Q&A Michal Jurewicz, Mytek Digital

Michal at Mytek R&D Studio in Brooklyn, testing the new DSD ADC prototype circuit (June 2014)

I first met Mytek's Michal Jurewicz at RMAF 2011. This event marked the official launch of AudioStream as well as the debut of Mytek's first consumer DAC. At that time, the Mytek Stereo192-DSD DAC at $1695 represented the least expensive DSD-capable DAC by a good margin and remains, imo, very competitive in terms of its price and performance. Michal was kind enough to participate in this Q&A where we focus on DSD, misconceptions, and Mytek's place in the world of hi-fi.

Could you tell us about Mytek's history and when and why you decided to get into consumer audio?
I arrived in New York in 1989 after graduating from the University in Electronics and Acoustics as an EE in my native Warsaw, Poland. That summer I got a job at New York’s largest recording studios; the Hit Factory. I later got hired by another major Manhattan Studio - Skyline, home to legendary producer Nile Rodgers. I started designing converters in 1990, while at Skyline, where there was this fantastic opportunity to test them on the spot, still hot from soldering. My early converter prototypes where used to mix down many major records of that time, for example Mariah Carey and James Taylor or the B52’s who worked at the studio.

Mytek 8x192 A/D/A

I formally registered Mytek in 1992, and started to run it full time after Skyline closed in 1995. I have since designed about 30 different A/D and D/A converter models for professional use. In the early 2000s we opened up the Mytek manufacturing and engineering facility in Poland. Our main strategic design and sales headquarters are in Brooklyn, New York, while most manufacturing and programming R&D is conducted in Warsaw (although the new Manhattan DAC chassis is still made in the US for ultra hi-tech reasons). Poland’s factory services Europe and Far East and does the majority of production and a lot of nuts and bolts product R&D. Poland’s office is headed by Marcin, with Adam conducting tech support and Ola, Adrian, Darek and the team of engineers and technicians making sure products are up to date tested and supported.

Mytek converters are considered one of the best sounding and are widely used in many recording and mastering studios. A number of large record label transfer facilities such as Universal, Warner and Sony use Mytek for PCM and/or DSD analog transfers. Countless major records were either recorded or transferred through Mytek converters. We have many major artists, Norah Jones, Metallica, Beck, Aerosmith, to name the few, owning and using Mytek.

I have kept my eyes on hi-fi market for a while, but it was the advent of computer audio that finally made us appropriate the Stereo192-DSD-DAC, originally designed as mastering DAC, for use in the consumer market. This DAC debuted at RMAF in 2011, and made a huge splash mainly because of it’s sound quality and DSD feature set. We have been very successful with this product, having since sold over 3000 units worldwide. This success made us realize the size of the consumer market which is substantially larger than the specialized professional niche we were servicing up until now.

Mytek Stereo192-DSD DAC

Mytek was one of the first companies to offer a DSD compatible DAC, the Mytek Stereo192-DSD DAC which shook up the market when it was introduced at RMAF 2011. The DSD market has since exploded yet there remains a general confusion about DSD and what represents "Pure DSD" playback. Some people have suggested that all current DAC chips convert DSD to PCM so from their perspective, DSD is pointless. Can you explain what goes on inside your Stereo192-DSD DAC and is it in fact playing back DSD as DSD?
I became familiar with the details of DSD technology in early 2000’s, when the Sony SACD Project commissioned a DSD Studio Master Recorder with Mytek. We have designed and built several prototypes of D-Master, in process learning about DSD and falling in love with DSD sound quality. DSD requires a different approach in handling production, it works well for all types of live or analog recordings, but it’s not the best for complex digital productions requiring large amounts of DSP. In other words this is "less is more format" that can deliver stunning sound quality if it’s used with care and professionalism.

"In other words this is 'less is more format' that can deliver stunning sound quality if it’s used with care and professionalism."

DSD and PCM were pitched initially against each other because of DVD-A vs SACD attempt of market control in early 2000s. There is a residue of that battle still lingering around. Mytek has decided to ignore politics of the format, instead focusing on its merits. Our strategy is to support both DSD and Hi-Res PCM in all our future products. My design philosophy is to always design the next product to sound better than the last. Currently we are designing a new generation of professional A/Ds which will do both 384k/32 bit PCM and quadruple DSD256. Both of these formats are excellent when used appropriately.

