DAC Manufacturer Aversion To External DSP

[barrows]

From the manufacturers perspective, i can totally understand why they want the customer to use their components as they are, and not do not external oversampling/filtering or even perhaps room correction/EQ.  Manufacturers work really hard to get their products to sound "right" and they have no way of knowing how experienced a customer is, and how much that customer may just screw up the sound by applying various DSPs, etc.  The manufacturer wants to be able to provide some assurance and consistency to customers that their products will sound good.  I do not think this means that (most at least) manufacturers are ignorant of the possibilities of DSP, I think it is just that they know, from experiences they have had over the years of building products and providing customer service, that often, the customer actually does screw up the sound sometimes by making changes without really being knowledgeable enough to do so in a coherent fashion.

Of course with software based changes, usually the customer can get back to the OE conditions if they screw up!  At least there is that.  Physical mods are a different story though!  I have seen some horrific attempts at doing mods to gear, even by so called "professional" modders, who charge a lot for their "services".

 

Also, we always have to remember, that most manufacturers these days, are still a bit "old school" when it comes to computer audio (also reviewers!).  So they really may not understand how much expertise has gone into  Jussi's work on HQPlayer, for the best example I know of.  Also when one suggests DSP to many manufacturers, they often assume this means going for analog, to digital, applying the DSP, and then going back to analog again!  They often do not realize that with computer audio, the customer is going to apply DSP in the computer, before any conversions.

 

I really think there is market, for example, for a DAC specifically designed to take advantage of HQPlayer.  A DAC which just accepts in DSD 256 and above, which has little or even no onboard oversampling of its own (or even better, maybe oversampling from DSD 256 to 90.3808 MHz).  This could be a simple DAC, with either a single Ethernet or USB interface, and perhaps a by- passable analog volume control, with a robust amplifier stage, for driving amp(s) directly.  I really, really want to make this product (I have heard the EC modulators, WOW!), but cannot find support for doing it, yet...

[barrows]

18 hours ago, bobflood said:

The Holo Spring 2 DAC and May DAC are pretty close to that ideal. The Dennefrips are also on that list.

 

Disclosure, I own and love a Holo Spring 2 KTE.

 

I like your project idea, good luck!

I am aware of what Holo and Denafrips are doing, but neither of these really meets my desires.  The Denafrips has no output stage and high output impedance, and as such is really not suitable for driving amps directly.  The Holo DACs have no volume control, and the output from the DSD section is too low to really drive amps directly.  Also both of these DACs include over samplers onboard (I understand that they are defeatable) and both include both R2R sections for PCM plus a discrete DSD converter, this is way too much stuff for me.

 

I would like to see just the input receiver (either USB or Ethernet, where the customer gets to choose one only) and the discrete conversion stage, followed by a volume control (bypassable at 0 dB) and a robust output stage capable of driving amplifiers directly (less than 100 ohms output impedance, with at least 4 V output at full scale, a little more would be better, and high current capability, say 100 mA or more).  Either no onboard oversampling, or maybe designed for DSD 256 input with further (onboard) oversampling to 90.3808 MHz to make the conversion stage even simpler.  No input switching, no switching to bypass the oversampler, just no other stuff at all.   With less "stuff" one has plenty of room in a standard size chassis for complete isolation of the input receiver, and very good multiple power supplies, with no need to go to the extra expense of multiple box chassis designs that would just raise the expense.

[barrows]

20 hours ago, Miska said:

 

And you can also consider T+A into similar category. They have their own filters for PCM side - that can be bypassed if you like. And separate pure bit-perfect discrete DSD side. They now even have built-in DSD1024 capable NAA in the SD(V) 3100 HV.

 

For my development headphone rig, I'm now testing the HA 200 headphone amp / DAC.

 

Jussi:  It appears that the newer T+A DACs use the new Thesycon U-HEAR USB receiver code.  Developers at Thesycon have confirmed that this code works with Native DSD to 1024 on Linux.  Can you confirm that the HA 200 works on Linux for Native DSD at 512 and 1024? 

[barrows]

6 hours ago, Miska said:

Just look for this badge:

Just remember, that this mark is a trademark.  This means that any company displaying this mark is paying for it, and adding that cost to the cable.  And some company which does the "certification" is likely getting a lot of money for it.

I am not suggesting the certification is not "real", and it likely does guarantee that the cable (or at least the sample which they tested) does meet specification.

 

As certification processes like this in the US (I cannot speak for what happens in the EU, I would expect that in the EU the Government is actually involved in certification) are done on a for profit basis by private certification companies, they really rub me the wrong way.  UL, for example, is basically a huge money grab-this very different from say, CE in Europe.

 

Every audiophile should keep at least one, properly verified, USB cable which meets the specifications, around for trouble shooting, and for reference vs. various "audiophile" options one might choose to try.  Certainly there are "audiophile" style USB cables which have proven to be problematic in some situations.  But my experience has also shown that some "audiophile" USB cables do have sound quality advantages (not all though).  I must say I am very skeptical of some of the "Audiophile" USB cables which appear to hand made (I mean the wire itself, not just the terminations): one cannot hand make a high speed data cable and expect it to meet stringent impedance specifications, as being tightly controlled for impedance requires very precise spacing of the conductors and geometry of the wires.  But there are plenty of "Audiophile" USB cables which do meet the impedance and performance specifications: I am also often more interested in "Audiophile" USB cables which actually do not adhere to the USB cable build specs (not the performance specs): as by using a different type of construction, these cables may be able to outperform cables which actually adhere to the build spec.  

[barrows]

28 minutes ago, Miska said:

 

Yes, of USB-IF, it is trademark to avoid people from randomly slamming it on something that has not passed the specified tests. Same goes for many other things like DLNA. Certification processes is not so different from FCC or similar... Or if you want to back yourself in electrical safety for CE marking, something like TÜV or FI approval. Similar approval / certification processes also apply to devices like mobile phones.

 

The certified cables I have cost between 10 and 20 EUR, so I don't hugely mind if part of the cost goes to the certification process. But when the cable has that badge, you know it complies to certain set of specifications.

 

Jussi, respectfully, I disagree.  I feel there is big difference between a government based certification, such as FCC, CE, vs. a private, for profit company which makes certifications like UL here in the US.  We have a lot of problems in the US when it comes to private for profit companies where profit is the only measure of success.  I consider, UL, for example, mostly a scam, although it is true that something which is "UL Listed" is likely safe to use in one's home, the way they collect fees for their "service" is highly suspect to me-many of their "requirements" appear to be arbitrary as well.  As an example, UL makes more money, the more times than can make a manufacturer "fail" testing before they achieve a passing grade and get "listed".  Of course manufacturers are allowed to self certify for CE, which is an interesting approach, but the EU sets the rules, not a for profit private company.  

 

What makes you sure that a USB cable bearing that badge actually meets the specification?  Are the cables individually tested and certified?  I would think not.  Given that many USB cables, bearing that certification, are manufactured in China, where I have had direct experience of cable manfacturing errors and outright scams (the sample cable for testing meets spec, but the subsequent production run cables do not), I would not be surprised if some cables bearing that  badge did not meet spec.

 

Of course for audio we only require USB 2 specifications be met.  And I am in favor of cables which meet the spec for impedance and bandwidth (the actual performance spec), I think it is possible that some "audiophile" USB cables which meet the performance spec (bandwidth and impedance) but do not comply with the construction spec, may perform even better sonically.

 

Perhaps there are a few made in the US USB cables (certified, High Speed), that are perhaps less suspect than ones which originate in China.  Does Belden actually make a USB bulk cable here, i would trust them.  I saw that Ghent markets a USB cable using Belden wire, but it is not USB cable, and does not meet the USB specification, I am unsure why Ghent chose to make a USB cable using that wire.

 

 

[barrows]

13 minutes ago, Miska said:

 

It is at least likely because the design has been tested to comply. Having proper QA is another matter.

 

iPhones are also made in China, as well as lot of other stuff.

 

 

I have Supra USB cable, but what I've tested I failed to hear any difference to a certified cable. So it is now in a storage box unused.

 

It is easy to get normal USB cable with ferrite beads, or have correct type of snap-on ferrites added on normal cable. For my Supra it is not possible because the outer dimensions are not standard. Almost all of my audio use USB cables have ferrites, except USB3 cables (Holo Audio and iFi gear, etc).

 

But I try to stick to objective measures. I'd be curious to see technical explanation what aspects and how they make it perform better sonically and/or measurements from the DAC output that show this difference.

 

Yes, me too!  I hear differences in USB cables, and these are distinct enough that once I know the difference, I can pick them out blinded as well, so the differences exist.  It is maddening, but true nonetheless.  I suspect it has mostly to do with how the signal and noise is interacting with the USB receiver.  Only with poorly designed DACs have I ever heard of any analog DAC output measurements which confirmed the differences in cables, but the differences in sonics are still there with well designed DACs (properly isolated USB inputs, etc).

HiFi News tested some audiophile USB cables a couple of times, and found correlation between listening tests and high precision eye pattern testing, I wish someone would do more testing like this.  I would really like to see really high precision spectrum analysis of the ground, power, and data lines, and of the ground plane on the USB receiver board with different cables. Maybe we would something!

 

The Lush USB cable, for example, sounded really different, and not too my liking (very soft and rounded), but it specifically, on purpose, does meet the USB impedance spec!  Biggest difference I ever heard, and anyone with access to these cables who does not believe in differences should test this, was between the Inakustik Reference and the Lush.  

[barrows]

41 minutes ago, R1200CL said:

I find this DAC discussion very interesting. 
 

Should manufacturers more use ethernet as signal in ?

What’s the benefits over USB as an example.  It seems to me USB creates support issues, will use of Ethernet remove those issues as one example ?

 

Will HMDI as transfer technology be a better option than USB ?

Or shouldn’t manufacturers at least incorporate a HMDI interface from your TV or player. 
 

What DAC technology is best for DSD input ?

 

Shouldn't a DAC with no internal up sampling and “perfect” clocks and power sound totally neutral, and the challenge for designer be how good they are to defeat jitter. (And SQ signature be decided by HQPlayer).

 

Will Ethernet to SPDIF converters be something that will better keep many of us using our old DAC with no USB interface, or poor usb interface be something to consider. And will this device still need support for RAAT and NAA, which I guess also mean a simple web interface, unless one create one Ethernet input (fiber) for each option ? (Or a simple switch to select).

 

Or could we place RAAT and NAA inside NUC, SonicTransporter, or whatever PC in use for the streaming purposes (if the DAC accepts Ethernet) ?

 

I wouldn’t mind if John S add some input here, but I suppose Superdad also can tell us something about an ideal technology for streaming and DAC’s. Maybe we need totally new standards ?

 

I think Barrows has some good ideas, but I understand from another tread he also is very keen on remove any preamp and go direct DAC to amps. How will this “requirement” affect SQ ? And cost ?

 

Am I correct that there is a consensus that volume control done right, now can be handled in the digital processing ?  And maybe even better than most preamps ? Hence one should assume preamplifier at least for music streaming isn’t needed. But maybe still for other digital interfaces connected to your DAC ?
 

 

 

1.  Most input options can be implemented to work "perfectly" with good engineering.  Both Ethernet and USB allow for any sample rate, which is an advantage as far as I am concerned.  The main advantage of Ethernet is that has no distance limitations, so the server computer on the Network can be far away from he audio system, and Ethernet has an optical fiber option, which allows for excellent isolation.  But, Ethernet requires more processing power than USB for the receiver, so it needs to be very carefully implemented inside a DAC, with its own separate power supply, and internal shielding.  Also, for DAC manufacturers, Ethernet requires some serious development to work really well, we are in the early days, and it is going to be awhile before many DACs have really god built in Ethernet interfaces, IMO, there are really only a couple of DACs with really good Ethernet interfaces now, and many of these are still limited for sample rate, and/or compatibility with various protocols (HQPlayer NAA for example).

