
A U
D I O... P A P E R
Will
digital ever sound like analog?
Development
of the Zen Triode NOS DAC
Aug 2009
by Steve Deckert
Of
course CD players have advantages and disadvantages
that we're all well familiar with...
Sometimes
the disadvantages can overpower common sense resulting
in total destruction of your player as shown below.
Sometimes common sense can have the same result,
sparing you from any further suffering, as was the case
here.

Can
digital sound like analog? If we use the CD player
shown above (before I smashed it) and a hi-end vinyl
rig (shown below) for the comparison then the answer
biases itself heavily towards NO, CD's can't sound exactly
like analog. But while CD's and digital are
one in the same to most people, CD's are nothing
more than plastic disks that are largely responsible
for the huge gap in sound quality when compared to vinyl.

This
can easily lead an audiophile to hate all things digital,
even conclude that the format itself is worthless. I
have been guilty of this myself. The problem with
this conclusion comes into focus when you listen
to well recorded LP's on the turntable and the thought
"analog kills digital" races through your
mind followed by the realization that the LP you just
listened to was mastered
in 20 bit digital back in the 70's.
IS
IT ANALOG OR DIGITAL?
Obviously
it sounds pretty damn analog coming off the turntable,
yet it was a digitally mastered recording. So
how can the LP sound so much better than say the same
recording dumped onto a CD? It's a combination
of things. The tonearm and cartridge have their
own colorations, usually desirable ones, the RIAA equalization
curve has it's own interplay on the dynamics resulting
in a certain signature which I'll get into in a bit.
The high end vinyl rig is a superior playback device
when combined with a hi-end phono stage even if there
were no colorations of any kind because there is no
Digital to Analog Conversion process happening. Also
popularly known as the DAC, this process falls far short
of having the same end result as when the process does
not need to be used.
IMPROVEMENTS
?
So
the focus has been on improving the DAC since it came
out. BTW, remember that the Analog to Digital
process (ADC) is not the same as the DAC process. This
is part of the reason why the LP's that were digitally
mastered didn't sound completely screwed up. For
playback the "Redbook" format became the standard
for consumer grade CD's and players that was 44.1KHZ
and 16 bit, often written as 44/16, even though it was
below the current technology of the day. Of course
later we started to see 24/96 and various products that
would upsample the data to 24 bit 192 KHZ. Each
of these increases puts more points on the line making
the top end sound smoother and more natural or phrased
differently, less grainy sounding.
Interestingly
enough, being old enough to have been listening to music
back when there was no such thing as a CD, it's been
my experience that even these higher sampling rates
do not make the CD sound like real music as was often
the result with LP's. CD's always have that "CD"
sound, where the music sounds like plastic. I'll
admit that it has more than irritated me over the years
that these increased sampling rates are not solving
the problem and that I basically could easily live in
a world without CD's.
My
biggest bitch with CD's is that it always seems to take
something like a Wadia DAC and matching transport to
get the CD's to sound acceptable. Yet you can
set up an entry level vinyl rig for a grand or two and
have potentially better sound. I don't remember
being told when CD players came out, that it would require
you to spend 10 times more money than you have in your
turntable to get as good a sound, yet here we are some
thirty years later and that is still the norm.
TRANSPORT
BLUES
CD's
require transports with lasers to read the data. This
process is full of errors in most consumer grade products. Even in a good transport, the CD itself will have
errors and scratches that cause more errors. This
digital data stream from the transport goes into the
DAC, where it's stored as read, errors and all, followed
by the DAC telling the transport to read it again and
again. This is called oversampling. Each
stored package of data is then processed and combined
together in an attempt to figure out what to put in
place of the errors. So we've got two rather negative
things happening in the playback process, jitter from
the transport that creates errors and oversampling and
processing that is done in the DAC.
TOO
EXPENSIVE
This
isn't to encouraging to a guy who's not made of money,
because the only way to get less jitter is to buy a
really good transport. Why even bother buying
a really good DAC if you can't also spend the same amount
of money on a transport? You'll just be feeding
the DAC tons of errors, and the resulting sound from
the DAC won't be that much better than the CD player
you tried to use for a transport.
NEW
HOPE FOR THE LITTLE GUY
Two
things have happened in the last 10 years that when
combined may bring some real hope to all of us who find
something almost subliminally irritating about the sound
from CD's. One is computers, and one is a man
called Ryohei Kusunoki
who wrote a paper in 1996 about a DAC concept where
oversampling and fir filters are eliminated. My only
regret is that I didn't run into his paper sooner. This started
what is popularly known today as Non-Oversampling DACs,
of which there are many available be they still preached
against by most engineers.
