Newbie at drum recording
All:
I miked up 8 mikes for the drums, then ran them through a Phonic MU244X mixing board. Each and collectively did not clip, but I made them as loud as possible. Then I ran 2 stereo outs into a Fostex recorder. I set the levels there so there was no clipping.
When I loaded it into n-Track and set the sound level to about 0. The clip indicator goes crazy on every beat, yet I do not hear any buzzing, popping, or distortion in general.
Am I really clipping?
Your recommendations are always very good – any suggestions?
Thanks in advance,
Beth
I can test this later tonight with some custom VST coding…but I think it just represents the amplitude hitting 1.0. Not necessarily clipping, but probably a very decent indicator. Since the amplitude cannot surpass 1.0, there is no other way to tell.
Digital Audio myth number 17: Clipping always means audible distortion.
Soundforge has a nice tool that shows all clipped peaks, and it has another tool that shows all audible clipped peaks, which I think it defines as any clip related plateau longer than 5 samples.
And there’s a big difference.
That doesn’t mean you should always ignore inaudible clips, processing the audio can make inaudible clips audible, but it does mean that you can’t always trust your ears.
Correct, you will not always hear clipping on fast transients that are clipped (like drums). So yes, you really are clipping and clipping = bad.
Scan the file for the number of overs. You can also visually inspect the file for them. Also, remember that the philosophy of recording as hot as possible that applied to tape does not apply to digital in the same way, you don’t really want to set it to record quite that hot - there are no tape saturation benefits to digital, e.g. Finally, even if you can’t hear them, that doesn’t mean that they are not having some kind of psycho acoustic effect - being able to “hear” these sorts of things has lots of diffferent meanings, it turns out.
Whoa, boys, we don’t have nearly enough data to answer correctly yet.
First, tell us WHICH clipping indication you’re getting. Recording meters, playback meters, or plugin meters?
For playback and plugin meters, “clipping” means “over 0dB” but there is NO sawing off of the waveform. It just means that a floating point value went above 1.0, which WOULD clip if converted to fixed point and sent to your soundcard.
For recording meters, the highest value possible is 0dBFS, which is either the highest positive or negative number in the bit width you’re using. A “clip” is generally registered when we see more than one of these in a row – the usual rule is 3 in a row should be called a clip. But as mentioned above, if there’s just one spike like this, it’s not audible. If there are a bunch of waves that clip like this in a row, it sounds like crap.
But you’re probably not talking about record meters, because those don’t move in this case (unless you have something goofy going on, like you’re using a laptop with built-in mic, and you’re monitoring the mic – you can test this by tapping at any time).
Also, note that it is possible to clip on record and for n-Track not to notice it. At least, it happens on my system, using n-Track V3. Not sure why, but I assume it doesn’t have time to test every single sample or something like that. This isn’t a big problem for me, since I generally leave enough headroom so any clipping I might get would be the ignorable kind (one spike that barely goes above zero).
If you’re seeing clipping on playback, in the playback meters, it’s because you’re doing a good job of recording! When you play all the tracks back with the faders set to unity gain (0dB), the result is clipping. Just lower the faders! (You can also reduce the master fader if you like.) For drums, I like to send all the drum tracks to a Group (right click on the track – in the wave display – and select Output To, and pick Group 1). Then you can control all the drums with one fader.
Anyway, enough speculation. Tell us WHERE you’re seeing the clip indications!
As Learjeff suspected I am seeing the clipping indicated on the playback VU-Mmeters. It also appears as red reference markers on the timeline above the tracks. I have each drum track set at 0 on the volume. If I lower the 2 stereo drum tracks to -2.5 each, the clipping stops. When I start add guitar tracks back in, the cumulative volume indicates clipping again. Again however, to my ears, I am not identifying any distortion.
I think I will experiment with some of Learjeff’s suggestions. Also, if I have to lower the cumulative volumes on all the tracks to avoid this, I am thinking it may not have that great an impact on overall volume for the CD. We are going to have it professionally mastered, and it is my understanding that, that will increase the volume on the final version.
I wish I was more skilled at explaining this, but I don’t have the experience of you folks yet. All your suggestions are truely appreciated.
Thanks for your help,
Beth
I still say you need to look at each individual file, given the unorthodox way you are recording them, to make sure.
Learjeff, If I have clipping during playback you mention to lower the faders. Would you recommend using a limiter or compression to eliminate the clipping?
| Quote (leker @ June 14 2006,12:23) |
| Learjeff, If I have clipping during playback you mention to lower the faders. Would you recommend using a limiter or compression to eliminate the clipping? |
If using a compressor or limiter for the sole purpose of not clipping, then 'no'.
Just to state the obvious - the sum of several tracks, all of which are below clipping, regularly results in clipping in the sum. How much depends on the whether the peaks in the waveform coincide.
To move on to the inobvious. The audibility of clipping depends on the severity of the clipping, its duration, and the amount of change in the spectrum of the resulting signal. A signal with a simple spectrum (such as a sine wave) makes detecting distortion relatively easy since the additional harmonic content added by the alteration of the waveform is not concealed by other harmonic content. In a complex signal like a cymbal hit, it is more difficult to identify moderate amounts of distortion products since the the presence of such a large number of frequencies in the undistorted signal conceals the presence of the additional frequency components not present in the input.
These additional components will be audible if they are louder than what is called a “masking curve” in psychoacoustics. Any sufficiently loud tone will obscure frequencies in its vicinity. The level required to make a nearby tone audible is described by the masking curve. It is shaped so that frequencies above the masking tone must be fairly loud to be heard, with the audibility threshold diminishing gradually as the frequency increases. Frequencies below the masking tone become audible more quickly (the masking curve is steep). The gradual slope of the curve above the masker means that 2nd harmonic distortion must be quite loud to be heard, 3rd harmonic is audible at somewhat lower levels, 4th lower still, etc. High-order harmonics are quite audible at very low levels.
