To DSD or Not to DSD, That Is The Question: PS Audio PerfectWave DirectStream DAC v Berkeley Audio Design Alpha DAC Reference Series

PS Audio has just announced a new product - the DirectStream DAC ($5,995). In essence, the DirectStream DAC converts all incoming data including PCM to 1-bit DSD using their own code implemented in a Spartan VI FPGA. Chips need not apply. Of course PS Audio claims that this conversion to DSD improves the resultant sound quality of everything that passes through it, especially PCM. According to PS Audio:
The problem is the PCM decoding process itself: whether a classic ladder-DAC or more modern multi-bit Sigma-Delta type, PCM processors universally mask some of the subtle cues in music, and no amount of upgrading, expenditure, tweaking or improvement can fix this fundamentally-flawed system.

In order to extract everything hidden in PCM recordings, a completely new processing method is needed.

Their solution? DSD.
Put DSD into DirectStream, you get DSD. Put PCM 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 -- and recycling bins or landfills.

Berkeley Audio Design Alpha DAC Reference Series

Marketing speak aside, I found this approach of particular interest especially when compared to another recent product announcement, this one from Berkeley Audio Design for their Alpha DAC Reference Series ($14,000):

DSD capability for the Alpha DAC Reference Series is provided by an included state of the art software application that provides either real time conversion of DSD 1x and DSD 2x to 176.4 kHz 24 bit PCM during playback or conversion to 176.4 kHz 24 bit AIFF or WAV files. The software application is included in the price of the Alpha DAC Reference Series and is compatible with either Windows OS or Mac OS based music servers.
According to Berkeley, its better to convert DSD to PCM in your computer before sending it to your DAC.
We could, like many other manufacturers, convert 1-bit DSD to multi-bit within the Alpha DAC Reference Series and show “DSD” in the front panel display. That would be the easiest approach from a marketing perspective. But that would also mean increasing the amount of processing in the DAC during playback which would degrade audio quality, and audio quality is the reason the Alpha Reference Series exists.

So here we have two manufacturers claiming similar results, i.e. great sounding DACs, using diametrically opposed approaches. Interesting no?

It would be very interesting to compare these two products and I hope to have the PS Audio DirectStream DAC in-house asap. Unfortunately, our repeated emails and phone calls to Berkeley Audio Design have so far gone unanswered.

Wavelength's picture


It would be relevant if DAC chips made today actually did native DSD. Since I think 2001 the TI/BB DSD1700 was the last true DSD chip made and it was only capable of DSD1 speeds.

Today all chip makers including ESS (used here) convert DSD inside the chip so that their digital filters and output stage works. These DAC chips made by ESS, TI/BB, AKM and Wolfson are all optimized for PCM output.

But in the end with any solution, the proof is in the pudding and how good it tastes. Maybe PS Audio's equipment sounds better this way? got me, everyone should try the Kool Aid before buying it.


Michael Lavorgna's picture

...with PS Audio saying their FPGA-based process handles DSD natively while Berkeley says since all chips convert DSD to multibit, what's the point. At least that's how I'm reading it.

But I agree, the proof lies in the listening.

Steven Plaskin's picture

Playback Designs also uses a FPGA in their DACs.

Michael Lavorgna's picture

Like in the recently reviewed Hugo.

bigrasshopper's picture

My guess is that until the Berkeley Dac emerges from the inner sanctum of Robert Harley and is heralded into the world with special language and beautific blessings, it will not be given up to the "common" people for consideration.    I would be surprised if it ever makes the pages of Stereophile.  Everyone likes their own pets.   The next stop will probably be Computer Audiophile, where there is very little danger of a sound actually being described in a way that has any relative meaning.  

Michael Lavorgna's picture


I enjoy Chris' site and think he does a great job. I also enjoy reading Ted Brady's reviews.

junker's picture

x2. He's one of the sane ones. Thanks Ted!

bigrasshopper's picture

I didn't mean to sound like I am bashing his site.  His enthusiasm for audio and the ingenuity of his frontiering platform have been extremely helpful in refining the tools of the trade by voicing the varied interests of his readers.  I read every review he writes.  If I have critisms I should probably take them there. I suppose, I just don't  respond to his writing, or rather what he is either unwilling or unable to describe.  I don't know if it's a strategic de-emphasis but I continue to glean all I can from the other valuable impressions that he shares.  

Mostly, I thought it interesting that you decided to mention that you got no response from Berkeley.   It's interesting where, how and why manufacturers launch their wares.  Obviously, a statement product from a company that only makes a very limited range will want the glossiest send out they can possibly get.  I'm sure they'll get it.  But I feel obliged to feel slighted by their silence.  A polite no thank you would not have caused a strain.  What I value is a sensitive, honest reveiw that balances the buyer and his choices with more weight than the reviewers flare for filigree and marvelous analogy, or the reverse that gives the reader no clue as to what the reveiwers preferences or biases are and whether and how those were fulfilled.  Or worst of all, a glorified product description.  I think Micheal and Steve manage to make short work in finding that Goldylocks zone.  I know it's tougher than it looks.

