the bunch) handles 6 channels or so just fine. The main differences are

- latency (how long does it take for the data to get from A to B)

- undesired data manipulation (e.g. sample rate conversion by the Windows kernel mixer)

- flexibility

- how easy is it to ship multiple channels

- can you trade-off bits against number of channels

- can you ship floating point, etc.

 

All of those are system design question and have only little to do with the physical layer (i.e. the cable). To give a complicated answer to the original question:

 

"Having a faster cable does not improve the sound quality in any way". For example using CAT6 (i.e. higher bandwidth) wires instead of CAT5 on a CobraNet does not improve sound quality, latency, or flexibility.

 

Cable specifications on well defined interfaces (such as Firewire, USB or CAT5) are pretty strict. If a cable is compliant, than you'll get 100% of the bits across and that's that. Improving the cable further doesn't do anything.

[Robman2]

This is great feedback and I am grateful for the sensible replies.

 

It's nice to know that Phil and his dogs have stopped watching his Ule's Gold DVD long enough to chime in...

 

A simple one line follow up, if the actual conductors were 7% less resistant then standard spec wires in other words if the 1's and 0's were delivered that much faster...would that enable both ends of the chain to complete their tasks sooner?

 

I sense from the explanations so far...not... but I'll have to re-read these and let it sink in.

 

Thanks so very much to all...

[deanmass]

I am still really curious why fiber has not replaced all of these. It has the most bandwidth, or so I think, can be run for MILES, has become fairly cheap and common....Maybe someone can chime in with the 'tech' reason for it.

[kevinnem]

I will state that I am certinly no expert. BUT I would assume that the devices on teh ends would certinly be the limiting factor.

 

I would think a less resitive calbe, may improve the Distance that you could run. But I don't think it would improve the data rate any.

 

I have found that with most thing digital (like bose eng said.) it kinda works or not. There is not a lot of grey area. The exception to this is wireless (radio) netoworks, that I have found exibit exactly this condition.

 

Kev.

[boseengineer]

 

Originally posted by Robman2 ... A simple one line follow up, if the actual conductors were 7% less resistant then standard spec wires in other words if the 1's and 0's were delivered that much faster...would that enable both ends of the chain to complete their tasks sooner? ...

 

 

The quick answer is "NO". Here are some of the less complicated reasons.

 

1)

Any modern digital communication system contains many components. There are the cable, connectors, PC-boards, line-drivers, I/O buffers, peripheral conditioners, clocks, software infrastructure, etc. To make things work, all of those have to match certain requirements. That's basically done by putting a stake in the ground of what the maximum speed will be. For USB 1.0 thats 12 MBit/s. IEEE 1394 runs at 100Mbit/s, 200Mbit/s or 400 MBit/s. Ethernet at 10MBit/s or MBit/s (MBit/s is in the works).

 

Everything bascially runs at a fixed "agreed upon" speed". The specification for each component in the chain are derived from that speed (and a bunch of other things of course). Everybody is certainly welcome to exceed that specification but as along as everyone meets them, the whole thing will work. Exceeding the spec on one component does not allow to run the whole thing any faster.

 

It's a little like saying: "I put extra thick AC wires in my house, now I can not only ship 110 Volts but 179 Volts or even 267 Volts". The point is, your wires could certainly provide that but all your equipment in the house will still only work with 110 Volts.

 

2)

Reducing the resistance does not make the wire any "faster". Information travels close to the speed of light (ca. 300.000.000 meters per second) regardless of cable resistance.

What you are looking for is "bandwith" and that's a pretty complicated phenomenon. The resistance of the cable has to match the input and output resistance of the line drivers (to avoid reflections). Other things concern wire thickness (because of "skin effect"), number of twists per meter, inductance and capacitance, etc. The actual DC resistance is much less of an issue.

 

Hope that helps

[boseengineer]

 

Originally posted by deanmass I am still really curious why fiber has not replaced all of these. It has the most bandwidth, or so I think, can be run for MILES, has become fairly cheap and common....Maybe someone can chime in with the 'tech' reason for it.

 

 

The simple answer is $$$. For an in-home application, copper is still cheaper.

 

It's a matter of finding the right tool for the right job. All high-bandwidth connections are fiber these days. But for the moderate requirements of a single house-hold, copper is still competitive and for a single device-to-device connections considerable cheaper and more convenient.

 

Eventually fiber will be the in-house connection of choice and it's already starting (e.g. http://www22.verizon.com/fiosforhome/channels/fios/root/about_fios.asp)

 

For device-to-device connections, optical is IMHO just a gimmick. Copper does the same job better and cheaper. The only advantage of fiber that it provides perfect ground isolation, but a good copper connection can do that too.

[Gulliverseyes]

There is a fairly new protocol called iSCSI, which uses standard Ethernet cabling/interconnect to move storage. It's as fast as 100MB/s using GigE cabling ... which is a little bit faster than USB 2.0 or Firewire 800. The cool thing is that it can be networked for shared storage using a standard GigE switch (ie. iSCSI SAN, but management s/w is needed) or used for storage transfers over long Ethernet lines (the internet would slow it down, though, and transfers would be about the same as TCP/IP). Users can use a standard NIC card at the host computer, but a TOE (TCP off load) card will take stress off the CPU. Windows provides the iSCSI initiator (host s/w to convert to protocol) for the PC world. ATTO makes an initiator for Macs and bridges to bridge Ethernet cabling using iSCSI back to SCSI or fibre channel for the target end (standard hard drives and tape libraries).

 

As far as optical, remember, it's not just the cable cost, but you need an optical interface at each end of each run to convert to an electrical signal = $$$.