16 bit vs 24 bit audio

Discussion of the mysteries behind bit-depth, sample rates

and sound quality
by Tweak

This article is going to be kept as simple as possible. It's just designed to

get the new person up to speed on the issues and provide a strong sense
of perspective on what really matters. We are going to talk about bit depth
and sample rates, how these translate into storage requirements, and then
talk about the subjective differences between the two methods of recording
your music. In short, what is the relation of 24 bit recordings to the "sound
quality" we all want.
When I first ventured into digital audio, it was a much simpler world.
Products that recorded and generated digital audio were all 16 bit.
Compact discs, the main method for music distribution, has digital audio
that has a bit depth of 16 bits and a sample rate of 44.1 kHz. Gradually,
products began to appear with a higher bit depth--an 18 bit drum machine,
20, then 24 bit effects processors. Then recorders made the leap to 24 bit.
Today, your audio interface is probably 24 bit and allows you to select
sample rate of your choice, 44.1, 48, 88.2, 96 and even 192khz. Multi track
recorders vary between 16/44.1 and 24/96. When you buy one you have to
decide which way to go and get it right the first time.
So what do all these number mean and how important are they? That's
where we are going to go. First, you have to have these definitions under
control.
Bit Depth - refers to the number of bits you have to capture audio. The
easiest way to envision this is as a series of levels, that audio energy can
be sliced at any given moment in time. With 16 bit audio, there are 65, 536
possible levels. With every bit of greater resolution, the number of levels
double. By the time we get to 24 bit, we actually have 16,777,216 levels.
Remember we are talking about a slice of audio frozen in a single moment
of time.

Now lets add our friend Time into the picture. That's where we get into the
Sample Rate.

The sample rate is the number of times your audio is measured (sampled)
per second. So at the red book standard for CDs, the sample rate is 44.1
kHz or 44,100 slices every second. So what is the 96khz sample rate?
You guessed it. It's 96,000 slices of audio sampled each second.

So lets put it all together now. This brings us to the Bit Rate, or how much
data per second is required to transmit the file, which can then be
translated into how big the file is. Your CD is 16bit, 44.1 so that is 44,100
slices, each having 65,536 levels. A new Audio interface may record
96,000 slices a second at nearly 17 million levels for every slice. If you
think that is a lot of data, well, you are right, it certainly is. The Bit Rate is
usually expressed in Mbit/sec. But you don't need to do all this math. I'm
going to do it for you. This is not an important area in the recording process
to get sidetracked on. What is important for you is how this translates to
your hard drive storage.

Space required for of stereo digital audio

File Size
Bit
Sample
of one
Bit Rate
Depth Rate
stereo
minute

File size of
a three
minute
song

16

44,100

1.35
10.1
30.3
Mbit/sec megabytes megabytes

16

48,000

1.46
11.0
33
Mbit/sec megabytes megabytes

24

96,000

4.39
33.0
99
Mbit/sec megabytes megabytes

mp3
file

128
k/bit
rate

0.13
0.94
2.82
Mbit/Sec megabytes megabytes

So you see how recording at 24/96 more than triples your file size. Lets
take a 3 minute multi-track song and add up the numbers. Just to put the
above into greater relief, I included the standard MP3 file's spec.
Hard disk requirements for a multi-track 3 minute song
number
Bit
size per
of
depth/sample
mono
mono
rate
track
tracks

size per
song

songs
per 20
gigabyte
hard disk

songs per
200
gigabyte
hard disk

16/44.1

15.1 megs 121 megs 164

1640

16/48

16.5megs 132 megs 150

1500

24/96

49.5 megs 396 megs 50

500

16/44.1

16

15.1 megs 242megs 82

820

16/48

16

16.5 megs 264 megs 74

740

24/96

16

49.5 megs 792 megs 24

240

Note that these size do not count any out takes you may have recorded but
didn't delete and they assume a linear recording method from beginning to
3

end of the song for every track. They also don't account for hard disk sector
size which will leave a lot of space on the drive empty. Real world mileage
may vary.

