You are here

Interface, dynamics, headroom, and ideal input level

For everything after the recording stage: hardware/software and how you use it.

Moderator: Moderators

Interface, dynamics, headroom, and ideal input level

Postby cashhewn » Fri Dec 18, 2020 4:55 am

I'm trying really hard to understand the concept here, it seems like it should be simple, but I just can't manifest the epiphany...can someone please help?

A person told me that in a DAW working with a 24-bit, fixed point WAV session, headroom can be calculated by subtracting the total dynamic range of your converter from the dynamic range of a 24-bit WAV file. He said that in my case this would be 144dB (dynamic range of the WAV) minus 118dB (dynamic range of my converter) for a total of 26dB of headroom.

That makes sense to me. But here's where I get lost:

Then he says that since my converter's maximum input is +18dBu on mic inputs and +26dBu on line inputs, that anything I record hotter than -26dBfs in the DAW I'm only amplifying noise.

I can't understand it, I would have thought that even recording a line input coming in at 0dBfs which would be +26dBu at my converter's input (right?), that I would have only eaten exactly the amount of available headroom, and probably not create any digital clipping. Yet does this amplify noise somehow? Does it have to do with the concept of digital "full scale" and my converter's inability to take advantage of that 26dB above its dynamic range somehow? Seems like that's where my mind wants to go with this to make sense of it but I can't quite grasp it yet...

In any case, suffice it to say that I usually try to stay under -20dBfs at all times with my DAW input levels, but I just want to understand the math behind what my "ideal" input level would be in the DAW, where I should cap the limit on my peaks, and why this person says that I'm amplifying noise with anything coming in above -26dBfs.

Thanks so much for your help!
cashhewn
Poster
Posts: 28
Joined: Mon Jan 20, 2020 5:43 am

Re: Interface, dynamics, headroom, and ideal input level

Postby ef37a » Fri Dec 18, 2020 9:39 am

I can give you my HO while the top boffs finish their brekky.

That figure of "144dB" is really a bit of a red herring I think, yes it is the theoretical dynamic range of a 24 bit system but since even the very best equipment only exceeds 120dB by a few dB it is indeed theory.
The dynamic range of an interface is determined by the difference between the noise floor and the clipping point, usually that of the microphone amplifiers.

But what is the worry? Even the humble Behringer UMC204HD that I had for few weeks was quiet enough such that I could record strummed acoustic guitar at about a foot from a dynamic mic of around 2mV/Pa (i.e. a wee bit more sensitive than the 57/58s) with signal averaging -20dBfs and peaks to -8dB or so and the noise floor was very acceptable. "Rocky/folky" singing would also give easily good enough results. The 204 is maybe not quite quiet enough for a 7b for speech in a VERY quiet room but it would be fine with almost any LDC!

Once you get to AIs costing 4 times that of the Berry, no worries.

Dave.
ef37a
Jedi Poster
Posts: 12834
Joined: Mon May 29, 2006 12:00 am
Location: northampton uk

Re: Interface, dynamics, headroom, and ideal input level

Postby Hugh Robjohns » Fri Dec 18, 2020 11:53 am

cashhewn wrote:A person told me that in a DAW working with a 24-bit, fixed point WAV session, headroom can be calculated by subtracting the total dynamic range of your converter from the dynamic range of a 24-bit WAV file.

Not really. I can see where he's trying to go, but he's got himself very confused and so it's not helpful... or useful, actually! What he's assuming is that the noise floor of the converter sits at the same place as the theoretical noise floor of a 24-bit (fixed point) DAW... but it doesn't! It's actually the clipping point that is the same... and the headroom is whatever you decide you want it to be!

Technically, headroom is defined in a couple of different different ways depending on the circumstances.

In professional circles, headroom is defined very simply as the signal range (in decibels) available between the system's 'alignment level' and the clipping point (or a set percentage of third-harmonic distortion in some analogue systems like tape recorders!).

In an analogue studio console the alignment level is typically 0VU (normally +4dBu) and the clipping level is typically +24dBu (sometimes higher), so there is usually around 20dB of headroom.

