|
Post by starandchlorisse on Oct 25, 2016 14:06:55 GMT
Is it safe to say whether or not clipping has occurred by looking the waveform ? How you do that ? And then edit it out ?
I have some nature recordings and was not careful enough about the levels.
|
|
|
Post by Ethan Winer on Oct 25, 2016 16:26:25 GMT
If you see 2-4 samples (or more) in a row that are at maximum level, it's safe to assume there was clipping. Note that clipping can occur earlier in the chain, before the converter's input, with the same appearance: multiple samples in a row, throughout the file, with some flat/maximum level continuing for several samples. I learned this the hard way many years ago when I first started recording in software. I was recording violin overdubs for a friend, and I noticed the record level was a little too hot. So I lowered the record level in the recording program. All that did was put the clipping at a lower level. In other words, the converter's input was already clipped, and the software recorded that clipped wave at a lower level. So Rule #1: always set software record volume controls to maximum. Or zero if gain is available - you don't want gain either!
|
|
|
Post by starandchlorisse on Oct 25, 2016 17:08:31 GMT
"always set software record volume controls to maximum." you mean to set the recording level high- 0 db- so you can get the maximum volume from the source ? Correct? many thanks
|
|
|
Post by Hexspa on Oct 26, 2016 6:00:46 GMT
I was recording violin overdubs for a friend, and I noticed the record level was a little too hot. So I lowered the record level in the recording program. All that did was put the clipping at a lower level. Ha. Guilty. So, my understanding of levels is like this: In a standard DAW, an armed track will instantiate with it's fader at unity (0dB). Leave it there. Ethan is mentioning a software mixer. Assuming that's not the DAW, but a software component to your audio interface, leave the faders at unity there also. Any gain change within an interface's software mixer (such as CueFX by MOTU) will be the equivalent of turning the knob on the interface's hardware. Whenever I record anything I turn the pad function off and turn the gain all the way down. I aim for an in-software level of -10dB peak when the source is playing as loud as it's going to (in the case of a live source/performer). That allows for headroom during recording and room for processing later. Plus, from what I've heard, some analog-modeled plugins operate better at a lower-than-0dBFS level. That optimal level corresponds to the analog 0dB which is actually -18dBFS (depending on calibration, manufacturer, etc). [There's more you can read about this but Bob Katz makes a strong case for it.] Long story short, use your preamp gain to get the correct level - peak somewhere between -20dBFS and -10dbFS depending on your preferences and needs. If you need, use the pad function so you're not clipping your preamp (which, in my experience, only happens when I'm clipping within the DAW as well). Additionally, you can use 32bit float to avoid clipping in software (I've heard, haven't tried). Keep your levels low, make sure to use a limiter with a half dB headroom when mastering and you should be fine. To answer your question - I'd say it depends. Many waveforms look clipped after mastering. They may have been or they may have been squashed by dynamics processing. If you know that no limiting has been applied then the square wave appearance may indicate clipping. Or, it may indicate you've recorded a square wave! So ya, common sense. Thanks, -m PS - You can do the "audio engineer move" and raise the level to where you want it then back it off a touch - that's good advice in general for setting anything from EQ moves, gain staging or distortion amounts. PPS - There's also something called "inter-sample peaking" or ISP. It's up to you how much you want to consider this. Often you'll need oversampling to detect it. I've found there to be ISP on Steely Dan's Aja (title track, towards the end) and it's considered one of the best-engineered albums ever.
|
|
|
Post by Ethan Winer on Oct 30, 2016 18:28:56 GMT
By "software mixer" I meant the volume control levels in the recording software. Volume settings in the control panel for a sound card might or might not be after the converter. It might actually control the converter hardware's input level, and in that case it would prevent clipping. But once the digital audio gets to the recording program, you can't correct the damage from clipping earlier in the converter.
As for 32-bit floats, as explained in my book all modern recording software uses 32-bit floats (or sometimes 64 as an option though that's just silly). So once the audio gets into the software safely, it's impossible to damage anything with levels that are too high or too low.
|
|
|
Post by Hexspa on Oct 31, 2016 18:20:17 GMT
By "software mixer" I meant the volume control levels in the recording software. Volume settings in the control panel for a sound card might or might not be after the converter. It might actually control the converter hardware's input level, and in that case it would prevent clipping. But once the digital audio gets to the recording program, you can't correct the damage from clipping earlier in the converter. As for 32-bit floats, as explained in my book all modern recording software uses 32-bit floats (or sometimes 64 as an option though that's just silly). So once the audio gets into the software safely, it's impossible to damage anything with levels that are too high or too low. There are certainly several gain stages. I think on my Ultralite, the input is pre converter. Anything else I'm unsure. "...all modern recording software uses 32-bit floats... so once audio gets in... it's impossible to damage anything..." Ok. Well, as per www.ableton.com/en/manual/audio-fact-sheet/ section 31.2.4, Live does use 32 and 64 bit processing. Maybe I'm misunderstanding (and I don't even understand that reference) but you can definitely clip audio in Live. These are clips recorded from the output of a track with Analog on it outputting a sine wave at -17dB. The left clip had 35dB gain applied, the right had no gain added. I adjusted the clip gain so they match in appearance but one's clipped and one isn't. That's with 24bits selected as the project depth in preferences. But I may be missing something. -m
|
|
|
Post by Ethan Winer on Nov 8, 2016 17:53:24 GMT
As I said, audio can clip going in or coming out of the software. But inside the software it won't clip other than with a plug-in that itself has volume limits. I don't know Ableton, but if it uses 32-bit floating point math as do all other programs, something else is going on in your test. How did you apply the 35 dB of gain? And how did you reduce it back down to -17?