So you should be noting two things now:

1. Recording at 24/96 yields greatly increased audio resolution-over 250
times that at 16/44.1
2. Recording at 24/96 takes up roughly 6 times the space than recording at
16/44.1
End of technical discussion. Are you still with me? Lets get into the thick
of it.

Should you record at 24 bits?

Now lets get to the subjective side of how music sounds at these different
bit depths and sample rates. No one can really quantify how much better a
song is going to sound recorded at 24/96. Just because a 24/96 file has
250 times the audio resolution does not mean it will sound 250 times
better; it won't even sound twice the quality. In truth, your non-musicially
inclined friends may not even notice the difference. You probably will, but
don't expect anything dramatic. Can you hear the difference between an
MP3 and a wave file? If so, you will probably hear the difference between
24 bit and 16 bit audio. It gets real subjective hear as our ears are
different. But lets try to be a little objective here.
Lets talk about sample rate and the Nyquist Theory. This theory is that
the actual upper threshold of a piece of digital audio will top out at half the
sample rate. So if you are recording at 44.1, the highest frequencies
generated will be around 22kHz. That is 2khz higher than the typical
human with excellent hearing can hear. Now we get into the real voodoo.
Audiophiles have claimed since the beginning of digital audio that vinyl
records on an analog system sound better than digital audio. Indeed, you
can find evidence that analog recording and playback equipment can be
measured up to 50khz, over twice our threshold of hearing. Here's the
4

great mystery. The theory is that audio energy, even though we don't hear
it, exists as has an effect on the lower frequencies we do hear. Back to the
Nyquist theory, a 96khz sample rate will translate into potential audio output
at 48khz, not too far from the finest analog sound reproduction. This leads
one to surmise that the same principle is at work. The audio is improved in
a threshold we cannot perceive and it makes what we can hear "better".
Like I said, it's voodoo.
So should you record at 24 bits? Its going to depend on who you ask.
Some people say "It's all going to end up as 16 bit any way" when the cd is
burned. Others will tell you that when an audio interface processes and
mixes sounds at 24 bit the result is better and remains better even after the
final conversion to 16 bit. And just about every other position is taken too
when you add sample rates. Some say 16/44.1 is good enough for CD its
good enough for me. Others say do 24/44.1 because it's not that much
more space and it increases the signal to noise ratio. There is one
argument that says 24/88.2 is superior to 24/96 because it is an even
number conversion going back to 44.1.
So what's that? You want to hear what _I_ think. Ok, but its one man's
opinion, not the gospel according to Tweak! I was a believer for the past
decade that 16/44.1 was the way to go and I have recently changed my
mind. It's really a matter of what you are recording and what your gear can
re-produce. If you have a nice mic, a very good preamp and a clean audio
system and are recording highly dynamic instruments such as acoustic
guitars, classical orchestras, acapella vocals, the difference will be there.
Quiet passages will be less likely struggling to stay above the noise floor on
your system. The Bass will be tighter, and the vocals may sound "airier".
However, if you are going to be doing a loud radio-ready, levels on the
ceiling piece, my opinion is that the benefits will be less audible. But there
are other benefits as well. Because I do all my recording on computers
with sequencers, 24 bit files seem easier to work with. They have more
headroom for tweaking. One can record with less compression. Once
inside the sequencer, audio files may be converted to 32 bit for processing
and converted back to 24 or 16 on the way out. My advice is to record at
24 bits/44.1 at minimum and go up to a sample rate of 88.2 or 96 if you
think your material warrants it (and you have the disk space).
So what is Sound Quality?

