PCM to DSD (DSC1) vs native PCM

[Lukaso]

On 11/16/2020 at 9:53 PM, Andy-HandyHifi said:

Are there ANY ADC's out there that directly output their 'multi-bit intermediary' format directly out for the ultimate in audio format quality? 

 

As far as I know, there are officially few audio ADCs with such features, the AD1871, AT1201, and PCM4222.

There is probably a custom Cirrus Logic ADC (similar to CS5381) that can output a raw multi-bit signal (unverified information). 

[Andy-HandyHifi]

I guess the question then should become does the gear exist with the appropriate software to receive multi level 1 bit oversampled data (unary/thermometer coded), recognize it as such as edit appropriately?  Considering the available processing power available these days, there is an incredibly untapped resource here.  

 

Miska has proven this fact that we all should have known from the early days that these intermediate formats are quite workable and editable.  Perhaps 20 years ago processing power was the big limitation.  

 

I have no idea about Sonoma/SADIE system which was IMO flawed from the start since the so called DSD-wide format should have been the DSD format from the get go.  Never should have been a single bitstream.  At least to start with, taking hardware and software limitations of the 1990's when these things were being hashed out, you could have at least have had a 24 to 32 level system that would have been editable and unquestionably higher resolution with greater linearity that 1 bit SDM filtered into 'DSD wide' which was 8 bit in 2's complement I believe at 64x.  The problem was why 1 bit SDM to being with as a bottleneck ?  

 

No. no No I am not trying to start a new format was here.  It just seems logical to have originally recorded in a multibit SDM signal that DOESN'T truncate and or decimate at the ADC output before sent to the editing system, where said signal is 'returned' to some kind of mult bit SDM format in the case of DSD-wide and Sonoma pre transmission to digital editor/mixer.   Then does whatever edits, mixes and or mixdowns need be done before once again returning to the lower resolution format similar to what we started with!

 

Pyramix has gone a different direction, and if it were my studio, it would be Pyramix all the way.  You still have the 'punch-in, punch-out' modes that allows for some basic edits like crossfades to momentarily live for a (seconds) in the PCM world where they are then seamlessly re-modulated back to 1 bit as it if never left.  X64 stil does this work at DXD 352khz.  

 

But when we go up to x128 x256 an x512 if it exists yet the edit system operates x64 files at 352khz, x128 files at 705khz, x256 files at 1.4MHZ, and x512 if ever available at nearly 3 megahertz!!!  These are not oversampled, noise-shaped numbers.  We are talking at least 24 bits per sample here.  If I am listening to a DSD256 project recorded in a native purist format with meticulous care at setup, perfect mix set before recording and only minor drops to PCM for editing, I would be in hog heaven.  On the OTHER hand I can't say that if I were listening to a DSD256 file simply as the final output format, while the recording was fully recording and mixed at 24bit 1.4mhz, then modulated to DSD256, I would also be in dog heaven.  

 

The only higher heaven would be to hear and playback that original 'DXD Mix Mode' recorded and edited on the fly with no other conversions at all.  Give me the 24 1.4MHZ PCM file.  

 

I mean seriously.. what is the argument even worth?? Pure DSD at 256, 512, 1024, 2048, etc. versus PCM with 1.4million 24 bit samples?  And as time goes, each side will throw out more and more gaudy number  

 

 

Now what in the hell did I even write all that???? Too late to click delete.  Suffice to say I suppose these are GLORIOUS times for high resolution audio and we need to enjoy the hell out of it.  Oh, and also to say even if the DSD camp is aware of it or not, their efforts are pushing the PCM camp to be even better as well.  

 

Keep competing, I say.  It is better in the end for all our audiophile ears!

[Miska]

On 11/16/2020 at 10:53 PM, Andy-HandyHifi said:

That being, when a DSD signal is present, the ONLY way it is processed is the direct method.  I am positive this is exactly what you were saying. 

 

Yes, indeed...

