[ddots-l] Re: question about microphone gain input
- From: "Omar Binno" <omarbinno@xxxxxxxxx>
- To: <ddots-l@xxxxxxxxxxxxx>
- Date: Fri, 24 Jul 2009 17:43:16 -0400
Bryan,
I do use 24-bit recording, and from what I've read, the RME Fireface's preamps
and converters are pretty high end. Yes, no? Having said that, I notice that I
need to turn the gain input on the fireface to about 75 percent in order to get
this female's voice to reach about -7 on the peak meters in sonar. Any thoughts?
Omar Binno
Website: www.omarbinno.com
AIM: LOD1116
Skype: obinno1
----- Original Message -----
From: Bryan Smart
To: ddots-l@xxxxxxxxxxxxx
Sent: Friday, July 24, 2009 5:22 PM
Subject: [ddots-l] Re: question about microphone gain input
Just wanted to point out that these sorts of situations are one of the
reasons why high-end preamps and 24-bit recording is superior to budget pres
with typical 16-bit recording. With a budget setup, even with a quality mic,
you're always fighting the noise floor (the background hiss/hum/buzz coming
from the analog part of your recording setup), and dithering artifacts (the
extremely quiet digital noise that is the result of attempting to record
signals that are more quiet than the sampling resolution can accurately
represent in digital form).
Here are a few interesting facts about levels and digital recording.
First, about levels. You probably know that the strength of a signal is
measured in DB, with 0 being the loudest possible signal that can be accurately
stored in digital form. As the number decreases, the level drops off on a
logarithmic, not a linear, curve. This is important to know. If, for example,
you lower a signal by 6 DB, the signal will sound half as loud as before. For
each additional 6 DB, the signal will be half again as quiet as before. The
same idea works the other way around: if you raise a volume control in Sonar by
6 DB, the track will sound twice as loud as before.
This fact about levels is directly related to how the signal is digitally
represented. As you might know, the computer represents all sound data in
binary (as a long pattern of 1s and 0s). If you are recording in 16-bit, then
the computer uses a group of 16 1s and 0s (or 16 bits) to describe each digital
snapshot/sample that it records from your audio interface's inputs. If you are
recording in 24-bit, then the computer uses a longer group of 24 1s and 0s to
represent each sample that it takes from the audio interface. Regardless of the
length of the group of 1s and 0s, the first few bits always describe the
portion of the sound that is the loudest, and additional bits describe parts of
the sound that are increasingly more quiet. Specifically, the second bit
describes the part of the sound that is half as loud as the first bit, the
third describes the part of the sound that is half as loud as the second, and
so on. Also, each additional bit has a more difficult time at accurately
representing its portion of the sound. Even cheap converters will encode the
first 12 or so bits accurately, but, beyond this point, the bits are
representing sound that is so quiet, that occasionally what should have been a
0 is stored as a 1, and vice versa. These little errors aren't usually a
problem, since they're so quiet, that you don't hear the errors in the
sound....until you amplify the sound, that is. Even if you have great mics,
great preamps, and great converters, there comes a point where there aren't
enough bits available in the sample to represent sounds that are quiet beyond a
certain point.
So, when you reduce the level of the signal coming in by half, that is a gain
reduction amount of -6 DB, and that also means that the first bit of the
digital sample is no longer used. If you reduce the incoming signal's level by
another half, to a quarter of full strength, that is a total gain reduction of
-12 DB, and now the first two bits aren't used. If you recorded a signal at
this level with only 16-bit resolution, the converters on your audio interface
would only be using the quietest 14 of the 16 available bits that you're using
to record the sound. It would be very likely that a part of what you recorded
at this level was so quiet that it was beyond the lower limit of the converter
to represent in only 16 bits. you could digitally amplify/normalize the
recording back up to 0 DB, and the signal would sound loud, but it is still
only recorded with 14-bit resolution. This is where the digital artifacts come
in. When Sonar amplifies the signal, it basically has to promote bit 3 to bit
1, bit 4 to bit 2, bit 5 to bit 3, etc; all of the bits get shifted to the
left/up. However, Sonar still only recorded 14 bits of signal for each sample,
so it no longer has the last two bits to accurately represent the quietest
detail of your recording. You'll hear some small amount of crackle/fuzz because
the bits that were less accurately represented have been promoted to represent
louder sound than they originally recorded.
This can be bad if you have good converters, but, if you have cheap
converters, it can be horrible. Remember that I said that many cheap converters
don't accurately represent data beyond the first 12 bits or so? Well, if you
have one of these, and you normalize the quiet signal as I described above, you
aren't turning a 16-bit signal in to a 14-bit signal, you're turning a 12-bit
signal in to a 10-bit signal!
I hope that this starts to make it clear why 24-bit is better than 16-bit. If
you recorded an extremely quiet signal, as we discussed above, and had to boost
it by 12 DB, then you'd still be giving up the same two bits. However, if you'd
recorded in 24-bit, instead of recording in 16-bit, then you'd be reducing your
recording with 24-bit resolution down to 22-bit resolution. This means that a
24-bit signal that is amplified by 12 DB still has dramatically high resolution
(about 64 times as much, in fact) than the original unamplified 16-bit signal.
So, in summary, recording in 24-bit allows you to dramatically amplify the
recorded signal without introducing a lot of grainy digital artifacts. When you
record in 16-bit, amplification really isn't a good option.
Now, armed with that info, back to recording technique.
