Word
Clock Tests
DAVID RIDEAU discovers how different clock sources
can alter converter sounds and how striking the results are.
I am an analogue man. I know this is a strange way to start an
article about digital gear, but I feel it’s an important point. I’ve
been in this business long enough to remember when the digital audio
medium did not exist and the professional recording process was a simple
one. These times fostered sensibilities that made the audio’s path top
priority and an engineer’s job (design or recording) was to enable the
music to speak without interference. Now with the digital revolution
there have been quantum leaps in the storage and manipulation of audio,
but not always with the highest regard for audio quality. With this
philosophy I am constantly looking for ways to improve the digital
signal with analogue serving as my holy grail. This quest fostered this
digital clock comparison.
We have all heard from the manufacturers of digital audio equipment
on how important an accurate clock is in the performance of their
products. Unfortunately, we have also all experienced the dreaded
‘ticks’ that occur when a device finds it’s unable to function due
to the quality of its digital clock source. Besides this point where
jitter is so high that a unit can no longer function properly, there are
also more subtle abnormalities that can harm a signal’s sound quality.
I have heard many levels of performance in the multitude of digital rigs
that reside here in the LA area, but how much of that can be attributed
to clocking issues?
Tests
Our idea (I enlisted a couple of colleagues) was to take a sampling
of the most popular clock sources and put them through a series of bench
tests by an independent lab, followed by a series of listening tests to
see what we could discover. The test consisted of four units: the
Aardvark Aardsync II, Digidesign Universal Slave Driver (USD), an Apogee
PSX100, and the new Rosendahl Nanosyncs.
We purposely ran our listening test before reading any bench test
results to avoid any bias. The test was held at Westlake Audio and it
began by creating a live mix from an analogue two-inch Studer 24-track
machine through an SSL G console. The mix included real drums, vocals
with dense reverb tails and lots of complex stereo information — all
the material that suffers most during digital conversion. At this point
we fed the stereo bus to one of the three A/D and D/A converter sets we
selected (Digidesign 888, Apogee PSX100 and the Sony 3348HR) that were
alternately clocked by our various test units. Using an external patch
point to return the twice-converted audio to the console made it easy to
A/B this signal with the live mix.
In all our listening tests we were careful to make sure that all
clock runs were as short as possible with the same high-quality 75 Ohm
cable. Every other variable, such as power source, was all uniform to
the best of our abilities. Keep in mind our goal was to simulate a
real-world professional working environment, so we approached the set-up
as we would for any other session, except to ensure that all clocks were
on a level playing field.
Results
The results were surprising in many ways. First, it was amazing how
different the clock source could make the converters sound and how easy
it was to hear these differences! These were so dramatic that we had
other engineers who were in the building listen and they too were
amazed.
I rated
my findings with the standard being how well the signal compared to the
live stereo bus (a perfect ten rating). For example, with the 888 being
fed 256kHz directly, the USD scored a 6.8; the Nanosyncs fared much
better with a 7.8; and the Aardsync was best with an 8.5 (the PSX 100
has no 256kHz output).
The conclusions were interesting. In general, the Aardsync
consistently performed best with the Apogee and Nanosyncs next and the
USD last. I realise that in some instances we are comparing apples and
oranges (ie. the USD does much more than generate word clock), but our
goal was to create a real-world environment that investigates how much a
better clock source means in the performance of hardware that our
readers may already own. As an added bonus, we had the opportunity to
compare converters under the same conditions producing a test within the
test.
The next step was to compare the labs’ bench tests to see if there
was any correlation between what we heard and the data received. At
first glance the test results were quite confusing. First, there seemed
to be no significant difference in jitter between all the units! They
all came in at around 200-240ps range. I called several manufacturers
and found out that jitter can be measured in countless ways: RMS with
many time interval options or peak readings, both of which can be done
at any number of frequencies. In general, the ear is more critical to
clocking errors at higher frequencies where jitter can cloud the sound
stage, so some sort of weighted curve most likely would produce results
that better relate to how we perceive clocking errors. Whatever approach
our lab used (unknown), the jitter results we’re non conclusive. No
wonder audio professionals are so confused; until there are standards,
jitter specifications are virtually useless.
Thankfully, our tech tested further to see if there were other areas
where she could find hard data of any variation between the units. Where
there seemed to be a significant difference in results could be seen on
a digital scope across the power spectrum. With the clock source fed
directly into a digital scope, it was possible to see exactly how clean
the signal was.
In the tests, we found that the output of all the clocks exhibited a
power spectrum, which contained, in addition to the primary component at
the base frequency, n = 48kHz or 12.288MHz, odd harmonic components
occurring at regular intervals of 3n, 5n, 7n, etc., which are directly
related to the square wave of the word clock (see Figure 1). Additional
smaller amplitude peaks, 60 to 70dB below the primary signal were also
observed. We found that the main difference between the different clocks
was in the relationship between these smaller higher-frequency
components and the main signal. In the case of the Aardvark, we found
that the smaller components appeared at regular intervals, at n/2, or
half way between the harmonic peaks (see Figure 2). Therefore, we
believe these to be harmonically related to the base frequency and
therefore inaudible. In the case of the other clocks, the USD in
particular, the smaller components occurred randomly between the main
harmonic peaks (see Figure 1, small peaks), which could cause audible
artefacts. These artefacts seem to have a direct relation to our
listening test results.
Conclusion
The manufacturers of all these products have created fine equipment.
The Digidesign Universal Slave Driver (USD) is a high-quality,
multi-purpose synchronisation device designed for use with TDM-based Pro
Tools systems. It supports all major industry standard clocks and
formats and can also act as a standalone timecode reader/generator. It
can provide solutions for many audio professionals’ problems that are
encountered every day. At the price list this unit is still a good buy
and fills a specific need in our industry. Obviously many hits have been
recorded on Pro Tools with this unit as the clock source, but I
personally would invest the extra cash to incorporate an additional
high-quality clock into the same rig to achieve the next level of
performance.
The Rosendahl Nanosyncs is a real contender in the field of digital
clock sources. With six BNC clock outs, Word Clock in, SPDIF and AES
outs, Video in and out, Blackburst generator, and the ability to resolve
to LTC all as standard features, the price makes this unit a perfect
solution for someone in need of a lot of quality WC outs at a great
price.
The Apogee PSX 100 also qualifies in the most value for money
department. It provides quality 24-bit 96kHz A/D and D/A converters,
ADAT and TDIF interfaces, AES SPDIF I/O, bit splitting and all with a
great individual clock source.
The Aardvark Aardsync II is an amazing unit! Time after time it
turned a good sounding signal into a great one. The common failures of
digital conversion were always minimised with the Aardsync leaving an
accurate, focussed, open, uncompressed, wide interpretation of the
analogue stereo picture. Not the most economical choice compared to our
other entries (four WC BNC outs but no LTC resolve, Video Lock, 96kHz or
Blackburst generator without considerably more of an investment), but at
that price the Aardvark provided excellent results with all of the
converters.
The purpose of these tests was to give readers an idea of how
important clocking issues are in our age of modern recording. I also
hope it can help encourage manufacturers to provide buyers with a more
standard jitter test that would only help clear up the cloud of
confusion that surrounds this specification. The listening part of the
test is subjective, and I encourage buyers to audition clocks in their
own environment. I guarantee that the time and money you spend is well
worth the investment!
|