Go Back   Cockos Incorporated Forums > REAPER Forums > Recording Technologies and Techniques

Thread Tools Display Modes
Old 11-04-2010, 09:53 AM   #1
Human being with feelings
zappsunzorn's Avatar
Join Date: Feb 2008
Location: Virginia Beach VA
Posts: 2,269
Default The McGurk Effect, a short video clip showing how our eyes can deceive our ears


Something to remember when comparing gear.
zappsunzorn is offline   Reply With Quote
Old 11-04-2010, 11:53 AM   #2
Human being with feelings
bls's Avatar
Join Date: Jun 2009
Location: Tx
Posts: 698

That's pretty cool. I had no idea about that.
bls is offline   Reply With Quote
Old 11-04-2010, 10:01 PM   #3
Human being with feelings
karbomusic's Avatar
Join Date: May 2009
Posts: 24,918

I played a private party in a living room tonight (acoustic). Lots of people moving, talking, lips moving etc. It was hard to ignore all the sounds and activity and play and my mind kept wandering off the music. Especially since everyting was brightly lit, IE: not a stage with a dark audience as usual. Only closing my eyes fixed the issue. Wonder if they are connected, I always play better with my eyes closed.

If it requires a null test to find it, it is by definition minuscule.
karbomusic is online now   Reply With Quote
Old 11-04-2010, 10:18 PM   #4
Human being with feelings
Join Date: Nov 2009
Posts: 642

Wild, LOL!

Makes a good case for closing your eyes once in a while when working with DAWs and VSTs that have a highly visual interface I guess.
flmason is offline   Reply With Quote
Old 11-05-2010, 03:31 PM   #5
Human being with feelings
Join Date: Aug 2006
Posts: 2,012

Originally Posted by flmason View Post
Wild, LOL!

Makes a good case for closing your eyes once in a while when working with DAWs and VSTs that have a highly visual interface I guess.
Makes a REALLY good case for:

1. disabling "draw media peaks" on recorded audio except when necessary for certain kinds of edits and navigation, and

2. Using generic UI mode for plugins, or at least "knob"-type UIs as opposed to a graphical representation, and

3. Avoiding spectrum analyzers as an aid to subjective sound-processing, and reserving them for more technical tasks such as noise removal, etc.

Something I'm increasingly considering is that a huge portion (maybe even all) of the widespread preference for hardware/analog setups (including my own) is that a mixing console and rack gear forces better work practices. Metering is the most obvious thing: analog meters are calibrated to show a target AVERAGE level, calibrated based on the amount of clean headroom available (i.e. 15 or 20dB below subjectively objectionable clipping). DAW metering, on the other hand, kind of totally sucks because the only obvious "target" IS the point of clipping (and clip detection is pretty piss-poor on almost all interfaces and DAWS, for a lot of reasons).

No matter how much you know about this, no matter how thoroughly you understand audio, it is still a challenge to avoid the siren song of metering based on PEAK level instead of average level when using digital. And almost impossible if you DON'T fully understand it. So people sit down in front of a DAW and push all their levels up to the point of DETECTED clipping (minus one dB or whatever), and now they're pushing ~10 volts through an analog front-end designed to handle 1.23 volts, so the front-end of the AD converter is crapping out, plus undetected inter-sample clipping at the AD, plus whatever circuit-protection or whatever is built into the AD to minimize overs is doing whatever it does, plus the resonant peaks of the sharp cutoff filters in front of the AD are driving the extreme lows and highs into distortion... and what do you know, it sounds brittle, harsh, flat, ugly (especially since they are doing the same thing on the output side)...

Then they sit down in front of an analog console, where the meters aren't showing them PEAK level, but instead showing zero as a TARGET to get the needle bouncing around 1.23 volts average, where the system has 8 or 16 times the power-handling of "zero dB" reserved for headroom, and what do you know? It sounds "smooth", "warm", "3-D"... all that analog goodness. Gosh, analog sounds better than digital...