In budget 16-bit world, you get your levels set for your vocalist by trying
to have them sing as loud as possible, and adjusting the gain on the input
channels so that the meters never go over about -3 or -4 during that absolute
loudest note. As you've discovered, if you attempt to boost the recorded signal
after the fact, you increase the level of the dithering noise. It is also
tempting to close-mic the performer since, if you don't, you must turn up the
input channel's gain in order to get the meters close to -3 or -4, and, the
more you turn up the input gain, the noise floor of the preamp also becomes
louder. You can get good results when recording at 16-bit resolution, but it
takes a lot of work to do it well. If you record too quietly, then you'll pay
for it with dithering artifacts when you amplify the recording during the mix.
If you record the signal as hot as possible, then you'll avoid dithering
artifacts, but you'll run a greater chance of clipping during a really good
take. At the same time, if you turn up your preamp to get a good hot level, the
noise floor of the preamp comes up and you hear hiss. If you keep the preamp
turned down, but close-mic the performer, you'll have mic proximity effects to
deal with, and, for vocalists, you'll risk popped Ps, breathing in the mic, and
other hazards that can ruin a take. This is even more difficult when you're
recording several people/instruments at once, since you have lots of input
levels that are near maximum, and could clip at any time. You can use hardware
limiters on the inputs to prevent digital clipping, but they'll still color the
sound. Basically, you must really work and/or spend a lot of money on gear in
order to avoid problems at 16-bit.
However, suppose you had a high-end pre, a high-end interface with quality
analog/digital converters, and recorded at 24-bit resolution. When you were
setting up levels for your vocalists, you could let the levels peek at -10 or
lower. Instead of only 3 or 4 DB of headroom, you'd have 10 DB to work with.
This means that, in order to clip, someone would need to sing or play a note
that was over 2 and a half times louder than the loudest note that they played
when you were setting up levels. This is highly unlikely. Once you've finished
recording, you'll still need to amplify/normalize the recordings, but, even
after boosting the signal by 10 DB, you still will be using a recording with
22-bit resolution.
24-bit recording, while it will allow you to avoid clipping by recording at
lower levels and later amplify the recording with almost no noticeable
artifacts, won't save you from a noisy preamp. Preamps seem to do an
increasingly worse job with cleanly amplifying a signal the more you turn them
up. Besides having a nicer tone (which is a concept that it is hard to
describe), higher quality preamps can produce higher gain for your converters
without yielding a lot of hiss.
You're using a Tascam FW1884. The FW1884's built-in preamps aren't that bad,
but they put out a lot of hiss when you crank them up. You have some options.
First would be to use better preamps. The FW1884 has an ADAT input that will
accept another 8 channels of digital signal from another piece of digital
equipment. You can get a dedicated high quality preamp/digitizer unit and
connect it to the FW1884 through the ADAT port. Presumably, you'd buy a device
that has preamps with greater dynamic range. Then, you could crank up the mics
without hearing lots of hiss from the preamps.
Second, you should reconsider your mics. In the perfect world, you'd use
condenser mics to record your vocals. Good condensers will put out a hot
signal, and have a very low noise floor, so you won't have to crank up the
preamps very far, and, when you do, you'll have to crank them up a long way
before you hear any hiss coming from the mic. Unfortunately, condensers are
very sensitive, and so they're not as nice when you can't isolate the
performers. If you have a vocal booth/closet, then a condenser mic might be
something to think about. If you'll be recording several people in the same
room, then condensers won't work, since they'll pick up everything in the room,
not just what you point them at.
Still, even among dynamic mics, there are levels of dynamic range. If you
must use dynamic mics, then better preamps are probably the way to go.
There is my essay for the week. *smile*
Bryan
------------------------------------------------------------------------------
From: ddots-l-bounce@xxxxxxxxxxxxx [mailto:ddots-l-bounce@xxxxxxxxxxxxx] On
Behalf Of Stacy Blackwell
Sent: Friday, July 24, 2009 2:51 PM
To: ddots-l@xxxxxxxxxxxxx
Subject: [ddots-l] Re: question about microphone gain input
Omar, in addition to the other comments, I have just recorded 4 lead singers
in my band with the levels all supposedly set the same. The male vocals did
not require much trim during playback, but the female singer, well, I am having
to really turn up the trim and now a high hiss is heard when soloed. I think
she was scared of popping the mic and held it too far away. Also, if a person
sings to the side or across the top (depending on the type of mic) instead of
directly into it, the gain and tone can be affected. Trial and error, live and
learn. But I do think that singers, male or female, differ in their abilities
to "project" their voices which can result in different gains. I have also
learned that I should also wear headphones to listen for a constant entering
volume from the singer instead of listening to it after the recording. Keep in
mind that I don't have a "real" studio and have this equipment mainly for my
own personal songwriting and recording, so this was my first time to record
other vocalists. See ya, S.B.
------------------------------------------------------------------------------
From: omarbinno@xxxxxxxxx
To: ddots-l@xxxxxxxxxxxxx
Subject: [ddots-l] question about microphone gain input
Date: Thu, 23 Jul 2009 18:00:26 -0400
Hello,
I have my vocal mic plugged into my Fireface Soundcard. When one of my female
clients records vocals, I end up having to turn up the gain quite a bit.
However, I notice when I speak or sing into the mic, I need the gain up only
half way. Is it usually the case that gain input on the soundcard or mixer
changes from singer to singer, or does it usually stay at the same level.
Thanks.
Omar Binno
Website: www.omarbinno.com
AIM: LOD1116
Skype: obinno1
------------------------------------------------------------------------------
Windows LiveT Hotmail®: Search, add, and share the web's latest sports
videos. Check it out.
Other related posts:
