What is the benefit of very expensive DACs?

Legolas

VIP/Donor
Dec 27, 2015
1,042
387
455
France
These are my findings, after having modded about 20 different DAC's for 10 years, attending shows for 15 years and now selling my own DAC for a few years:

Above about $10K, most of them sound very good. There are some giant-killers out there, even at this price range though. Going to $20K or $50K will not necessarily get you better SQ IMO. More features like MQA, DSD, network access, and prettier chassis/displays, maybe. Features and aesthetics are nice, but I prefer the SQ to lots of features. Just my take.

Achieving great SQ is mostly a matter of implementation in the DAC, not so much the parts selection. PC board design, power supply design, power delivery design (decoupling, wiring, regulation etc,..) and the simplicity of the signal path are all critical. Choosing op-amps over discrete transistor design can be beneficial, however there are some really good op-amps out there that sound amazing as long as you don't use too many and feed them properly. I do stay away from really old D/A chips however because they tend not to reproduce the HF transients very accurately. One critical thing to consider when buying an expensive DAC is how are they doing the digital filtering. Aside from jitter, digital filters are the bain of good SQ.

Steve N.
Empirical Audio

Indeed, that is why IMO it lis best to do away with them altogether (the filter). Not sure I like pop amps at all. Much prefer I/V conversion with a transformer.
 

Empirical Audio

Industry Expert
Oct 12, 2017
1,169
207
150
Great Pacific Northwest
www.empiricalaudio.com
Indeed, that is why IMO it lis best to do away with them altogether (the filter). Not sure I like pop amps at all. Much prefer I/V conversion with a transformer.

The problem with op-amps is that most designers don't know how to feed them power properly. Likewise with D/A chips. Each is a little different, but the power delivery is the key. Also, board layout is critical to reducing distortion and noise. Do these things properly and good Op-amps from Analog devices and Burr-Brown can actually beat discrete implementations.

How is that possible? Two things: 1) the circuitry is more compact on a chip, eliminating board effects (from ground-planes and adjacent traces) 2) the power delivery to a distributed discrete implementation is problematic because of the large real-estate that it takes

I did mods for 10 years, ending in 2009. I have done mods on both implementations and the result is that the discrete implementation sometimes sounds a bit more analog but ultimately lacks dynamics/linearity. I found that the Op-amp could be made to sound better with mods in all cases.

One must be picky about what Op-amps one uses because some of them have poor power delivery implementations on-die, making it virtually impossible to get them to sound really good. Others are just mediocre designs.

Steve N.
Empirical Audio
 

morricab

Well-Known Member
Apr 25, 2014
9,391
4,988
978
Switzerland
I heard a M2 Vaughan. Sounds very good, not as good as the best analog gear but anyhow.

Now my Question: What is the benefit of a DAC lets say for over 20k? The last 5% or less or just the name?

I am using a Windows 8 with Foobar and Jplay (please do not start any discussion). I do not think that JRiver 19 is better.

Josef

In Germany it is impossible to get one of the big DACs for testing

If you are willing to drive to Switzerland (near Zurich) then you can hear the Aries Cerat Kassandra Ref. MkII against whatever you want to bring for comparison.
 

opus112

Well-Known Member
Feb 24, 2016
462
4
148
Zhejiang
Indeed, that is why IMO it lis best to do away with them altogether (the filter). Not sure I like pop amps at all. Much prefer I/V conversion with a transformer.

Using an opamp for I/V conversion is perhaps one of the worst possible applications for it. Transformers provide the advantage of galvanic isolation which is jolly useful in the output stage of a DAC - it minimizes common-mode currents which are jolly hard to avoid when running unbalanced cables. I'm aware that Audio Note claim I/V conversion with a transformer but its unclear to me how a transformer can perform this function. I prefer when designing DACs to use a discrete I/V circuit and then feed the derived voltage into the output transformer. A single transistor (bipolar or MOSFET) does a surprisingly good job as current conveyor when you're not too worried about THD measurements.
 