 

On 11/16/2020 at 10:53 PM, Andy-HandyHifi said:

giving you the the 32 values one would expect of  two's complement 5 bit PCM.  (which of course becomes MUCH MUCH more efficient a way to represent the values once we give to millions of levels in 24 bit and greater ;)

 

Note that switching over to two's complement, while seemingly straightforward, you would loose some of the important presentation power of SDM. Where for most intermediate values you have many possible bit combinations to represent the same value. For example in 4-bit scrambled thermometer code you can represent value "1" in four different ways: 0001, 0010, 0100 and 1000

 

On 11/16/2020 at 10:53 PM, Andy-HandyHifi said:

If we thought of it as we commonly do in 2's complement PCM with negative values, we get this... (assuming we have negatives in unary code... we dont)

 

Most DACs are current output, like DSC1 too, so they don't really have negative values. Just current from 0 to some other value. A chip may have current to voltage converter built-in though. Some chips have this as optional extra, on choice of designer.

 

On 11/16/2020 at 10:53 PM, Andy-HandyHifi said:

My only minor disagreement in what I have independently discovered is that the Signalyst DAC is a 32 level DSM.  When it comes to DSD, it is just a 32 tap FIR filter.

 

No, it is 33-level... ;) With one bit you get two levels, with two bits you get three levels, etc...

 

The register always holds 32 bits, which when latched to output represent once of the 33 possible output levels. The converter itself doesn't understand more. This is always the case, so to get the converter correctly running needs 32 clocks before output is unmuted or otherwise you get loud pop because of uninitialized bits. But this applies to all DSD converters, because you need some time to establish the initial state (some DACs fail to do this and as result you get pops or clicks).

 

On 11/16/2020 at 10:53 PM, Andy-HandyHifi said:

Thank you Miska.  Although the FPGA in my iFi iDSD Pro sounds exceptional, the Signalyst simply by is nature is more capable and I will be using it from now on with Roon/HQPlayer and iDSD direct DSD decode at DSD 1024.

 

Thank you! Great to hear! Modulator design is pretty time consuming and complex task, while it has big impact on the overall performance.

 

And making more and more advanced modulators also make them exceptionally heavy to process. Doing lot of calculations at very high sampling rates means high loads. ASDM7EC being an example of the heaviest so far.

 

On 11/16/2020 at 10:53 PM, Andy-HandyHifi said:

My agreement is why the heck don't we do this all the time??? Why can't be have SDM modulators (other than the Grimm 1 bit converter, discounted here for obvious reasons) output their 4, 8, 16, 32, 64 or 128 level modulator output as a standardised file to be edited, mixed, etc, and THEN downsampled or decimated to the appropriate final format?

 

I guess it is mostly due to convenience factor and legacy. In the past, not very long ago, dealing with something like 96 kHz 24-bit content was heavy and storage consuming enough. So the ADC chips just run the SDM data through PCM conversion, which is also one reason we need to do some fixing with DSP at playback time as well.

 

Most ADC converters use very few levels. If you look at noise floor profiles from ADCs running at 192k or higher PCM output you may notice it looks almost exactly like DSD128 converted to PCM. Because many of the converters run the conversion stage at 5.6/6.1 MHz and only few levels.

 

OTOH, you have multiple DSD128 recorders available (from Korg and TASCAM for example) that use TI's PCM4202 or PCM4204 chip. Which is true 1-bit ADC. Sure it has PCM output mode too, where the SDM stream is converted to PCM as in other chips too. My RME HDSPe AIO cards have PCM4202 ADC (like I believe bunch of other RME converters), but unfortunately lack capability to output the raw DSD data. So you can find many "PCM recordings" on the market, made for example with RME converters, where the data began it's life as real "DSD" and got converted to PCM by the on-chip conversion.

 

In addition, I'm pretty happy with my two RME ADI-2 Pro's and their DSD256 ADC output. Although alternatively you can get very powerful 768/32 PCM output as well.

 

[Andy-HandyHifi]

4 minutes ago, Miska said:

 

Yes, indeed...