Mytek Stereo192-DSD-DAC (as well as the new Manhattan DAC) use the excellent Sabre D/A converter. This chip, if properly implemented (which btw is not trivial), can sound excellent. When playing PCM, the chip works in PCM mode which basically means this signal path: PCM>reconstruction filters> optional 32 bit volume control> modulator > 6 bit “quasi DSD DAC”. In DSD mode the signal path is DSD>32bit vol control in DSD domain>DSD LPF in DSD domain>modulator>6 bit DAC. The signal path in DSD mode inside Sabre is simpler, no surprise it does sound better. The same goes for majority of the DAC chipsets out there with the exception of ladder PCM1704. We are not using 1 bit conversion, instead we are using 6 bit DSD conversion because it offers better performance. 6 bits offers 64 steps in place of 1 step of 1 bit but at the same speed of 2.8Mhz up to 11.2Mhz (for quad DSD).

"Orthodox purists will argue it’s not 'pure 1 bit', but in practice it’s like arguing that the number 64x1/64 is NOT equal to 1. Multibit DSD is better than 1 bit DSD..."

Orthodox purists will argue it’s not "pure 1 bit", but in practice it’s like arguing that the number 64x1/64 is NOT equal to 1. Multibit DSD is better than 1 bit DSD, that’s why Sony used 8 bit DSD called DSD wide in their Sonoma DSD workstation and that’s why we use multibit DSD too. There is an AES paper on DSD-Wide by Peter Eastty explaining multi bit DSD which some people may confuse with PCM. That confusion is not really a matter for debate, it’s a matter of understanding how PCM vs 1 bit DSD vs multibit DSD work. The bottom line is that Mytek DACs when playing DSD are “Pure DSD”, there is not a “mystery” PCM step involved.

I've often commented about the sound of DSD as being different from PCM mainly in terms of a dimensional quality I rarely hear with PCM. Its as if music sounds like its coming from a real 3-dimensional space. Do you believe there is a difference between DSD and PCM playback and what would you say accounts for this difference?
In reality the origins of today’s DSD and PCM are the same. 99.9% of chipsets used for production and playback use some form of DSD at either end, even if they process PCM. DSD was originally invented by Sony engineers in the late 1990’s when they realized it’s possible to throw out bunch of digital processing (decimation and reconstruction filters) from an ordinary A/D D/A signal chain and leave only what is essential- front and back DSD modulators. As with anything else in high end audio simplifying this signal path resulted in better sound quality. Removal of filters increased the bandwidth and removed rounding errors in filters which happened to truncate this low level information you call "real 3-dimensional space”.

Resolution Magazine recently published an article I wrote, in which I argue that 24 bit PCM is not enough and compromises the sound quality. In our recent tests of our new prototype ADCs we have observed that reducing the resolution of decimation filters from 32 bit (fixed) to 24 bit results in significant drop in sound quality, particularly that “3D” perception where a lot of low level information is involved. It’s not surprising that properly executed high speed PCM (let’s say 192k and above) with enough bit resolution (32+) can approach DSD in sound quality. But that’s again under the condition one knows what they are doing. It’s really easy to mess up PCM recording, and harder to mess up DSD.

"It’s not surprising that properly executed high speed PCM (let’s say 192k and above) with enough bit resolution (32+) can approach DSD in sound quality."

There is also the question of what constitutes a "pure DSD recording". I am of the opinion that a DSD recording can originate as DSD, as with the Channel Classics recordings for example, as well as recordings sourced from analog tape. Others point out that most DSD recordings have at some point been converted to PCM for editing so they are no longer DSD. Is there such a thing as a native DSD recording and can an analog to DSD transfer be considered DSD?
Generally, I would describe a recording claiming to be "pure DSD” as a recording that went through one generation of live analog to DSD conversion, then was edited, preferably with no gain changes or processing, and put out as an SACD or DSD download.

These generally will be live recordings or “semi-live” when the mix is performed on the spot, usually classical or acoustic recordings (for example Channel Classics or Sound Mirror, some of Blue Coast). Then there are all other types of recordings that require post-production, either mixing or mastering, that may add another layer of conversion or processing veil. In practice, outside of these classical/acoustic recordings, it’d be only a small fraction of records that one could claim are “Pure DSD”. Typically it’d be impossible to even know that without knowing all details of production. IMO labels who claim “Pure DSD” could benefit from describing their production techniques in detail on their album covers.

