Topic: Recording levels for loud-ish concerts (aka, 24 bit is just alright with me)

[heathen]

How hard is it to properly set levels when recording in 24 bit at an even moderately loud concert?  We definitely don't need to worry about getting the levels exactly as high as possible but no higher, which was always a worry back in the cassette days.  So, with 24 bit we can err on the side of lower levels.  But how low is too low?  I decided to test this out with a recent recording.

The Zoom F8 has a dual channel record feature that I'll often use when I can spare the channels.  I'll set the main channel to where I think I'll get close but not exceed clipping.  Then I'll set the dual channel very low as a safety.  I did that at the recent JRAD show in Vermont and decided to compare them after-the-fact.

The main channel peaks at -3.9, and the dual channel peaks at -16.9.  When you're in the moment, peaks that low look REALLY low and they look even lower when viewing the waveform in Audacity.  But does it actually matter to the final sound?

Using Audacity I first amplified (using the built-in "Amplify" effect) the dual channel to peak at -3.9, just like the main channel.  Then I inverted the dual channel (using the built-in "Invert" effect).  Finally I mixed the two together to a new track ("Mix and Render to New Track").  What I got was silence.  I even used the "Sound Finder" feature in the "Analyze" menu with the settings maxed out and that didn't detect any sound.

These results shouldn't be surprising.  We've known that 24 bit has more than enough dynamic range for recording the concerts most of us record.  But it's good to see some objective evidence to confirm that.

I'm not knocking 32 bit, or arguing that 24 bit is just as good or better.  If I were in the market for a new recorder right now I'd absolutely go 32 bit.  But this should help the FOMO that some may have while still using 24 bit.  There's no need to worry about trying to get the levels as high as you can get away with...err on the side of caution safe in the knowledge that it's going to be okay.

[Rairun]

Bitrate is definitely not the bottleneck there. AD converter self-noise is what we're are fighting against most of the time - all gain staging we do is so that we are able to provide the ADC with a healthy signal that towers above its self-noise. Multi-ADC designs do help a lot when the audio you're recording has a high dynamic range (i.e. so that the quieter parts don't get too close to the noise floor set by the ADC).

If the sound source is quiet enough, of course, then the microphone's self-noise becomes the bottleneck, and gain staging isn't going to fix it, let alone 32-bit float.

Re: your test, I wouldn't expect a recording peaking at -16.9 to display any perceivable noise. If you recorded a super quiet acoustic number right after the rock song that peaked at -16.9, though, using the same settings, you'd likely start hearing some noise that a multi-ADC design might have prevented.

[heathen]

If you recorded a super quiet acoustic number right after the rock song that peaked at -16.9, though, using the same settings, you'd likely start hearing some noise that a multi-ADC design might have prevented.

Agreed, or something like wildlife field recording.  That's why I made sure to limit what I was talking about to loud-ish concerts.

[Gutbucket]

Nice test.  Not surprised.  What model F8 are you using?

I ask because I think 24 bit recording with F8N-Pro works differently than F8 / F8N.  I think F8N-Pro always uses the 32-bit switching ADC architecture and saves 24-bits worth of the output.  (I'm an original F8 user myself)

Rairun is likely right in that a significantly quieter acoustic noise floor is likely required to see a difference.  Forgive one persnickety correction though- "Bitrate" is sample-rate frequency measured in kHz and would not be a bottleneck in either case.  "Bitdepth" was the intended term I suspect (16/24/32bit).

[heathen]

Nice test.  Not surprised.  What model F8 are you using?

Just the bog standard F8...not even the F8N.

[Gutbucket]

Great, that means your test basically applies to my setup as well.

What mics were you using? Would like to look up their sensitivity to compare with the mics I'm using.

[heathen]

Great, that means your test basically applies to my setup as well.

What mics were you using? Would like to look up their sensitivity to compare with the mics I'm using.

In this instance I was using the AT4050ST (with the pad on, and in mid-side mode, if either of those matter).

[morst]

In this instance I was using the AT4050ST (with the pad on, and in mid-side mode, if either of those matter).

Mic pads add self noise, as they lower the output level, but not the noise level.
I've been advised that separate resistive pads do not have such an effect.
Cool test.

[heathen]

Mic pads add self noise, as they lower the output level, but not the noise level.
I've been advised that separate resistive pads do not have such an effect.

I didn't know that!

From a quick scan of the manual it looks like the F8 doesn't have any sort of pad built in either.

I'm guessing that for loud-ish concerts whatever amount of noise there is won't make a difference anyway (within reason, of course).

[morst]

Mic pads add self noise, as they lower the output level, but not the noise level.
I've been advised that separate resistive pads do not have such an effect.

I didn't know that!

From a quick scan of the manual it looks like the F8 doesn't have any sort of pad built in either.

I'm guessing that for loud-ish concerts whatever amount of noise there is won't make a difference anyway (within reason, of course).

I think that's why I never noticed the noise back when I ran the Beyer MV100 which had a minimum gain setting of +20dB which caused me to need some sort of pad for super loud shows.
By the time the show was loud as F to need the pads, the music overshadowed the noise by a lot.
Recording nature sounds would likely be far more revealing of the noise levels.