Legolas

VIP/Donor
Dec 27, 2015
1,042
387
455
France
These are my findings, after having modded about 20 different DAC's for 10 years, attending shows for 15 years and now selling my own DAC for a few years:

Above about $10K, most of them sound very good. There are some giant-killers out there, even at this price range though. Going to $20K or $50K will not necessarily get you better SQ IMO. More features like MQA, DSD, network access, and prettier chassis/displays, maybe. Features and aesthetics are nice, but I prefer the SQ to lots of features. Just my take.

Achieving great SQ is mostly a matter of implementation in the DAC, not so much the parts selection. PC board design, power supply design, power delivery design (decoupling, wiring, regulation etc,..) and the simplicity of the signal path are all critical. Choosing op-amps over discrete transistor design can be beneficial, however there are some really good op-amps out there that sound amazing as long as you don't use too many and feed them properly. I do stay away from really old D/A chips however because they tend not to reproduce the HF transients very accurately. One critical thing to consider when buying an expensive DAC is how are they doing the digital filtering. Aside from jitter, digital filters are the bain of good SQ.

Steve N.
Empirical Audio

i think there is a different approach with DS designs and conversion. The noise created by up sampling in the chip and the subsequent filter that is required to deal with that by-product, seems to go against the pure straight through nature of digital to analogue conversion if that is possible. The cluster of NOS designs (TotalDAC, Audio Note, Zanden, Aries Cerat, Ypsilon) that do not upsample and do not mess with the data (avoid any logic in the chip) or go chip less in the case of TotalDAC, seem to sound the nearest to real music or very good analogue IMO. Most of these DACs also have no filter or do the filtering in the analogue domain I believe. I am confused by the Select II, does it upsample and have a filter? Anyone know?

Then the DAC in the 10K+ price bracket should, ideally have a great power supply and line stage as well. My opinion is, look at the manufactures other products, and if they also make very good pre-amps then there is a very good chance they have implemented those skills and tech in their DAC as well. Lets face it, a DAC is IMO essentially a pre-amplifier with a digital PCB inside. I don't think anyone could suggest a pre-amp is top level with a basic power supply and opp amps for the gain stage.
 

morricab

Well-Known Member
Apr 25, 2014
9,391
4,988
978
Switzerland
Using an opamp for I/V conversion is perhaps one of the worst possible applications for it. Transformers provide the advantage of galvanic isolation which is jolly useful in the output stage of a DAC - it minimizes common-mode currents which are jolly hard to avoid when running unbalanced cables. I'm aware that Audio Note claim I/V conversion with a transformer but its unclear to me how a transformer can perform this function. I prefer when designing DACs to use a discrete I/V circuit and then feed the derived voltage into the output transformer. A single transistor (bipolar or MOSFET) does a surprisingly good job as current conveyor when you're not too worried about THD measurements.

I/V conversion is what transformers do, no mystery there. Audio Note uses it, Aries Cerat uses trafos too. Monarchy Audio used a simple resistor, which worked well for a high current chip like the PCM63 but less well (so it is said) with the PCM1704.

You are right about opamps being horrible for the purpose of IV conversion because their slew rate is totally inadequate for the pulses coming from a DAC chip. Doesn't matter how good it is as an amplifier per se...it matters what it can do in the use intended and for this use they all suck.
 

morricab

Well-Known Member
Apr 25, 2014
9,391
4,988
978
Switzerland
The problem with op-amps is that most designers don't know how to feed them power properly. Likewise with D/A chips. Each is a little different, but the power delivery is the key. Also, board layout is critical to reducing distortion and noise. Do these things properly and good Op-amps from Analog devices and Burr-Brown can actually beat discrete implementations.

How is that possible? Two things: 1) the circuitry is more compact on a chip, eliminating board effects (from ground-planes and adjacent traces) 2) the power delivery to a distributed discrete implementation is problematic because of the large real-estate that it takes

I did mods for 10 years, ending in 2009. I have done mods on both implementations and the result is that the discrete implementation sometimes sounds a bit more analog but ultimately lacks dynamics/linearity. I found that the Op-amp could be made to sound better with mods in all cases.