This is replicated in the professional digital world by establishing a notional working alignment level of -20dBFS, and the clipping point is obviously 0dBFS -- so again, we have 20dB of headroom.

The noise floor of a good analogue console will be around -95dBu, giving a total system dynamic range between noise floor and clipping of 95+24 = 119dB.

The noise floor of a modern converter is also around -119dBFS (around 20 bits are carrying useful audio, the other four are just random noise...). So both analogue and practical digital systems have directly comparable dynamic ranges.

Now, while 20dB headroom has been established over the last century as a good arrangement when recording unpredictable sources and for mixing, it's not always necessary for finished/mastered material, or when working with heavily compressed music or professionally controlled broadcasts. In those cases the system maybe configured intentionally to have less headroom.

The EBU broadcasting spec calls for 18dB of headroom instead of 20, for example, and streaming online services typically work with around 14dB of headroom at the moment. It applies to hardware too -- RME interfaces usually have three different I/O configuration options, each with differing headroom margins, for example.

An alternative but common use of the Headroom term applies in the context of a recording. Here, the term is often used to refer to whatever 'unused space' remains between the highest recorded sound and the clipping point of the digital converters. So after a recording you might say you have 6dB of headroom in that recording...

He said that in my case this would be 144dB (dynamic range of the WAV) minus 118dB (dynamic range of my converter) for a total of 26dB of headroom.

No... this is nonsense. The clipping point of the (fixed-point) WAV and the clipping point of the converter are exactly the same 0dBFS. They have exactly the same headroom.

What differs between the theoretical 24-bit WAV and your converter is where the noise floor sits, and what your friend is actually describing is how much of the potential 24-bit system's dynamic range is effectively lost to the converter's noise floor!

In a perfect 24-bit system, the noise floor is defined by the essential dithering of the LSB, and that's typically 3dB above the minimum quantising level, so actually -141dBFS.

Point of information: you only get a dynamic range of 144dB if you don't have dithering... but then you get quantising distortion instead which we don't want... so the 144dB figure is not actually relevant in the context of a quality audio system.

Unfortunately, in the real world there are no practical converters that have achieved a noise floor of -141dBFS -- and there probably never will be. Most decent budget gear has a noise floor around -118dBFS or so. Really good high-end converters are typically around -122dBFS now, and some clever 'stacked' converter systems are claiming to push around -130dBFS.

But to put that into a practical context, most real-world acoustic recordings have an acoustic noise floor that sits somewhere between -60 and -90dBFS....

Then he says that since my converter's maximum input is +18dBu on mic inputs and +26dBu on line inputs, that anything I record hotter than -26dBfs in the DAW I'm only amplifying noise.

Nope. Utter nonsense! Oh dear... the poor chap is a bit befuddled, isn't he? :ugeek:

Seems like that's where my mind wants to go with this to make sense of it but I can't quite grasp it yet...

The only thing to grasp is that he doesn't understand what he's talking about.

In any case, suffice it to say that I usually try to stay under -20dBfs at all times with my DAW input levels...

Very good idea! When tracking and mixing, it's a good idea to keep the average levels around -20dBFS, but you can happily allow transient peaks up to -12 or -10dBFS without any problems.

I get a bit squirmy if there's more than the odd one or two as high as -6dBFS. :shifty: And I tend to err on the low side because there are no penalties in bringing the level up -- the noise floor is defined by the acoustic recording, not the digital system. But clipping sounds horrid and is difficult to correct afterwards! So I've made lots of recordings that don't peak much above -20dBFS without any noise problems at all in the final mastered version.

Headroom is always your friend!

Plus, these guidelines exactly mirror traditional analogue practices. (The headroom margin wasn't displayed on analogue meters, but it is on digital meters... which causes a lot of confusion!)

I hope that's cleared things up. This analogue/digital levels comparison might also be a handy reference:

analogue_digital comnparison levels.jpg
User avatar
Hugh Robjohns
Moderator
Posts: 30630
Joined: Fri Jul 25, 2003 12:00 am
Location: Worcestershire, UK
Technical Editor, Sound On Sound
In my world, things get less strange when I read the manual... 