THE
COMPUTER
By
using the computer hard drive to store digital music
we now have the ability to store error free files thanks
to programs like EAC (Exact Audio Copy) available free
on the web. These programs will read the data
from the disk and compare it against an international
database of CD's to see if each track from the CD is
now an error free copy on the hard drive. You'll
be shocked at how many errors are on your CD's, but
this type of program will read the disk millions of
times if need be to get an error free read.
USB
PORT
Disabling
the Windows Kernal that handles the sound buss in a
PC makes it possible to send a bit for bit output direct
to the USB port on your computer. That means that
if you get a USB DAC with a good reclocking scheme and
super accurate clocks you can pass almost error free
data to the DAC. This would be an ideal alternative
to spending thousands of dollars on a hi-end transport.
When
you combine this with a DAC that's non-oversampling you've
got a listenable source that no longer sounds like plastic
so this is the approach I have taken. Since that
time my focus has been on the NOS DAC technology itself
to figure out what the best implementation of it is.
And since USB has this potential, what chipset's
and re-clocking schemes will give the best result. We
already know that many early USB sound devices had issues
that resulted in less than stellar error performance
so this part obviously has to be right.
THE
ZEN TRIODE NOS DAC APPROACH
Since
the rest of this paper will focus on my DAC and output
stage design and since
it's a non-oversampling DAC I want you to understand
why oversampling and fir filters found in almost all of
today's DACs are poison to music playback. I've
been asked when or if Decware would come out with a
Zen DAC for the past 10 years and until I heard this
poison removed from the playback I had no intension
of pursuing any type of Decware DAC... basically just
washed my hands of the whole thing. For me hearing
my first couple NOS dacs left me realizing two things;
A) There is twice as much information on a standard
redbook format as I would have ever guessed, and while
not an LP, we suddenly have usable playback. B)
The timing and pitch and timbre are now correct, hence
the plastic sound is reduced to the point of being no
longer irritating, if even on a subconscious level.
MAKE
CD SOUND LIKE VINYL - NO
It
doesn't have to sound like a record to be good, and
that's my contention. All I've ever wanted was
to get digital sound to be a usable source in my listening
room and it's not usable if there's a subliminal and
audible irritation factor in listening to it. Digital,
even poisoned, has always had certain things that could
be argued were better than LP's. Let's get it
to sound like real music with the proper weight and
interplay of dynamics throughout the frequency range
and you'll see the listening experience will become
enjoyable like it used to be when vinyl and tape was
all we had. This all starts with a non-oversampling,
non-filtered DAC design so I refer you to summary of
Kusunoki's excellent paper on the subject: Non-oversampling Digital filter-less DAC Concept
SOME
ENGINEERS
You'll
conclude after reading his paper that during the development
of DACs engineers came to a fork in the road where they
had to choose between frequency domain or time domain.
Since they could only have one of the two, they
decided that screwing up the time domain was the lessor
of two evils. However, the only thing wrong with the
frequency domain was that it contained noise and garbage
above the 20 kHz hearing limit of our ears, in some
cases out to several hundred kilohertz. That would show
up on any decent scope and on any frequency response
test and since the whole push with CD's in the first
place was less distortion they couldn't bear to have
it there despite the fact that we can't hear it. We
can hear changes to the time domain but either
it went unnoticed or the politics of having clean specifications
on paper left no alternative.
EARS
OVER RULE THE EYES
I'm
certainly not bothered by content above the 20kHz limit,
because with my tube gear it's not going to irritate
my amplifiers, my speakers or my ears. The amp will
filter it away, as will the speakers that can't reproduce
it even if it is there, and of course my ears which
can't hear anything above 20kHz. A well designed analog
filter could be used with minimal negatives making it
compatible with any amplifier or preamp.
TAKES
MORE THAN JUST THE DAC
But
the DAC itself is only part of the equation. The analog
output stage(s) and the impedance/gain relationships
between them will determine how real the output sounds.
This is one reason why a vinyl rig always sounds different
than the average DAC. Most of the time, the output stage
of any given DAC is designed to accomplish a 2 volt
output with the lowest possible distortion, period.
In the case of a vinyl rig, you have the tiny signal
from the cartridge of just a couple millivolts that
has no bass in it. The phono stage uses the industry
standard RIAA equalization curve to put the bass back
in and add enough gain to bring the output level up
to a couple volts. This ends up being somewhere between
40 and 60dB of gain, more than any other part of your
playback system by a huge amount. The phono stage then
drives a preamp that in turn drives your amplifier.
This gain from the phono stage is controlled by
the preamp, in most cases reduced to achieve the desired
listening level.
If
you could somehow get the sound to come out of the preamp
without the phono stage driving it, it would not sound
the same. You can think of gain as voltage or pressure.