One must be picky about what Op-amps one uses because some of them have poor power delivery implementations on-die, making it virtually impossible to get them to sound really good. Others are just mediocre designs.

Steve N.
Empirical Audio

The problem is that opamps suck for the purpose of I/V conversion...if you look what the opamp has to deal with coming off the DAC chip and you look at slew rate specs it becomes pretty easy to see the issue. Now, after the I/V conversion you might be right that an opamp can beat a discrete transistor configuration for the analog filter/buffer but not a good tube stage IMO.

The simplest implementations I have heard done right, resistor or transformer I/V conversion followed by a single tube output stage, with or without output trafo has consistently sounded better to me than other configurations.
 

morricab

Well-Known Member
Apr 25, 2014
9,391
4,988
978
Switzerland
i think there is a different approach with DS designs and conversion. The noise created by up sampling in the chip and the subsequent filter that is required to deal with that by-product, seems to go against the pure straight through nature of digital to analogue conversion if that is possible. The cluster of NOS designs (TotalDAC, Audio Note, Zanden, Aries Cerat, Ypsilon) that do not upsample and do not mess with the data (avoid any logic in the chip) or go chip less in the case of TotalDAC, seem to sound the nearest to real music or very good analogue IMO. Most of these DACs also have no filter or do the filtering in the analogue domain I believe. I am confused by the Select II, does it upsample and have a filter? Anyone know?

Then the DAC in the 10K+ price bracket should, ideally have a great power supply and line stage as well. My opinion is, look at the manufactures other products, and if they also make very good pre-amps then there is a very good chance they have implemented those skills and tech in their DAC as well. Lets face it, a DAC is IMO essentially a pre-amplifier with a digital PCB inside. I don't think anyone could suggest a pre-amp is top level with a basic power supply and opp amps for the gain stage.

I mostly agree; however, I still think DACs can sound very good with even integer oversampling (like 4x or 8x). I have two nice sounding DACs in addition to the Kassandra (of course they aren't in this league), the Monarchy M24, which has the BB DF1704 digital oversampling filter (8x) and Kinergetics Research KCD-55 Ultra (with Sony 8x oversampling filter) that I have also Lampizated with an Amber DAC output stage. Both are very nice for the money and require extreme efforts to top with redbook cd.
 

opus112

Well-Known Member
Feb 24, 2016
462
4
148
Zhejiang
I/V conversion is what transformers do, no mystery there. Audio Note uses it, Aries Cerat uses trafos too.

Certainly mysterious to me how it can be done - got any links or even explanations of your own?

You are right about opamps being horrible for the purpose of IV conversion because their slew rate is totally inadequate for the pulses coming from a DAC chip. Doesn't matter how good it is as an amplifier per se...it matters what it can do in the use intended and for this use they all suck.

Slew rate is but one of the reasons, but its a major one for sure. The other biggie in my experience is that opamps operate in classAB.
 

Empirical Audio

Industry Expert
Oct 12, 2017
1,169
207
150
Great Pacific Northwest
www.empiricalaudio.com
Certainly mysterious to me how it can be done - got any links or even explanations of your own?



Slew rate is but one of the reasons, but its a major one for sure. The other biggie in my experience is that opamps operate in classAB.

True, however it is a simple matter to bias Op-amps into pure class-A. This is what I do in my DAC, which uses a single OP-Amp and a discrete class-A driver in each signal path. Cannot get much simpler than one stage. That's why it sounds good enough to get awards. Simpler is better. The circuitry around it is very complex though, to make it run linear, set gain etc... I run it with 24/192 filter for all sample rates. This is what makes it sound so good for 44.1 tracks.

Steve N.
Empirical Audio
 

opus112

Well-Known Member
Feb 24, 2016
462
4
148
Zhejiang
True, however it is a simple matter to bias Op-amps into pure class-A.