 

2.  HDMI sucks for audio, nuff said...

 

3.  For DSD conversion to analog, a discrete approach can be used, and the data can be kept single bit this way.  Which approach is "best", hmmm...  Implementation is very important, saying one approach is "best" is fraught with problems, as no one can guarantee "best" implementation with a given approach.  There are quite a few DSD DACs around these days, but many which do not keep things single bit all the way through.  T+A, Bricasti, Holo Audio, Denafrips, LampiZator, are some commercial DACs which use discrete single bit converters for DSD.  Also EMM labs and Playback Designs.  And there are various DIY variations on Jussi's DSC approach available as well.  

 

4.  Nothing is "perfect", remember also, a DAC is both an analog and a digital component, and analog circuitry is always going to add its sound signature to a DACs output.  But the approach of dong all the oversampling/processing in a computer and not the DAC does give the user more control over the sound, and frees up the DAC designer to make things as "perfect" as possible in the DAC hardware.

 

5.  As my desires are for single bit DACs and DSD, SPDIF has no place in my world.  SPDF is an ancient interface technology, originally designed for simplicity and ease of use, not ultimate sound quality.  While with heroic engineering SPDIF can be made to sound very good, i am of the opinion that it should be avoided for future development.

 

6.  There is nothing "wrong" with USB.  In fact, USB is one of the best digital audio interfaces we have.  Like anything, it can be done well, or not.  IMO the best option, if done well, is an optical Ethernet interface, with no sample rate limitations, and supporting all the relevant protocols (DLNA, RAAT, and NAA).  but USB is really the best option for most existing DACs right now, with a separate Ethernet Renderer, preferably optical, which supports all sample rates and Ethernet protocols.

 

7.  i prefer a simple system, and there is no reason to have an additional box (preamp) if one does not have multiple sources to contend with.  For me, that means a DAC with optical Ethernet input, going directly to an amp, or even a single box DAC with amplifier inside.  While a lot of audiophiles believe in many boxes, this is old skool thinking.  With the technology available now, it is possible to build things much smaller, and the need for many boxes is mostly inside peoples' heads.  The simpler the system, the less loss one can have, and the better fidelity.  Consider that every connection is a point of loss, every additional circuit is a point of loss...  Not every DAC built now can perfectly drive every amplifier directly, but it is very, very easy to insure that one's DAC design has adequate output capability to drive amplifiers directly.  Digital volume control with DSD is tricky (HQPlayer can do it well).  I think the DAC should have an analog volume control onboard, which is bypassed at 0 dB, in order to allow for volume control for those who might be nervous about just relying on volume control in software (HQPlayer).  There are very, very, very good resistor ladder analog volume control chips, which make adding an analog volume control to the DAC relatively trivial, without compromise.  For example, Pass Labs' most expensive mega buck preamp uses resistor ladder volume control chips.    

[barrows]

3 minutes ago, hopkins said:

Can you give me an example of a USB implementation "done well" were  swapping a USB cable or tweaking with the source has no impact on the sound quality? 

Source and cable still matters, why would it not?  This is no indication of there being anything "wrong" with USB as an interface. In fact, USB is the best digital interface we currently have for audio, yes, some USB inputs are better than others, as is the case with any input.

SPDIF: cables matter

Analog inputs: cables matter

Ethernet inputs: apparently, cables matter

 

just like any input, it can be done well, or not, thta is up the engineering.  But there is nothing inherently wrong with USB.

[barrows]

4 minutes ago, hopkins said:

So what is your explanation for the fact that even with a "well implemented" USB input there is still an impact of cables and source tweaking? 

This has been discussed ad infinitum at other places on this site and I feel no need to repeat it here.

 

And remember, my best case scenario DAC project outlined here would utilize an optical ethernet input.  This does, however, require a very sophisticated onboard Ethernet receiver implementation and isolation.

[barrows]

6 minutes ago, R1200CL said:

Ethernet seems like a better option than USB in order to avoid these cable discussions, and it already have a good standard for fiber, and remove the cable issues. 
 

Anyway, the manufacturers could make a module based DAC, so one could choose preferred interfaces. 

