by Roger Nichols 11/1/92
Every once in a while there is a quantum leap in a particular field because
some one wasn't paying attention when he was told that it couldn't be done
or there wouldn't be a use for what he was building. I guess that is what
happened when Edison said that someday every household would have a telephone
and when Tesla said that fluorescent lights would provide for a more efficient
use of electricity.
Well there is another one of those guys, a Quantum Leaper if there ever
was one, John Meyer. He was told that nobody would want a perfect speaker
because it wouldn't sound good. He came up with the HD-1. Most recently
he decided that there should be a way to test the frequency response, signal
to noise ratio and the distortion of an audio component by playing music
through it instead of tones or pink noise. People laughed. They not only
told him that it couldn't be done, but that there was no reason for doing
it. Well, it turns out that the reason some people didn't want things tested
this way was because their devices would fail the scrutiny of this new test
instrument.
Enter SIM(TM), Source Independent Measurement. At first glance, you ask
yourself why. After a short exposure to what the system reveals, you ask
yourself why it took so long before there was a SIM machine.
Ever since I started measuring pieces of audio equipment, I've had the sneaking
suspicion that the test equipment wasn't telling me everything I wanted
to know. How come two pieces of gear could measure identically and yet sound
so different? The systems look perfect when they are fed pink noise, but
that is not usually my favorite thing to listen to, except at the end of
a long party when I can't get all the guests to leave.
John Meyer has been thinking about the same thing for a long time too. The
difference was that John was smart enough to be able to figure out how to
measure audio components while music was playing through the system instead
of test tones. I, however, am at least smart enough to know this is the
machine to use to find out how a piece of equipment is really performing.
I wonder. If John Meyer had been around before the distribution of electricity,
would he have developed a gas powered SIM machine?
The process of testing audio gear can be very simple. When you test an amplifier
for frequency response using tones, you compare the level of the original
tone going in with level of the tone coming out. If the comparison is identical
over the entire audio band, then the frequency response is flat. The tone
sweep only tests how the gear responds to one frequency at a time, and not
necessarily how the device will react when subjected to music. Music contains
many frequencies at many different levels in various harmonic relationships,
all happening at once. Pink noise provides us with a noise source that has
energy spread over the entire audio spectrum. You could put pink noise in
one end and measure the energy spectrum on the signal coming out, but that
wouldn't give you much more of an insight into the equipment's reaction
to music.
As the name indicates, the SIM machine doesn't care what you give it, the
measurement process is independent of the source. It compares the GOZINTA
with the GOZOUTA. It analyzes the information presented and displays the
results on the video screen.
In the Lab mode SIM compares two signals, the input and the output, of the
device under test. The on-screen menu lets you select the reference channel,
either the internal generator or the front panel input, and the measurement
channel, which could be the output of the piece of gear under test or the
microphone input in the case of speaker testing.
The screen is divided into three major areas. The top two displays depend
on the measurement being performed.
In the spectrum mode, the top graphic display shows the spectrum of the
source and the lower graphic display shows the spectrum out of the device
being measured. During any measurement you can select the amount of data
averaging in the display.
In the frequency response mode, the top graphic display shows a frequency
response trace and signal to noise ratio or coherence trace. The S/N (coherence)
display is graphed against frequency so you can see exactly where the noise
lies. The bottom graphic display shows the overall phase response of the
device under test. There is a screen cursor that can be placed on any trace
in the graphic area. The cursor values are displayed in the data area of
the screen.
The data area at the bottom portion of the screen displays setup information
such as channel gain, cursor data, name of file being displayed, display
offset values, reference source and number of vectors averaged for the particular
test being performed. All parameters necessary to re-create the test are
displayed in this area. When the graphs are printed out to a printer, all
of the test parameters are printed as well.
A unique measurement that SIM performs is automatic delay finding. In the
delay finder mode, SIM compares the source and measurement channels and
figures out how much delay there is in the signal path. The range is 0-983
milliseconds in 20 microsecond steps. It has never been wrong yet. It does
it by performing FFT/IFT pairs on the data. It requires either pink noise
or music to figure it out. It wouldn't be fair sending it a sine wave, now
would it? After SIM determines the delay, one key transfers the setting
to the internal delay line. The SIM has now synchronized the reference and
measurement channels.
During frequency response measurements the data can be averaged over multiple
samples. The choices are 1, 2, 4, 8, 16 and continuous. An additional choice
is whether the averaging is RMS (level only) averaging or vector averaging,
which takes phase relationship into account during the averaging. The data
is processed in band as narrow as 1/24th of an octave, so small anomalies
will show up readily.
SIM is based around the AT&T DSP32C floating point signal processing
chip. This is the same chip that is used in the DisQ digital console, the
SADiE hard disk editing system and the CEDAR single ended noise reduction
system. There are three DSP32Cs in SIM. The DSP housekeeping is done by
an 80486 computer running at 33mHz with 16 megabytes of memory, a 109 megabyte
removable hard disk and a Super VGA graphics display. Meyer has custom built
the A/D convertor boards that supply data to the DSP32Cs. That data is supplied
directly to the DSP chips without passing through the PC buss.
The computer is shock mounted and housed in an industrial strength enclosure.
The cooling fans have been replaced with quiet ones so that there is minimal
additional noise generated in the measurement environment. There is a special
card brace that holds the cards in their slots so that you don't have to
open up the box and re-seat the cards every time you move the SIM machine
to another location.
