updated 31 March 2003

 

Configuration of Non-Linear Digital Video Editing Systems, Part Two
an article by Charles F. McConathy

In the second half of this article (check out the first half here), Charles continues with various video storage hard drive options, explains the advantages of a DV VCR, and concludes with some answers to frequently asked questions.

ATA to SCSI Storage Systems:
Medéa and other companies build storage systems that use less costly ATA drives converted to SCSI. These drives need a PCI SCSI Host Adapter. They are less costly than SCSI to SCSI. If the systems support Ultra SCSI 160 connections they can sustain from 65 to 95 MB/sec. The systems use hardware RAID-0 with large buffers dedicated to each drive. These drive systems are excellent for DV/DVCAM and uncompressed standard-definition video.

SCSI to SCSI RAID Storage Systems:
SCSI is about twice the cost of ATA storage. Ultra SCSI 160 is best used for uncompressed SD/RT and HD video systems. For high-demand RAID storage, SCSI technology is best since it imposes minimum overhead during the connect-burst/ disconnect/ connect-burst routine of RAID-0. ATA systems, by contrast, tie up the bus during the entire data transfer cycle. SCSI cables have to be reasonably short, so storage systems must be installed within a few feet of the workstation. For some, however, this is not acceptable due the noise the drives and fans make.

At ProMax we use eight drives striped as RAID-0, as four and four on each channel of a dual channel Ultra 160 PCI host adapter. We recommend disk drives that include a 16 MB buffer, as these are especially good for SD/RT and HD applications. For added performance we often add a 9th drive for audio. Multiple tracks of audio creates a lot of head movement thus keeping audio on a separate drive causes less latency to video thus less chance of dropped frames during complex areas on the time.

Many have asked if adding a high performance RAID storage system to a DV/DVCAM workstation will improve functions like rendering. The answer is no, it's a waste of budget. Rendering is computer intensive, not I/O intensive. DV/DVCAM uses only 3.6 MB/sec thus adding a RAID system that sustains high rates like 65 to 170 MB/sec is overkill.

We think that SCA (80 pin Single-Connector-Attachment) drives in removable bays are superior to fixed 68 pin drives. The reason is that replacing a drive is very easy. Removing disk drives from the cabinet and packing them in foam before shipment keeps them safe. A cabinet with 8 to 15 internal fixed drives is very heavy and hard to ship without damaging the drives and the cabinet.

RAID levels:
RAID-0 adds performance by striping the data across two or more parallel drives, reducing the workload for each individual drive. RAID 0 is often done with software, but hardware RAID-0 is better especially if the hardware adds additional memory buffers for each drive. Note however that RAID-0 offers no redundancy: if one drive crashes, all the RAID's data is lost.

RAID-1 mirrors one set of drives to a second set. The problem with mirroring is that it takes twice the time for each write since the data must be written twice. And mirroring is expensive since you need twice the amount of drives. RAID-1 thus provides redun-dancy for reliability purposes, but at the expense of performance, although hardware RAID-1 controllers minimize the penalty.

RAID-3, 4, and 5 stripe data over multiple drives, perhaps 4, 5, 6, or more. One drive is a "parity drive" providing redundancy, thus you "lose" storage space equivalent to one drive. If any single drive fails, you replace it and the system rebuilds the data on it.

Some systems take a lot of time to rebuild especially if done while the storage system is under use. Some RAID-3, 4, 5 systems have very fast reads with slower writes, since it takes time to scatter the data over multiple drives. Other systems have overcome this with special hardware. You need to determine which RAID system best fits your requirements. Don't just take the word of manufacture with published data rates, but try to get real-world results from other users. Benchmarks don't always tell the whole story. An old timer in the storage industry once said, "there are liars, damn liars, and benchmark liars!"

Some companies quote the data rate of the storage interface, like Ultra 160 as 160 MB/sec. No single drive can sustain this rate: it's the burst rate, not a sustained rate. To get a sustained 160 MB/sec you typically need 8 drives striped as RAID-0 across a dual channel host adapter.

Fibre Channel RAID Storage Systems:
Early Fibre Channel was based on one-Gigabit technology. Two-Gigabit Fibre Channel systems have started to deliver, with data rates high as 190 MB/sec. These are excellent for use in editing SD/RT and HD video. It is possible to mix 1-Gbit and 2-Gbit Fibre technology, but the results will be performance at the 1-Gbit level. Optical Fibre Channel cables can be 100s of meters long, so noisy storage cabinets can be placed in a machine room away from the edit suite. Fibre Channel storage can be configured to safely share files with multiple workstations in a SAN (Storage-Area Network) configuration, though the ease and flexibility of this sharing varies by vendor.

Fibre Channel cost from 35% to 50% more than SCSI, due to the added costs of PCI host adapters, software, hubs, and switches. Added productivity is the key consideration and often has fast payback. Fibre Channel SANs are best suited where multiple editors need to share the same material in order to finish projects in a timely manner.

Much study must be made before installing a Fibre Channel disk array. Take care in picking the right host adapters, cables, software, and switches before ordering. Switches are expensive but are highly necessary. A switch isolates each station from the others, guaranteeing uninterrupted bandwidth. Also the switch protects each station from "LIP" (Loop Initialization Process). LIP is the process of polling the Fibre Channel bus each time a workstation starts up, which interrupts data transfers for a split second. During LIP dropped frames will take place on active workstations, which is eliminated with a Fibre Channel switch.