If your aim is classA operation then yes, for sure a CCS to either of the supply rails achieves it. However the weakness of opamps in my experience isn't quite solved by that measure. That's because there's still a push-pull output stage rather than the (much more desirable for SQ) single-ended OPS. No amount of external biassing can fix that. Discrete seems to be the only solution at least as far as I'm aware, for opamps that don't bring an external comp pin out which allows complete bypass of the OPS.
 

opus112

Well-Known Member
Feb 24, 2016
462
4
148
Zhejiang
Certainly mysterious to me how it can be done - got any links or even explanations of your own?

I did some of my own research into this and found a very helpful page over on Sowter's site : http://www.sowter.co.uk/dacs.php

Its clear from the schematic shown that the transformer's not doing I/V conversion as there's still an I/V resistor on the secondary. The trafo is doing impedance transformation plus some filtering by means of its leakage inductance.
 

LL21

Well-Known Member
Dec 26, 2010
14,411
2,509
1,448
If your aim is classA operation then yes, for sure a CCS to either of the supply rails achieves it. However the weakness of opamps in my experience isn't quite solved by that measure. That's because there's still a push-pull output stage rather than the (much more desirable for SQ) single-ended OPS. No amount of external biassing can fix that. Discrete seems to be the only solution at least as far as I'm aware, for opamps that don't bring an external comp pin out which allows complete bypass of the OPS.

Always enjoying reading and [trying] to understand your posts as I am not a techie. I probably have asked you this before, but it has been a while. Do you have a general approach to DACs you favor? ie, R2R, etc? Would love to read more about that, and perhaps others like Empirical Audio could do likewise in layman's terms so those of us who understand very little could at least try to keep up.
 

opus112

Well-Known Member
Feb 24, 2016
462
4
148
Zhejiang
Always enjoying reading and [trying] to understand your posts as I am not a techie.

The enjoyment is mutual Lloyd - always have appreciated your questions on technical issues.

I probably have asked you this before, but it has been a while. Do you have a general approach to DACs you favor? ie, R2R, etc? Would love to read more about that, and perhaps others like Empirical Audio could do likewise in layman's terms so those of us who understand very little could at least try to keep up.

Yes I'm still a devotee of multibit DACs even though I've had a brief fling with PWM. I might indeed go back to PWM at some point when I'm wearing a firmware hat. I've stuck with NXP/Philips chips for all the time I've been doing DAC development (which is at least 7 years now), TDA1387 is still far and away the best bang-for-the-buck in RBCD compatible chips.

If you want to read more in detail about my deliberations into my latest DAC design (its called '0DAC' or 'lingDAC') which is aimed at being the cheapest DAC with high-end pretensions, go over to Hackaday where I'm still updating the progress of the project : https://hackaday.io/project/27001-audiophile-sounding-dac-for-almost-no-money

The TL;DR version is - its using an array (typically 4) TDA1387s followed by a steep passive filter, then an active discrete I/V stage. This approach is I believe original in the sense that very few DACs are using steep passive filtering (5th order in this case) and none that I'm so far aware of are using a passive filter before a discrete active I/V. The advantage of doing it this way around is the active circuitry is protected by the filter from the very wideband output of the DAC chip itself. Previously I'd been doing passive I/V followed by passive filtering - the passive I/V places more demands on the DAC's output stage and I found it called for too extreme measures on the power supply.

There's also this interview which contains some more background info about my research into DAC SQ - http://fairhedon.com/2017/10/02/an-interview-experienced-digital-and-analog-designer-richard-dudley/
 

LL21

Well-Known Member
Dec 26, 2010
14,411
2,509
1,448
The enjoyment is mutual Lloyd - always have appreciated your questions on technical issues.



Yes I'm still a devotee of multibit DACs even though I've had a brief fling with PWM. I might indeed go back to PWM at some point when I'm wearing a firmware hat. I've stuck with NXP/Philips chips for all the time I've been doing DAC development (which is at least 7 years now), TDA1387 is still far and away the best bang-for-the-buck in RBCD compatible chips.

