Ok, time-out for you to do more homework.
-
WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!
Best No oversampling dac to buy??
- Thread starter KehaDNb
- Start date
Ok, time-out for you to do more homework.
I removed the run-in to this final sentence because there were so many unfounded a priori conclusions in there that were in the Not Even Wrong category. You did invoke Nyquist however, and he was no dummy—although I own a DAC capable of the highest sampling rates and stream from hi res services, I’m under no delusion that this “hi res” obsession is anything but in large part illusory, because few of us (especially those over age 30) can detect anything close to 20,000hz, and you barely need much over 16 bit to resolve our empirically demonstrated 96dB dynamic range capacity. The only empirically demonstrated benefit of hi res formats is for mixing and mastering engineers and music catalog archivists.
Some of the fundamental purposes of oversampling DACs are to spread out or shape away quanitization noise from the audible range, and to push the noise floor into higher, inaudible frequency ranges above what we can detect—in other words to solve the problems you keep maintaining that oversampling causes.
There are a few NOS DACs, such as the Holo May, that perform comparably to DS DACs on the bench, but the majority we’ve tested here do not. There are some folks that seem to prefer NOS DACs because their flaws are experienced by them as “warmer” or “more natural”—although digital noise is pretty much established as unpleasant, and that warmth can be added at the amp stage through tubes, PEQ or modeling software in a much more satisfying, nuanced way.
The entire purpose of DACs is to recreate the sound of the original master as accurately as possible, not to cause errors and digital noise, or even warmth or colorization for that matter. That means reconstructing what was sampled with the lowest SNR, highest dynamic range and least distortion possible. Amir always gives his subjective opinion on which post-conversion filter sounds “best”, but I personally struggle to hear the differences among any of them and I’m not alone.
We’ve achieved the limits of what DACs can do functionally awhile ago—the SINAD differences we’ve been tossing around toppled the limits of transparency arguably before this site was founded. We’re mostly amused at this point by the pricey gear (including NOS or R2R ladder DACs) that barely manage to clear Red Book CDs in quality. So it’s highly unlikely that NOS DACs are going to make a comeback in the future, especially for masters with high sampling rates. That would be progress going in reverse.
DACs should be the least of your worries and your least costly expense at this state of the art. If you truly want to achieve the qualities you seem to aspire to, I’d suggest you redirect your energies on your headphones/speakers, DSPs, valves if you like, or even vinyl. If your goal is digital fidelity and precision however, just pick the cheapest of DACs on Amir’s top ten and you’re there. Most of what you’ve been parroting is marketing woo, and if there’s any concept that would take you the farthest on your journey from this site, it’s cognitive bias.
Last edited:
Its the missing harmonics of the square wave that show up as 'ringing', its not real ringing, it just looks like that
I currently use the second best measuring NOS DAC known to be confirmed by measurements aside from Holo Audio May which is the Schiit Yggdrasil More Is Better where you can completely bypass digital filtering hence it’s NOS and filterless (only analog filtering after conversion) and I seriously disagree with the NOS being warmer and more natural. That’s pure bullschiit IME. I have extensively sighted testing the Moffat’s custom closed loop filter against NOS mode and I certainly couldn’t tell the difference between the two whatsoever. I left the MIB on the OS mode since I paid for that 4x oversampling Schiit Burrito filter that’s included in the DSP circuit board inside Yggdrasil
Tell me you've never learned about the Nyquist-Shannon sampling theorem without telling me that you've never learned about the Nyquist-Shannon sampling.
OK, but you said "reconstruction filter". Terminology: the reconstruction filter reconstructs the continuous analog signal from the discrete digital signal. It's just a lowpass filter, but it's called this to indicate its job. There must be one, and it must be analog. You can see why people reacted.
And, I don't want to be too picky, but as long as I'm typing, "Filtering in the digital domain is different to time domain filter."—digital audio is in the time domain, and most of the digital filtering done is indeed time domain filtering—frequency domain much more rarely. I'm not sure what you intended to say here, maybe you mean continuous domain, I don't think you mean frequency domain.
