But the baseline Sabre DAC is still not good enoughLOL
With your way if thinking, everything should be good enough. Anything that uses more than an untouched ASIC chip is a scam. Bruno's Mola
Mola DAC must be a real scam as it uses no ASIC at all. Why do you suppose he designed it that way?
Not according to my friends in the hard-core DSP industry. One of them stated "there does seem to be people out there who don't know how to use the technology, but that's another story".
What Charles Hansen is saying about DSD is pretty much the same thing I also said.
"The only difference in the algorithms used to process the raw data"
And those algorithms make a difference. They aren't all equal. The better ones require much more resources to process.
All of the rest of the issues he mentions become null and void once you upsample to 11.289Mhz or beyond. For example our Purestream DAC out of band noise peaks at 180Khz, and is still -70dB below the noise floor @ 180Khz. At 100Khz the noise floor is still -120dB down. And this is regardless of the source audio format, or resolution played back. Scarletbook DSD 64 for SACD only calls for -30db @ 50Khz to deal with the noise. And yes I agree single rate DSD is sub par compared to DSD128 and higher.
The bottom line is with a properly built setup that doesn't rely on off the shelf ASIC's for the SRC/SDM, there's no issues with either format. The DSD/PCM debate is only a battle for owners of DAC's that use ASIC based chips to process their audio. Because there's no ASIC on the planet that can do both equally as well. Even the AKM's with DSD bypass need to have an external filter after the DAC specifically designed for 1 frequency if you want the best of it. And if you do that, it's only the best with 1 format. That's what we do with our Purestream DAC. And it works very well that way. We utilize none of the internal digital processing at all. The discrete analog filter after the chip is tuned for DSD 256 only, with external filter cutoff point set @ 180Khz. With PCM or DSD lower than 256, it sounds lousy. This is why Boulder uses the exact same chip in DSD bypass mode for their $60000 2120 system. They do the same thing. Only they use a low powered ARM based SRC/SDM engine before it, and we use Intel I7 quad core processors.
Last edited:
There's not a whole lot you can do to not use them properly. Besides a bad board layout, poor filter design, bad power supply, and bad clock. You can't screw up a DSP section that has no way to mess with. Just copy the reference board and there you go. You're a superstar DAC manufacturer. The reference board is as good as it gets.
Can I quote you on that?
Well it's what some believe anyways. Just look at the spec sheets of the reference boards. Anything better is beyond the threshold of human hearing. All a waste of money.
Wow $950 for the ESS 9038 evaluation board. You could buy a complete Oppo Sonica DAC cheaper and just rebox it in a thicker aluminum case and just be done with it.
http://www.shawelectronics.com/es9038pro-8-chnl-eval-bd-4868-9038pro-8.html
http://www.shawelectronics.com/es9038pro-8-chnl-eval-bd-4868-9038pro-8.html
Are you sure about this? AFAIK Sabre still runs asynchronously in DSD mode,
if this is the case, there must be some form of re-sampling going on.
Most ppl I have spoken to that know this chip well would say difference
between asynch and synch operation is subtle, ie; with / without SRC going
on but I cant say I've done this comparison myself. Also it's worth
remembering that Sabre generally run off a 100MHz clock and as we both
well know most 100MHz clocks won't get anywhere near the phase noise
performance of these nice 'selected' clocks, especially where it matters.
WRT direct DSD decoding, one observation with the various discrete resistor
based DSD decoder DAC's is they don't seem to measure very well. I don't
know if this is just poor implementation? Did you end up measuring your
own, I'm interested to see the results - as we discussed elsewhere.
If your DAC is truly superior it should measure well or at least have some
component of measurements that tell the story.
cheers, Terry
When you upsample to the highest frequency an oversampling SDM chip is capable of processing, the internal oversampling section no longer upsamples anymore. So you end up effectively bypassing the processing. This has nothing to do with the internal Async. But with the Sabre, there's no way to bypass the modulator no matter what.
Regarding our Purestream DAC, With our favorite 7th order modulator, it has a dynamic range of 121dB up to 100Khz. Noise peaks at 180Khz at -70dB down. Anyways industry standard numbers only tell so much. The Oppo Sonica is enough to impress most numbers mongers. Until you actually compare it to much better DAC's with actual listening tests. I prefer to share specs that actually matter for sound like the actual measured specs of each clock that goes in the units. And since we clock direct off the pin with no prescaling or PLL's, we have very close to the same phase noise on the BCLK, and DAC inputs as well. Example here:
Attachments
Last edited:
Mike doesn't believe in measurements. Words are a more useful measurement according to Mike
The way you talk is that the difference is night and day and justifies the extra dollars is essentially what you keep arguing !
Mike doesn't believe in measurements. Words are a more useful measurement according to Mike
I'm more into real world results. Like when my $2000 DAC was compared to the $90000 MSB Select 2:
https://www.miveraaudio.com/forum/superstack-feedback/superstack-first-impressions
All depends on your hearing ability. Like I said for most people that $9 Apple lightning DAC would be all they need.