Then we start processing with our DAW's super-powerful parametric eq with a pleasant and tidy little graphical display of the frequency spectrum, and we start making our cuts and boosts, and whoa-- that can't be right--look what we just did to the frequency curve! That graph basically shows that we killed all the highs!(or lows!) Or that we just overlapped our last cut with a broader cut that cut it again! We want this track to be sparkly and airy (or deep and powerful)! You can't put a curve that looks like that on it!

So we start second-guessing our decisions based on how it "looks". (That compressor knee can't be right-- look how it looks on the graph!) We start "thinking" about how it's supposed to be. It's very hard to ignore those visual cues. Even if we're determined to make decisions by ear, how do you ignore that little window that says you've just cut out all the highs? Moreover, I am increasingly convinced that even many supposedly floating-point nigh-infinite-headroom plugins crap out with high internal gain. Doing basically the same mix twice, except doing one where I keep the average per-track signal level at -20dBFS or so, I'm finding way better results with the "quiet" mix, and ugly, "digital" brittleness when I disregard gain-staging even when it's supposedly irrelevant.

Then we sit down in front of a mixing console and a rack of hardware, and all we have is knobs, and often cryptically-labelled ones at that. We start making decisions based solely on how it sounds, and what do you know? Our mixes our sounding better than ever! I guess analog DOES sound better!

The more we alternate between the two, the more we try out new super-duper digital stuff that's supposed to be like analog, the more certain we become that, while it might get close, it's just not quite there. To prove it, we set up a shootout. We set a digital system and an analog system to the exact same settings and track or sum or process something through it... and hmm... there doesn't seem to be much difference. In fact, the lower-noise digital might actually sound a little crisper and cleaner and more realistic... I know, it must be that the test is wrong!

So the answer must be that you can't just do some clinical A/B shootout because (check all that apply: "you have to do the whole record from start to finish to get the cumulative effect", "analog is organic/musical and you can't just isolate the differences like that", "the magic of analog is how it overloads/saturates/distorts and just because you can get a more clinically-accurate recording from digital doesn't mean you can make a great album in digital", "until the Beatles reunite and record an album that I like better than I like Sgt Peppers in all-digital it proves analog is better", "lol newbs think U can get neve result form mackie mbox and pod I have 2 hunderd thousand les pual and TONE"...)

Anyway, definitely food for thought.

Last edited by yep; 11-05-2010 at 03:37 PM.
yep is offline   Reply With Quote
Old 11-05-2010, 04:08 PM   #6
run, megalodon
Human being with feelings
Join Date: Apr 2010
Posts: 2,860

Although I can't imagine that the effect from the original post would translate that well to the daw world, since I think processing speech has it's own special ways of working, I'm all for discussion about the visual impact of daws. I think this is one of the most important usability issues for daws and I don't think anybody really cares enough. Concentrating on how an eq curve looks rather than how it sounds is definitely something that I have been guilty of doing, it'd be nice to get and option to remove the display from reaeq.

I'd like to take this opportunity to once again plug some visual related FRs that are pretty important to me (for Reaper).

I have a running FR for a toggle "display peaks" action in my signature, and also created an FR to expand Reapers ability to use generic UIs for plugins when I was new to the forum, but it didn't really take off: http://forum.cockos.com/showthread.php?t=58962

Most importantly, see the link for visual simplification options in my sig. If all of these options were implemented the complexity of the interface could be greatly reduced. It's kind of sad to see all the votes against it though. I guess those people think that the visuals are not affecting them.

Last edited by run, megalodon; 11-05-2010 at 06:12 PM.
run, megalodon is offline   Reply With Quote
Old 11-05-2010, 04:54 PM   #7
Human being with feelings
kelp's Avatar
Join Date: Sep 2009
Location: Milwaukee, WI, USA
Posts: 395

To yep's point -- can REAPER's track meters be made to behave like VU meters? Ala PSP's Vintage Meter?
End of Kelp.
kelp is offline   Reply With Quote
Old 11-05-2010, 06:10 PM   #8
Human being with feelings
Join Date: Mar 2010
Posts: 4,713

Not yet they can't, but I would absolutely love it to be an option.

