Meeting held May 21, 2003
On May 21, 2003 the PNW Section met at Glenn Sound Studio in Seattle for a presentation on mLAN and Firewire from Mike Overlin of Yamaha and John Strawn of S Systems Inc.
Section Chair Rick Chinn opened the meeting with greetings and upcoming Section election info. Committeeman Bob Gudgel, the meeting organizer, had the 40 attendees introduce themselves.
Mike Overlin had a career as a professional musician, recording studio owner and engineer, and instructor at TreBas Institute in Hollywood, California. His career with Yamaha began in 1994 doing tech support for computer based digital recording products and as a product specialist for Yamaha pro audio products. He is currently the manager of Yamaha's mLAN licensing program in the US.
John Strawn was originally trained as a performing musician, and is now an independent consultant specializing in DSP for audio and music. He received a Ph.D. from Stanford University Center for Computer Research in Music and Acoustics (CCRMA) in 1985. A Fellow of the AES, he chairs the AES standards committee SC-06-02 on Audio and Music over 1394, and was the papers co-chair for the October 2002 Los Angeles convention. Having led an R&D group for Yamaha for several years, he continues to consult for Yamaha, especially relating to 1394.
First, Mike Overlin spoke on the basics of mLAN and its relationship to Firewire/IEEE-1394. mLAN stands for Music Local Area Network and is a Yamaha trademark. Starting over a decade ago Yamaha became interested in an audio network and control system that could work in any size installation, from homes to large venues. Firewire, a trademark name for the IEEE1394 standard, had the desired characteristics - it was hot pluggable, high speed, low latency, AV media ready and cost effective. You can connect many processors, mixers, keyboards and other devices with one small cable and connector. However, it needed some improvements to be a better fit for professional audio. He showed slides of a very large installation at Miami's Pro Player Stadium.
Next, John Strawn continued with a detailed overview of protocol layers of Firewire/IEEE-1394 and mLAN. Firewire was invented by Apple Computer in the early 1990s, and is now a public standard (IEEE-1394) with a trade association (the "TA") and patent pool with small licensing fees. It is a serial bus and hot-pluggable - the network reconfigures itself when devices are added or subtracted to a live network. Up to 1023 buses are allowed, with up to 64 nodes on each bus, each with up to 2E48 bytes of memory. The only topology not allowed is a ring or circle - the startup protocol will fail. A bus Master device is automatically determined on startup, which can change if the need arises.
The original 1394 specification cable was copper, with at least 2 pairs of twisted cable plus 2 power wires. Sony's similar connector (iLink) offers no power wires but the same protocol. Originally, a network allowed 16 hops over copper wire of 4.5m each, for a maximum overall distance of 72m. Newer 1394 specs allow for optical, 800 Mbps CAT5 and even Gigabit (in the works). Various bus speeds were shown, S100 (100 Mbits per second), S400, S800 and so on, with various speeds possible over various conduits, such as copper wire, POF (plastic optical fiber), HPCF (hard polymer clad fiber), and GOF (glass optical fiber).
1394a made small improvements to the original specification, and 1394b was a major rewrite. A bridge device is needed to interconnect buses, and the standards for that are being worked on now.
1394 does not have all the layers of a typical network protocol. There is the physical layer, PHY (cables, connectors), the link layer (packetizing data, handling isochronous and asynchronous data), transaction layer (asynchronous management), and serial bus management. There are two ways of moving 1394 data - asynchronous, when you are not so concerned about the time the data moves, but you are concerned about reliability and accuracy, such as disk transfers; and isochronous data for AV media. With isochronous data, a node sends data packets out, not quite knowing or caring if it gets to the destination. Missing an occasional packet is acceptable. Packets are sent at a regular 8 kHz rate for nodes with bandwidth reserved. 80% of the bandwidth is typically reserved for isochronous data. Nodes can also send asynchronous packets in a managed scheme.
Yamaha proposed additions for audio and music to 1394, which the TA, being unable to make public standards, submitted to the IEC. John described how the Audio and Music Protocol, also known as IEC 61883-6, accomodates MIDI data in the quadlets (32 bit data portions) of Common Isochronous Packets (CIPs). A clever timestamp system allows audio data clock recovery from the CIP without the need for a dedicated clock line.
Manufacturers may need to create a custom "Protocol Engine" - a term John attributed to (AES Western VP) Bob Moses, which describes proprietary interfaces between a Link layer and Hardware. This would deal with issues such as synch, FIFOs, clock recovery, headers, etc. plus the audio and MIDI. He mentioned products from MOTU and Metric Halo as examples of devices that solve these problems.
mLAN may be described as extensions built upon the foundation of 1394 and the IEC 61883 standards. Any mLAN device is compliant with 1394 and IEC 61883, but can perform connection control and network topology configuration that regular 1394 normally does not. For example, a large distributed audio system with mLAN devices can retain all of the patching and network architecture needed after a system reboot, which regular 1394 would not normally do.
Finally, Mike spoke again on mLAN hardware. He showed a slide of a 63 node lab demo with 350uS maximum latency. Major installs were mentioned. Many products with mLAN other than Yamaha were mentioned. Licensing is free, and Yamaha is not the sole vendor of the chips. It is incorporated into Apple's OSX, and of course, also works with Windows and soon, Linux. mLAN improvements have made it more versatile, customizable, scalable, faster and cheaper.
Many question were asked on hardware configurations. The presenters complimented the group on the high quality of the interaction, due no doubt to the presence of many in the group actually working on Firewire projects.
We adjourned for cookies and soft drinks in the lobby. Special thanks to Glenn Lorbiecki of Glenn Sound Studios in Seattle for generously allowing our use of the studio for the meeting.
Reported by Gary Louie, PNW Section Secretary
Last modified 2/4/2004.