Disclaimer: Aurora Video Systems graciously lent a new Fuse-X card to me for this evaluation but provided no other compensation. All observations and measurements herein are entirely my own and are offered simply as one man's opinions. This article is primarily of interest to a Pro Tools user.
Aurora's press release dated July 25, 2005:
“The Fuse-X is the latest example of Aurora's continuing commitment to professional audio editors, mixers, sound designers and composers, offering the only zero-latency, in-sync playback solution for compressed video in pro audio environments…With Interceptor mode enabled on an Aurora Fuse-X or Pipe-family card, D.A.W. users are able to accept video files in a variety of QuickTime compression formats, and play them as-is within Protools™, Digital Performer™, Logic Pro™, and other audio packages.” I couldn’t help but be intrigued by the promise of “zero latency” and unfettered playback of a variety of QuickTime codecs “as is”. I just had to investigate!
Review in a nutshell: Overall the Fuse-X package may be just what you need for OSX. Macintosh only. The card is designed for NTSC and PAL output and is genlockable. Scrubbing performance is quite good. Fuse-X hardware relieves your computer’s processor of much data crunching. It’s low demands result in good D.A.W. performance. It is capable of near-perfect video/audio lock under the right circumstances. The card can play PAL files to NTSC monitors and vice versa, although sync tightness is reduced. Aurora’s half-bandwidth playback mode trades some picture sharpness for lower PCI requirements and helps reduce picture stutters during times of very high system loads. We saw it work as promised on a G5 but less effectively with some codecs on the G4s (a target for a future update, perhaps). Our tests were done on TDM systems, as we do not own LE systems. Based on the card’s reasonably low cpu demands I expect any recent model Mac can handle it well. You may not get perfect sync performance with a slower machine that does not meet Aurora’s minimum specifications. DV movies worked well on the three systems we tried, even the slow G4. Some other codecs may require a very fast G4 or G5 to work smoothly with medium or better quality settings. The card comes with a CD containing various installers and “MediaGrabX”, a basic and reliable capture app with some sliders to alter input video settings. We used it to digitize a few of the files for our tests and found it is easy to use. In short, Fuse-X is very good but there are some things to consider before purchasing one.
Basic setup is not hard but may take a couple of reboots. Apparently, many preferences are stored on the card itself in NVRAM and require a reboot to take effect. Hopefully, Aurora can tighten up this process a bit. Plunk the PCI card in your system, install the drivers (We experienced some exceptionally long wait times as the installer from CD, 4.2, chugged away. A 4.4 driver update seems to have fixed that.), reboot, launch the AuroraAV preferences window, set Desktop/Interceptor Control preference for “Desktop & Interceptor”, reboot, and finally go to the Monitors preference page to set the card for PAL or NTSC, just like you would set a computer monitor for different screen resolutions. While you’re setting preferences, initially set the card’s Interceptor Mode for full bandwidth/odd field if you are using NTSC, or full bandwidth/even for PAL. Aurora’s Desktop/Interceptor technology is cool and under the right circumstances your video monitor can become a second or third desktop monitor. For use with Pro Tools you must have Interceptor turned on but you will not see or be able to use the additional desktop space while Pro Tools is running. The Interceptor allows you to play almost any QuickTime supported movie through your monitor; you are not limited to one or two specific codecs. You can, for instance, display a PAL movie on your NTSC-only video display. Sync might be a bit wobbly, but hey! It will get you through in a pinch.
Those of you from the Digidesign User Group know that I am very concerned with accurate and predictable video synchronization for Pro Tools. I checked performance on three systems using a test session designed to stress a system’s media drives and PCI bandwidth. Aurora does not offer guidelines for which PCI slot to install the card in so the slots we used were somewhat arbitrary. I noted the slot in each system’s description at the end of this article. The Aurora card can be set to clock from its own crystal or genlock to a video source applied to its composite input. For our tests everything was always genlocked to a house reference blackburst. All sync tests were performed with a standard CRT video display, the kind with no extra latency, to make measurements easier.