Installing RAM:
The cost of memory is at an all time low. The minimum we suggest is 512 MB. 768 MB and more can be helpful especially with applications like Photoshop and AfterEffects. Not all RAM is created equal, so make sure you pick a known proven source when purchasing RAM. Be sure to handle memory by its edge and ground yourself before installing modules. Do not touch the contacts or parts on the SIMM. This could cause damage that might not show up for months.

Software Conflicts:
Conflict Catcher, Virex, and other such utilities can cause conflicts when used with video editing systems. It is best to create a complete set of Extensions and then by trial and error eliminate troublesome extensions if conflicts become a problem. Sherlock's indexing feature (on Macs) and Microsoft Office's FindFast indexer can bog a machine down and should be turned off while editing. Appletalk networking, especially on slower Macs, can also cause dropped frames or interrupted playback.

Partitioning Data and Boot Drives:
It is best to run video editing applications from their own boot partition and to keep that partition (and computer memory) free of applications such as web browsers or games that are memory- and CPU-hungry. It's OK to run web browsers and such from their own boot partition (remember that even if they aren't being run, installing such applications on your video partition can add extensions or services that run in the background, reducing the efficiency of your machine). AfterEffects, Photoshop, Commo-tion, and other video editing applications can be installed in the video partition.

Large data drives do not have to be partitioned. It might be advisable to partition large data drives especially if your projects are small and you work more than one at the same time. The reason is that it is advisable to start each new project on a freshly erased partition. If you partition a large drive you can erase any partition without harming data on the other partitions. Erasing clears the directory; thus no fragmentation is left on the partition.

Data Backup:
Backing up large disk storage takes a great deal of time even with the best backup devices. With video editing the data set is constantly changing, making it even more difficult since even if one wanted to there might not be enough down time to perform a backup. And remember that it takes as much time to restore as it does to backup.

Many editors simply keep their EDLs and project files backed up on a ZIP disk or a floppy, and keep the original videotapes. It's also a good idea to keep graphics, titles, and other files that take only a small amount of room, but take a lot of time to create, on ZIP disks or other removable media as well. By doing so, should a failure take place, getting a project going again will take much less time.

Adding a Digital Video Source ­ Deck or Camcorder?
Many people begin editing using their camcorder as a deck. However, once an editor uses a deck like a Sony DSR-11 he or she will never want to use a camcorder as an editing deck again. Why? A camcorder transport is slow. With a camcorder it will take a lot more time to log a series of tapes, and you're placing additional wear on its heads.

A DSR-11 will rewind a 60 minute miniDV tape about 3 times as fast as camcorder and will record and play standard size tapes up to 3 hours. In fact, if you place a 3 hour DVCAM tape in a DSR-11 and set it to record DV, you can record 4.5 hours on it. If your budget allows a deck you will not be sorry, since your productivity will be much higher than when using a camcorder. Note that some of the less costly walkman-style decks have camcorder transports, so will not be as productive as a DSR-11, DSR-20, DSR-30, or other more professional decks.

Frequently Asked Questions:

Q: Can I use an older computer to build a video editing system?

A: Yes, but be aware of the bottlenecks and conflicts when attempting to use older computers.

Q: Can I use an older computer like a Beige G3/266 Power Macintosh?

A: Yes, but you will need to select a compatible PCI FireWire card and ATA storage along with compatible Mac OS, QuickTime, and FireWire drivers. Final Cut Pro 1.2.5 will work with a G3/266 but Final Cut Pro 2 and higher requires a G4/400 or better yet a DP model. And be aware that the newer G4's are many times faster, and the built-in FireWire ports tend to be more trouble-free and more compatible than those on a PCI card.

Q: Do I need SCSI storage for DV/DVCAM editing?

A: No, ATA or FireWire drives are satisfactory.

Q: What is the best way to convert Hi8 to DV/DVCAM?

A: Copying a Hi8 or SVHS analog source via S-Video to a DV/DVCAM deck offers very acceptable results. Often the digital copy looks better than the original since the DV deck's comb filter eliminates noise. New time code will be created on the digital tape which can be used for batch capture. Some decks offer E-E ­ direct analog to FireWire conversion without making a copy, so you can capture the analog video directly (although without its timecode). Sony, Canopus, PowerR, and others make converter boxes that take S-Video or Composite along with RCA audio connections and transcode to DV over FireWire.

Q: Can I make BetaSP dubs to DVCAM with Time Code?

A: Yes ­ Excellent quality dubs can be made with a properly equipped Beta SP deck like a UVW-1800 or BVW-75 and a DVCAM deck like a DSR-80 or DSR-1800 via component YUV and balanced audio connections. For direct input into DV/DVCAM editors, it's also possible to use media converters such as the ProMax DA-MAX+ with the following connections: component, S-Video, composite, balanced and unbalanced audio, and FireWire ports. The DA-MAX+ also converts FireWire machine control to RS-422Sony protocol, so you can control a BetaSP deck and transfer its TC as well.

Q: Why do some want to work with uncompressed video using AJA Kona or CineWave cards instead of DV via FireWire?

A: Simply put, uncompressed video offers 4:2:2 color sampling instead of DV's 4:1:1 which means that chroma-keying is easier. Also, added graphics, text, and animation will not have edge problems and will end up much cleaner with uncompressed video, without needing careful filtering and preprocessing.

Back to Part One.

Copyright 2003 by ProMax Systems, Inc., Irvine, California.
Charles F. McConathy is the President and CEO of ProMax Systems, Inc. which is a leading provider of Turnkey solutions for NLE workstations.

Written by Charles F. McConathy.
Thrown together by Chris Hurd.