By creating a high pressure with the phono stage
against the front door of the preamp, the signal enters
the preamp with purpose. If you reduced the pressure
from the phono stage by say half, the signal entering
the preamp would have less pressure behind it and the
preamp would have to add gain to increase the signal. Since
the signal entering the front door of the preamp was
somewhat whippy when the pressure was reduced by half,
the preamp simply made the whippy sound louder. The
end result is playback with a lack of dynamics, weight,
and warmth.
COMBINATIONS
Playing
with different cartridges and phono stage / preamp combinations
made it possible for the audiophile to get this sense
of liveliness and energy with the correct frequency
balance and weight. To replace this with a CD player/DAC
removes the cartidge and phono stage leaving the preamp
to deal with a specified output from the DAC of 2 volts
with no pre gain or post gain strategies to control
the dynamic interplay between the source and the amplifier.
So just having a DAC by itself will depend on the type
of preamp and amplifier you have and if there is any
synergy between those components.
COMPONENT
SYNERGY
Synergy
between components is partially dictated by impedances,
such as the output impedance of one gain stage against
the input impedance of the following stage. And
of course the pressures developed in each stage. Left
to chance it's a guessing game what component combinations
will have a synergy similar to the phono stage/preamp
combination. I'd like to involve pre and post gain strategies
with the output of the DAC so that nothing is left to
chance and the level of control is there to achieve
the signature and presentation that best serves the
music. After all, musicality is what brings long term
satisfaction from good audio gear. I always enjoyed
being able to adjust the signature of my vinyl rig via
fine adjustments to the tonearm and by adjusting the
output level from the phone stage against the input
level of the preamp to get the right tone and presence. I
also enjoyed using two or three different cartridge/tonearm
combinations for different recordings. One was great
for rock, one was unsurpassed on female vocals, and
one was great on everything else.
THREE
BLOCK DESIGN
It
would be nice to maintain this kind of control if we're
going to get a similar kind of enjoyment from digital,
so the output stages in this design are absolutely paramount
to how good the DAC itself is going to sound. To achieve
this I have experimented with countless combinations
and come up with a three block system, with four
stages of gain. The first block will be the output from
the DAC chips themselves which will be the equivelant
of a single stage of gain with a 2 volt output and low
impedance. The second block will be a de-coupled single
triode gain stage with it's own tube regulated power
supply. The third block will be a de-coupled solid state
gain stage with two stages driven by a large self charging
battery supply. This brings in the opportunity
to manipulate the interstage pressures and impedances
in front of and between the two final stages. The
output from this final stage can be de-coupled or direct-coupled
to drive the amp or preamp of your choice. This
is where you'll be able to adjust the inner dynamics
and presence of the music similar to the way small adjustments
to your tonearm and cartridge changed the sound of your
vinyl playback.
Having
this three block design makes it possible to change
the analog output from the unit so that it comes directly
off the DAC chip, or the tube stage, or the final stage.
Both the tube stage and the final stage have external
gain controls, aka volume knobs, that allow you to play
one control against another when all three blocks are
used. This will let you adjust for the proper weight
and frequency balance when you get recordings that are
too thin or too fat to be useful. Of course some
might say you can do the same thing with a bass and
treble control, but why kill the transparency when you
can make it happen naturally by manipulating impedance.
You can also adjust both volumes to the same output
as the DAC chips and seamlessly go back and forth between
the three while you listen. All three outputs
sound very different from each other, to the same degree
that phono cartridges sound different from each other.
GOAL
If
the goal was to get the sound as close to vinyl as possible
a reverse RIAA passive network would have to be added
to the output of the DAC chips and then a 47dB gain
stage with an RIAA network would have to be used in
front of the second block, the tube stage. This
would give the unit a vinyl signature, duplicating the
bass and shimmer of LP's almost convincingly. But
with two networks and two more gain stages or basically
the equivalent of a phono stage with a passive reverse
RIAA on it's input, we trade transparency for a less
resolved but juicer sound. It would sound like
a low grade cartridge lacking detail and air. My
goal is not to make the DAC sound like a vinyl rig,
but rather get the digital music free of time domain
distortions that create that plastic sound AND make
it usable by adding a considerable degree of control
over all the subtleties, similar to what I can do with
a good vinyl rig.
PROS
& CONS
My
implementation of all this has pros and cons. Pretty
much all pros for the buyer, but serious cons during
assembly. A single rack mount chassis houses the
three main components and their individual power supplies
all mechanically decoupled from the main chassis
and hand wired to the mryiad of led's and switches.