So how do they handle the two different clock rates, ie; the OP samples are
re calculated according to a completely different clock rate (usually 100MHz)
Yes that sounds like what you would expect from quad DSD.
Another way to do it is use something like HQplayer and feed DSD256 from
compupu to DAC. If the DAC is implemented well should get a similar result.
Haha - you are funny!
This reminds me of discussions with Joe Rassmussen, (who is a friend). He
was always in sales pitch mode (as you are) pushing his SAW resonator
superiority and I said show me some phase noise / A.V. plots to support
these claims! His reply was always 'it sounds better' ... but I was actually
genuinely interested how these SAW clocks perform WRT LF phase noise.
So this is a similar scenario, I am genuinely interested if you managed to
build a discrete DSD DAC that has very nice linearity and FFT harmonic
spectrum (which IME does affect sonic performance). But it seems this is not
the case, I'm sure you would show us all if it was. That's OK, as you say,
the sonic result is most important but I always try to get both.
It's kind of like zero feedback analog design. Many, including myself, believe
it just sounds better, but it almost always measures much worse to high OLG
closed loop (opamp) design. So I have this little (or you could say big)
challenge to make zero FB circuits (I-V's etc) that also measure as good or
in some cases even better than closed loop designs. So far I've been pretty
successful.
WRT the clocks.... Yes, I am using the same clocks. Very good phase noise
performance and cheap too. He is a good man.
cheers, Terry
So how do they handle the two different clock rates, ie; the OP samples are
re calculated according to a completely different clock rate (usually 100MHz)
Yes that sounds like what you would expect from quad DSD.
Another way to do it is use something like HQplayer and feed DSD256 from
compupu to DAC. If the DAC is implemented well should get a similar result.
Haha - you are funny!
This reminds me of discussions with Joe Rassmussen, (who is a friend). He
was always in sales pitch mode (as you are) pushing his SAW resonator
superiority and I said show me some phase noise / A.V. plots to support
these claims! His reply was always 'it sounds better' ... but I was actually
genuinely interested how these SAW clocks perform WRT LF phase noise.
So this is a similar scenario, I am genuinely interested if you managed to
build a discrete DSD DAC that has very nice linearity and FFT harmonic
spectrum (which IME does affect sonic performance). But it seems this is not
the case, I'm sure you would show us all if it was. That's OK, as you say,
the sonic result is most important but I always try to get both.
It's kind of like zero feedback analog design. Many, including myself, believe
it just sounds better, but it almost always measures much worse to high OLG
closed loop (opamp) design. So I have this little (or you could say big)
challenge to make zero FB circuits (I-V's etc) that also measure as good or
in some cases even better than closed loop designs. So far I've been pretty
successful.
WRT the clocks.... Yes, I am using the same clocks. Very good phase noise
performance and cheap too. He is a good man.
cheers, Terry
The Sabre chip only uses the master clock as a reference clock. It generates PLL's internally for all of the sample rates.
Yes you can use Hqplayer just fine with our DAC. But there's also 3 other options that work well too.
I'm not really looking to impress anyone with the numbers of our prototype Purestream DAC. Mivera Audio is a beta testing company. We weren't looking for approval from numbers mongers about how the product actually sounds without any listening tests. More for real world feedback from the audience that buys DAC's. That being said we are very well aware of how the unit measures, and before we release our OEM system for active speakers we will have a full gamut of measured results to share with interested OEM's if they care to see them.
Regarding the clocks, the NDK 2520SD's are good only in the 22/24Mhz versions. We had to have 600 clocks sorted to get a measly 25 clocks that measured better than -112dbC@ 10Hz phase noise. Just don't get the 45/49's. Because they are only fit for low end applications. Much better off with a Crystek 575 or 957.
But now that we have the SDA's in hand, there's no turning back for us regarding budget clocks. Have you tried these yet?
Attachments
Last edited:
I see they finally have the datasheet publicly available:
http://www.ndk.com/images/products/catalog/c_NZ2520SDA_e.pdf
http://www.ndk.com/images/products/catalog/c_NZ2520SDA_e.pdf
I just realized at 1K the 22.579Mhz NZ2520SDA's even have lower phase noise than the Pulsar!
http://www.pulsarclock.com/ds/Pulsar_Clock.pdf
Well soon I'll be taking actual measurements of both right down to 0.01hz. Then we will see how they compare in the real world.
So far I only swapped 1 untested unit out with a standard 2520SD that I had measured and confirmed to have -113.8dBc 10hz phase noise. Right away I could notice a big improvement to the realism and soundstage presence. And who knows the actual phase noise, as it was randomly selected from the reel.
http://www.pulsarclock.com/ds/Pulsar_Clock.pdf
Well soon I'll be taking actual measurements of both right down to 0.01hz. Then we will see how they compare in the real world.
So far I only swapped 1 untested unit out with a standard 2520SD that I had measured and confirmed to have -113.8dBc 10hz phase noise. Right away I could notice a big improvement to the realism and soundstage presence. And who knows the actual phase noise, as it was randomly selected from the reel.
Then I think you should show your complete product measurement instead of the osc PN. It is a design challenge to utilise the potential of a good clock...
//