People ask me all the time, and I wince every time they do "What gear (fill
in the blank...soundcard, preamp, cable, recorder, sequencer) will give me
the "best" sound quality?" I wince because I imagine the person going out,
buying a $1000 preamp and the stick it in their studio and connect the mic
and crestfallenly realize it "sounds the same" "maybe a little better" um,
"hard to tell". Why the heck is that? The main reason is that it is all a big
system and its only as pure as its dirtiest pan pot. One humming cable can
obliterate the gains made by otherwise great sounding gear. A pristine mic
preamp connected digitally to a soundcard with the jitters will be defeated.
And it is not just gear. It's technique. Someone with a $2,000 Neumann
U87 and a top of the line Great River preamp will still have an awful
recording if the mic is not placed well, or if the room brings out undesirable
characteristics. If you apply to much compression on a vocal while
recording its going to sound bad no matter what gear you have. This is all
to say, positively, that when you know how to apply eq, processors and
plugins you will be on your way to achieving a mix of high sound quality.
This brings me to the point. A person who really knows and works with
their gear can deliver greater sound quality at 16 bit with lesser gear than a
person who just plugs in and goes on top of the line 24/96 recorders.
Technique is more important that bits and rates. And there is only one
thing more important that technique.
The most important point is saved for last, and I've said it before, and will
again. Great music recorded on a crappy cassette deck will win more
hearts than a turd polished at 24/192. Your talent is more important than
anything else, and you can't buy that. So remember to practice, work on
your technique, and when the pearly gates of Sound Quality open, then
consider 24/96 as you would a nice finish on a well-crafted piece.

FORUM:
Ok, short version. Yes, 24/96 or even 24/44.1 is better then 16 bit anything.
Its purely mathematical and since it is late I do not want to get into all the
particulars but if you are able to capture at 24 bits then I say definitely do
that and only dither down to 16 bits for the final master. This means if you
are having someone else do the mastering, send him or her the 24 bit files.
6

As far as sampling rate goes, going above 44.1 KHz has its advantages.
For instance if you would like to have your music ready for the DVD 5.1
surround audio standard then 24/96 is definitely the way to go or even
24/192 for DVD stereo. These sampling rates arent practical for everyone
in a home studio environment due to HD space limitations, hardware and
software limitations in some cases.
However, this is not necessary to make great sounding music since the
human hearing range does not extend that high into the frequency range
anyway. So 44.1 KHz is acceptable since the actual audio frequency this
will produce tops out at close to the limit of human hearing on the high end
of the frequency spectrum. Again, I do not want to get too technical this late
at night. However, dynamic range is very different. Granted 24 bits
represents a noise floor well below the hearing range. 16 bits has a noise
floor down at the bottom of the range of human hearing as well. However,
this is closer to the border of human hearing. Now, since everything you are
recording and mixing is digital, this means everything is reduced to
mathematical equations. Every time you make any gain change (this
includes normalization, EQing, faders, compression,etc) the dynamic
range is being recalculated and a small degree of error is being introduced
to the pristine and pure signal you started with.

Everything below the LSB (least significant bit) is truncated and that bit is
rounded up or down. Now you can imagine with all the FX you apply that
alter the gain of the original signal as well as any fader changes etc that
this number gets recalculated and rounded a lot. Eventually this can start to
affect the higher order bits and you can start to hear audible artifacts as a
result. Now if your LSB is the 16th bit then you are losing very valuable
data as the rounding errors increase and each effect is recalculating off the
previous effects result, which only compounds the errors until you get
artifacts. However if your LSB is the 24th bit then things change. Yes you
will still have rounding errors but first off, they will be below the range of
human hearing and second, these last 8 bits will be gone anyway once you
dither down to 16 bits.
What does this mean? It means that all the rounding errors are taking place
below the 16th bit and your higher 16 bits, the ones that will actually be
burned onto you final CD, will be much closer to the actual number (which
7

is what digital music is) they should be because they will not have been
rounded. These pristine 16 bits need to be preserved as much as possible.
So, to shorten this up, use 24 bits if you can and stay there until the final
pre has been mastered and then dither the final master to 16 bits (meaning
nothing else has to be done to it).
*************************************************************
*************************************************************