 

 

Note that switching over to two's complement, while seemingly straightforward, you would loose some of the important presentation power of SDM. Where for most intermediate values you have many possible bit combinations to represent the same value. For example in 4-bit scrambled thermometer code you can represent value "1" in four different ways: 0001, 0010, 0100 and 1000

Yes absolutely.  Which allows much greater linearity since you can scramble the code, unlike the resistors in something like R2R

 

 

4 minutes ago, Miska said:

 

 

Most DACs are current output, like DSC1 too, so they don't really have negative values. Just current from 0 to some other value. A chip may have current to voltage converter built-in though. Some chips have this as optional extra, on choice of designer.

 

 

No, it is 33-level... ;) With one bit you get two levels, with two bits you get three levels, etc...

 

 

 

 

Yes 33 levels.  That is what I meant. Sorry.  32 switches, taps or whatever will always in unary code be 33 levels.  Now that we have 'stacked' 32 times the same 1 bit signal offset by clock cycle of choice, the result is a 'thermometer' (I am a very visual learner so the idea of calling it thermometer code speaks directly to my visually dominated brain) 

 

Those 32 bits which we will now speak of a bit switches, since they are all equally weighted unlike PCM ladders, can give us a total of 33 values.  Its value 0 through value 32.  Being able to turn all the switched to 'off' seems to slip some folks minds.  Including mine for years. I am not a math guy.  I have quite high IQ but strengths are all artistically weighted to the creative audio-visual fields.  This the master degree in Piano Performance :)  Your LED light example is a terrific one that explains it very well.

 

To clarify my argument about the DAC and whether is is truly a multi bit DSD DAC or not.  It is almost like an oversampling PCM DAC of 1.4mhz.  Just because you are playing 44.1 upsampled through it, doesn't mean you are listening to true 1.4mhz PCM.  Same thing here.  Basically we have a digital DSD FIR filter implemented via discrete analog components.  It will to me always be a 1 bit FIR filter, until the day we get real multi-bit Delta Sigma source material to send to it :)

[Andy-HandyHifi]

which CAN be done, and hopefully will be.  Its this very, very best way IMO to convert to analog.  Unary coded pulse streams which can vary in number of levels as needs and computing power can provide.  

 

It just all seem so SIMPLE now.  

 

I apologize for writing this off for so long, Jussi.  What you were doing was already going on deep in the 'bowels' of places like Burr Brown but you were one of the major internet 'prophets' who from early on spoke on it to the masses.  

 

Job well done, and as for the haters and disbelievers, it really isn't about faith or belief.  The science and the data is right there in front of you if you really want to look.  

[Miska]

59 minutes ago, Andy-HandyHifi said:

Basically we have a digital DSD FIR filter implemented via discrete analog components.  It will to me always be a 1 bit FIR filter, until the day we get real multi-bit Delta Sigma source material to send to it :)

 

Yes, now we send it 1-bit data, but the converter itself doesn't care if we chose something else... :)

 

[Andy-HandyHifi]

1 hour ago, Miska said:

 

Yes, now we send it 1-bit data, but the converter itself doesn't care if we chose something else... :)

 

fair enough :)

 

Indeed all it sees is 32 levels of data.  Doesn't care if it came from a 1 bit delay shift register that is the same 32 'bits' in sequence for all 32, just offset by a clock until, or if it's 32 discrete bit's from the output of a multi level SDM.  All the filter sees are the numbers and does the multiplication and summation just the same.  

 

I really, really do like the sound of the best true PCM converters, like the Pacific Microsonics 2.  Played back via a NOS bitpefect well resistor matched R2R DAC sounds insanely true.  

 

I do believe, however, had we started digital audio NOT with 2;s complement exponential PCM dependent so heavily on the matching parameters of the PCM ADC and DAC resistors, and used something more akin to 32 to 64, even 124 level unary code at higher sample rates, we would be way ahead of the game when it comes to digital sound quality. 

 

Just one man's opinion, though.  Keep up the good work.  HQPlayer is going onto my Mac Server tonight permanently for use with Roon.  

 

Thanks for your contributions! 

[Rexp]

On 11/19/2020 at 7:00 AM, Andy-HandyHifi said:

fair enough :)

 

Indeed all it sees is 32 levels of data.  Doesn't care if it came from a 1 bit delay shift register that is the same 32 'bits' in sequence for all 32, just offset by a clock until, or if it's 32 discrete bit's from the output of a multi level SDM.  All the filter sees are the numbers and does the multiplication and summation just the same.  

