Recording - March 2016 - 63

so that high-frequency interference from
the digital circuits doesn't creep into the
audio chips with bad results.
Let's get back to those pro studios...
why couldn't they choose their favorite
outboard mic preamps, then run the outputs into one of the all-in-one interfaces?
After all, most of the interfaces have
"mic/line" inputs, don't they?
Well, not really... and therein lies a tale.
The line that wasn't
Some interface boxes have completely
separate microphone and line inputs;
MOTU's Ultralite AVB, for example, has
two mic inputs along with six separate
balanced line-only inputs. (It also has
eight line-level outputs, which allows for
some flexible and creative signal routing.) Many other lower-budget interfaces, however, don't handle mic and

and you get the possibility of galvanic
isolation of instuments or amps from the
console or preamp's circuit (see the sidebar "Going Direct" in article 6 of this
series, back in June 2015). That's a
worthwhile tradeoff.
Line-only inputs are usually recognizable by not having a gain control that
goes with them, as with the line-only
A/D converters described earlier; a
given level at the line input generates a
predetermined level in the digital signal.
(For more on this, see the sidebar "Yes,
But HOW Big?")
The reduced levels in a padded-down
"mic/line" input, however, offer very little advantage other than low cost. And
running every signal through the mic
preamps means that coloration or noise
in the preamp circuits will be superimposed on any line-level signals that pass

Yes, But HOW Big?
I've mentioned in passing that the "nominal" level of professional analog audio signals is +4 dBu, which works out to
0.775 Volts RMS. That's the level that appears at the output of
our big console when the VU meters read zero.
You may be wondering how that translates into a digital signal level-in other words, what the standard is for converting
a nominal professional-level analog signal into a digital one.
The answer is very simple: there isn't any standard.
Digital signals are usually measured in dBFS, which translates
as decibels below full scale. Full scale is the maximum level that
can be recorded; the digital equivalent is 111111111111111
in a 16-bit system (add eight more ones for a 24-bit system).
Everything has to be below the maximum, because once a digital signal is maxed out, there ain't no more.
But back to the original question: what digital level will be
produced when you feed, say, a sine wave at +4 dBu into an
A/D converter? Again I tell you: there is no standard for that.
Each manufacturer is free to set its own standard, so that +4 dBu
line inputs separately, but double them
up in a "mic/line" input. Why do I keep
using quotation marks around that
phrase?
Many interfaces (and not a few consoles) don't really have separate linelevel inputs. Instead, they take a linelevel signal and pad it down, using an
attenuator, to microphone level, then
route it through the preamp circuits just
as if it were a mic signal. Why is this so
widespread? The three resistors that
make a line-to-mic-level pad only cost a
few cents, whereas a real line-level input
circuit costs more. But...
There's a basic rule in mountain-climbing,
quoted by the great Colin Fletcher:
"Never lose altitude unnecessarily." I feel
the same way about audio signal level: I
don't like to reduce it unnecessarily. Of
course, sometimes it's unavoidable;
when connecting a keyboard or amplifier to a direct box, the signal is typically
reduced to mic level. But in that case,
along with the level drop, you get a balanced signal that can travel down long
cables without losses or noise pickup,

the Auroras. (There are also versions of
the Aurora 16 available which connect
to computers via USB and Thunderbolt.)
I recorded the signal in 32-bit floatingpoint (which is really 24-bit audio formatted in a way that makes processing
easier), with a sample rate of 44,100
samples per second. Why didn't I go to
a higher sample rate? Well, I knew that
the product of the recording would be
an audio CD or a download using
44,100 samples/sec; if I recorded at a
higher rate, I would either have to convert it back to analog and mix it through
the console, then convert back to digital,
or do a sample rate conversion.
I didn't really want to do either-and
there, perhaps, I was working by force
of habit. Until recently, A/D and D/A
conversions were the least audibly transparent processes in audio, and sample

signal might produce -14 dBFS on one maker's converter, -18
dBFS on another maker's, and -20 dBFS on a third maker's
box. The Audio Engineering Society's Standards Committee has
been wrestling with the question for a couple of decades, with
no decision. The result: total anarchy.
There are a few local standards. If you're producing something
for National Public Radio, for example, they insist that 0 VU, the
nominal level which translates to +4 dBu at a mixer's output, must
be represented on a digital recording by a signal with a peak
level of -20 dBFS. That informal standard holds through most of
the film-sound world, too, as a generally-accepted best practice.
It's actually a pretty sensible level; research has shown that audio
peaks very seldom reach more tham 20 dB above nominal level,
so a nominal level of -20 dBFS means the digitized signal will
hardly ever reach the maximum. And since digital maximum represents a particularly ugly-sounding form of clipping, that means setting a converter's internal scale to make +4 dBu in (0 VU on most
pro gear) translate to -20 dBFS is likely to produce good results
when recording to a digital multitrack... but sorry: it isn't official.

through it. I mentioned several times that
with mic preamps, you really get what
you pay for-and a pro studio probably
won't want a cheap circuit's coloration
or noise added to the sound of the higher-quality mic preamps they choose. If I
was buying an all-in-one interface, I'd
be sure to get one that offered dedicated line-level inputs separate from the mic
inputs.
Figure 2 shows a block diagram of an
interface with separate inputs for line
and mic signals, while Figure 3 shows
an interface with "mic/line" inputs.
Recording the First Steps example
session
I've left this until last because it was
pretty straightforward. The university's
studio has a couple of Lynx Aurora 16
boxes, which incorporate both A/D and
D/A converters for 16 channels of
audio. These boxes, which accept balanced line-level signals only, talk to the
computer via digital cables using the
AES/EBU standard; the computer has a
Lynx AES16 PCIe card in it to talk with

rate conversion added painfully apparent artifacts to the signal. Neither of
those criticisms is entirely valid these
days; A/D and D/A through the Lynx
converters is the best PCM digital I've
ever heard, while companies like
iZotope make sample rate converters
that are well-nigh inaudible.
Still, there's an old habit dating back
to the analog days: do as little to the signal as possible, and do only what's necessary. Did I want to end up at 44,100?
Very well, I'd start at 44,100, and stay
there. And I would convert between analog and digital as few times as possible,
since that's arguably the most vulnerable
point in the recording process.
So that's what I did on the Ottoman
Underground session. And, after converting the signal to digital, I recorded it on
a computer. Next month we'll talk about
that; stay tuned!
Paul J. Stamler (stamler@recording
mag.com) is a recording musician, engineer, educator, and collector of bizarre
vintage recordings, living in St. Louis.
RECORDING March 2016

Table of Contents for the Digital Edition of Recording - March 2016