Thanks Mark, points noted. Need time to think about future plans🙂
Btw, we are talking about DSD only, so the scope accordingly.
Btw, we are talking about DSD only, so the scope accordingly.
My definition of the best sounding sounding Dac is that it should be completely transparent.
When playing LP's with various types of music, preferably recorded analogue, sound should be indistinguishable from when digitized and played over the Dac.
This definition excludes personal preferences by just comparing A to B.
This of course means you will need a very good ADC.
So far, in a first attempt, Marcel's 74AHC02 solid state Dac seems to come closer to that definition than anything auditioned before.
Hans
When playing LP's with various types of music, preferably recorded analogue, sound should be indistinguishable from when digitized and played over the Dac.
This definition excludes personal preferences by just comparing A to B.
This of course means you will need a very good ADC.
So far, in a first attempt, Marcel's 74AHC02 solid state Dac seems to come closer to that definition than anything auditioned before.
Hans
Andrea looked at Marcel's design and maybe learned something from it. At least that's what led to support for RTZ. Other than that, Andrea's dac is designed like Marcel's on steroids. The output stage we are using here in Auburn nobody else has. It alone would have to retail for several $k. Speakers matter too, and I think you know cables can have some effect as well. Does it sound transparent, more so than what you have probably heard so far? Well, you just expressed an opinion about Marcel's dac. What if somebody could do the same only kick it up another notch or two?
Regarding LPs, I think we are doing better here than what you can probably do today. If we were using MC, maybe you would be around right (except for any difference in mastering). If you were using the optical like we are, then LPs would get another step or two up as well.
Overall though, I agree with you about comparing digital to really good vinyl. A good analog source, tape or phono, is necessary as a kind of reference.
Last edited:
@mark, Well, you just expressed an opinion about Marcel's dac. What if somebody could do the same only kick it up another notch or two?
What said, like always, was my very personal opininion, the only opinion that counts for me when listening to music, an opinion that may be interesting to read for other people.
But when someone else comes to a different conclusion, that’s great but it most likely won’t change my opinion.
But kicking the Dac up a notch or two sounds to me as if it were a very mediocre product, which is obviously way beside truth.
Hans
What said, like always, was my very personal opininion, the only opinion that counts for me when listening to music, an opinion that may be interesting to read for other people.
But when someone else comes to a different conclusion, that’s great but it most likely won’t change my opinion.
But kicking the Dac up a notch or two sounds to me as if it were a very mediocre product, which is obviously way beside truth.
Hans
Okay. Maybe a bad choice of words. Do you suppose, say, for example, an $8k class dac design could be improved to move it up into the $12k-$18k class?
I expect you probably know Andrea believes in ultra low phase noise clocking? You may also know that shunt regulator output impedance goes down as frequency goes up, although at some frequency that benefit starts to diminish. You also probably know about ground bounce and other effects when logic chip outputs are switched at the same time in different combinations? What if every single thing you know about that could be improved in Marcel's dac was improved? Would you expect it to become less transparent?
I expect you probably know Andrea believes in ultra low phase noise clocking? You may also know that shunt regulator output impedance goes down as frequency goes up, although at some frequency that benefit starts to diminish. You also probably know about ground bounce and other effects when logic chip outputs are switched at the same time in different combinations? What if every single thing you know about that could be improved in Marcel's dac was improved? Would you expect it to become less transparent?
So true. Which is a good reason why it shouldn't matter whether I reveal testing details or not. You already admit it won't change your opinion.
Last edited:
Those points you mention are all important aspects in a Dac’s design.
I was a bit sceptical about sharing analogue and digital ground, but so far I haven’t been able in this case to prove by measuring that this was making things worse.
The only way to prove that separating both worlds makes a difference is to completely split analogue and digital on two different boards and judge by listening tests whether this makes any difference.
The other thing I did was measuring base noise as a function of signal amplitude. What came out was that Marcel’s Dac has a negative noise modulation, meaning that noise even drops by 1dB at very high signal amplitudes, an excellent achievement IMO.
Then last but not least, when feeding the Dac with various levels of DC in PWM mode, it does not generate spectral content in the audio range, which is what many Dac’s with straightforward SDM’s have as a nasty habbit.