The rear of the machine contains connectors for power, video display, printer,
mouse, optional microphone switcher and the main test I/O. A military style
multi-pin connector attaches to an eight line snake interface to the external
audio equipment. These connections go to console out, EQ in, EQ out and
monitor in. These connections are switched internally to the correct reference
and measurement inputs of the system. There is also an XLR output on the
rear from the internal signal generator.
The front panel displays metering for reference, measurement and microphone
inputs. There are also input connectors on the front panel for reference,
measurement and microphone signals. There are connectors for the internal
generator, a set of headphones so you can listen to what is being measured,
and for the computer keyboard. Controls on the front panel include the power
switch, computer reset, 48 volt phantom supply to the microphone, headphone
level, internal generator level and sine wave oscillator frequency. The
hard disk is removable by pulling it out of the front panel, thus allowing
multiple users to have their own data and test setups without disturbing
anyone else's.
SIM was originally designed for the sound reinforcement business. In the
past the sound reinforcement companies would equalize the venue during the
day, long before any music was played. During sound check, a few final "tweaks"
would be done "by ear" based on what the sound technician heard.
When the show started, the room sounded quite a bit different because the
hall was now filled up with sound absorbing people. It took a few songs
before the sound technician could get things back under control, and it
had to be done "by ear" with maybe the aid of a spectrum display
to show how much energy at what frequencies was reaching the reference mic.
With SIM, there are three major advantages. One; you can measure the system
with speakers on only a short time and store the resultant curves. With
the speakers turned off, you can still send signal through the EQ and display
the 1/EQ curve on the display. Now you adjust the EQ curve to match the
response curve. When the speakers are turned back on during sound check,
the system is already set to go. Two; when the audience fills up the hall
the sound changes. Even the background music played before the concert starts
is enough to enable you to adjust the EQ. When the show starts, you are
already 99% there and with just a couple of twiddles (technical term) of
the EQ knobs you're done. Three; The top display on the graph is a red line
labeled "Coherence". If any anomalies show up during the show,
such as a power amp crapping out or a speaker clipping, it will instantly
show up on the display. Since SIM is comparing what is being sent from the
console with what is coming out of the speakers, a distorted keyboard or
guitar on stage will not show up as a problem in the PA system.
I have been SIMing everything that moves. The ADAT passed with flying colors.
It looks a lot better than the 3M digital machines that I love so much.
The 3M machine has a 180 degree phase shift at 18kHz and is 270 degrees
out by 20kHz. The Sony 3348 has the same phase shift, but it doesn't happen
until up above 20kHz. Analog EQs really show their phase shift when you
crank in a lot of boost in the high end. The Meyer CP-10 EQ was flat as
a ruler, no matter how much boost or cut was cranked in.
When looking at digital machines, you can see the noise increase as you
get up toward the nyquist limit. You can see how different dithering schemes
used by convertor manufacturers effect the high frequency noise. In some
cases, the S/N ratio at the higher frequencies drops to only 40dB while
in the midrange the S/N ratio is better than 100dB.
To measure THD (Total Harmonic Distortion), you select the spectrum display
and the internal oscillator as the source. The cursor will jump to the peak
of the spectrum display and the THD can be instantly read in the bottom
data window.
Walter Becker played the guitar solos on Donald Fagen's new album. We didn't
want to spend half an eternity getting just the right sound on the right
guitar amp and have Walter be to wiped out to play the solo. We also didn't
want to decide later that the guitar amp sound wasn't quite right and get
stuck with what we recorded. We chose to record the guitar direct and to
run it out through an amp later, after more of the other overdubs were done.
I ran the guitar signal out through an amp. In the control room we listened
to a mixture of the direct signal and the microphone in front of the amp.
Both signals sounded the way we wanted by themselves, but sounded phasey
when mixed together. Flipping the phase button didn't work because the signals
weren't 180 degrees out, they were something less.
SIM to the rescue. Using the delay finder, I fed the direct signal into
the reference input and the mic return from the amp into the measurement
input. Bingo, SIM told me that the mic signal was 1.48 milliseconds behind
the direct signal, just enough of a delay to cause problems in the high
frequency range. I have tried to use a dual trace scope to find the delay,
but it is to hard when the signal is not repetitive.
Now that I know the delay, I can move the amp recording earlier to make
it coincident with the direct signal. I could have laid the amp track off
to time code DAT dialed in the correct offset and bounced it back, but I
chose to use the "digital advance out" feature of the Sony 3348
to move the amp track earlier. I just selected 1.48 millisecond advance
out (047hex) and bounced it to another track. The whole process, including
measurement, took 15 minutes.
The same problem exists when recording the amp and direct sounds together,
there is a delay because of the speed of sound through the air. Once you
measure the delay, then you could also run the direct signal through a digital
delay to null out the time difference. Just this little difference makes
the guitar sound a lot better.
Oh, by the way. Steve StCroix says that his cat is one of his favorite pieces
of test equipment in the studio. He says that if his cat leaves the control
room then something is wrong with the sound. I SIMed my cat. She has a hair
ball around 12 kittyHertz. My wife has an effect on the sound when she is
in the control room. I SIMed her. It turns out she has acute phase on the
top, a couple of humps in the upper midrange and a nice round bottom end.
I have to go now, I'm late for an appointment with my doctor. I'm having
a SIM machine permanently grafted to my hip. Do you think anyone will notice?