There's a lot of confusion about "why" you would record, mix, and process
at 24/96 when CDs are just 16/44. The simple answer is "numbers that are
more accurate." Filters and processors are merely mathematical
calculations that alter the digital data that represents your music. If you
process a file that is 24/96, you will receive a more accurate result than
performing the same process on a 16/44 file.
So, instead of processing a 16/44 file numerous times (which would require
rounding off "numbers" each time) you process at 24/96 and then convert
down to 16/44; this limits data accuracy loss.
There's more to it than that; a sampling rate of 44 does not accurately
capture or re-create the complex frequencies generated by real guitars.
Look around this site for some of the other explanations and reasons for
recording at high bit and sample rates.
*************************************************************
*************************************************************

I normally work in 44/24 and dither to 44/16 as the final step. I recently took
some samples from an old vinyl record, and after recording at the usual
44/24 I decided that now was the time to start working in 96/24 so I rerecorded it.
I asked my girlfriend to listen to both versions (not telling her which was
which) and she complained that the 96k version hurt her ears with too
much high end. I had to agree that the 96k version did seem to be "too
clear" in the highs, but I think it's probably just a matter of getting used to it.
8

nonetheless, I switched back to 44/24 for now, because i don't want any
surprises with it sounding different after I dither it down.

*************************************************************
*************************************************************

I would simply like to point some things out for consideration. What type of
guitar are we talking about here and what frequencies are we missing and
should we even care?
Right now even industry pros are not in total agreement as to whether
sampling rates higher then 48 KHz are of any benefit at all since the actual
frequency range reproduced goes above that of the human hearing range
anyway (statistically speaking), and the standard for audio is still the CD
standard (44.1 KHz, 16 bits). Now every human being is different. There
may be some mutants out there that can hear above 20 KHz but not by
much usually. The range of human hearing is generally considered to be 20
20,000 Hz. With a sampling rate of 44.1 you have an effective frequency
response of up to 22.05 KHz (above what most people will statistically be
able to perceive as sound). 48 KHz is considered a safe sampling rate
since this takes us up to 24 KHz of actual reproduced frequency before we
get a severe roll off. Frequencies this high are generally considered to be
inaudible (this should even be outside the range of our mutant friends but
you never know). The general argument is whether or not people will even
hear the difference sampling at a higher sampling rate will provide. If
there is a noticeable, audible difference at 96/24 it is most likely due to an
exaggeration of the frequencies in the upper mid and high band frequency
range due to the movement of the steep low pass digital filter to a higher
cut off frequency, and the added dynamic range, which gives the recorded
material a more natural sounding decay and a lower noise floor (I wont
bother getting into the actual electronics and theoretical versus real world
performance right now).
Now, a lot of people embrace 96 KHz because they believe it helps to
reproduce that certain quality of air that is usually associated only with
good high quality analog tape. Analog tape can capture well up past 40
9

KHz. Your high end goes up about an octave and you get a smoother roll
off. Since 96 KHz has a Nyquist frequency of 48 KHz you get a frequency
range closer to analog tape and because the filter is higher up you get a
smoother roll off up to well past the inaudible range of human hearing
(especially with 24 bits of headroom). What is it about this that seems to
sound better? Did our hearing improve? Most likely the reason higher
sampling rates sound better is due to the filter being placed at a higher
frequency and thus moving certain problems such as ringing and phase
shifting way outside the range of human hearing. Since the filter is so high
it allows for a smoother roll off of the higher end of the frequency spectrum
and it is this smoother roll off that is actually producing the sense of air,
brightness and overall better sound quality we seem to be hearing. The
interesting thing here is that many have said that even upsampled material
from 48 KHz to 96 KHz sounds just as good as a sample actually recorded
at 96 KHz with no noticeable difference. Once again this shows that the
increased frequency is not responsible for better sound since we cannot
hear it anyway. This lends credence to the idea that a gentler filter with a
smoother roll off and the fact that artifacts are moved way up into the
inaudible range are responsible for the difference in quality we detect.
Especially when you consider the fact that upsampling does not restore lost
frequency.
Now all that being said, lets be practical. As of now, the standard sampling
rate for audio is 44.1 KHz. So we have to ask ourselves if our end result will
be a CD that is 44.1 KHz/16 bits, is there really any benefit to higher
sampling rates and word lengths? Im not going to go into dynamic range
since I already offered an explanation for that earlier. To help determine
whether or not higher sampling rates are better we have to look at the gear
we will be using among other things. Since recording guitar is the thing that
sparked my response in the first place, I am going to focus on this.