Loud signals can also have temporal masking whereby the masker obcures signals which follow closely and even those which precede the event by a few milliseconds. In essence a sufficiently loud and sudden signal can make you “forget” a sound you have already “heard”.
Add to this the fact that even when the signal is theoretically audible, an individual’s ability to recognize distortion is also a function of their familiarity with the undistorted signal and it is not surprizing that modest amounts of distortion can be inaudible, especially to untrained listeners.
You can easily determine whether the signal is clipped by mixing down the tracks then importing the mixdown into a new song and visually inspecting the waveform in the vicinity of the clip indication. If there are flat tops to any of the peaks clipping has occured.
The abillity of one sound to conceal another is the basis of most of the “lossy” low bit-rate codecs such as MP3, WMA, AAC, Ogg, etc. The fundimental concept involves first analyzing the signal to determine which portions of the signal are likely to be audible then only transmitting data which allows the reconstruction of the audible portions while eliminating the data which would otherwise simply describe inaudible portions of the signal. Why transmit it if you can’t hear it?
Of course this process is not perfect and to achieve very low bit rates you have to discard some “not very audible” stuff along with the inaudible and do other tricks that may compromise the reproduction of the clearly audible parts as well.
Jim
<!–QuoteBegin>
| Quote |
| Learjeff, If I have clipping during playback you mention to lower the faders. Would you recommend using a limiter or compression to eliminate the clipping? |
This is one of the primary uses of limiting and compression but if someone else will be mastering your project it may be best to allow them to do the compression required for loudness maximization.
Compression can be a great tool for getting individual tracks forward in the mix by altering their dynamic characteristics but it has a definite effect on the “sound” when used aggressively. By playing with attack, release, ratio and threshold you can get a variety of effects that are useful to creatively modify the sound.
Limiting is specifically intended to prevent clipping and when done correctly has less obvious consequences than compression. A limiter is basically a compressor with a very fast attack and release time and a high compression ratio (it typically also has a high threshold). It is intended to act on just the peaks of the signal, allowing a higher average level without much sonic consequence. If overused it can be quite audible and will in fact introduce distortion. Actually any change in the shape of a waveform introduces some amount of distortion and very fast limiting can be thought of as “soft” clipping which generates fewer high-order distortion products. Since high-order products are more audible than low-order distortion (see above) it has fewer negative effects than “hard” clipping. Slowing the attack and release of the limiter reduces the waveform modification but may allow some peaks to “hard” clip before the limiter fully reacts.
A compressor with very slow attack and release works as an AGC (automatic gain control) and mimics manual volume adjustment. In that configuration it will do little to stop clipping. Some AGCs have fast attack and very slow release which works better to supress clipping but has severe effects on musical dynamics.
Jim
| Quote (TomS @ June 14 2006,12:03) |
| I still say you need to look at each individual file, given the unorthodox way you are recording them, to make sure. |
Is my recording method unorthodox? I am always open to suggestions. Here are my limitations. I have a Fostex recorder with 4 inputs A, B, C, D and tracks 1, 2, 3, 4, 5/6 & 7/8.
I wanted to mic up 8 drums so I ran them through the mixer, then out in stereo (this includes 2 overhead condersor mics). I connected the 2 stereo outs, into iputs C & D, recording onto tracks 3 & 4. I used inputs A & B to lay down scratch guitar and vocal tracks 1 & 2.
After moving all tracks to the PC in n-Track, I bounced drum tracks 3 & 4 in the Fostex recorder over to 5/6. This freed up tracks 3 & 4 to add permenant guitar and vocals. I can then discard the "scratch" guitar and vocals. I keep doing this process, recording individual tracks and adding them into the pc, keeping only the ones I like.
The result is, synced up individual tracks for all vocal, instruments, etc. Only the drums are 2 stereo tracks due to the limited inputs on the Fostex.
If anyone has other options you think I should consider I would greatly appreciate it. I know there is always a better mouse trap out there. :-)
Thanks again, Beth
Indeed, the gun has been jumped. My assumption (ass-u-me) was that a single track was soloed and the channel and master faders are at 0db. If the playback meter clips, indeed your file is clipped. If the recording meter clips … ever… you have digital clipping on the input.
Beth,
What is “unorthodox” is using an external recorder rather than going straight into the computer. What you are doing is exactly correct if you are limited to 4 input channels but have extra playback channels. Your approach may also be more reliable than depending on a general purpose computer to not crash or glitch during a recording.
How do you move tracks from the Fostex to the PC?
Obviously you can get more inputs by buying appropriate hardware, either in another dedicated HD recorder or a multi-input card for direct recording to the PC. That would let you multi-track the drums as well but I don’t see anything odd about your method other than the “indirect” recording method.
Jim
I believe she already said she moves them using a flash card. No worries about that Beth. In the long run you’ll probably want to go direct to the computer (and ideally, one input channel per mike), but let’s bypass that for now.
Jimbob is right that it’s best not to compress the mix if you’re going to have it mastered. Mastering engineers hate that because you’ve then done part of their job, and done it poorly.
Better to just turn up the volume while mixing. Avoid clipping indications as a general rule – you may not hear it, but those signals are getting leveled off before being sent to your soundcard, and it does affect the sound if it’s more than a little clipping. And you want to hear what you’re mixing, right? Mix down to 32-bit format and send that to the mastering engineer. In 32-bit format, the overall level won’t matter a bit (hard to believe, but true).