26th June 2014 -
PS Audio DirectStream DSD DAC. NOW ON DEM.- Author:
Pure DSD done properly
Introducing DirectStream, one of the most remarkable DACS ever built. Hand written, discrete, perfection based conversion that uncovers all the missing information hiding in your digital audio media for all these years. CD’s, downloads, high-resolution PCM or DSD based media are expertly upsampled in the DirectStream to ten times DSD rate and output as pure analog directly into your amplifier or preamplifier. Classic PCM based DACS, including many of today’s DACS that can also process DSD, tend to cover up some of the subtle musical details buried deep within digital audio music; a problem inherent in their architecture. DirectStream solves this problem by employing a pure DSD single-bit approach for both PCM as well as DSD media. This means that your entire library of music can finally reveal all the music and subtle low level details buried deep within its core. Imagine going to a recording studio and listening to a master tape of any recording made. This is the experience DirectStream provides its owners who enjoy a renewed sense of enjoyment and discovery when listening to everything in their library: CD’s, downloads, DSD. It is likely that whatever DAC you are currently using to decode your music may be covering up some of the finer details such as room reflections, upper harmonics, inner details and proper tonality. Removing the barriers inherent in PCM based processors can reveal more in your library than you might have thought possible.
The back story
In 1981, music was reproduced exclusively on analog-based turntables and tape decks with limited dynamic range. Despite the very listenable music coming through those grooves, the audio world wanted more: greater dynamic range, lower noise, extended frequency response. No wonder that when, in 1982, Sony and Philips announced they had achieved “Perfect Sound Forever”, music lovers around the world waited in eager anticipation of analog’s promise finally fulfilled. Unfortunately those early CD’s sounded less musical than analog. In some cases, much worse. The launch of the PCM-based Compact Disc would set in motion a 30-year war amongst Audiophiles, decimate the analog format, cover up subtle musical details for decades and turn the music and audio industries upside down –and not necessarily in a good way. There had to be a better way to fulfill the promise of analog’s century of dominance for sounding like music and PCM would not be it. That answer is DSD.
The poor results of the first CD’s might not have been a problem if Sony and Philips had waited another decade, passing over the classic PCM delivery system in favor of PDM (DSD). The problem is the PCM decoding process itself: whether a classic ladder-DAC or more modern multi-bit Sigma-Delta type, most PCM playback processors mask some of the subtle cues in music. Fortunately music’s details have been more faithfully recorded than we ever suspected, but the cover-up of those recorded details has been with us for more than 30 years. Happily, the missing musical information is still buried deep within our audio libraries—waiting to be revealed. In order to extract everything hidden in PCM recordings, a completely new processing method is needed.
Pure DSD is the answer
Put DSD into DirectStream, you get DSD. Put PCM (how CD’s are recorded) into DirectStream, you get—DSD. DirectStream converts all digital inputs, including PCM, to pure 1-bit DSD, in an elegantly-simple path. In the process, the PCM feed becomes more linear, less edgy, and never-before-heard musical details are released from all digital audio recordings. Billions of CDs and high-resolution downloads worldwide will gain new life, and be saved from obsolescence. There’s a notion floating around high-end audio circles that DSD may be nothing more than a passing fad. In fact, we would suggest the opposite may be true. PCM based music decoders are nearing the end of their lifecycle in favor of a significantly more musical sounding format, DSD.
A better format
DSD is a high sample rate single-bit format while PCM is a lower sample rate many-bit format. DSD sample rates are 64 times higher than those used to create CDs and the output of a DSD stream can be placed through a simple filter to produce pure analog ready to feed your audio system. PCM, on the other hand, can produce great sounding music but to do so requires a very complex and technically daunting conversion process to produce something you can play on your system. Sometimes the simplest path is the best sounding one. DSD is a far more analog-like encoding and decoding process than classic PCM. In fact, most modern A/D converters (used to make the recordings) are DSD based already and that movement is growing. There are millions of PCM based recordings that will be in our libraries for years to come, but with the introduction of DirectStream, you no longer need listen to them with a PCM based processor.
The secret is in its simplicity
The PCM processor’s tendency to mask music’s subtle details is largely due to its complex needs, and the technical requirements of PCM processors.
Converting PCM to DSD
Converting PCM to DSD can be an easy exercise using any number of computer software programs. Converting PCM to DSD properly, and in such a way as to reveal missing details in the music, is a serious technical challenge. To handle this properly DirectStream utilizes 66-bit fixed-point FIR filters, eliminates headroom issues common to PCM, flattens noise response in the 20-100kHz region, uses coefficients optimized for best sound rather than faster processing speed and optimizes hardware-specific operations not possible in software. Unlike other processors available today, DirectStream unifies all inputs (PCM or DSD) at 10 x DSD, then uses a true single-bit double-rate DSD core engine. True DSD core engines (compared to the standard multibit Sigma-Delta converters followed by random lower quality multibit converters) offer advantages in simplicity, linearity, and in analog-like overload characteristics that avoid PCM’s “hard clipping” potential and a PCM processor’s propensity to mask subtle details.