"In practice, outside of these classical/acoustic recordings, it’d be only a small fraction of records that one could claim are 'Pure DSD'"

For technical reasons (high frequency noise) double DSD or even quadruple sound better and are easier to handle, if one can afford dealing with the file size.

In terms of analog to DSD transfers, one could ask, is a scan of an excellent negative, a digital scan or analog photo? I think it’s both, it could be that the digital scan captures the majority of the “analog film” feel of that photo. Analog to DSD transfers are perfectly legitimate for this reason. I often use digital photography for audio format analogy - think of digital camera RAW file as being “DSD”, then a TIFF being “PCM” and JPEG being “MP3”

"The main benefit of quad DSD256 other than slightly better sound quality when compared to regular DSD, is that it fixes the issue of high frequency noise residue present in regular DSD."

We're starting to see quad-rate DSD capable DACs. Do you think we'll also see the equivalent source material? Or is quad-rate DSD more concerned with filtering and processing lower resolutions?
Our next professional A/D converter to be released later this year can do quad DSD. Typically a mastering or transfer studio will make several passes of the same recording in different formats that they can later use for different things. I’m sure, some of them will make quad DSD files. The main benefit of quad DSD256 other than slightly better sound quality when compared to regular DSD, is that it fixes the issue of high frequency noise residue present in regular DSD. My guess is that the usage of quad DSD will be determined by practicalities, the file is 4x the size of regular DSD. I suspect double DSD is for the moment a good practical compromise.

Mytek Stereo192 ADC

In terms of sound quality, are there differences when designing components for the Pro Audio community as opposed to the consumer community?
The main difference is the budget available. An Audiophile can often spend more on a stereo system than a musician on an entire studio. Studios generally use equipment up to several $K in value. This defines the budget available for design. Over last 25 years Mytek has learned how to maximize sound quality with a defined budget, we are very good with this, as you can experience with Stereo192-DSD-DAC that sound-wise challenges $10K+ DACs. When one designs a dedicated Hi Fi unit, such as our new Manhattan DAC, the focus shifts from pure practicalities only onto design and finish which can be very expensive. A lot of Hi Fi components out there are about the looks, more than the sound. With the Manhattan DAC we tried to cover both ends very well, I think we succeeded, this new DAC is an extremely attractive offering in it’s price range for both reasons, it sounds and looks very good.

"An Audiophile can often spend more on a stereo system than a musician on an entire studio."

Again in terms of sound quality, what would you say are the three most important elements of a DAC and why?
Low level resolution that defines the 3D space and somewhat related distortion specs which define the veil. Then the dynamic performance which is hard to measure but extremely important, tightness of bass, transients etc, this is often related to design of the power supplies and choice of analog signal path. Then the quality of digital section, clock jitter, digital filter design and resolution etc.

Mytek at RMAF 2011

In your Stereo192-DSD DAC you offer both digital and analog volume control. What is the reason for offering this option?
We use a 32 bit fixed digital volume attenuation which has theoretical dynamic range of 192dBs. It ’s very transparent, but still when you significantly turn down volume let’s say -30dB or more you do start hearing its effect on sound quality, very typical digital effect of losing the low level resolution, signal is still clean but more “digital” so to speak. The analog fader that we also offer is not as transparent as digital, it’s in fact slightly euphonic. The resistors we use produce some amount of warm 2nd harmonic. This (very small) euphonic character seems to be often desirable, most of our customers do prefer the analog volume control. In addition, with analog volume control there is no loss of low level resolution. You could say that the analog fader sounds more “analog” while digital sounds like clean “digital”. We also offer relay bypass option for the clients with their own preamp.

Now that you've been involved with the audiophile community for a number of years, what would you say are the most common misconceptions you've encountered?
I have kept an eye on and used hi-fi grade equipment for years. We have dismantled and analyzed countless pieces of hi-fi equipment in the past to try to understand the correlations between sound quality and engineering. Because of larger budgets and unusual creative ideas, hi-fi designs are often very inspiring. A lot of audiophile rumors and myths often do have merit.

"The biggest issue in the audiophile world are not myths but the inability of the end user to quantify their importance."

I generally don’t outright dismiss unusual claims, but try to verify them in order to turn them into benefit in the next design. The biggest issue in the audiophile world are not myths but the inability of the end user to quantify their importance. There is a huge market offering $1000 fixes outside the box for something that could be addressed inside the box for additional $5. It’s not unusual for a $1000 power cord to be used on a $1000 DAC. This is still a legitimate pursuit, but how much it does matter is very subjective. When it comes to misconceptions, there are a lot of these about digital audio in general. You mentioned one earlier: the fact some people think 6 bit DSD is PCM. Most audiophiles listen first, some of them have time to try to understand the theory behind, but not all of them. These are origins of rumors. I still like that environment, as it gives me, the person with a scientific background, the opportunity to analyze and understand things behind the rumors to try give Mytek products competitive edge.