Re: Interface, dynamics, headroom, and ideal input level

Postby Arpangel » Fri Dec 18, 2020 12:09 pm

You’ll all be pleased to know I now record in 24/48 that’s the default setting in my Reaper template.
It took me a long time to change from 16/44.1 but it was psychological, I could see all those extra bits taking up space, and it freaked me out.

:)
User avatar
Arpangel
Jedi Poster
Posts: 7208
Joined: Sat Jul 12, 2003 12:00 am

Re: Interface, dynamics, headroom, and ideal input level

Postby The Elf » Fri Dec 18, 2020 12:50 pm

Arpangel wrote:You’ll all be pleased to know I now record in 24/48 that’s the default setting in my Reaper template.
24-bit is fine, but unless you're working for video there's little reason to change rate from 44.1kHz.
User avatar
The Elf
Jedi Poster
Posts: 16543
Joined: Tue Aug 14, 2001 12:00 am
Location: Sheffield, UK
An Eagle for an Emperor, A Kestrel for a Knave.

Re: Interface, dynamics, headroom, and ideal input level

Postby merlyn » Fri Dec 18, 2020 4:06 pm

It's an epiphany you want, eh?

A dB is a unit, like a metre or a kilogram. You can't add a metre and a kilogram.

1m + 1kg = 1m + 1kg

So, in the hope of causing an epiphany, what is a dB measuring?
merlyn
Regular
Posts: 236
Joined: Thu Nov 07, 2019 3:15 am

Re: Interface, dynamics, headroom, and ideal input level

Postby Hugh Robjohns » Fri Dec 18, 2020 4:21 pm

merlyn wrote:A dB is a unit...

A decibel is a logarithmic ratio of units.... :ugeek:
User avatar
Hugh Robjohns
Moderator
Posts: 30630
Joined: Fri Jul 25, 2003 12:00 am
Location: Worcestershire, UK
Technical Editor, Sound On Sound
In my world, things get less strange when I read the manual... 

Re: Interface, dynamics, headroom, and ideal input level

Postby cashhewn » Fri Dec 18, 2020 4:52 pm

I love this site. For me it’s hands down the greatest resource out there. Thank you all so much for your help.

@merlyn I thought I had the epiphany when I discovered the idea of -20dBfs = 0VU = +4dBu, but even this seems quite elastic when I’m measuring AC voltage RMS and meter calibration between my devices, so I still have some way to go as an “engineer”...in any case, none of this is hindering me from making music!
cashhewn
Poster
Posts: 28
Joined: Mon Jan 20, 2020 5:43 am

Re: Interface, dynamics, headroom, and ideal input level

Postby merlyn » Fri Dec 18, 2020 5:19 pm

According to Wikipedia :

The decibel (symbol: dB) is a relative unit of measurement corresponding to one tenth of a bel (B).

In a ratio of units e.g. (Volts out)/(Volts in) the units cancel out and we have a dimensionless number.
merlyn
Regular
Posts: 236
Joined: Thu Nov 07, 2019 3:15 am

Re: Interface, dynamics, headroom, and ideal input level

Postby Wonks » Fri Dec 18, 2020 5:24 pm

So it’s not like a metre or kilogram at all.
User avatar
Wonks
Jedi Poster
Posts: 11215
Joined: Thu May 29, 2003 12:00 am
Location: Reading, UK
Correcting mistakes on the internet since 1853

Re: Interface, dynamics, headroom, and ideal input level

Postby merlyn » Fri Dec 18, 2020 5:33 pm

Yes, the dB is. The number is not.
merlyn
Regular
Posts: 236
Joined: Thu Nov 07, 2019 3:15 am

Re: Interface, dynamics, headroom, and ideal input level

Postby Kwackman » Fri Dec 18, 2020 6:51 pm

merlyn wrote:According to Wikipedia :
Wikipedia is a great resource, but not the gospel!
I'd prefer a definition from a more reliable source.
AES.org for instance, or even this magazine's excellent glossary.
User avatar
Kwackman
Frequent Poster (Level2)
Posts: 2186
Joined: Thu Nov 07, 2002 1:00 am
Location: Belfast
Cubase, guitars.