|
|
|
Post by Hexspa on Nov 8, 2016 23:32:28 GMT
Could've just explained it but.. in this day of video and high speed connections.. youtu.be/EXoLgMD9XTMThe only thing that's weird is that the sine wave clips at a lower level yet there are still the substantial peaks in the sample. -m
|
|
|
Post by Ethan Winer on Nov 10, 2016 16:25:22 GMT
To really get to the bottom of this I'd need to do a Skype video session with you to be able to ask questions back and forth to understand what's going on. It looks to me like you raised the gain 35 dB somewhere, but I can't tell where? On a buss? On the track?. The overload was obvious on the VU meters, but then you routed that back to a track with the gain lowered again. So I guess it clipped somewhere, but I don't know Live well enough to understand where you patched in the 35 dB of gain and how that affects things.
|
|
|
Post by Hexspa on Nov 10, 2016 20:34:42 GMT
If that would help you I'm available for that. Please feel free to PM details.
Here are notes on the video: 1. I instantiated a virtual instrument on a track followed by a plugin called Utility. Incidentally, that track is going to a Group (buss) but it's irrelevant. 2. I then routed the output of that track ("Mel" in the video) directly to an audio track called "resampling". 3. First I played a sine wave which peaked around -17dBFS and recorded it to the resampling track. *Please note I did not use the proper "resampling" feature in Live to accomplish this. That would route the audio through the master buss. I just set the input of the audio track called "resampling" to the output of "Mel". It's directly routed. 4.After recording the initial sine signal, I then applied 35dB of gain using the Utility plugin which I had inserted directly after the Analog VI on the "Mel" track. That means the output of Mel was boosted prior to the track's output and therefore prior to the input of the "resampling" track. Redlining on the meters happened not only there but also on the output of the Utility plugin but that wasn't shown in the video. Exactly what type of math is used within Live's Audio Effects is unknown to me. I guess if it's anything less than floating point then the clipping occurred there. I then recorded the boosted signal from "Mel" on the "resampling track. This signal peaked around +17.82 dBFS both on the output of Mel and the input of Res confirming the direct routing. 5.Lastly, I turned off Warp modes on the clips (probably also irrelevant) and then used the Clip Gain function on the second, boosted, recording so it'd fit in the waveform viewer for visualization purposes.
The weird thing is that, though the sine is clipped, there are also huge peaks which go beyond the clipped level.
All I know is keep the levels down and I won't clip. However, it's certain that I can clip audio in Live by driving levels up.
-m
|
|
|
Post by Ethan Winer on Nov 13, 2016 20:44:38 GMT
Try this:
* Record a sine wave at some level below 0 dBFS, and put that file onto a track that's routed to the master output bus.
* Insert a plug-in onto the sine wave track that raises the gain enough to cause severe clipping.
* Lower the gain of the master bus by the same amount, or use a plug-in if the master bus has no volume control.
Is the sine wave at the master bus output clipped? If so, then you have a legitimate complaint to bring to Ableton. If not, then something else is going on in your other test.
|
|
|
Post by Hexspa on Nov 14, 2016 2:07:20 GMT
Hmm. Doing what you said when using the resampling function produces no clipping. Performing that test when recording the master buss directly produces no clipping. But, when receiving audio directly from the boosted, unattenuated track, I get clipping. So, idk. That's the screenshot with the results of the three passes. Interestingly, those apparent peaks in the bottom waveform disappear when zoomed in upon. As a side note, though the red track appears to say "resampling" that's because the name of the first track is "resampling". That is not the proper Resampling function of Live. Thanks for the test. -m
|
|
|
Post by Ethan Winer on Nov 16, 2016 17:48:08 GMT
Yeah, I have no idea. Does the Ableton manual include a flow-chart? The SONAR docs show a highly detailed block diagram of signal flow through the program, duplicated on Page 140 of my Audio Expert book to explain the concept. So that's the sort of thing that would tell you what's going on in your program.
|
|
|
Post by Hexspa on Nov 17, 2016 10:26:10 GMT
I saw one somewhere but I don't think it's official. There is this: www.ableton.com/en/manual/routing-and-i-o/ which describes how audio moves around and this: www.ableton.com/en/manual/mixing/ which states: "Because of the enormous headroom of Live’s 32-bit floating point audio engine, Live’s meters can be driven far “into the red“ without causing the signals to clip. The only time that signals over 0 dB will be problematic is when routing to or from physical inputs and outputs, like those of your sound card, or when saving audio to a file." ...but that's obviously false in this instance. That is unless, since my project is set to 24 bit, the Utility plugin which provides gain is internally clipping. Maybe if I set the project to 32 bit then the plugin will operate using floating point math. I'd have to check. No apparent qualification has been made regarding inter-track routing or the internal processing of any native plugins afaik. But you see it. It's replicable and I'm willing to share the project. -m
|
|
|
Post by Ethan Winer on Nov 21, 2016 20:40:17 GMT
Well at this point you should probably ask Ableton support. I'll be interested in hearing what they tell you!
|
|