No off-the-shelf DAC IC’s
A key design feature that makes the DirectStream’s performance possible is the hand written code that makes it work. Seven years in design by our lead designer, Ted Smith, every operation and process was written from scratch to make sure the best possible audio performance was achieved. Most DACS are designed around off-the-shelf DAC IC’s. DACs designed around dedicated IC (integrated circuit) chips are limited in the amount of signal-processing and math they can perform, due to their restricted ability to dissipate heat. Their limited “real estate” also forces digital circuitry to be located directly adjacent to analog circuitry, which can introduce noise and jitter into the signal output. Instead of being based upon a packaged PCM-based DAC chip, with all of its inherent compromises and limitations, DirectStream utilizes an FPGA (Field-Programmable Gate Array). An FPGA is capable of dissipating large amounts of heat, enabling intensive levels of signal-processing that are impossible to achieve with an off-the-shelf chip.
The DirectStream DAC is entirely DSD based, even for PCM inputs. DSD was chosen as the core engine for this instrument for a number of compelling reasons: Simplicity. DSD is simple to convert to analog: just low pass filter it. In the DirectStream no limiting active output electronics are used thus making sure all details come through without any degradation. Linearity. DSD is inherently linear: it’s nearly impossible to build a PCM DAC that always takes the same sized step in the output for any possible unit increment of the PCM value because of component matching challenges. DSD doesn’t need this level of component matching and this means linear performance for all frequencies. Soft clipping. Like magnetic tape, DSD soft clips when overdriven: signals which exceed the nominal full scale value (by less than 3-4dB) only get slightly compressed if at all. With PCM the consequences are flat tops which induce extra energy at the squared off edges, or worse, wrap around which is very audible: the results can be an audible glare not present with DSD. Common architecture. Ironically most reasonably priced PCM players actually use DAC chips that utilize a sigma-delta modulator to get a DSD-like signal anyway. Similarly most A/Ds are sigma-delta based. Matching the recording process to the playback process offers sonic advantages. Immunity to cable and input differences. DirectStream handles the PCM conversion from all inputs without recovering a clock, by simply watching for the edges and making decisions about what they mean in context. The result is that any jitter present on the input clocks is lost entirely in DirectStream. There is little audible difference between TOSLINK and, for example, I2S because the output clock’s rate only depends on the long term average rate of the inputs. This means even poor quality sources sound musical. Unmasking of subtle musical details. Separating the digital and analog stages from each other, operating from a single master clock, 10x DSD processing and a purely passive output stage reveal musical details formerly buried within the digital audio media and masked by classic PCM processing and decoding techniques.
The DSD Engine
The heart of the DirectStream DAC is its DSD engine: 10X DSD rate. Regardless of input format, whether PCM or DSD, all data are upsampled to 30 bits running at 10 times the standard DSD rate and then back down again to double rate DSD for noise shaping. This critical step sets the DirectStream apart and is responsible for much of the uncovering of details in the music. No rounding. The internal volume control keeps complete precision: every bit in the input affects the output of the DAC for any volume level. Except for the sigma-delta modulation process itself there is no rounding, dither or other trimming. This means that perhaps for the first time, the DACS volume control has no degradation or data loss at any volume setting, eliminating the need for a preamplifier if the user wishes. 30 bit frames. The design utilizes a minimum width of 24 bits and achieves at least 144dB S/N in all of the upsampling filters. Use of full precision in the upsampling filters and significant number of “guard bits” in the IIR filters and the sigma-delta modulator help maintain our goal of perfecting the audio output. Music has never been this clean and high-resolution analog-like. Headroom. Depending on the source material some designs may run out of headroom or approach saturation levels. The DirectStream design opts for at least one extra top bit everywhere in the digital path, coupled with an extra 6 dB of head room in the analog path above and beyond the 6 dB of headroom that SACD needs. This means no PCM or DSD program material will ever have dynamic restriction from the DAC itself. Transformer coupled output. The output of the DSD engine is fed directly into the output stage which is based on high speed video amplifiers, passive filtering, including a carefully crafted high bandwidth audio transformer at the output. Unlike conventional DACS that depend on active analog output circuitry and the problems associated with them, DirectStream is purely passive. This approach is one of the critical success stories of DirectStream. Because the output DSD stream is running at double DSD rate (5.6mHz) the ability of active analog electronics to handle these tremendously fast frequencies without adding any sonic artifacts is a serious challenge, one handled without incident when passive means are employed as found in DirectStream.