If you want to read more in detail about my deliberations into my latest DAC design (its called '0DAC' or 'lingDAC') which is aimed at being the cheapest DAC with high-end pretensions, go over to Hackaday where I'm still updating the progress of the project : https://hackaday.io/project/27001-audiophile-sounding-dac-for-almost-no-money

The TL;DR version is - its using an array (typically 4) TDA1387s followed by a steep passive filter, then an active discrete I/V stage. This approach is I believe original in the sense that very few DACs are using steep passive filtering (5th order in this case) and none that I'm so far aware of are using a passive filter before a discrete active I/V. The advantage of doing it this way around is the active circuitry is protected by the filter from the very wideband output of the DAC chip itself. Previously I'd been doing passive I/V followed by passive filtering - the passive I/V places more demands on the DAC's output stage and I found it called for too extreme measures on the power supply.

There's also this interview which contains some more background info about my research into DAC SQ - http://fairhedon.com/2017/10/02/an-interview-experienced-digital-and-analog-designer-richard-dudley/

Wow! Thanks! I just started to read it...and am flattered i actually made it into your postings!
 

Empirical Audio

Industry Expert
Oct 12, 2017
1,169
207
150
Great Pacific Northwest
www.empiricalaudio.com
If your aim is classA operation then yes, for sure a CCS to either of the supply rails achieves it. However the weakness of opamps in my experience isn't quite solved by that measure. That's because there's still a push-pull output stage rather than the (much more desirable for SQ) single-ended OPS. No amount of external biassing can fix that. Discrete seems to be the only solution at least as far as I'm aware, for opamps that don't bring an external comp pin out which allows complete bypass of the OPS.

The downside to discretes is the real-estate they take. A bad result of this is that power delivery to discrete stages is difficult and never as optimum as a chip.

The sound quality I get with the class-A biased op-amp is excellent. I used to mod DACs, lots of them. Some had discrete output stages. Even with a lot of mods to these, I could never get them to sound like my op-amp/discrete output stage.

Steve N.
 

Empirical Audio

Industry Expert
Oct 12, 2017
1,169
207
150
Great Pacific Northwest
www.empiricalaudio.com
Always enjoying reading and [trying] to understand your posts as I am not a techie. I probably have asked you this before, but it has been a while. Do you have a general approach to DACs you favor? ie, R2R, etc? Would love to read more about that, and perhaps others like Empirical Audio could do likewise in layman's terms so those of us who understand very little could at least try to keep up.

Every approach to the DAC has positives and negatives. I have sold ladder-DAC with Tube output stage and now I sell a Sigma-Delta with selectable digital filters and SS output stage. The ladder DAC was based on an older chip and was very musical, but ultimately lacked the kind of focus, dynamics and detail that I get with the Sigma-Delta DAC. It also did not support hi-res. I listened to a lot of DAC chips before I chose Analog Devices. Some chips, like the AK4396 chip that uses switched-capacitors always sounded electronic to me. I modded a proto board with them and was never able to make them musical.

My strategy is to make the signal path as short as possible, feed each stage with its own voltage regulator and never create breaks in the ground-planes. The analog part of the circuit board is for traces less than 1". Anything longer gets a jumper comprising silver wire with cotton insulation. PC boards are one of the things that impacts SQ. I discovered this doing mods on Mark Levinson gear. Very dark sounding because of long PC board traces.

My DAC has a selectable digital filter, so that one can select the 192 filter when playing 44.1 tracks. Then the analog filters are mainly in play. This is what makes it sound so good for 44.1.

My DAC volume control varies the reference voltage to the D/A chip. Using this technique avoids changing the data, changing gain or signal path impedance when adjusting volume. It avoids bit-decimation that software volumes cause. It avoids additional stages that add distortion and noise. As the volume is decreased, the distortion actually decreases. All other volume technologies make it increase.

