-20db or best signal

I have heard more than once on this forum that it’s best to record at -20dB or thereabouts. I always try to get the strongest signal. I run into clipping quite often. Still, -20dB seems extremely low. What factors should be considered in setting my level?

For a home studio set-up, that’s a bit low. We generally recommend aiming for a peak level of around -6 dB (about half the track height).

The single most important rule is: Keep below 0 dB at all times during recording.
The point about aiming for -6 dB is to allow a bit of headroom to avoid accidentally clipping.

You can get away with most of the work at -20 in a recording studio where everything is quiet and the microphones, mixers, and recorders are well behaved.

Home recording should be louder because of microphone limitations and unfortunately, that makes it more likely to have accidental overload damage at the exact time home performers are least likely to pay attention to it.

Everybody wins.

Koz

It also depends on what you’re recording… If you are recording a live vocal or a live acoustic instrument you need more headroom for insurance against clipping with unexpected peaks. If you are digitizing analog records or tapes the levels are a lot more predictable and you can get-away with less headroom. And if you are digitizing an analog recording it’s not the end of the world if you have to go-back and start-over at a lower level.

Headroom is a funny thing… If you don’t use it you didn’t need it and if you do use it, it’s no longer headroom!

I always try to get the strongest signal.

If you’re old enough to remember analog tape, you needed a “hot” signal to overcome tape noise. But with digital, no tape noise!* We’ve got lots of dynamic range so digital recording levels are just not that critical, as long as we avoid clipping. Also, tape is more forgiving of overloading. It begins to soft-clip when you go over 0dB and the recording/playback EQ further “softens” the clipping. So it was common to occasionally go “into the red”. Analog-to-digital converters (and digital to analog converters) cannot go over 0dB and they will hard-clip if you try.

Although the digital levels are not critical, it’s still important to get a strong analog signal, and a good acoustic signal if you’re using a microphone, for a good signal-to-noise ratio. And, if you can’t get the digital levels above -20dB there might be a problem on the analog-side.



\

  • There is something called quantization noise and you can hear it at 8-bits but you can’t hear it at 16-bits. Like regular (analog) noise it’s most noticeable at low levels but “rides on top of the signal” so quantization noise goes-away completely with “digital silence”.

“Full disclosure” - You do loose resolution at lower levels (relatively more quantization noise). Every bit gives you 6dB of dynamic range so If you are recording at 16-bits and peaking at -6dB, you are only using 15-bits. At -18dB (or -20) you are only using 13-bits. At 13-bits you’re still probably OK. The quantization noise probably isn’t audible and any analog noise is probably higher anyway. Or, if you have a 24-bit soundcard/interface it’s no issue at all.

I’m interested in exploring distorted sounds, like guitar. But that involves clipping, no? I wonder if I can clip and then just reduce the level to where it would be without clipping? Is it only in the final mix that we should not clip? Otherwise I don’t see how distortion is achieved.

Digital clipping is not nice, not musical, and if played too loud through speakers can fry the tweeters (due to excessive energy at extremely high frequencies).

“Distortion” means simply that the waveform shape is modified (“distorted”) in some way - usually by squashing down the peaks.

Audacity has a distortion effect that can modify the waveform in various ways, including “hard clipping” (like digital clipping), “soft clipping”, and many others: https://manual.audacityteam.org/man/distortion.html

Hey Steve. OK. I’ll try that Distortion effect. What about using Filter Curve to remove light clipping? I’ve been using that and it removes the red line for clipping, if the clipping is light.

Clipping is where the sound system stops following the show. It’s also the place where the system starts making up its own sound. The difference between classical guitar and fuzz guitar. Good Fuzz Guitar is a tube amp in overload. Remember when efficient, cool-running transistor amps came out? Everybody took one strum and said ewwwwww. I think you can still buy tube amps. I know you can buy transistor amps with a button to make it sound like a tube amp.

Transistor amps don’t overload sloppy and gracefully like tubes. They just stop dead which makes your ears bleed. That’s clipping. Take the gently rounded tops and bottoms of waves and just whack them right off.