Critical design choices
One master clock. Two clocks would mean more jitter. Even being able to shut one of them completely off while using the other adds stubs and possibly worse, logic or switches between the clocks and the final retiming-flip-flop. PCM architecture DACS almost universally require more than one clock and thus mask subtle details in the music. 10X sample rate. All input sample rates supported are synchronously upsampled to 10x the standard DSD sample rate and then back down to double rate DSD. There’s no need for other clocks to interpret the inputs, regardless of their sample rate, because of the instrument’s single clock input sampling architecture. This means music played through DirectStream is open and full of details formerly masked by other methods of decoding. Low noise transfer. Connections between sections of the design have large impedances to slow down the edges and lessen any noise transfer, in addition to significant power supply isolation. This technique improves musicality and reduces, if not eliminates, the audible glare associated with traditional DACS. Low interference. Control signals run as slowly as they can possibly go, plus control of their transition times limits the amount of induced noise (and hence possible jitter) into the main digital processing area. By slowing these control signals, music remains pure and unaffected by internal commands necessary for the DirectStream’s operation. Balanced architecture. Balanced signals are used throughout reducing radiation and noise in the ground and power rails. A balanced architecture is largely responsible for cleaning up the background and lowering noise floors to black levels.
Every decision for sound
Non-saturated logic. Non saturated logic is used throughout much of the device to improve the predictability of the transition timing. Coming out of saturation is an undesirable statistical process that introduces noise and jitter. Every effort to reduce jitter has been lavished upon this design and the increased level of available details in the music result from these efforts. Careful terminations. High rate signals (and/or signals with fast edges) are isolated from control signals and especially each other. If they have to be fast they are terminated appropriately to help keep jitter low. Audible glare and harsh artifacts riding on musical signals are reduced by these careful steps at every juncture of the design. Low temperature coefficient parts. 0.1% precision thin film low temperature coefficient 1/8W or 1/4W resistors are incorporated everywhere in the audio path, lowering the temperature coefficient of the components below standard practices. This enables each and every DirectStream to be hand built and nearly identical in sound. Resonance matching. Low noise techniques are employed, such as the liberal use of low inductance capacitor bypassing, each of which has a self resonance frequency at the main clock rate to keep as much noise from ever getting into the voltage rails. Consistency throughout the instrument is a critical feature found in the perfection of DirectStream.
Every trace on the PC board is hand routed: even for the digital sections. No autorouting is employed, so that each trace, each critical path, is calculated and designed for lowest noise, jitter and isolation. Every unit is delivered by the production team with pride because of the extraordinary attention to detail in the design itself. Go direct into your amplifier or preamplifier DirectStream provides a volume control that is perfect and never loses any resolution at any setting. This, coupled with two choices of output level, permits DirectStream to be connected directly to either a power amplifier or preamplifier for best sonic advantage. Our typical recommendation is to take DirectStream directly into your power amplifier, thus avoiding any potential sonic degradation that might be caused from the extra circuitry and volume control within a preamplifier. In some cases it is preferable to include a preamplifier in the signal path. DirectStream can then be adjusted to maintain a fixed level appropriate for the preamplifier’s input.
One of the handiest features on the DirectStream that will see a great deal of use is the polarity (phase) control. This touch screen option (available on the remote as well) allows users to select normal or inverted polarity for every input and for every performance. Sources and music have no standard for absolute polarity, a fact easily demonstrated by switching polarities on the as the music plays. Many Audiophiles mark their discs in phase and out of phase and are very careful to select the proper polarity before playing each disc. Some CD players and sources themselves invert polarity so that a disc you believe is inverted actually is correct played on a specific device and incorrect on yet another. To fix this problem you can assign each input on the DirectStream to switch to the proper polarity when you activate that input. Touching this control is the same thing as reversing the loudspeaker connections of both loudspeakers (+ to -) at either the amplifier or speakers. All live acoustic music is in absolute polarity but some recordings may accidentally be made inverted and should sound more realistic and pleasant when you correct their polarity with the switch.
Inputs and outputs
DirectStream has seven digital inputs which include: two I2S, one TOSLINK optical connector, one coaxial S/PDIF connector, one balanced XLR AES/EBU, one asynchronous USB input and one optional Network Streaming Bridge slot. DirectStream has both single ended coaxial and fully balanced XLR outputs.
Built in Boulder
The DirectStream is assembled, programmed and tested at our production facility in Boulder Colorado. There is a measure of pride of workmanship that goes into every one of these PerfectWave products and it shows from the moment you open the unit up and plug it in. This is one gorgeous piece of equipment and just the beginning of the most beautiful natural sounding audio equipment you have ever had the privilege to own. Add a PerfectWave PowerBase and together we believe you’ve never heard anything quite like it in your system.