I'm pretty sure it's part of Dannii's epic metering FR, and likely to appear sometime during version 4
timlloyd is offline   Reply With Quote
Old 11-05-2010, 09:18 PM   #9
Human being with feelings
Join Date: Aug 2006
Posts: 2,012

Originally Posted by kelp View Post
To yep's point -- can REAPER's track meters be made to behave like VU meters? Ala PSP's Vintage Meter?
Yes and no. You can make REAPER's meters show pretty much any scale you like under Options>Preferences>Appearance>VU Meters/Faders. There are plenty of options, and you can customize the meters pretty freely.

However, that does not, strictly speaking, solve the problem. Let's forget about digital for a minute and pretend that the whole world is analog...

In this all-analog world, "clipping" and "distortion" are somewhat subjective concepts, and not always altogether bad ones. Moreover, "noise floor" is not even necessarily a rigidly-defined concept, since the gain and processing you apply can change it.

A dedicated designer of high-end analog gear is looking to get the best-sounding, widest "sweet spot" they possibly can, where the signal is both well above the intrinsic noise floor of analog circuitry, but also below the threshold of obnoxious and ugly "clipping" from grossly-overloaded circuitry. (this plus all the usual stuff of accurate frequency response, dynamics, etc).

Now, there are basically four measures of "accuracy" in audio reproduction: frequency response within the range of human hearing, dynamic response, noise, and total harmonic distortion. Unless you genuinely believe in the outright supernatural, every effect that any processor can have can be measured in terms of strict accuracy in relation to the above four.

High-quality digital (once encoded as digital) can be safely said to be transparent (i.e. perfect) according to all four measures. That is, every digital copy is identical to every other digital copy. There are, broadly speaking, two contested points: one is the process of conversion to digital and back to analog, i.e., whether anything meaningful is "lost" in that process. The second is whether digital processors/amplifiers/summing stages/etc are capable of achieving the same subjectively-pleasing results through calculated processes that analog devices have achieved through trial-and-error design.

Let's skip over the contentious worlds of, say, digital guitar effects, synthesizers, and so on, and let's suppose for the moment that such deliberately imprecise and distorted devices are best left to trial-and-error approaches (whether analog or digital). Let's stick with the basic task of actually RECORDING GOOD SOUND, and let's pretend that we still live in an all-analog world when it comes to recording technology (whether or not we are recording guitars through PODs or VST synthesizers, or whatever).

So in this all-analog world, we need to design mixers, recorders, preamps, etc, that have some kind of usable input and output meter. We need to give the user some sense of when the signal is too high (i.e. when it is "distorting", or in danger of it), and when the signal is too low (i.e. when it is too close to the intrinsic and unavoidable noise floor of random electron movement in copper wiring).

Neither of these are necessarily "hard" measures. Noise is an analog system is somewhat dependent on gain, wire gauge, etc, and "distortion" creeps in early. And very often, the bigger and more powerful the wires, power supplies, etc are (i.e. the more "headroom" before distortion), the more noise is introduced by the system itself.

In this purely analog world, we have a whole bunch of both objective and subjective trade-offs to consider, even disregarding anything like "magic" or "mojo". The more headroom and clean gain we supply, the more we increase and amplify the intrinsic noise caused by the random movement of electrons in the system.

The very best designers and manufacturers of analog systems often do a bajillion little things, from limiting the extreme low- and high-frequencies, to devising circuits that saturate or "clip" only gradually, and often pleasingly, so that brief "clipped" portions sound more like extra "punch" or "crunch" instead of sounding like fizzy trash, and so on. Their objective is to create the widest and most flexible "sweet spot", so that you could, for example, plug in an electric bass and get a full, deep, transparent sound with tight, solid lows, OR you could push the device a little harder and get a thunky, churning growl and burp on the attacks... this may very well be a nonlinear, frequency- and gain-dependent system response that is very different from "perfect" digital gain.

Back to the point about metering, the meters on such high-quality analog devices are necessarily imperfect and imprecise. In fact, they are very often calibrated by the user-- that is, you plug in your electric bass (or mic, or whatever), and turn it up until you think it's "clipping", then set your "zero" level some 14 or 20 or 24 dB lower than that.