Steve N.
Empirical Audio
 

LL21

Well-Known Member
Dec 26, 2010
14,411
2,509
1,448
Every approach to the DAC has positives and negatives. I have sold ladder-DAC with Tube output stage and now I sell a Sigma-Delta with selectable digital filters and SS output stage. The ladder DAC was based on an older chip and was very musical, but ultimately lacked the kind of focus, dynamics and detail that I get with the Sigma-Delta DAC. It also did not support hi-res. I listened to a lot of DAC chips before I chose Analog Devices. Some chips, like the AK4396 chip that uses switched-capacitors always sounded electronic to me. I modded a proto board with them and was never able to make them musical.

My strategy is to make the signal path as short as possible, feed each stage with its own voltage regulator and never create breaks in the ground-planes. The analog part of the circuit board is for traces less than 1". Anything longer gets a jumper comprising silver wire with cotton insulation. PC boards are one of the things that impacts SQ. I discovered this doing mods on Mark Levinson gear. Very dark sounding because of long PC board traces.

My DAC has a selectable digital filter, so that one can select the 192 filter when playing 44.1 tracks. Then the analog filters are mainly in play. This is what makes it sound so good for 44.1.

My DAC volume control varies the reference voltage to the D/A chip. Using this technique avoids changing the data, changing gain or signal path impedance when adjusting volume. It avoids bit-decimation that software volumes cause. It avoids additional stages that add distortion and noise. As the volume is decreased, the distortion actually decreases. All other volume technologies make it increase.

Steve N.
Empirical Audio

Very, very interesting. Thanks!
 

Empirical Audio

Industry Expert
Oct 12, 2017
1,169
207
150
Great Pacific Northwest
www.empiricalaudio.com
I see a lot of designers chasing new technologies, such as DSD, MQA and new digital filter technologies. They want the latest Saber chip in their new DAC. They believe that they will ultimately achieve performance by doing this, as well as checking some boxes that many consumers want.

I, on the other hand, spend most of my time optimizing the D/A technology I choose because it's the novel circuits and careful implementation that ultimately delivers the analog sound, not new technologies IME. Older D/A chips can sound amazing. Novel features also provide Trade Secrets that are difficult to copy (particularly by the Chinese).

Any designer can select the newest DAC technology. It's difficult to distinguish your products doing only this. Designing novel circuits and using optimum implementation is not something in the "cook-books" or datasheets either. A lot of experience and know-how is required.

The fact is that it requires one to wear a lot of hats to do an excellent DAC design:

digital design
analog design
grounding and shielding
transmission-line effects and termination
circuit board layout and stackup
power delivery/decoupling
power supply
passive parts selection
voltage regulator design
firmware design
S/W driver design
chassis/mechanical design

I would guess that 0% of DAC designers are experts in all of these critical areas. That is why I contract with other experts in some areas for their design expertise or license their technology. I licensed for instance; Paul Hynes regulator technology that I use in all of my products. I know which areas I am weak in, or those that are not worth my time getting up to speed on. I think it is important for a designer to know his limitations.

Steve N.
Empirical Audio
 

About us

  • What’s Best Forum is THE forum for high end audio, product reviews, advice and sharing experiences on the best of everything else. This is THE place where audiophiles and audio companies discuss vintage, contemporary and new audio products, music servers, music streamers, computer audio, digital-to-analog converters, turntables, phono stages, cartridges, reel-to-reel tape machines, speakers, headphones and tube and solid-state amplification. Founded in 2010 What’s Best Forum invites intelligent and courteous people of all interests and backgrounds to describe and discuss the best of everything. From beginners to life-long hobbyists to industry professionals, we enjoy learning about new things and meeting new people, and participating in spirited debates.

Quick Navigation

User Menu

Steve Williams
Site Founder | Site Owner | Administrator
Ron Resnick
Site Co-Owner | Administrator
Julian (The Fixer)
Website Build | Marketing Managersing