Everything below is the same whether using the [only] drivers suggested for Duo in Win2k (old 1.26 from summer 2002) or the latest 2K/XP Quattro drivers.

• The line output's left channel is down about -16dB* in both consumer and pro modes. (tested my amp with another source so I know it's not that) The outputs' noise floor is also at least 10dB if not 30dB higher than my AP2496 was.
• The headphone output works fine, other than the noise at high gain.
• The line input's left channel is also down about -16dB in both modes.*
• I really, really miss the Delta Monitor Mixer. Even with just 2 analog channels, it would've made these tests among other things a million times easier to conduct. I used to be able to route different applications to different virtual outputs (well, the SP/DIF output was real, I just never used it as such) and mix them at will. As it is, there's not even a frickin' main volume control.
• Recording is ridiculously sensitive. Surfing the Net while capturing -- despite using different physical disks -- is asking for glitches, though even then I'm not sure some haven't happened entirely on their own. The glitches typically aren't simple 0dB spikes, either. They last up to 1ms and therefore require hand editing + prayer for a locally repetitive waveform.
• The above on steroids: sometimes a recording will contain 20-50 minutes patches turning into pure 0dB noise. When capturing unattended, this is actually more frequent than "glitching" as above, to the extent that for every capture I do, the first step is to extract the sound, load it into Sound Forge, and zoom out to look for ruined sections.
• Similar: one time at a recording session, the stream switched suddenly to 22KHz (from 96KHz). Of course, the host app didn't know this and so we got the last hour recorded in chipmunk-style. It was fixable (throwing away lots of quality), but still not pleasant. Listening to the raw feed we determined that loading up Finale on the laptop was the cause, but that's no excuse for supposedly "professional" drivers.
• Continuing in spirit: the drivers just don't like multiple apps. Finale doesn't even use audio, I don't think (I have an M-Audio external 1x1 box handling MIDI, not the MS Mapper), but it causes problems. If I have Vegas open I can't play a movie in the background. Etc.
• Complete application failure: Sonar and its much-vaunted WDM architecture doesn't work at all. It was great with my AP2496, but with the Duo I get "device does not support selected format" on startup and a play button that does nothing. (Stepping down to 16/48 gets rid of the message but not the problem.)
• And new for today, another weird one: during a capture the audio skipped almost 10 seconds ahead, just randomly. This was completely apart from the normal A/V synchronization, as no frames were dropped and no resampling performed. It may have been a Time Warner thing, stand by... update: the audio sped up drastically for a few seconds (sort of like a short form of the sample-rate glitch), causing it to be ahead. hopefully this was isolated to this one piece of the cap that was before the show... update #2 it definitely drifts further out (and in?) of sync with time. assumesamplerate(47000) showed promise in a short clip but we'll have to wait for a full test-render to see for sure... update #3 worked many hours with a Q2 render (which itself didn't complete instantly) -- messing with offsets and resamplings always broke another section. confirmed that even after the glitch the rate was nowhere near constant; for example, setting the offset to get frame ~20,000 in sync might result in 10,000 AND 30,000 being behind the video. and of course, my backup capture had the 0dB bug of doom
• Another new piece of fun fun fun, 7/12/04 - the box starts & stops working randomly. When it's in a "mood," Windows cannot find an output sound device. Cycling the power does not help. Then someone will ICQ me hours or days later and I suddenly notice that it made a beep. Standalone mode (monitoring inputs with headphones) seems unaffected, so tally yet another USB fuckup.

*When I first starting collecting these data, the left output was down "only" -7dB and the outputs varied thusly (directly copied; too weird to paraphrase): The line input's left channel is down about -7dB in pro mode. in consumer mode the right stays the same but the left is now 7dB *louder* than it. tested with two sources. (perhaps the left is actually working correctly and the right is not?)


Back to StuffToDo