Now, when we talk about recording guitars, we are talking about a lot of
things. One of the main things that must be considered is the mic being
used (in addition to its placement among other things). The point of
concern here is the frequency response of the mic. I have not seen many
mics that respond to a frequency range up to 48KHz. Even some of the
best mics (i.e. Neumann U87) have a frequency response of 20 Hz to 20
KHz. After this the frequency response is attenuated and rolls off out of the
10

human hearing range. Now if I am micing a cab, what frequency range do

you think I am capturing? It doesnt matter if I set my parameters to a
higher sampling rate (such as 96 KHz) or not, once the frequency is lost, it
cannot be regained simply by sampling at a higher rate. Not to mention that
most of the sound we associate with good sounding guitar in todays power
rock is the sound of a miced cab. Many mics color the sound a certain way
and often it is the way they capture and color the mid range frequencies
that determine whether or not the guitar tracks will have that pro sound.
The natural filtering of mics helps to create that sound we all want when
recording guitar. A staple and favorite for a lot of people, the SM57, has a
frequency response of 40 Hz to 15 KHz. That tops out well under the
highest frequency the human ear can detect, and that is definitely under
22.05 KHz (the highest frequency represented in a 44.1 KHz sample). This
is not even taking into account any additional frequency filtering that will
take place once you start to EQ the tracks.
Another link in the chain is your outboard gear. If you are using a pedal of
some sort (usually digital) then you must consider its sampling rate and
word length in order to determine what frequency range and dynamic range
you are actually capturing. Most of these pedals sample at 44.1 KHz and
most newer ones have 24 bit AD/DA converters. Once again, this gives us
a frequency response that tops out at 22.05 KHz. Anything above this is
lost and cannot be regained simply by setting our DAW to capture at a
higher sampling rate.

What about running direct? Most people do not like the sound of direct
recorded guitars and there are a few reasons for this. One of them is the
over exaggeration of the higher end frequencies. Since you are not micing
a cab you are losing a lot of frequency filtration and coloring that happens
at the different stages (the amp, the speakers and the mic). So you end up
with a squeaky, gritty, sharp, dry and generally unpleasant tone, and all of
this with a top frequency of 22.05 KHz. What most people then do to try to
get around this is to use some sort of amp emulator and cab emulator to
add the missing warmth lost due to recording direct. These are merely
filters attempting to recreate the color that real amps and real miced cabs
add to recorded guitar tracks. This means emulating the frequency
response and filtration of a real mic. Once again, what frequencies do you
think you will end up with?
11

So the real question is how much of these sparkling high frequencies do I

need or want? Will it color the sound favorably or even noticeably? How
much of the frequency spectrum will still be there once the track is
converted to 44.1 KHz? And perhaps more importantly, how much of these
high frequencies am I actually capturing? The debate goes on as to
whether there is any real benefit to higher sampling rates (especially as
long as CD audio is the standard). Whether or not there is the future seems
to be heading in that direction with DVD audio. There is nothing wrong with
setting your DAW to capture at higher sampling rates if you are able to do
so. Whether or not there is truly a noticeable difference due to the higher
sampling rate, well thats another story.

12