Analog "clipping" is not the hard cutoff that digital clipping is (and even that is poorly-reported, more often than not). Analog gradually flattens and saturates. Often, it sounds best when pushed a little past the point of "perfection": with a little bit of high-quality saturation, drums get punchier, vocals get bigger, guitars get louder, horns get brassier, bass gets growlier and fatter, piano gets plunkier and more resonant, strings get woodier, and so on.

A bit of saturation tends to suppress the "note" and bring out the detail and immediacy. Analog circuits often sound "better" when pushed harder, plus they get further away from the noise floor.

Coming back to the original point, there is not the same, obvious, "clipping" point with analog that there is with digital. If I'm setting up an analog studio for the primary purpose of recording, say, punk/metal/industrial music, then I might want to get a console with a wide, saturated "sweet spot" like a Neve or API that will give crunch and punch to the drums and breath and fire to the vocals and growl and presence to the bass and so on, and the fact that it becomes somewhat inaccurate, distorted, and nonlinear might actually be an asset. OTOH, if I'm recording something like top-40 pop or orchestral music, then I might want to go with an extremely linear high-headroom system like an SSL, whose saturation "sweet spot" is much airier, deeper, more transparent, and "lighter" than the thick, heavy-iron sound of Neve transformers.

Getting back, by layers, to the original point, in the above all-analog world, the threshold of "clipping" is unavoidably a somewhat subjective one. There is no obvious and fixed point at which the gain knob on a Fender Twin goes from "clean" to "distorted"... it's a continual progression. Give 100 Fender Twins to 100 guitar players and ask each of them to mark the spot where it becomes "distorted", and you'll get 100 different answers, and half of them will require an outboard distortion pedal.

There is no bright-line mathematical distinction that separates analog "clipping" from softer "harmonic distortion." Where that threshold is set is a matter of opinion based on subjective and objective measurements. And more importantly, how close to "clipping" the average signal level can get while still sounding "good" varies quite a bit from one device to the next.

Now, with digital, there are not the same kinds of gray areas. Digital is generally either perfect, or else clipped in pretty ugly ways. With a digital system, you don't get sonic improvements by "pushing it hard". If you want a saturated or distorted effect in digital, you have to get the sound right before converting to digital, or else do it deliberately with some kind of plugin.

Going all the way back to the original post, the serious problem with digital is not that it's flawed, but that digital METERING is flawed, on two counts:

- one is that digital "clip" detectors are almost universally imperfect, and have tremendous potential for "missing" a lot of digital overs, for a lot of complicated reasons.

- two is that, even if the digital system and metering is internally perfect, there is an analog front-end before the AD converter, and an analog back-end after the DA, and THOSE systems are very often designed/calibrated to approximate analog dBu or dBV mixer/recorder levels, which are designed for optimal sound at something like 20dB BELOW what the digital meter shows as "clipping." So even if your interface or DAW has perfect digital clip-detection (and none of them do), all of your metering is still ignoring the poor, overloaded analog front-end of your AD converter (never mind your preamp, which was designed to output 1.23 volts steady-state, not 9+V...)

Answer is to PAY ATTENTION TO GAIN-STAGING. When you change a knob, you change the sound. Forget the meters and follow your ears. Target your recordings for -15 or -20dB or so per track, and KEEP THEM THERE throughout the plugin and processing chain.

Stop looking for meters to tell you what to do. Set your record levels too low to clip, even accidentally, and then let sound be your guide. If you are turning down your track faders, or fighting with the master volume, you're almost certainly recording too hot.
yep is offline   Reply With Quote
Old 11-06-2010, 03:29 AM   #10
Human being with feelings
beingmf's Avatar
Join Date: Jul 2007
Location: Jazz City
Posts: 3,943

Feel free to comment on this thread of mine (of 2008)...
beingmf is offline   Reply With Quote
Old 11-06-2010, 06:24 AM   #11
Human being with feelings
ugh's Avatar
Join Date: Sep 2007
Location: earth
Posts: 1,379

Originally Posted by zappsunzorn View Post

Something to remember when comparing gear.
an interesting video, thanks for posting the link.
ugh is offline   Reply With Quote

Thread Tools
Display Modes

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off

Forum Jump

All times are GMT -7. The time now is 09:10 PM.

Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2019, vBulletin Solutions Inc.