Yes, but USB is currently available, and optical ethernet in a DAC is in its infancy.  It will be quite a few years before we see really well implemented optical Ethernet interfaces become any kind of standard in DACs.  Right now there is one DAC with this I ma aware of.  

If i was developing the proposed DAC for commercial release, it would have a modular input section, and the customer could choose between optical ethernet and USB inputs at purchase.  The requirements would be to support up to DSD 1024 and for ethernet, DLNA, RAAT, and NAA.

 

Jussi, I just do not consider multi channel a viable thing for commercial high end audio products.  there is so little interest in it in the high end, that it is a niche I would avoid.  Not that I never appreciate a really good MC set up, with a few titles...  It is just not very viable commercially or domestically.

[barrows]

19 minutes ago, R1200CL said:

What is your best guess cost estimate for such DAC ? (I think XLR out).

HW cost. (As engineering may be harder to predict). What is your preferred DAC chip ?

 

Can we at AS challenge some DAC manufacturers to build a DAC according to such specs ?

Maybe some sort of group buy or crow founding ?

impossible to predict due to development costs, and engineering time.  My goal would be to have it be reasonable, as I have no interest in making 5 figure DACs.

 

This would be for DSD 256 input and above only, no PCM, US made,  I would like to target $7K or less, hopefully...  But remember this approach is all about less features: single input only, designed for oversampling in software, etc.  Sales would be somewhat limited, as this is special use case product, so development costs would only be spread over a handful of units.

No DAC chip, my preference would be for discrete DSD conversion stage, running at a high rate of DSD.  the exact topology of this conversion stage would be TBD, I have a heard a few different discrete DSD DACs here, and some of these approaches have some "problems" which would need to be ironed out.

 

Existing manufacturers want to build their own products, as well they should, and are unlikely to build something that few folks on the Internet want.  So this would require serious backing, and at least a year long development cycle, probably hiring at least two really sharp engineers-doing something like this is not simple, especially doing it right.

 

And there are plenty of good DACs out there already, which are close, like T+A and Playback Designs, etc, etc...  So I am not sure it would be a wise move to invest a ton of $$ into.

[barrows]

2 hours ago, Albrecht said:

And if I was a manufacturer like Ed Meitner, - I"d NEVER put an USB B input on any of my DACs. 

I disagree with this.  The same could be said for any digital input on a DAC.  I mean, there are plenty of terrible sources with Toslink output, or SPDIF or AES... (and not even considering how bad HDMI is by design...)

If you are making a DAC, you have to put some kind of input on it.  Of course it is probably wise to advise your customers that source quality will make a difference in the ultimate performance of the product.  Despite every effort of every DAC manufacturer who claims that their DAC is immune to source quality, i have never experienced such a DAC, and I have played with a lot of DACs over the last 15 years or so.  Some DACs are certainly better in this regard than others, but none of my experience have been immune to the source quality.

[barrows]

6 hours ago, Summit said:

HDMI doesn’t sucks for audio. If it would no manufacturer would use it for AV.

When I say HDMI, I mean HDMI.  It is compromised for audio, if you do not believe me, do a little research.  Charlie Hansen (RIP) has explained why fairly well and his thoughts on this are available somewhere on the Internet, a Google search will find it.

My opinions are formed from technical facts, these are not things I am just making up.  HDMI was developed in order to provide a simple, easy to use, interface for home theater use, it was not designed to deliver the best audio performance possible.

I would suggest that anytime one combines a Television with an audio, one is making a compromise on audio performance, unless one unplugs all TV from the wall and connections between it and the audio system before listening to music.   

[barrows]

35 minutes ago, 4est said:

I wouldn't get too hung up on having an uber expensive machine for starters. Even then they last for a long time if you are willing to stay put. Until my recent upgrade I was getting by with a i7 3770k with a 1060 video card. It would play- poly sinc/DSD7 at 512 just fine. The new filters are even better, but I didn't know what I was missing until I upgraded. It sounded great, and one can always upgrade computers as they go. When using an NAA one has already separated the functions and there are plenty of lower cost computers than can serve as an endpoint.