 

I really, really do like the sound of the best true PCM converters, like the Pacific Microsonics 2.  Played back via a NOS bitpefect well resistor matched R2R DAC sounds insanely true.  

 

I do believe, however, had we started digital audio NOT with 2;s complement exponential PCM dependent so heavily on the matching parameters of the PCM ADC and DAC resistors, and used something more akin to 32 to 64, even 124 level unary code at higher sample rates, we would be way ahead of the game when it comes to digital sound quality. 

 

Just one man's opinion, though.  Keep up the good work.  HQPlayer is going onto my Mac Server tonight permanently for use with Roon.  

 

Thanks for your contributions! 

If I record at 16/48 using a laptop/Audacity and playback in Audacity it sounds like the original. The fidelity is far higher than in commercial recordings. So I'm not convinced high rates are the answer. 

[Andy-HandyHifi]

On 11/20/2020 at 12:15 AM, Rexp said:

If I record at 16/48 using a laptop/Audacity and playback in Audacity it sounds like the original. The fidelity is far higher than in commercial recordings. So I'm not convinced high rates are the answer. 

 

That is fine.  We were having more a theoretical discussion on the actual method of operation in a discrete analog 'DAC-less' converter that can convert 1 bit Delta Sigma (Pulse Density form branded as DSD by Sony/Philips) and can easily covert to again with no DAC or hardware DSP etc with some slight changes in configuration also any unary/thermometer coded Multi-Bit oversampled files, the faster the better.  

 

When you recorded at 16/48 in audacity, whatever you recorded first in the analog domain was first something completely different than that.  It was 99 percent chance to be exactly multiple one bit unary coded on/off square waves that had to at some point be decimated from millions of samples per second down to 48,000 samples per second.

 

It is a VERY complex subject that gets even the best in the industry confused; not to mention the rest of us.  But it really wasn't per say any about anyone's take on subjective quality.  

 

MY irritation in the way things have 'always been done' is why we throw away resolution so quickly and decimate, and on a related note apply so much DSP the original recording session and the final distributed file are scantly alike in reality if you could look at all that happened to the digital file along the way. 

 

 

[Andy-HandyHifi]

On 11/18/2020 at 3:23 PM, Miska said:

In addition, I'm pretty happy with my two RME ADI-2 Pro's and their DSD256 ADC output. Although alternatively you can get very powerful 768/32 PCM output as well.

 

You just finalized my Christmas Present to myself.  I regret selling my RME DAC only version.   I am buying a package of the Audio Precision software that allows third party AD/DA support on AISO.  That endorsement will mean it will do quite well until I can afford to upgrade to even more sensitive equipment.  I am tired of the 'SINAD Monopoly' being played on some measurement sites :)  haha i do kid.  I just want to do my own measurements, customized for me and what I want to see.  

 

Thanks!

Andrew

[Andy-HandyHifi]

On 11/19/2019 at 5:03 PM, Miska said:

 

Which way?

 

On my 9038Q2M based DAC (Pro-Ject PreBox S2 Digital), I have confirmed results of improved digital filter performance and some other aspects. You can find the results from the relevant thread here.

 

 

It goes through it's DSP with volume control and remodulation. So it doesn't help working around it's modulator. But it helps working around shortcomings of it's digital filters. Or put other way, DSD goes through much simpler DSP chain in the chip than PCM does.

 

 

Yes, down side of Sabre is that it doesn't have Direct DSD mode. While AKM, Cirrus Logic and Wolfson have. And TI/BB chips don't even have anything else.

 

You can find some DSC1-style DACs on the market that have a separate discrete DSD section. First and most well known is T+A with DAC8 DSD and the newer more expensive models. I guess Denafrips is closest to DSC1. Then there are Holo Audio DACs which I believe are also similar and (AFAIK) also feature AKM SRC chip you mentioned (just like TEAC UD-50x series).

 

Thread here discusses some of these aspects (but also about R2R PCM implementations):

At the end you can find fairly up to date list of DACs.

 

 

you are the master at this, but I have to disagree with you on the BB implementation for how it processes DSD.

 

They are still using the venerable DSD1700 model, at least in the versions of the chip from the pre Texas Instruments days and I don't THINK it has changed much afterwards, but I default to you and others. 

 

While true the DSD and PCM sections are not 100 percent separate systems in say, the DSD1793 used in all of iFi Audio's products, the conversion works the same as most all other 'true' or 'native' DSD DACS.  

 

The BB DAC has 63 equally weighted thermometer code switches for the 6 Most Significant twos complement PCM bits.  of course as you know, but for the edification of others, it is 63 switches in thermometer code because all switches in off position gives you level 0, therefore 64 levels of conversion.  

 

You are correct that DSD does not bypass the segment DAC in the Burr Brown implementations in something like a direct mode.  BUT, it is really semantics because when DSD is present, those same segments are grouped together to form 8 taps out of them.  The tap groupings are one of the major thing that changes when you select a different DSD filter.

 

(Unlike the other DAC's you mention including your own excellent design, the BB design choses to use unequal weighting for cutoff and slope purposes.  It in no way changes the fact that dynamic element matching or scramble code is used... its just a bit more complicated but in the end its all the same.)

 

The 1 bit- DSD bitstream then is treated like the other native DSD DAC's.. there is an 8 bit delay line creating 8 levels of the 1 bit DSD stream offset in time by clock as desired for moving average filter.  

 

So for DSD the Burr Brown converts 1 bit inputs as it would a 9 level Delta Sigma input.  Something akin to 3 bits if not inaccurately 3.5  just to make a point in the difference from the PCM world.  That isn't enough for the ultrasonic noise output so there is specified or in some versions, I think already in place a second gentler filter after the FIR conversion just described. 

 

I am still planning on trying out HQPlayer with the iFi IDSD at DSD512 or DSD1024.  I want to see what your filters I can come up with and how they compare to the filters on the iFi FPGA, ESPECIALLY when going from lower rate DSD to higher rate DSD.   Like DSD64 TO DSD1024.  I have an idea how the iFi filters work and and sound.  Would love to compare with your software.  And considering in the end the DSD1793 works essentially the same as most all other native DSD filters, i am very excited to hear it this way.  Next in line for me to test though is the T+A HA200.  

 

The iFi sounds as good as one would expect at that price; for me, it as a voicing that matches my personal preference almost exactly.  So its even better for the price to me.  But, always have upgrade fever, right??? :/

 

Anyway that is the the best I have got, and my math is likely off somewhere, but I hope the 'jist' of it makes sense.  

[Miska]

11 minutes ago, Andy-HandyHifi said:

You are correct that DSD does not bypass the segment DAC in the Burr Brown implementations in something like a direct mode.  BUT, it is really semantics because when DSD is present, those same segments are grouped together to form 8 taps. 

 

I meant that the TI/BB only has "direct mode", nothing else for DSD. While some others have switchable choice between DSP and non-DSP paths.

 

 

[Andy-HandyHifi]

1 minute ago, Miska said:

 

I meant that the TI/BB only has "direct mode", nothing else for DSD. While some others have switchable choice between DSP and non-DSP paths.

 

 

 

 

hahahaa... sorry then I wasted a lot of time on agreeing with you :)  I feel dumb.  Maybe some one else will find it educational.  

 

I also heard a little bird mention there has been a major behind the scenes shakeup in the audio world.  I am not going into it; it is no one's business until it is.  But it may have way down the road some interesting new developments in the DSD world... we will see!

[Miska]

21 minutes ago, Andy-HandyHifi said:

I want to see what your filters I can come up with and how they compare to the filters on the iFi FPGA, ESPECIALLY when going from lower rate DSD to higher rate DSD.   Like DSD64 TO DSD1024.

 

At least you have multiple options for doing it. At the moment I'm personally sticking to DSD256 using ASDM7EC modulator. Since there are no computers available at the moment that could run this modulator at higher rates... (in real-time, of course with HQPlayer Pro you could run offline conversion to DSD1024 or DSD2048)

 

[asdf1000]

16 minutes ago, Andy-HandyHifi said:

I also heard a little bird mention there has been a major behind the scenes shakeup in the audio world.  I am not going into it; it is no one's business until it is.  But it may have way down the road some interesting new developments in the DSD world... we will see!