Re: Interface, dynamics, headroom, and ideal input level

Postby Hugh Robjohns » Fri Dec 18, 2020 6:56 pm

We risk disappearing down a pedantic wormhole here... It comes down to what meaning you associate with the term 'unit'.

The volt is a unit. You can have a battery that produces 6 Volts, say. Or it can drive 3 Amps around a circuit. Or deliver 5 Watts of power. Or it might reach a temperature of 28 Celcius ... All recognisable SI units.

But the Bel is actually just a ratio of two quantities. And the decibel is a tenth of a Bel, but still a ratio of two quantities. So although it has a name, and is often described as a unit, it's not actually a specific 'thing' -- It's just a ratio of things, and it can be used to compare the values of anything you like. We typically use it to compare acoustic sound pressures, signal voltages, digital quantising levels, amplifier powers, and much more besides...

But nothing produces '6dB', for example.

You could, though, have an amplifier that delivers an output voltage twice as big as the input voltage. In which case it would have a gain 6dB... Or if it was twice the power it would be 3dB bigger.

It can become an absolute value in some contexts, too, and take on an equivalence to a conventional unit if the denominator of the ratio is a specified reference value... So +4dBu -- the u defining the specific reference of 0.775 Vrms -- is exactly the same as saying 1.223 Vrms

But the suffix denoting the reference is essential and absolutely critical to understanding what the decibel term means in that situation, whether it's dBu, dBV, dBFS, dBTP, dBm, or whatever...

Frustratingly, a lot of people who don't understand the true nature of the decibel are prone to omitting the suffix and imagine the rest of us can mind read.... ;-)

6dBu has an entirely different meaning to 6dB or 6dBV....
User avatar
Hugh Robjohns
Moderator
Posts: 30630
Joined: Fri Jul 25, 2003 12:00 am
Location: Worcestershire, UK
Technical Editor, Sound On Sound
In my world, things get less strange when I read the manual... 

Re: Interface, dynamics, headroom, and ideal input level

Postby CS70 » Fri Dec 18, 2020 8:18 pm

merlyn wrote:According to Wikipedia :

The decibel (symbol: dB) is a relative unit of measurement corresponding to one tenth of a bel (B).

In a ratio of units e.g. (Volts out)/(Volts in) the units cancel out and we have a dimensionless number.

To me the essential point of using decibels (or any nonlinear scale) anywhere is because you may need to talk of both small numbers and very large quantities (numbers) in the same sentence/context. Linears scales become very unwieldy when the numbers get big, while a nonlinear one allows you to keep the size of the integers you use small enough.

Logarithms just happen to be a good curve, at least in sections, for a lot of physical processes. The price is that they aren't so intuitive - the same numerical difference between two points, taken at different point of the scale, mean very different quantities.
User avatar
CS70
Jedi Poster
Posts: 7718
Joined: Mon Nov 26, 2012 1:00 am
Location: Oslo, Norway
Silver Spoon - Check out our latest video and the FB page

Re: Interface, dynamics, headroom, and ideal input level

Postby merlyn » Fri Dec 18, 2020 8:44 pm

Hugh Robjohns wrote:We risk disappearing down a pedantic wormhole here... It comes down to what meaning you associate with the term 'unit'.
There's a well accepted definition of 'unit'. The decibel like the litre is a 'Non-SI unit accepted for use with SI units'. :)

How did the OP's friend come up with such a load of rubbish? I would say it's because his maths isn't good. If he had thought about dB as units he wouldn't have made so many mistakes.

dB -- ratio
dBu -- Volts
dBFS -- a sample value

If the OP's friend had been aware of that he may have realised you can't add and subtract them as you please. :)
merlyn
Regular
Posts: 236
Joined: Thu Nov 07, 2019 3:15 am

Next