But we all know that going for technical perfection doesn’t always go hand in hand with sonical improvements.
Listening will always be needed as final proof of the pudding.
Hans
Hans,
Good points, thank you.
Regarding technical 'improvements,' I would agree they are only experiments to see what might or might not work. A lot of experimental work has been done with Andrea's dac already. Voluntarily spent many hours of my own time and my own money finding out some things that I didn't think worked for the better. Yet the experiments had to be done. Otherwise there is no way to be sure.
As you say, the proof is in the listening.
Very much looking forward to hearing Marcel's dac. Whatever the results, I will report them truthfully and in as much detail as I can (have already discussed some possible limitations with Marcel, and we have agreed as to how that should be handled).
Mark
Good points, thank you.
Regarding technical 'improvements,' I would agree they are only experiments to see what might or might not work. A lot of experimental work has been done with Andrea's dac already. Voluntarily spent many hours of my own time and my own money finding out some things that I didn't think worked for the better. Yet the experiments had to be done. Otherwise there is no way to be sure.
As you say, the proof is in the listening.
Very much looking forward to hearing Marcel's dac. Whatever the results, I will report them truthfully and in as much detail as I can (have already discussed some possible limitations with Marcel, and we have agreed as to how that should be handled).
Mark
My 74AHC02/74AHC08 DAC and my RTZ shift register DAC are different designs anyway, even though they are both RTZ single-bit FIRDACs. To name a few differences:
0. The shift register DAC uses shift registers as DAC and the 74AHC02/74AHC08 DAC uses 74AHC02/74AHC08 logic gates
1. The 74AHC02/74AHC08 DAC includes a digital part that converts PCM to a single-bit signal, while the RTZ shift register DAC doesn't. The results of the shift register DAC may therefore depend on how the single-bit signal is generated.
2. The 74AHC02/74AHC08 DAC has an LT3042 as voltage reference, the RTZ shift register DAC has a very low noise Hilbiber-style bandgap reference
3. The dimensioning of the reconstruction filters is different.
0. The shift register DAC uses shift registers as DAC and the 74AHC02/74AHC08 DAC uses 74AHC02/74AHC08 logic gates
1. The 74AHC02/74AHC08 DAC includes a digital part that converts PCM to a single-bit signal, while the RTZ shift register DAC doesn't. The results of the shift register DAC may therefore depend on how the single-bit signal is generated.
2. The 74AHC02/74AHC08 DAC has an LT3042 as voltage reference, the RTZ shift register DAC has a very low noise Hilbiber-style bandgap reference
3. The dimensioning of the reconstruction filters is different.
A quick update that Mark and I have exchanged PMs and I now have his address to send him Marcel's RTZ prototype.
I have to say that I admire Mark's commitment to these evaluations but, personally, I can't get too excited as I've learned that life is to short to spend it chasing mirages (especially as I have other interests); there are just too many variables at play, as Mark has outlined above and even my mood can influence my perceptions and I know that getting older has blunted my acuity. My own comparisons of kit are pretty superficial and subjective based on a simple criteria like 'am I enjoying listening to the music' and , unless significant, I tend to think of perceived differences as, well, differences, rather than better/worse. I have had lots of listening to implementations of Marcels's RTZ and Valve DACs as well as ppy's DSC2, they all sound a little different but they're all very listenable - I'm sure the same would apply to Andrea's DAC if I were to build one. That brings me to my other joy, which is simply to build something and get it to actually work!
In broad terms my opinions are, assuming a reasonably competent implementation, that DSD-based playback sounds more convincing that PCM-based and that discrete designs, such as being discussed here, sound better than DSD-capable chip-based DACs but I readily admit that my opinoons are based on a very limited sample.
Anyway, I'll look forward to watching how this progresses.
I have to say that I admire Mark's commitment to these evaluations but, personally, I can't get too excited as I've learned that life is to short to spend it chasing mirages (especially as I have other interests); there are just too many variables at play, as Mark has outlined above and even my mood can influence my perceptions and I know that getting older has blunted my acuity. My own comparisons of kit are pretty superficial and subjective based on a simple criteria like 'am I enjoying listening to the music' and , unless significant, I tend to think of perceived differences as, well, differences, rather than better/worse. I have had lots of listening to implementations of Marcels's RTZ and Valve DACs as well as ppy's DSC2, they all sound a little different but they're all very listenable - I'm sure the same would apply to Andrea's DAC if I were to build one. That brings me to my other joy, which is simply to build something and get it to actually work!
