Owners Manual

Digital Converters
There is a lot of discussion on internet bulletin boards regarding bit
depth, sample rates, human hearing, dither, jitter and word clocks.
Many people claim to hear a difference between 48K, 96K and 192K
and those that percieve those differences seem to generally prefer
192K. Some don't hear the difference. Yours truly has listened to
digital for many years, and hosted big A/D and D/A shoot-outs and
heard various jitter effects and all the subtle differences between
most converters - but - it wasn't until comparing the SLAM!'s DAC
to 'conventional' DACs with CDs that a particular problem with
those conventional DACs became obvious. The SLAM! DAC was
'faster', and had way better imaging and depth while our conventional
DACs sounded slurred, drums were un-focussed, and more effort
was required to hear details and mix values. Not quite subtle.
After many hours (weeks) of rigorous calibrated A/B and A/B/X
comparisons it became clear that conventional DACs have a ramped
-up effect on transients, especially obvious on big bangs out of a
black background. In 15 years of listening to digital, I had never
noticed this effect, nor had seen any article, heard any discussion,
nor had it been suggested to listen for it. With more listening, it
became obvious that this effect messes up imaging width and depth
& groove or feel. It adds to the harshness of digital and could be a
major cause of 'digititus'. Hell, even tapping ones foot to the music,
or just enjoying the music was a little more difficult with conventional
DACs. We started out calling the effect pre-echo but settled on
'time-smear' as months wore on. There is no official and universally
accepted name for this effect. Then came the hard part. What is the
cause, who has researched this, why haven't we been told? The cause
is half easy and half incredibly difficult. It is easy to demonstrate that
the problem is caused by digital brick-wall FIR filters, because one
can readily hear the differences as filters are shifted higher and/or
made less steep. The difficult and frustrating part is that nobody
seems to know why we hear the filters. According to the theories that
digital designers live by and our present knowledge of human
hearing, we shouldn't be able to hear the difference between a FIR
with a 45K corner frequency and one with a 90K corner frequency
or one with a more gentle 90K slope - but we do, or at least those who
listen do. Pay attention to the leading edge of drums and percussion.
Further discussions with a variety of digital gurus, all seemed to
indicate, that they are aware of the 'time-smear' but also have no
solid scientific reason for why we hear it. Occasionally somebody
with interests in not discussing time-smear will mention "the FIR
impulse response ripples are at 22K or higher and we don't hear
that". OK, but we DO hear the FIRs all the same. Maybe distortion
in tweeters and electronics is a factor and maybe not. Nobody
mentioned jitter years ago until help arrived. I don't know why we
hear the FIRs either, but because we consider the problem to be very
significant for music, we think that you should know it exists and
that you should be concerned too. 'Groove' and 'feel'
are kind of
important in music and time-smear doesn't ever help.
FIR filters have compelling advantages, but perhaps inherant
problems that a few expensive converters have minimized. The
good news is that a lot of CDs that we blamed mastering or digital
for sounding bad, come back to life. Also good news is that time-
smear may be reduced to acceptible levels at 192K sample rates (and
48K data rates with SLAM!'s converters). In fact, the SLAM!'s
converters have less time smear than other converters running at
192K and are probably the first with this kind of timing performance
near this price range.
The Quantum Converters
When we decided to offer digital converters, the idea was to provide
a convenient method to 'insert analog' into a digital studio. Part of the
goal was to avoid Phase Lock Loops (PLL's) to recover and clean up
the clock signal encoded in AES data streams. Two reasons for this:
we feel that most PLLs in use are truly inadequate and it requires a
great deal of work, time and intelligence to design a PLL fine enough
for 24bit audio. If a converter improves sonically by using a word
clock input, this is an indication that the clock recovery is insufficient
and that jitter has been audibly attenuated but most likely still a
problem. Then we met a new company with a better answer.
The converters in the SLAM! were were co-developed by a group
of very clever engineers in Switzerland called Anagram Technologies
along with bits from us. Anagram mostly design converters for hi-
fi companies like Manley, Nagra, Audio Aero, Camelot-Tech and
Talk Electronics. "Anagram Technology DSP filters seem to be the
hot new numbers in the digital world", (Stereophile, Apr 2002).
Anagram had developed probably the best Asynchronous Sample
Rate Converter based on software running in a high speed SHARC
DSP chip. The process essentially eliminates jitter and provides the
best audible aspects of 192K converters at sensible data rates like
48 or 96K. We can't claim zero jitter, but it's damn close and difficult
to measure. The jitter is lower than most converters' internal crystals.
The DSP process borrows concepts from quantum physics. Quantum
Theory says we can never know both the mass and energy (speed)
of sub-atomic particles at the same time. In digital audio with any
amount of assumed jitter, likewise, there is always some uncertainty
of the actual sample value because we are unsure of the clock. What
makes Anagram's process unique is that both data and clock are
treated as a linked pair, and treated with parallel algorithms. Thus
the name and a hint of how jitter can be eliminated in software.
In our research and with input from Bob Katz, it became obvious that
besides jitter and analog audio implementation, that some of the
biggest differences between converters was due to the filters - both
the digital and analog varieties. This is why you have 3 choices for
both the ADC and DAC. It is unlikely that we can hear beyond 20K
or 25K but there do seem to be differences caused by ultra-sonic
noise and stray signals that may affect 20-20K audio. Some digital
filters and all analog filters, even with 80K 3dB points will create
some phase shift within the audio band. This is usually somewhere
between subtle and inaudible and may even be desireable. We might
prefer the 20K filters when we want some tape-like emulation
(especially the DAC) or just 'smooth'. There are also situations when
one needs a 20K filter to prevent problems further down the digital
chain. Some plug-ins or data-compression algorithms are known to
to misbehave with ultra-sonic harmonics or noise. 80K is flattest.
The combination of the SLAM! and Quantum was not intended to
be the absolute in clinical or sterile conversion. There was some
effort to make these into "warm converters" (hot, even). There are
no ICs or op-amps from the XLR input to the ADC just class-A low
distortion tube circuits. In fact, the input stage of the ADC is totally
passive; just a transformer (the warmth), a few inductors and
capacitors. With 192K sampling for speedy transients, tubes and
iron for the "phat", and limiters for color and ballz, it all should be
fun & useful and provide 'personality'. These converters may be
even cleaner, crisper and more accurate than most too. You'll just
have to compare them to find out.
19