The Fuse-X card performed brilliantly in our G5, but our first victim was our slow G4/867 single proc. This system is so old it does not even meet Aurora’s minimum specs (G4/dual 867 or better), but I’ll wager there are many machines in use that are even slower and I think it provides a good real world balance for this review. I should mention that we saw a problem on this machine that we did not experience on the others. Motion jpeg A movies often stopped playing out of the video monitor while Pro Tools was running. If re-imported they would play again for perhaps several start/stop cycles, but then disappear from the video monitor. No other codecs we tried had the problem so we pushed on. We determined that a Pro Tools movie offset value of 5 quarter frames put the video in sync with audio. We also saw that playback synchronization can be consistently perfect if you do all the right things! I personally do not care if the card does not keep its promise of “zero latency” as long as it behaves in a totally consistent, predictable way. Because of its genlocked output there was no drift during playback. Read on, my friends, because some of our same old sync problems I’ve discussed while examining DV/firewire converters came around to bite us again!
There was no drift but there was not always perfect synchronization either. The card is a good piece of engineering but Aurora cannot overcome problems outside of their control! As I’ve discussed previously, unless you buy Digidesign's video hardware the only way to guarantee perfect audio-to-video synchronization is to follow a two-step approach. First, genlock all audio and video hardware to a single video reference, a.k.a. “house sync”. Doing so prevents any drift between them and allows your hardware to identify “frame edges”, the boundaries between one video frame and the next. In it's price range, Fuse-X is the only choice I know of that is capable of genlock. Second, you must ensure that your D.A.W. time-aligns audio “frame edges” with the video frame edges. To do so, Pro Tools requires a SyncIO (or USD with Mix hardware) to be attached! You must set your SyncIO or USD to generate SMPTE at the same frame rate as your video and place Pro Tools software “online”. (It is possible to "chase" SMPTE rather than "generate" it, with equal results, as long as the SMPTE is frame-edge aligned. Most professional video decks do so.) Pullup or pulldown settings do not matter, only the SMPTE and video frame rates. Generating the code forces Pro Tools to align audio frame edges with video frame edges. If you do not have a SyncIO or USD, or you choose to skip the second step, Pro Tools will not attempt to resolve those frame edges, meaning Pro Tools will simply start the audio anywhere within a video frame’s worth of perfect, which means that unless you take both steps your audio and video will most likely be out of sync by up to a full frame! To add insult to injury, without frame edge alignment the sync error changes randomly each and every time you press play. Fuse-X’s genlock capability does not make it immune to this behavior. Aurora can hardly be faulted when Pro Tools does not resolve frame edges, but by taking the proper steps you can enjoy perfect audio and video synchronization, better than DV/firewire converters, better than the old Miro card, rock solid and reliably in sync. Well, almost rock solid and reliable. On our slow G4 there was an occasional one frame error. No drift, no partial frame errors, but once in a while after pressing play the video started up a full frame behind audio. I suspect a computer that meets Aurora’s minimum speed specs will not have trouble. Our other two (faster) machines did not.
While I’m harping on the subject of aligning your frame edges, it is worth noting that when Pro Tools is looping in playback or while recording, the video sync will change on each loop around unless the loop selection is perfectly set on frame boundaries. This is Pro Tools/QuickTime behavior and is not limited to Aurora, but working with the Fuse-X’s genlocked output makes the phenomenon easy to see. Non-genlocked devices, as nearly all freestanding firewire/DV converters are, allow so much sloppiness that this specific error is obscured.
Scrubbing was smooth and responsive even on our slow G4. While scrubbing back and forth on our test files, Syncheck™ showed the video was generally one to two frames behind audio. Picture accuracy FELT tight and accurate as we scrubbed. If you have been putting up with scrubbing through a firewire/DV converter you will be absolutely thrilled.
We tried playing different frame rate movies through a single video standard and discovered that you can get away with it as long as the movie’s frame rate is equal to or less than your video standard. NTSC movies playing through the PAL setup resulted in some skipped frames. With QuickTime trying to stuff 30 frames per second through a 25 frames per second output, this should not shock anyone! Syncheck™ showed some sync errors as well, also to be expected. A 24p test file through the PAL setup worked usably well. With a movie offset value of 3 quarter frames we saw very consistent plus/minus ½ frame errors (to be expected), but no lost frames. Playing our PAL test files while Pro Tools and card were set for NTSC (did I mention how cool that is?) resulted in sync errors generally less than a frame. We also checked a 23.976p test file through the NTSC setup and found very consistent results with low errors. Again, the movie offset value needed to be 3, not 5.