In broad terms my opinions are, assuming a reasonably competent implementation, that DSD-based playback sounds more convincing that PCM-based and that discrete designs, such as being discussed here, sound better than DSD-capable chip-based DACs but I readily admit that my opinoons are based on a very limited sample.
Anyway, I'll look forward to watching how this progresses.
Last edited:
Thank you, Mark
You have a really good setup for testing these type of DACs. For starters, just focus or the DAC ‘engine’. You can look at other improvements later and agree costs go up but that is something to consider later. For now you have a setup with JLSounds usb driven by external clock, going into ESLs and already tested AKM dac(EVM) and Andrea’s DAC. So just use the same setup with Marcel’s and see.
Of course. I will have to adapt the pinout of my external clock, dsp, and reclocking board to Marcel's dac I2S input. Straightforward to do, but takes a little time to set up for Amanero pinout format. Will put it all in a shielded case too, because experience has shown it can make an audible difference.
Thanks again Mark, sounds like a plan🙂
Actually, I am a bit tempted to get Andrea’s production version for these tests also but not sure if Andrea is planning more improvements or what modules are needed? Do you have any further information that would help?
Actually, I am a bit tempted to get Andrea’s production version for these tests also but not sure if Andrea is planning more improvements or what modules are needed? Do you have any further information that would help?
PM sent.
Basic points are that an open GB exists for people who may be interested in present offerings for diy projects. Andrea has also announced plans to come up with with a higher level series of products starting off from what he has learned during this phase of development.
IMHO, what he has now is very good, but its still a set of boards for a diy project. That is to say, its not a commercial dac in a box at six times retail markup. IME with these high end dac projects every little detail matters. It means if the diy'er wants to get the best out of it, they will have some work to do. Or they can make something as good as they know how, and or have the time for, and enjoy what they have accomplished. Most people appear to find their projects with Andrea's products are better than anything else they have ever been able to before so they are happy. I personally don't believe its imaginary either. Most of the people interested in Marcel's dac probably already believe some dacs are audibly superior to others, so I am likely preaching to the choir on that.
EDIT: For anyone interested in a set of boards to make a diy dac, Andrea has offered a couple of discount packages in the present GB. IMHO the prices are very good. OTOH, someone who wants to spend more for very best may want an even lower phase noise clocking option. But trust me, staring out with 5/6MHz clocks and using frequency doublers to bring that up to 22/24Mhz is more costly than just settling for 22/24Mhz SC cut crystals in a pair of Driscoll oscillators. Is there an audible difference? Yes. Sound stage is improved with the lower close-in phase noise option, that's pretty much it. (Which is not to say that's all low close-in phase noise is good for, its just the difference between Andrea's clocking offerings for his DSD dac. IOW both options are very good; its that at that level audible differences are relatively modest.)
Basic points are that an open GB exists for people who may be interested in present offerings for diy projects. Andrea has also announced plans to come up with with a higher level series of products starting off from what he has learned during this phase of development.
IMHO, what he has now is very good, but its still a set of boards for a diy project. That is to say, its not a commercial dac in a box at six times retail markup. IME with these high end dac projects every little detail matters. It means if the diy'er wants to get the best out of it, they will have some work to do. Or they can make something as good as they know how, and or have the time for, and enjoy what they have accomplished. Most people appear to find their projects with Andrea's products are better than anything else they have ever been able to before so they are happy. I personally don't believe its imaginary either. Most of the people interested in Marcel's dac probably already believe some dacs are audibly superior to others, so I am likely preaching to the choir on that.
