Shouldn't 24-bit sound better than 16-bit?

Tonight I recorded an FM radio show from my tuner onto the computer as a
.wav/uncompressed PCM using Sound Forge.

Before the show started, I changed a setting so that it would record in
24-bit rather than 16-bit, thinking this would improve the quality.

Richard Crowley...

There is almost nothing any of us can do to "improve"
the quality of anything. You can put a quart of water into
a gallon jug, but it is still only a quart.

However, the end result sounds muddy, completely lacking in top end - in
fact the online stream would have sounded better.

Adam Sampson...

As a first shot at figuring out what's wrong, what happens if you
convert your 24-bit recording to a 16-bit file then play it back?

Arny Krueger...

We've done that with his online file, and found that the problem was indeed
in the recording.

That'll at least tell you if it's recording or playback (or both)
that's broken. You could also try playing back the 24-bit recording
using a different bit of software.

Arny Krueger...

We've done that with his online file, and found that the problem was indeed
in the recording.

I've certainly seen (and written!) software that's screwed up 24-bit
playback by using the wrong sample format for output, but the
resulting distortion is usually a lot less subtle than you describe...

Arny Krueger...

More likely - the card itself has a serious problem when run in 24 bit mode.

Richard Crowley...

OK. Are you saying that it does NOT sound "muddy"
when you record in 16 bit? If you don't do all the
corners of the experiment, you have a non-sequitir.

Arny Krueger...

Does putting muddy water into a clean glass make the muddy water
automagically become pure? Of course not, and that is a good metaphor for
what you did. The muddy water is the FM broadcast, which is full of noise
and distortion on the best day of its life. The clean glass would be your

P...

Maybe to your ears, as an audio perfectionist (and this is not meant as a
putdown!) I just wanted a reasonably acceptable recording at the equivalent
of a 160kbps mp3, or slightly higher. What I got sounded lower than 32kbps.

recording technique - compared to the usual FM broadcast, 16 bits done right
is a very clean glass.

P...

I know what you're saying, but to my ears there was absolutely nothing wrong
with the signal. The resulting recording had very little top end at all,
totally different to the broadcast.

Geoff...

Well you stuffed something up, or your soundcard and/or drivers is broken.

P...

I didn't stuff anything up, I changed one setting. The soundcard and its
drivers are in perfect health, I've since done test records at 16-bit, at

Richard Crowley...

Clearly the soundcard and/or its drivers are NOT in
"perfect health" or you would have never posted your
question in the first place.

P...

Geez, talk about sinking the boot in. Elsewhere in the thread we've now
established what you're saying to be true.

Did you ever mention whether you have done "test
records" at 24-bit? Useful differential diagnosis
procedures suggest trying various combinations to
isolate the specific problem area.

P...

Fair enough, but when one has had a product for a number of years with no
failures of this sort there's no real reason to imagine there'll be one.

OTOH, unless you have very good sources, (and FM
radio is clearly NOT one) dunno what recording at
24-bit benefits you anyway? It will get chopped back
to 16 bit when you make your CDs in any case.

P...

OK, OK. It was a whim. One that I regret, but that I would not in my
wildest dreams have ever thought would result in a totally substandard
recording. I can't imagine any reason why I'd deliberately try and sabotage

Richard Crowley...

The problem you are seeing is surprising to all of us.
It is very unusual and I don't think any of us have seen
anything like it before.

a recording that was very important to me.

Geoff...

On the other hand, given a non-broken system, there is no point in NOT
recording at 24 bits, if nothing else to make level-setting a non-issue
(within reason).

dpierce...

In actual practice the advantages to this are substantially
less than one might think. The noise floor of a large number
of digital system sporting both 16- and 24-bit modes is
often not much different, effectively limiting the dynamic
range of the 24-bit mode to maybe 3 dB better than 16-
bit mode.

P...

It's not the 24-bit mode that's the problem. Further testing has revealed
it is perfectly OK at 48000kHz/24-bit but, as we know, hopeless at
44100/24-bit.

Eeyore...

Sample rate conversion error ?

Thus, the "level setting" advantage is at best a minor
one. For that you get to have files that are 50% or 100%

Geoff...

Yes

larger for the same effective audio.

Geoff...

... which is a total non-issue.

dpierce...

I think if I had to ship a 150 MB or a 200 MB file to a remote
location via a network connection that is less than reliable
or is less than lickety split lightening fast, this total non-
issue would start ot become an important issue. If I had to
ship a couple of voice-overs to a mailbox that might have
a size limit imposed on it (as many do). this total non-issue
would become a significant issue.

Geoff...

His problem is related to recording for a specific purpose, not shipping or

dpierce...