Here's DSatz weighing in with his explanation, from a different thread (bolding is mine)

> The analog front end of the recorder is separate from its ADC architecture and file storage format.  If an analog input to ANY recorder is hot enough to overload the recorders analog input circuitry, the recorder's input sensitivity will need to be adjusted, a pad-engaged, or limiting applied to prevent it from overloading.

This is a key point (aside from the fact that past a certain point, limiting can't prevent overload in an analog front end; it protects only the stages that come after it, such as the A/D. But the possible benefit of a limiter is, uh, limited since it is just as easy to overload the input of an analog limiter as any other analog circuit--and limiting changes something essential about the sound of a recording, and requires judicious application).

-- Just wanted to add that throwing the pad switch on a condenser microphone reduces the signal/noise ratio of the recording by the amount of the pad, since it reduces the signal levels without lowering the noise floor of the electronics. Those switches are for emergency use ONLY. If you reasonably anticipate that a microphone's output levels could overload your recorder or preamp's inputs, get some in-line resistive pads (typical values are 10 and 20 dB) and put them at those inputs when setting up for a loud event. They knock down noise exactly as much as they knock down signals, thus preserving your s/n ratio. They're cheap, they're rugged, and they don't interfere with phantom powering.

Or depending on your recorder, it may or may not have settings for input sensitivity that work like the resistive pads that I'm talking about.

There's a ~40 dB difference in sensitivity between the most and least sensitive microphones on the professional market, and the maximum SPL of live sound itself varies by (say) another 40 dB from situation to situation. The gain setting of an analog mike preamp MUST be at least approximately correct to avoid clipping and/or the addition of extra, potentially audible noise. That's regardless of whether the rest of the recorder is 16, 24 or 32-bit.

[heathen]

Thanks again for educating me about that!

[Gutbucket]

From a quick scan of the manual it looks like the F8 doesn't have any sort of pad built in either.

Not a stand-alone pad per say.. however a -10dB pad is engaged if you switch on the input limiters.

Other than using an external pad or one built-in to the microphone, there are two ways of accommodating overly hot input signals into the original F8:
1) Use line-in, which reduces input sensitivity, however doing so on the original F8 requires use of the TRS inputs rather than XLRs, and no phantom power is available.
2) Stick with mic-in (via XLRs with phantom) and engage the digital Advanced Limiter, set to 0dBFS.  This reduces input sensitivity by 10dB (equivalent to a 10dB pad).  The 0dBFS setting is not the threshold engagement value which a traditional limiter would have, but rather a "never exceed" value.  The limiter engages progressively above some lower threshold, using a variable ratio that increases as the signal approaches the set limit.  I think the engagement threshold is -10dB below the "not to exceed" setting, at which it begins engaging at 1:1 and progressively increases to infinity:1 just prior to reaching the set limit (which is why setting it to 0dBFS makes the most sense for our use).  Using sensitive mics I keep the Advanced Limiter engaged on all channels at all times, and aim to set input gains so as to peak below -10dB, just prior to limiter engagement.  That way the limiter is there to catch any wild overs but doesn't usually get engaged at all.  I've had no noise floor issues doing that. When recording on stage with sensitive mics I often need to set input gains on the F8 to minimum and let the Advanced limiter catch a few peaks on a couple channels.  Fortunately it's quite transparent sounding.  Otherwise I would have moved on to F8N which would allow for line-in via XLR with phantom.

[heathen]

From a quick scan of the manual it looks like the F8 doesn't have any sort of pad built in either.

Not a stand-alone pad per say.. however a -10dB pad is engaged if you switch on the input limiters.

Other than using an external pad or one built-in to the microphone, there are two ways of accommodating overly hot input signals into the original F8:
1) Use line-in, which reduces input sensitivity, however doing so on the original F8 requires use of the TRS inputs rather than XLRs, and no phantom power is available.
2) Stick with mic-in (via XLRs with phantom) and engage the digital Advanced Limiter, set to 0dBFS.  This reduces input sensitivity by 10dB (equivalent to a 10dB pad).  The 0dBFS setting is not the threshold engagement value which a traditional limiter would have, but rather a "never exceed" value.  The limiter engages progressively above some lower threshold, using a variable ratio that increases as the signal approaches the set limit.  I think the engagement threshold is -10dB below the "not to exceed" setting, at which it begins engaging at 1:1 and progressively increases to infinity:1 just prior to reaching the set limit (which is why setting it to 0dBFS makes the most sense for our use).  Using sensitive mics I keep the Advanced Limiter engaged on all channels at all times, and aim to set input gains so as to peak below -10dB, just prior to limiter engagement.  That way the limiter is there to catch any wild overs but doesn't usually get engaged at all.  I've had no noise floor issues doing that. When recording on stage with sensitive mics I often need to set input gains on the F8 to minimum and let the Advanced limiter catch a few peaks on a couple channels.  Fortunately it's quite transparent sounding.  Otherwise I would have moved on to F8N which would allow for line-in via XLR with phantom.

Thanks for the insight!  I'm learning a lot in this thread.