EDIT: For anyone interested in a set of boards to make a diy dac, Andrea has offered a couple of discount packages in the present GB. IMHO the prices are very good. OTOH, someone who wants to spend more for very best may want an even lower phase noise clocking option. But trust me, staring out with 5/6MHz clocks and using frequency doublers to bring that up to 22/24Mhz is more costly than just settling for 22/24Mhz SC cut crystals in a pair of Driscoll oscillators. Is there an audible difference? Yes. Sound stage is improved with the lower close-in phase noise option, that's pretty much it. (Which is not to say that's all low close-in phase noise is good for, its just the difference between Andrea's clocking offerings for his DSD dac. IOW both options are very good; its that at that level audible differences are relatively modest.)
Last edited:
That reminds me: whatever you do, don't generate the bit clock that drives my DAC with clock doublers, especially not a cascade of more than one clock doubler. They cause subharmonics in the bit clock that mix down out-of-band quantization noise and frequency-modulated idle tones into the audio band.
Without FIRDAC (such as some noDACs) or with a two-tap RTZ FIRDAC (such as my valve DAC), the worst effect would be the mixing down of frequency-modulated idle tones around half the sample rate (that is, half the bit clock frequency). Fortunately, a four-tap RTZ FIRDAC like the one in this thread suppresses that. The suppression is finite, though, maybe 40 dB...60 dB at best.
My DAC has no suppression at all for mixing down quantization noise around odd multiples of one quarter of the bit clock frequency.
There is no problem when you first multiply the clock by n and then divide it by n again. Sounds silly, but that could happen when there is some circuit somewhere that needs a (master) clock at n times the bit clock frequency.
See post 2850 of https://www.diyaudio.com/community/...jitter-crystal-oscillator.261651/post-6373471 for a simulation of the effect of one clock doubler when you have no FIRDAC or a two-tap RTZ FIRDAC. The four-tap structure should improve it substantially, but not entirely eliminate it.
Without FIRDAC (such as some noDACs) or with a two-tap RTZ FIRDAC (such as my valve DAC), the worst effect would be the mixing down of frequency-modulated idle tones around half the sample rate (that is, half the bit clock frequency). Fortunately, a four-tap RTZ FIRDAC like the one in this thread suppresses that. The suppression is finite, though, maybe 40 dB...60 dB at best.
My DAC has no suppression at all for mixing down quantization noise around odd multiples of one quarter of the bit clock frequency.
There is no problem when you first multiply the clock by n and then divide it by n again. Sounds silly, but that could happen when there is some circuit somewhere that needs a (master) clock at n times the bit clock frequency.
See post 2850 of https://www.diyaudio.com/community/...jitter-crystal-oscillator.261651/post-6373471 for a simulation of the effect of one clock doubler when you have no FIRDAC or a two-tap RTZ FIRDAC. The four-tap structure should improve it substantially, but not entirely eliminate it.
Could some kind soul post a link to Andrea's current group buy? I'm failing miserably at searching this evening.
Marcel,
By default my clock board (which would be used with your dac) is fitted with 22/24MHz Crystek clocks. Normally the USB board would divide that down to 11/12 MHz for a DSD256 bitclock. The bitclock would then be reclocked by the master clock before being sent to the dac board.
OTOH, a pic of Andrea's DSD dac can be seen at: https://www.thewellaudio.com/twsdac-dsd/
Bitclock for DSD256 is 11/12MHz, depending on clock family.
For RTZ at DSD256 it could be that the undivided master clock is used for returning to zero.
For Andrea's dac I am currently using 11/12MHz clocks with one doubler. Tried it first with a 22MHz clock for 44.1kHz family sample rates. Didn't notice anything getting worse when switching to the doubler.
Mark
By default my clock board (which would be used with your dac) is fitted with 22/24MHz Crystek clocks. Normally the USB board would divide that down to 11/12 MHz for a DSD256 bitclock. The bitclock would then be reclocked by the master clock before being sent to the dac board.
OTOH, a pic of Andrea's DSD dac can be seen at: https://www.thewellaudio.com/twsdac-dsd/
Bitclock for DSD256 is 11/12MHz, depending on clock family.
For RTZ at DSD256 it could be that the undivided master clock is used for returning to zero.
For Andrea's dac I am currently using 11/12MHz clocks with one doubler. Tried it first with a 22MHz clock for 44.1kHz family sample rates. Didn't notice anything getting worse when switching to the doubler.
Mark
You need a copy of the GB email for links. I will see if I can PM it to you.