The build is brutle and after the first production
prototype I didn't want to ever build another one. The
solution to this is of course re-design the chassis
to accommodate all three main blocks and their power
supplies, switches, jacks, connectors, and lights on
a single circuit board. This would make building
it take almost 30 minutes instead of the 30 hours
it actually takes. But there would be no vibration
control possible, and singles sound far better going
through silver wire than they do through circuit board
traces. I just can't see it sounding as good,
not to mention the size of the board could invite damage
from cracks during shipping.
COSTS
Besides
the technical downfalls of doing the whole thing on
one giant circuit board and wanting to get the
best sound I can out of this thing, I know it's going
to be expensive to hand produce and I'll likely have
to make all of these myself personally. So with
that in mind, this DAC is no help to the guys who can't
afford it. My plan would be to follow this model
with a two block system using the first two blocks of
this design. Then eventually a one block system
using the first block in this design. This would
hit three substantially different price points,
and it's possible to use only the first block of this
system with a good tube preamplifier and have something
good enough to bring big smiles.
FEATURES
AND SPECIFICS
BLOCK
ONE - THE NOS USB DAC DESIGN
I've
always enjoyed the transparency you get directly off
a DAC chip before you feed it into analog filters and
output stage. Problem is the output of the chip
is too low to be of much use by itself. The output impedance
is too high to drive anything with real authority other
than an ideal load. Somehow I wanted to have the option
to hear the raw output off the DAC chip but with the
same output and low impedance you would normally get
from the output stage. The best solution for this seems
to be the Texas Instruments 1543 chip. By itself it
will give about a volt with the ability to drive an
amplifier directly with pleasing results. This is about
a 20 year old chip design, still in production that represents
the most straight forward DAC design I've been able
to find so far. I don't want the DAC to have DSP
capabilities or anything in it that is not being used.
When you parallel these chips together the output impedance
gets lower and the power output gets higher. You also
reduce errors, because if one chip has an error and
the other ones don't, there is no error. This increases
the true bit rate over a single chip. I've spent some
time listening to the chip by itself, and in groups
of 4, 8 and 12 chips in parallel. Of course you can
go much higher, but money doesn't grow on trees and
the biggest performance change seems to come from comparing
one chip to eight. So at this point, I'm using 8x1543
NOS with no output stage for block one.
The
Inputs are USB, coax, optical I2S to the 1543 chips.
Both the USB and Coax/Optical are independently re-clocked
with 1PPM TXCO clocks. Input frequency will be
USB 32kHZ, 44.1kHz, 48kHz and coaxial & toslink 32kHZ, 44.1kHz, 48kHz, 96kHz.
The paralleled 1543 chips will be at 2 volts with no
fir filters or output stage.
BLOCK
TWO - THE TRIODE GAIN STAGE
The
triode gain stage will consist of a 12AU7a tube being
fed by a tube regulated power supply with a starved
AC filament. It's lower voltages will increase tube
life and just accommodate the 2 volt voltage swings
from the DAC. It will increase the voltage to around
10 volts with a variable output that lets you control
the gain or loss from this stage to any level you want.
The tube regulation in the power supply is done with
an OC2 tube that is in series with the high voltage
in place of a choke or dropping resistor. This has the
effect of isolating the power supply from the audio
circuit and filters off all the nasty harmonics off
the power grid that came in through the power cord.
BLOCK
THREE - THE FINAL STAGE
The
final stage consists of a pair of Burr Brown OP amps
being fed by an isolated DC battery supply for the ultimate
in clean power. The battery offers some 400 hours of
use on a single charge and re-charges itself when the
unit is turned off. This block features an adjustable
input impedance, and input gain control, defeatable
crossfeed for headphone use, a Hi-Z direct coupled output
or Lo-Z de-coupled output, and an inter-stage gain switch.
It's noise spec is insanely low and it will drive world
class headphones from 32 ohms to 600 ohms and a line
level output into impedances up to 100K at the same
time with world class sound quality.
You
can select one two or all three blocks at the same time
on the fly while you listen. Using two or all three
blocks gives you volume control and becomes your preamp
and headphone amp all in one. You will also be able
to select from two sources, the DAC and an AUX line
input in place of the DAC. This makes it pretty easy
to eliminate your preamp in most cases, reducing interconnect
cables and further clutter in the signal path between
the DAC and the amplifier. The reduced voltage tube
stage stays on all the time to increase the tube life
of both the 12Au7a and the OC2 regulator tubes.
You
can select between Coax or Toslink operation, or use
USB which takes priority when hooked up. There
is no remote to control volume because the further complexity
and cost just aren't justified and likely wouldn't help
the sound quality anyway. It is possible to remote
control the volume and track selection on the computer
if you use it's USB or digital output.
Here's
a shot of the unit on my bench.

-Steve Deckert
HOME
BACK
Decware is a trademark of High Fidelity Engineering
Co. Copyright © 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006
2007 2008 2009 by Steve Deckert
|