At what stage am I clipping? How does headroom help?

[Lunch Money]

Hardware:

Guitars and microphones into a UB802 mixer, then into a 24/96 Audiophile PCI. Recorded in Tracktion at 16/44.1 (or is it always 24-bit by default? I don't see where to set it).

I've noticed in several threads people talking about headroom and stuff, but I just don't get it. I play my guitar (for example) in Tracktion, and when the signal smacks into 0db, I get digital clipping. Now, I'm not complaining. This seems normal to me. To me, this is how it should work-- if you don't set your levels right, you clip and things sound like shit. Hence, I check my levels carefully and pray that I don't get over-ambitious and smack my guitar too hard and clip my beautiful take.

So where's this safety zone? How is it that people are getting away with instrument abuse and not really caring about their levels? What part of this equation am I missing?

Greg

[BONES]

24 bit gives you loads of extra headroom, assuming you are working towards CD quality audio. Bit depth provides an exponentially greater range of values so the difference between 16 and 24 bit is considerable. So you should be able to record in 24 bits with loads of headroom and then downsample when you have done all your processing, etc. without any noticable quality loss. Of course, if you're recording analogue sources there will inevitably be noise in the signal so you want to keep the level reasonable but it doesn't need to be perfect.

[IIRs]

24 bit gives you loads of extra headroom,

er.. no it doesn't

24 bit just allows low-level signals greater resolution.

Recording at 24 bit only gives more headroom if you leave a greater safety margin (as your low-level signals won't get so grainy)

32 bit float gives extra headroom (and is used internally in Tracktion) but there is no substitute for setting a suitable input level.

(or is it always 24-bit by default? I don't see where to set it).

Select the input: bit depth setting is just below the Wav/Aiff field

[Lunch Money]

OK, I suppose I could just test this, but I'm about to run out the door so I'll throw it out there anyhow:

Let's say I'm at 16-bit and I'm recording a hot signal that clips (audibly and also on the meters). Is it the case that without changing my setup at all except for changing to 24-bit that I won't audibly clip (even if the level meter is showing 'clip')? Now, my card doesn't do 32-bit, but would it be the case that in the same setup with 32-bit input, I'd have even MORE input room before clip?

That's the part that's confusing me.

Greg

[BONES]

24 bit gives you loads of extra headroom,

er.. no it doesn't

I suppose it all depends on perspective. I see the extra 8-bits sitting on top of the first 16, so I can get pristine 16-bit recordings with another 8-bits of headroom.

Lunch Money, If you change to 24-bit you can reduce the input level without sacrificing quality, which is the point of using a higher bit-depth. You should be able to normalise in 24 bit then reduce it to 16 bit without any noticable loss in quality. BTW, if you take the mixer out of the loop you should get a cleaner signal for recording.

[IIRs]

Let's say I'm at 16-bit and I'm recording a hot signal that clips (audibly and also on the meters). Is it the case that without changing my setup at all except for changing to 24-bit that I won't audibly clip

No. If you clip at 16 bit you will clip at 24. The trick is to leave loads of headroom when using 24 bits.. unlike 16 bit you won't loose any audible quality by not driving the converters hard.

I don't think there are any soundcards with 32 bit converters.. 32 bit float seems to be the standard for internal audio engines, but it always has to end up at 16 or 24 bit integer for a soundcard to deal with it..

[xoxos]

What part of this equation am I missing?

a h/w compressor? :p all i can advise is stay away from cheap dbxs.. imo 266xl 'does what it's supposed to' but is a signal crappifier. it makes me hate every sound that comes out of it.

[P.T.]

Just record at a lower level so that there is no chance of hitting 0db and clipping.

The reason they used to record hot in the analog days was because a recording at a low level would have a lot of noise int it. Tape hiss and such.

Digital has much less noise so your signal doesn't need to be so hot to overcome the noise.

Set your levels by playing as hard as you might ever do and get it to about -3db on those peaks.

Then when you are mixing you can set the level how you need it and compress it etc.

[Lunch Money]

Thanks, that clears it up. Just to consolidate what other people have said-- it's not that I can run a hotter signal with a higher bit depth, but rather that I can run a LOWER signal and still be able to bring its level up to a proper level later, without sacrificing sound quality.

FWIW, I'll probably be getting an external compressor anyhow, though I haven't the money for "quality" per se. That's all right with me, though, because I do my music for myself and whoever else cares to listen, and don't mind if it doesn't come out sounding like a top-40 production.

Cheers!
Greg

[mindless]

Hardware:

Guitars and microphones into a UB802 mixer, then into a 24/96 Audiophile PCI. Recorded in Tracktion at 16/44.1 (or is it always 24-bit by default? I don't see where to set it).


Greg

You're running too high of levels into that soundcard. I used to record with an Audiophile 24/96 and got clipping all the time. I switched to a Delta44 and rarely get clipping now. I still use the Audiophile for MIDI ins and outs.

[Lunch Money]

Cheers, good to know. Since I won't be switching soundcards any time soon, I guess I'll have to start running quieter signals.

[bool]

Yes, it's good to be careful with input levels - once you have clips, you can't get rid of them without using some sort of restoration - and even then it won't work satisfactory every time.

[Lunch Money]

Yes, it's good to be careful with input levels - once you have clips, you can't get rid of them without using some sort of restoration - and even then it won't work satisfactory every time.

Or in my case, ANY time. I have some clipping in a few songs that I'd like to clean up, and I've never figured out how to do it.

[egbert]

Thanks, that clears it up. Just to consolidate what other people have said-- it's not that I can run a hotter signal with a higher bit depth, but rather that I can run a LOWER signal and still be able to bring its level up to a proper level later, without sacrificing sound quality.

This is pretty close to the (lunch) money. It is still worth paying attention to gain levels at every level of the signal path though if you want to optimise recording quality and minimize noise levels.

Bear in mind that most decent mic pres have signal to noise ratios of the order of 100dB - and that is if your levels are optimal. Even the best A/D boxes make no claims of S/N better than 120 dB and some pretty pricey ones are 110 -117 dB.

This means that the effective bit depth of these "24 bit" converters is really 19 - 20 bits. The last 4 or 5 bits of your 24 bit samples are noise.

What this means is that if you record with 6 dB headroom (peaks 6 dB below digital zero) you have at best 18 -19 bits of resolution. This is probably OK with high level signal inputs. Certainly if your destination is a CD.

In the case of a recording obtained via a mic/mic pre - you will probably have at best 94 dB signal/noise if the mic pre was producing recorded peaks at -6dB at your converters.

If you then apply gain to your signal inside your DAW - eg using a compressor with autogain - you will bring noise levels up higher again. This noise could be quite noticeable in some situations - eg quiet acoustic tunes/sections.

If you have noisier signals - from electric guitars picking up dimmer hash, from old synths, whatever - you will want to make sure you are recording as hot as possible without clipping if you want to keep noise on the final recorded tracks under control.

Eg

[mauseoleum]

!!! CLIPPING IS GOOD !!! SOFTCLIPPING IS GOODER !!!