If you catch clipping during the recording and reduce the volume so it stops, that’s usually good to go as long as you didn’t need the performance that overloaded up until then. If you have an overloaded recording, reducing the volume will carefully preserve the original performance and the clipping distortion.

Even with the ClipFix tool, you will probably not get the original performance back. ClipFix tries to reconstruct the original wave from the healthy waves before and after—and only does it to one clip. If you have a highly distorted and clipped performance, you have a rehearsal.


There is a promotional and marketing gimmick. Home microphones almost always come out of the box with low volume. Modest, clear volume makes the customer think they did something wrong and post a question to the Audacity forum. Higher volume microphones run into evil-sounding clipping distortion which makes the customer send the microphone back.

No contest.

It got so bad that Windows started adding a “Microphone Boost” setting to make up for the super quiet shows the customers were making.

One other note. It’s not the worse idea to watch the volume lights on preamps and interfaces. Some of them warn you slightly ahead of actual clipping. It’s “You’re getting mighty close,” and not “Oops. Your show is trash.”

Koz

I wonder if I can clip and then just reduce the level to where it would be without clipping?

This a little tricky… Audacity shows red for potential clipping and you can get a false positive or false negative.

Your ADC (analog-to-digital converter, recording), DAC (digital-to-analog converter, playback), regular WAV files and audio CDs are all hard limited to 0dB and they will clip if you “try” to go over.

Audacity shows red if there are a few 0dB samples in a row, or if the audio goes over 0dB.

Audacity uses floating-point representation internally/temporarily so there is no upper (or lower) limits and it can go over 0dB without clipping.

If you record and you “see red” before any processing your ADC has hit 0dB and the audio IS clipped. If you reduce the volume Audacity will no longer show red but of course the wave shape has not changed and the audio is still clipped.

Or, you can clip your analog amp/preamp and if the ADC isn’t clipped the digital data won’t hit 0dB and Audacity won’t show red. There is actually a common situation where an audio interface clips at -6dB (50%) and again Audacity won’t know. (But the clipping LEDs on the interface WILL will light-up.)

If the audio is not clipped and you do something that pushes the peaks over 0dB (like boosting the bass, etc.) Audacity will show red but the audio isn’t actually clipped. As long as you bring the levels down before exporting everything will be OK.

Returning for a moment to the original post:

I wonder if you were hearing about RMS or perceived loudness (see Effect > Loudness Normalization) for which Audacity’s defaults are -20dB and -23LUFS, respectively.

Yeah, I remember tuba amps. I used to play my Strat or Les Paul through a pre-CBS Fender Twin Reverb. But that was ages ago. I just want to experiment with distortion without ligging all the equipment with me. I’m traveling with a little yamaha GL1 acoustic. Doesn’t have much output, but I can make decent sketches of songs. And with audacity I can make it sound electric by adding reverb, tremelo, echo and also get synth sounds simply by transposing the guitar. It sounds like asynth then. I’m using it to generate bass tracks too using change pitch. I don’t care too much about studio quality. I just want to be able to make some music wiith whatever I have at hand. my drum kit is a palying card tapped an the mic guard (to get a bass drum,) writing pens tapped on water glasses, hand drumming on the table or on an upturned drawer from the dresser.

I’ve been able to manage clipping by turn my twin mic slightly so as to catch the sound waves at different angles. This works great especially with -06 setting. I can select whatever of the four channels is least problematic. At this moment I have two channel vocal performance I’m editing. There are all kinds of pops, hissing and nuclear explosions. But it usually can be fixed with 100hz rumble or telephone in Filter Curve. If not, then one of the channels has less noise and I copy and paste the better parts from one channel to the other. I’m managing.

So I think the best option for me is to learn the basics of the Distortion Effect. I’ve got a long way to go learning the basic, then I see that there are fine tunings for all of these effects. I’m now getting a finer sensitivity to appying the Filter Curves. Just telephoning precisely where the noise occurs.