 

Hi Andy

 

Without mentioning any company name - can you explain the technicals of what this shake-up involves?

[Andy-HandyHifi]

39 minutes ago, asdf1000 said:

 

Hi Andy

 

Without mentioning any company name - can you explain the technicals of what this shake-up involves?

I really have to choose to let this company and this engineer do that on their own terms.  I am sorry, but I have slipped info in the past before that wasn't mine to slip and lost trust and lots of inside info.    

 

I will say as far as DSD goes, more research and looking at truly viable ways to filter the 1 bit DSD square wave as it is... with no need for delay lines, overlapping signals into complex analog realized complex FIR filtering to do the job of taking out the high frequency.  

 

It has been talked about before on the DIY forums but it has detractors.  Single bit switch from a single USB I2S (or not USB, just I2S) output to a specially designed analog output stage that would handle the noise issue. 

 

At this time, basically, the current most pure DSD conversion is really a digital FIR filter created by analog discrete components.  At the output of the summed taps, you get the actual voltage and get the music like 'voila!' instead of getting another set of binary numbers.  Some think we can do better and push past this model which every native DSD DAC I know of uses in some form. 

 

But alas....

 

Some think that a single bit switch RC filtered DSD solution is a pure pipe dream.  Even if theoretically it is sound and viable, in practice, not so much.  I will let the real experts explain why if they care to.  

 

And WAY out there, in the opposite of DSD land, work on true 20 to 24 bit converters, non delta-sigma but at high sample rates like double and quad DXD.  In both the ADC and DAC worlds.  Continuing and building on the legacy of the no longer made Pacific Microsonics AD/DA's that many still think are the best in the world even as they age another decade.  

 

 

Most I can say or will say.  If anyone can figure anything out, I didn't spill it to anyone :)  At least this time I KNOW I am not the leaker hahahaha.... 

 

 

[Andy-HandyHifi]

9 hours ago, Miska said:

 

At least you have multiple options for doing it. At the moment I'm personally sticking to DSD256 using ASDM7EC modulator. Since there are no computers available at the moment that could run this modulator at higher rates... (in real-time, of course with HQPlayer Pro you could run offline conversion to DSD1024 or DSD2048)

 

 

also, I just learned something really interesting about the iDSD PRO.  The SNR ratio is actually slightly better in DSD512 compared to DSD1024.  Not enough for anything other than expensive test gear to notice, but the reason is the logic and gate speeds cannot quite keep up at 1024 compared to 512.  Of course everyone needs to be reminded that even though the FPGA does the upconversion, the actual DAC is still the DSD1793 made in the late 1990s :D  It is amazing it can keep up, period.  

 

Once again, though we see how a discrete DAC-less converter specifically designed for these speeds would be the best option.  

 

I am increasingly convinced in this new world of computing in which we live, the best 'DAC' will be no traditional DAC at all as we know them.  

 

Discrete logic bitstream converters up to as many as current computing power can handle with the source computer handling all the 'heavy lifting'.  Perhaps one day a dedicated interface other than USB for consumer level products.  

 

 

[asdf1000]

25 minutes ago, Andy-HandyHifi said:

also, I just learned something really interesting about the iDSD PRO.  The SNR ratio is actually slightly better in DSD512 compared to DSD1024. 

 

Can you share your source for this? Or the measurements themselves?

 

[barrows]

7 hours ago, asdf1000 said:

 

Can you share your source for this? Or the measurements themselves?