Time out for a word about Pro Tools’ movie offset value. As just noted there does not seem to be a single setting that is best for all situations, even on the same machine! In each of our systems when we followed the rules for perfect synchronization, DV movies (PAL or NTSC) played with consistent near-perfect synchronization with an offset of 5 quarter frames. If we ignored the rules so that Pro Tools was not actually resolving audio to the frame edges, the best setting was 3 quarter frames, so that errors were distributed equally ahead and behind perfect. Some highly compressed codecs, like motion jpeg A, Sorenson, or mpeg4 seemed to require 7 quarter frames. Of course I cannot guarantee that my numbers will work for you, especially if you are using a latency-inducing solid state display. At the risk of tooting my horn too much, this kind of crazy stuff is why I think everyone should own a Syncheck™. How on earth can a normal person be expected to keep up with all the variables, especially as other software and hardware changes?
Fuse-X was put into our midline G4. During driver installation to OSX 10.3.5 we had to allow nearly 8 minutes for the installer to finish! For most of the wait there appeared to be no activity, as though the installer had crashed. A similar delay occurred while installing to the G5 running OSX 10.4.2 but, curiously, there was no wait to install on the slow G4 (same version OS as the midline system). Aurora is aware of the problem and expects a fix. After installing, before we did any serious testing, we noticed that QuickTime 6.5.1 was prone to crash while we tried to set export parameters for certain codecs (Sorenson and Aurora’s uncompressed being among them). We had seen the same problem on the slow G4 so we decided to upgrade this machine to 10.3.9 and QT 6.5.2. The crashes persisted until we un-installed Aurora’s software but did not otherwise interfere with our tests. Most of our codec comparisons were made on this system. DV movies worked well and while we could see the effects of cpu drain on Pro Tools’ counters we got excellent playback performance. We also saw surprisingly good performance with Aurora’s 10 bit Extreme uncompressed codec. It looked MUCH better than DV but uses 7 times as much disk space! For moderately sized sessions on a big system we think it is a serious contender, but hesitate to recommend it because half-bandwidth mode does not work properly with it. (If your session were to outgrow your system’s speed, you could not simply switch to half bandwidth mode and go back to work.) The other codecs we tried all had various troubles unless the capture quality (data rate) was low or lower. Some of this may be due to the single hard drive used for video, some due to the midline system’s speed. Half bandwidth mode did not work correctly for Mpeg4, so we cannot recommend it either. Medium quality Sorenson with the card in half bandwidth mode worked about as well as DV movies in full bandwidth mode, but Sorenson in full bandwidth mode showed some stuttering and glitches in a heavy session. Lowering Pro Tool’s processor setting to 40% helped only a little, but naturally made Pro Tools more sluggish. The movie offset value required for DV movies (5 quarter frames) did not always hold for the other codecs. 7 quarter frames gave proper results for them when the video was good enough to measure. What does it all mean? If you’ve got a midline machine you may want to consider using only DV movies.
Time out for a word about DV movies! I do not like the quality of DV in general, although I use it regularly, it always looks grainy and soft to me. I especially do not like the DV movies made by MediaGrabX! It retains DV’s softness but for some reason makes images much more contrasted and slightly darker than the original. We noted some other minor banding and picture displacements (wiggly vertical lines that were straight in all other codecs.) I preferred the look of medium or better quality Motion jpeg A and mpeg4, by far, as they looked much closer to the source’s contrast and sharpness. We also compared MediaGrabX’s DV movie to a DV movie exported by Avid Express Pro/MOJO from the same videotape source. It was also grainy and soft, like DV, but at least retained the original’s contrast.