Mytek Manhattan DAC Headphone Preamplifier

Does Mytek have any new consumer products in the works and if so can you give us some details?
Yes, at the recent Munich Hi Fi Show in May we announced the new true Hi Fi product, the new Mnahattan DAC which will soon start shipping. We have a huge interest, with many pre-sold units and are very happy how it came out. Later this year we will be offering our new professional ADC which we hope will be the best ADC in the world in terms of performance. We expect some of these to be used for digitizing vinyl. And there will be more products next year, we are here to stay and to push the performance envelope further, in this now rapidly changing high resolution music market.

COMMENTS
mlgrado's picture

I think it is a valid argument to say that un-adulterated 1 bit DSD is superior to any multi-bit conversion from our original 1 bit signal.

Why?

Consider the great pains engineers who record in true DSD go to to NOT have conversions to multi-bit. I think of Jared and Tom at Channel Classics, and Michael Bishop formerly of Telarc. In the case of the latter, he would rather mix DSD in ANALOG rather than use the DSD-wide conversion of the Sonoma workstation. If DSD-wide was so much better, then why do the top DSD engineers avoid it?

Because there are inevitable losses to be incurred. Convenience is increased at the expense of absolute fidelity. To covert to DSD-wide in the Sony e-chip, the 1 bit signal is converted to 8 bit in a presumably non-decimating fashion, uses 32 bit and greater math for its various DSPs, and is re-modulated back to 8 bit delta sigma as necessary before finally being remodulated to 1 bit.

That is a LOT of processing! And the result is inevitably inferior to the original. The result is fundamentally different than the original 1 bit signal, therefore the analogy of 64x1/64 =1 really doesn't apply at all.

Now I know we are talking about the ESS chipset here, not Sony DSD-wide. But if the ESS indeed do something similar, that is, process DSD in a non-decimating fashion to a multibit intermediate followed by a final modulation,

that is A LOT of processing on our orginal 1 bit signal. Other than convenience, I fail to see how this multi bit conversion is 'better' than native DSD.

Now I am sure the end result sounds fine and well. Afterall, lots of ears enjoy DSD on these ESS based DACS. But I will stick to DACS that pass the DSD signal untouched. For maximum fidelity, the 'orthodox' way is the way to go.

mytek's picture

Hi Migrado

There is merit to what you say in the sense that remodulation introduces some veil in DSD. It's also true that remodulation in DSD, which is a process corresponding to dithering and rounding in PCM, introduces even larger degradation. So for example E-chip volume change will produces more severe relative degradation than volume change in 24 bit PCM. (Btw we have used E-chip in D-Master DSD recorder). However, if one uses E-chip or corresponding process with care DSD sound quality being superior at the start can be largely preserved. For example short edits only, where it MUST be used anyway. If there is DSD editing , there must be a remodulation of crossfades there.
I think we both agree less is more for DSD.

As far as why multi bit DSD is better it's a big topic and anyone interested should start with Peter Eastty paper here:

http://www.aes.org/e-lib/browse.cfm?elib=9998

This paper specifically addresses the DSD processing challenges for which DSD -Wide was proposed.

However, the performance challenges for converting from/to analog are similar and adding bits helps to get better performance from these front and back ends.

It's true that you can use 1 bit quantizer which theoretically should be a good approach as at the first glance it's perfectly linear etc. However, because of implementation challenges, it's not physically possible to get as good performance from a 1 bit d/a as you can for example from 6 bit saber.

So while it's true that you do loose some quality in remodulation , you gain more in conversion, the overall balance being substantially positive.

Same goes for AD. I'm pretty sure that all future ADC chipset will use mutlitbit modulators. If you want 1 bit, there is the PCM4202 which is not bad, but there are better ones, multi bit again.

I hope this clarifies my views on the subject a bit more.

Best Regards, Michal at Mytek

Rick Tomaszewicz's picture

isn't always very profitable.

"There is a huge market offering $1000 fixes outside the box for something that could be addressed inside the box for additional $5."

I commented on one-box solutions in the previous article.

X