It should be noted that your statement was, exactly:

"a total non-issue"

which can be reasonably assumed to mean that it
is not an issue under any circumstances. There are
many circumstances under which it is an issue, even
for the original poster. Consider his statement:

"I should also tell you that the resulting file is
968MB for 63'56" of recording. "

Now, precisely how much of a total non-issue is it
if he wantted to archive this file off to a CD-ROM?
It becomes a MAJOR issue, i.e., impossible.

Consider also several comments made later in
this very same thread:

"Don't upload sizeable files - nobody will bother
to download them. Twenty seconds is plenty to
establish the nature of the problem. Even the
most limited web space should handle that
comfortably."

20 seconds of 24 bit, 44.1 kHz audio takes up 5.3 MB
at packed 3 bytes/sample, 7.0 MB at 4 bytes/sample.
yet only 3.5MB at 16 bit.

Laurence Payne...

Why? It's not any harder to read "23, 45, 57" than it is to read "1,
3, 6".

Then,

I wouldn't upload more than a few seconds of .wav
audio but much of my existing webspace is taken
up with other stuff, and I only get 10MB.

So, it would seem, even for the original poster, it is
far more than "a total non-issue" even in the most
limited sense of this specifc thread's context.

So, IN THIS CONTEXT, you have shown no real
upsides do using 24-bit recording, even ignoring
the OP's apparently broken sound card/driver,
considering that there is no dynamic range
advantage to speak of, with downsides of
substantially larger file sizes, copy times and
transfer limitations, it is most assuredly NOT
"a total non-issue."

Geoff...

As you like.

I find it a total non-issue to me. Which includes regular backing up of
large multi- \track projects. That OK ?

networking.

In your scenario, I would still be inclined to record at 24 bits myself,
then reduce to whatver format is desired (often multiple formats in that
scenario). Easier to set level for no overs, and I don't have to dick with
my standard setup. Rendering to whatever delivery format/s is trivial.

Again, something is broken. The thing is to isdentidy what is broken and
fix it (procedure, settings, software, hardware).

Richard Crowley...

Never a good idea to try a new mode (etc.) on an important
recording. At least run a test recording first to confirm that
everything is running as it is supposed to.

full quality.

What did I do wrong? Is it salvageable or am I stuck with it sounding worse
than an AM radio show? I would be extremely grateful for any help on this.

dpierce...

If you had asked me (well, you kinda did, since I was
listening), I'd actually take your question:

"Shouldn't 24-bit sound better than 16-bit?"

and divide it into two questions:

"For recording high-quality FM radio, shouldn't 24 bit
sound better than 16 bit?"

and the answer to this question is very easy: no, it should
not, assuming proper attention is paid to recording
techniques in both 16 and 24 bit modes. There is simply not
anywhere near enough dynamic range in FM recording to
make 24 bit worth the effort. No, that means, in BOTH cases,
you need to pay attention to your gain structure to make sure
the levels are high enough to capture everything without
dropping into the noise floor of 16 bit or 24 bit (which is,
in practicaility, not much lower than 16 bit), yet not so high
as to risk overload on peaks. And, yes, if your tuner has
too much 19 kHz pilot tone leakage,m rest assured that
your linear uncompressed PCM recording will do an
excellent job of recording it as well, without even breaking
a mild sweat.

So, in brief, for FM radio, use 16 bit. Period.

Your second question should be:

"Why did my attempt at doing at 24 bit fail?"

And the answer to that is MUCH more complicated,
primarily becuase we just don't have enough information.

But, given the answer to the first question, answering
the second question now becomes a low priority, academic
pursuit, one which, if you never get an answer, you'll still
make fine recordings of your FM programs using 16 bits.

But, that's only if you asked me.

Richard Crowley...

Only you can do the differential diagnosis with YOUR
hardware, YOUR software, YOUR source(s) and YOUR
procedures. Experiment with your setup and see where
the problem lies.

Hint: It is extremely unlikely that recording 24-bit causes
something to sound "muddy" where recording 16-bit is fine.

You also didn't mention what bitrate you are recording.
Is the broadcast stereo? Are you using an adequate
filter to remove the 19KHz artifacts of the stereo carrier?

It is also likely useless to record over the air FM in 24 bit.
It was likely 16-bit before it went through their modulator,
transmitter, feedline, tower, antenna, air, and your antenna
and receiver.

isw...

No idea what went wrong, but if everything was set up properly and
worked *perfectly* you should not be able to hear any difference between
the 16 bit and 24 bit versions.

P...

Exactly. That was the only setting that was changed. Of all the responses
so far yours and Don's are closest to the mark. More details to come, when
I respond to earlier threads!

Geoff...

It shouldn't be noticably different at all. You must have had something
stuffed up.

16 bit if recorded properly should be way better than the FM anyway !