Fuse-X next went into the fast G5 running OSX 10.4.2 and QuickTime Pro 7. We had no QuickTime crashes. DV movies, motion jpeg A, Sorenson, and Aurora’s 10 bit uncompressed were tested, mpeg4 was not. We saw no stress from the Aurora card and playback looked good! That’s not to say the video was perfect, it was not. We began seeing video hits at around 75% of maximum system stress. We set the card for half bandwidth mode and BINGO, no more picture problems. This is a nice feature! Thank goodness it does not need a reboot! In half bandwidth mode all of the codecs just mentioned played smoothly, including Aurora’s uncompressed, without hiccups or sync problems as we increased system stress until Pro Tools gave errors. That is very good. The same system with a MOJO is able to do just as well, though. It is probably a testament to the additional capabilities of the G5 design and our very fast video array. There was a small bug, however. This system, unlike the other two, uses two desktop monitors. As I said earlier, the Interceptor feature basically makes a third desktop monitor. On our system, the Interceptor inserted a third desktop between the two regular monitors. In order to cross from the left monitor to the right you must travel through the video screen. When Pro Tools is active the “third” desktop is replaced by your movie image but the hidden desktop space must still be there in some way. We discovered it is possible for the cursor to get stuck when crossing the boundaries between the two regular desktops. There were times when we could not get our cursor out of one display to the other, and one time it got stuck in the third desktop area while Pro Tools was running. We had no cursor at all. We could see it when we switched to the Finder but we could not get it out of the video display until we quit Pro Tools. Aurora is aware of these bugs. We did not see them when using a single monitor system.
Compared to Digidesign’s least expensive video hardware, MOJO, Fuse-X is substantially cheaper. While Digidesign’s hardware in a supported configuration will always provide near-perfect synchronization without having to generate time code, Fuse-X takes a bit more effort, sometimes a lot more, to get equal sync performance. Digidesign provides a picture that scrolls with your cursor as you drag. The Fuse-X card relies on QuickTime. As you drag, QuickTime’s picture stays strictly with the insert cursor. Picture scrolls when you click and drag to the left, but freezes when you drag to the right of the insert point. This behavior is shared by every video solution except for Digidesign’s own. Scrubbing with Digidesign’s hardware can feel sluggish on even the fastest computers with the fastest drives, with 8 to 12 frames or more of delay between what you hear and what you see on the video screen. Fuse-X is fleet-footed and scrubs with only a frame or two of delay. (To be fair, Digidesign provides an alternative “scrub window” which opens on your desktop during scrubbing if you enable it. Scrub sync of the window is nearly perfect.) Digidesign’s video solutions play only Avid media and resolutions, and require QuickTime codecs to be converted (using MediaStation|PT or an Avid system, and extra cost). Once converted, playback quality and consistency are very high. Fuse-X plays many QuickTime codecs without conversion but with results that may disappoint on slower machines and/or at higher quality settings.
What I like: Fuse-X is a very good value for the cash, supports most of the commonly used video codecs, displays any QuickTime-supported frame rate on any PAL or NTSC monitor, has great scrubbing performance, has reasonably low cpu demands, is genlockable, and can provide near-perfect synchronization in a properly configured system. It is affordable and useable in any professional audio situation.
What Else: Fuse-X requires a PCI slot, sync may not be perfect unless you follow all the steps mentioned earlier, your movie sync offset value may need to change based on codec or genlock settings, and picture scrolling follows the rules of other QuickTime-based solutions. Aurora’s 4.2 drivers may cause QuickTime Pro 6.5.1 and 6.5.2 to crash under some circumstances.
Please visit Aurora Video Systems for more information.
We tested Fuse-X with driver 4.2 on these three Macintosh systems.
Slow: G4/867 single proc, OSX10.3.5, QuickTime 6.5.1. Pro Tools 6.4.1. Mix+ hardware and dual channel ATTO UL3D in a 13 slot Digidesign expansion chassis. G4's slot order: chassis host in 1, open 2 and 3, Fuse-X in 4. Single desktop display. Drives: audio 7200rpm, video internal 7200 ATA.
Midline: G4/dual 1G MDD, installed to OSX10.3.5, then upgraded to 10.3.9 and QuickTime 6.5.2. G4's slot order: HD+ in slots 1 and 2, Fuse-X in 3, and ATTO UL3D in 4. Single desktop display. Drives: audio 10K SCSI, video internal 7200 ATA.
Fast: G5/dual 2.5G, installed to OSX 10.4.2, QuickTime 7. Pro Tools 6.9.2 HD++ in 64 bit/7 slot Magma expansion chassis. G5's slot order: chassis host closest to graphics card, Fuse-X next, ATTO UL3D third. Dual desktop displays. Drives: audio 10K SCSI, video 4 drive hardware RAID SCSI.
|