If you are ever offered that Dac for review I'll give it more credence than I do many others.  Ultimately they'll fals or stands on it own merits -  I hope that wasn't a completely useless post!


rexp's picture

The details of the thinking behind this DAC are an interesting read and can be found in a white paper link at the bottom of this press release:

dallasjustice's picture

That's the question. There are no solutions, only trade-offs. EMM and Playback Designs have been converting everything to 2xDSD for many years. So it's no surprise DSD makes those DACs sound their best.  

John Grandberg's picture

Does anyone else notice that graphic being a little odd? Either the PS Audio solution doesn't need any power whatsoever, or else they've deliberately left it out of the picture to make their design look that much more streamlined.

Also, I'm not sure which Sabre chip that might be, as it doesn't match anything I can find online - but "Multiple multibit DACs used at random"? Really? I'm sure that's copied verbatim from ESS literature right?

music or sound's picture

It would be interesting not only to compare single bit to sigma delta multibit but also a multibit R-ladder design. Total-dac now has also a DSD option (conversion to PCM). Multibit sigma delta is the standard compromise between single and full multibit designs.

But this difference not only applies to dacs but also to source material. So pure DSD recording should be compared with standard multibit sigma delta conversions and pure multibit conversions (are there any recent existing?).

What matters more: the dac or the recording?

barrows's picture

It is interesting hear PS is doing here.  But to call this approach simple is really misleading.  They convert everything to a very high rate, multi bit format for processing (volume control and such), then convert it to a single bit format at ultra high rate, and then convert to analog.  This approach makes the final conversion to analog very, very simple, and is probably where the "advantage" of their approach lies.  I also applaud the fact that this approach allows them to have a very precise and no loss digital volume control, even for DSD input.  But the processing required is far from simple to do this, hence the large FPGA. The good news is that it is apparent that Ted Smith has been working on these algorthms for many years, so perhaps he really has perfected things. This approach is not that much different than what goes on inside the ESS chips, except that PS makes a final conversion to a single bit format at the end, before the conversion to analog.  As with anything aspiring to achieve very high performance, the end result will depend on the implementation and the details paid to things like power supply implementation, board layouts to reduce the affects of noise on the final output, clocking, etc.   I would suspect that with that much processing going on inside, keeping that very high speed noise out of the clock ground and the analog output is going to be critical: probably why they ended up transformer coupling the output.  

labjr's picture

I know how they can end the debate. They could invent a third kind of digital audio  and transcode both PCM and DSD!

Seriously, I wonder why didn't they output quad rate DSD which would have completely separated the quantization noise from the audio?  Are they saving that for the next Perfect Wave? or maybe Perfecter Wave?

Ted Smith's picture


Yeah, the slide was getting full and I dropped the individual power supply regulators to each box. The intent of the slide was that we can avoid some inherent problems that chips have to deal with: analog power next to digital power, analog next to digital, clocks next to analog or digital, power dissipation, trouble doing component matching etc. I'm sorry if I didn't make that clear, but the slide was prepared for a talk and I did make it clear when I was giving the talk.

I'm a software guy so to me the approach is simple: all inputs directly to a FPGA, FPGA and clean clock -> reclocker, reclocker -> switch, switch -> low pass filter and clean power thruout. Heck when I built the prototype I only intended to do a DSD DAC, but after I decided to use a FPGA so I could change part of the hardware into software, I thought I should add some more input connectors while I was spinning the hardware. Thank goodness that the Spartan-3E on the prototype was big enough to do the PCM upsampling or I might not have noticed how much better PCM sounded than I expected.

I didn’t mean to claim that anything was unique to my design, just that by using solid engineering and concentrating on the things that matter most you can get something more than was expected. I knew that CD’s could sound much better than most of us have heard, but I had no idea how much better.

I’ll stop rambling now, but if it’s in the spirit of this site I’d be happy to answer any questions.


Michael Lavorgna's picture

Thanks for commenting and doing so is very much in the spirit of the site (we have an "Industry Voice" section to allow for more extended conversations including Q &As). I look forward to listening to the DirectStream DAC.


ball3901's picture

It is my understanding there is an upcoming interview with one of the heavy hitters in the audio industry about this very subject. It will be a very, very interesting one! Well, more interesting will be how people respond! This person will be shooting straight about DSD and PCM, and what it really important and what isn't.

Can't wait!!

Michael Lavorgna's picture
....your wait is over - Q&A with Thorsten Loesch of AMR/iFi.