 

On a related note, the data sheet for the new(ish) AKM 4499 DAC chip shows best performance on distortion and noise (for the direct DSD mode) with DSD 256, and that performance starts to degrade a bit with DSD 512 input.  I have heard Andreas Koch mention that DSD 256 is often a preferred conversion rate, because the switches are not fast enough to take full advantage of higher rates: his explanation suggested that the waveform (ideally a square wave, but in reality the transitions are sloped) spends a higher percentage of time in the transition zone with higher rates, leading to less accuracy.  So higher rates are not always better, depending on hardware (switches) capability.  Of course there probably are faster switches, although higher rates generally will require a higher clock rate, and higher clock rates come with a phase noise penalty as well (although clock doublers are now nearing perfection it his regard, so perhaps clock rate is not so much of a problem)

Given the superb sound quality with HQPlayer's EC modulators, and the limitations on (some) switch speeds, perhaps DSD DAC development should concentrate on optimizing for DSD 256 conversion.  From a commercial standpoint (outside of DIY approaches) this makes a lot of sense as well, as DSD 256 is easily achieved these days by many computer based playback systems, and many different playback softwares.   

[CJH]

Has anyone here tried the STAR Pure DAC http://www.audiodesignguide.com/DSC2/index.html .

CJH

[barrows]

26 minutes ago, CJH said:

Has anyone here tried the STAR Pure DAC http://www.audiodesignguide.com/DSC2/index.html .

CJH

I have the DIY predecessor to the Star Pure, Pavel's DSC-2 design, adapted to an isolated USB input (amanero/Cronus combo, with a single NDK SDA series clock) and Lundhal output transformers.  It is very good, but really needs an active stage after it (a really good preamp) as the output impedance is rather high and the Vout rather low (especially considering the need to attenuate the signal at least 3 dB, sometimes more, in software before oversampling to DSD).  This approach sounds really good, but I do not think this is the best implementation of a DSC style DAC, the measurements are not as good as they could be, with a 7th harmonic above -120 dB.

I suspect that DACs such as the Holo Audio May do quite a bit better, and it has an active output stage of very high quality as well.  I would really like to see Holo Audio do a single box, DSD only DAC, at the same quality level of the May, with perhaps a bit more gain in the output stage to reach 4 V out on XLR at full scale input, that could be a giant killer DAC for use with HQP.

[ericuco]

59 minutes ago, barrows said:

I have the DIY predecessor to the Star Pure, Pavel's DSC-2 design, adapted to an isolated USB input (amanero/Cronus combo, with a single NDK SDA series clock) and Lundhal output transformers.  It is very good, but really needs an active stage after it (a really good preamp) as the output impedance is rather high and the Vout rather low (especially considering the need to attenuate the signal at least 3 dB, sometimes more, in software before oversampling to DSD).  This approach sounds really good, but I do not think this is the best implementation of a DSC style DAC, the measurements are not as good as they could be, with a 7th harmonic above -120 dB.

I suspect that DACs such as the Holo Audio May do quite a bit better, and it has an active output stage of very high quality as well.  I would really like to see Holo Audio do a single box, DSD only DAC, at the same quality level of the May, with perhaps a bit more gain in the output stage to reach 4 V out on XLR at full scale input, that could be a giant killer DAC for use with HQP.

 

I have the Holo Cyan (DSD only version). No idea how it stacks up to the other Holo DACs but I prefer it over my Twisted Pear Buffalo III DAC.

[Andy-HandyHifi]

18 hours ago, asdf1000 said:

 

Can you share your source for this? Or the measurements themselves?

 

 

Can't share.  Comes from impeccable source.  I have not seen the measurements but would not worry about it.  We are talking SNR differences so small why worry at all???  Since especially these SNR differences are way way way above our ears ability to hear... it is only the smallest of technicalities and in no way means people should all of a sudden switch to DSD512.  

 

Not what I was trying to say.  I was really just pointing out how amazing the engineering crew at iFi that chose and developed the iDSD products around the DSD1793 made major accomplishments.  Even though the upsampling is done on FPGA, the actual conversion is still done with the FIR filter on the DSD1793 which is what, a 20 year old DAC now?  It was never designed for things like this; the engineers at Burr Brown Japan would have been blown away if they knew then what is being done with it now.  