Except that I would mention, at least for me, once I heard the EC modulators at DSD 256 through my (DIY) DSC-2 DAC, I would not be satisfied building a machine which could not run them.  This is actually why I am still using Roon for conversion to DSD, as I still have not gotten around to putting together a machine powerful to run the EC modulators at DSD 256 (my previous auditioning of them was brief, they would play on an I9 machine, not "K", for 20 seconds or so before stuttering).  

Ultimate goal became for me a non-compromised machine running I9-9900K or better so that I can run the EC modulators at DSD 256.  For me the EC modulators were so good, that I would not even bother building something which cannot run them reliably.  Jussi says for these modulators the base clock rate is the important factor, not so much Turbo boost, or number of cores.

 

I am looking forward to what Chris builds for the "Twenty" but the 10th Gen Intel processors are getting really, really expensive!  and then throwing in a Jcat NIC and, well...   I am hoping the price of I9-9900K might come down a little once the 10th Gens are established...

[barrows]

54 minutes ago, R1200CL said:

Maybe save the server discussion and HQPlayer requirements to CAPS Twenty part 2, or continue on the HQPlayer  tread 😀

Seems pretty relevant here?  Sending a DAC high rates requires oversampling in a server, so...

[barrows]

5 minutes ago, wklie said:

 

I've given some thought about choice in the context of ESS DAC.  ESS offers multiple filters and a variety of register settings, many of which might affect SQ.  A whole spectrum of the range of choices would be possible.  At one extreme, I've seen an ESS DAC exposing almost the whole register set configurable via the front panel.  Except for the DIY community, it would be too confusing for most users.  A simple and valid question like "what settings sound best" would be unanswerable.

 

At the other extreme like our products, we leave no choice to users (other than DSD upsampling, which marketing considers to be important).  We never had selectable filter in our product line.  Before the release of our ES9038PRO based flagship product, I read a number of reviews of ESS DAC and gathered filter preferences from multiple reviewers.  I researched the public discussions of filters, read opinions from experts like Miska and other manufacturers, and consulted an expert in private.  I also had a group of users did a blind filter test for me and have come to the conclusion that a selectable filter would be worthwhile.  Instead of exposing all the ESS filters, in an attempt to balance between choice and confusion, I proposed that we should have two filters.  Unfortunately, it was unanimously banned by all our decision makers.  Anyway, we keep our "house sound".

Thanks for providing a manufacturer's perspective.

 

This leaves me, personally, even more convinced that there could be enough demand in the marketplace for a very simple DAC, designed specifically for users who desire to oversample in software.  I am thinking a DAC designed for DSD 256 input (and up) only, which would use a discrete DSD conversion stage, include enough oomph in the output stage to drive amplifiers directly, and have a bypassable analog volume control (for those who might not be willing to fully trust VC in software).

While HQPlayer would be recommended, outstanding performance would also be available form other players, such as Roon, and Daphile, which both have good conversion to DSD (unfortunately Audirvana still does not allow for DSD-DSD oversampling, although Audirvana would work for those people who have no DSD files). 

[barrows]

15 minutes ago, One and a half said:

By simple DAC, do you mean an R2R type?

No, not for me, R2R is for PCM only, i prefer DSD conversion, as DSD conversion (at a high enough input sample rate) can be achieved by a fairly simple circuit, and have excellent linearity (unlike r2R for PCM with is fraught with linearity problems).

There have been a number of DSD conversion approaches done discretely, but most of them use some kind of high speed switch, followed by a filter circuit.  I do not know what the "best" approach of these is, figuring that out would likely take some development.  A version of Jussi's DSC approach is likely to produce very good results with enough attention to the details and excellent engineering.

 

When I say simple, mainly I am referring to not having any (or possibly just a simple bit to a much higher rate from an already high rate input) processing on the signal in the DAC, so no need for FPGAs or DSP chips doing all kinds of digital manipulation, instead, just the receiver circuit (USB or optical Ethernet) and the re-clocking stage, then input direct to the discrete converter.