[Andy-HandyHifi]

10 hours ago, barrows said:

On a related note, the data sheet for the new(ish) AKM 4499 DAC chip shows best performance on distortion and noise (for the direct DSD mode) with DSD 256, and that performance starts to degrade a bit with DSD 512 input.  I have heard Andreas Koch mention that DSD 256 is often a preferred conversion rate, because the switches are not fast enough to take full advantage of higher rates: his explanation suggested that the waveform (ideally a square wave, but in reality the transitions are sloped) spends a higher percentage of time in the transition zone with higher rates, leading to less accuracy.  So higher rates are not always better, depending on hardware (switches) capability.  Of course there probably are faster switches, although higher rates generally will require a higher clock rate, and higher clock rates come with a phase noise penalty as well (although clock doublers are now nearing perfection it his regard, so perhaps clock rate is not so much of a problem)

Given the superb sound quality with HQPlayer's EC modulators, and the limitations on (some) switch speeds, perhaps DSD DAC development should concentrate on optimizing for DSD 256 conversion.  From a commercial standpoint (outside of DIY approaches) this makes a lot of sense as well, as DSD 256 is easily achieved these days by many computer based playback systems, and many different playback softwares.   

 

Bingo. You have a different impeccable source that is explaining the exact same phenomenon.  My source is closer to the actual design of the iFi, and says that we do have the technology to improve logic speed and switches in carefully specified discrete designs.  (Much of what I just said is my personal understanding and paraphrase of what I was told, so keep that in mind.)  Clock speed likewise not a real issue anymore.  Yes, it is, but again, the clever engineer taking advantage of modern technology can exceed what we have been seeing out of DAC performance.  Discrete is way more expensive though.  Much easier for 99 percent of industry to buy best available DAC's in bulk and run.  

 

The AKM fire has really hurt this, and in more ways that you think.  Said source was having a novel DAC design prototyped at the factory that went up in flames.  So it isn't just a setback in the same ol' chips and shortages.  Real R&D that could push the industry forward, with possibly an actual 'leap' has been lost :(  

[Andy-HandyHifi]

On 11/21/2019 at 9:44 AM, numlog said:

What is the resulting difference (subjective and objective) between having 32 elements and 64?

The reason I ask because the DSC1 boards you can get from china are the updated version with differential output and 32 elements per phase.

It uses XOR gate to get the inverted DSD signals and then a transformer to convert the hot and cold output to single ended...

I am not sure the reasoning behind this but doubling the component count plus adding an XOR gate and especially a transformer to the signal path just to end up with the same single ended output as the original design seems counter intuitive.

So how about combining all 64 elements per channel and leaving out the XOR gate and transformers? this would closer resemble the 9038 SDM conversion

 

Alternatively what about having far less, 4 or 8 elements? I am weary of quality of chinese kit so also considering building a simpler version from scratch... it would cost basically nothing so no reason not to.

 

Thank you for this, but not in the way you may imagine. 

 

I have been banging my head proverbially on my computer desk everytime I read the specs of Chinese DAC's like Denefrips and similar OEM clones for sale on places like eBay say their 32-level FIR filter native DSD converter is a 6 bit DSD converter.  But, 6 bit means 64 level.  I am certain you just answered my question.  The are differential output with 32 level level FIR filter x 2, one for each 'side'.  You might then call it equivalent to 6 bit in 2's complement PCM.  I cannot believe this simple answer has eluded me for so long.  None of their sales guys nor 'tech' guys could get me the answer.  I am like 95 percent positive that this is the case.  

 

They all seem to advertise it as 32 level delay line FIR, but then immediately call it a 6 bit DSD converter.  That in itself would indicate a 5 bit converter, errr.... if we were talking about PCM but we aren't so I am still not sure why some companies insist on using 2's complement PCM numbers where they don't apply.  But now it makes sense.  Differential would be 2 32 level FIR, therefore 64 level and 6 bit PCM equivalent.  

 

Well, even if we forget about the 5 vs 6 bit thing, it is still somewhat inaccurate.  DSD processed via 32 tap FIR filter in unary code is actually 33 levels.  32 1 bit streams yes.  Every one seems to forget that all 32 streams in an 'all off' or, 'all switched to 0 state' is that 33rd level.  Same goes for 64 level, I mean 65 level filters.  

 

Thanks so showing me the light and you didn' even know it!  :) :)