I would say that up sampling is also a (digital) reconstruction filter?
Yes it is. It does the same conceptually as what happens in a DAC that performs oversampling. In fact, the filter discussed here is purely in the digital domain. The analog filter after the DAC chip is only a low-order simple filter at high frequency and therefore has no impact on the phase of the audible frequency band.
I don’t agree with it either, it’s just what some folks claim. The Holo May is an extraordinary DAC, and it was not my intention to claim that all NOS or R2R ladder DACs are poorly executed—as I said just about all DACs these days surpass transparency and there are a few non-delta sigma DACs that perform comparable to DS. They do seem to be more costly on average, but I fault no one for buying any gear they enjoy listening to.
For anyone who truly has a curiosity about this dichotomy, the Gustard x18 is an extraordinary DAC (at $749), that is both DS and NOS, selectable via its GUI. I don’t think Amir tested its NOS stage, but at the time of its release in December of 2021 it was neck-and-neck with the d90se (still in the top ten), and built with extraordinary heft and loving detail. It’s an option that is priced with at least some restraint that allow folks to take advantage of both approaches, for those who wish to experiment with either strategy (although I doubt there would be audible differences between the two).
audiosciencereview.com
EDIT: Ignore the above, the NOS stage in this DAC is an artificial simulation attempted by bypassing the first FIR filter. It is NOT NOS. It’s a perfectly terrific DAC, although there are more recent offerings from SMSL and Topping that measure even better on the bench and cost less. Thank you @voodooless and @solderdude!
Gustard X18 Review (Stereo DAC)
This is a review and detailed measurements of the Gustard X18 Balanced USB DAC with Bluetooth. It was sent to me by their seller, Shenzhenaudio and costs US $749. As you see style is very much the same as other Gustard products. I like the high resolution display with larger volume...
audiosciencereview.com
EDIT: Ignore the above, the NOS stage in this DAC is an artificial simulation attempted by bypassing the first FIR filter. It is NOT NOS. It’s a perfectly terrific DAC, although there are more recent offerings from SMSL and Topping that measure even better on the bench and cost less. Thank you @voodooless and @solderdude!
Last edited:
Except that the NOS mode, isn't really NOS:
This is the impulse response. That's certainly not what a NOS impulse should look like!
Well you’re right—my mistake for not revisiting his review. I only recalled that they had an NOS option in their menu. I wonder what the hell it does?Except that the NOS mode, isn't really NOS:
This is the impulse response. That's certainly not what a NOS impulse should look like!
solderdude
Grand Contributor
Yep, it is merely an imperfect emulation of a filterless NOS R2R DAC with some (inescapable) filtering in it.
Ironically it is emulated using an upsampling DAC-chip used in a way it should not be used.
yes... a Dirac pulse looks more like a Dirac pulse but that pulse is never present in any recording (it can't even) but sine-waves (what actual music consists of) is actually reproduced poorly.
But people want it because they think it is a more 'pure' way of making a DAC based on the idea(fear) of the word 'filter' which to some implies something is 'removed' that may be essential or it adds something that should not be added (ringing, which is totally misunderstood).
It brings a smile on my face and whenever I see a DAC with that option I quickly set it to a proper reconstruction filter.
Ironically it is emulated using an upsampling DAC-chip used in a way it should not be used.
yes... a Dirac pulse looks more like a Dirac pulse but that pulse is never present in any recording (it can't even) but sine-waves (what actual music consists of) is actually reproduced poorly.
But people want it because they think it is a more 'pure' way of making a DAC based on the idea(fear) of the word 'filter' which to some implies something is 'removed' that may be essential or it adds something that should not be added (ringing, which is totally misunderstood).
It brings a smile on my face and whenever I see a DAC with that option I quickly set it to a proper reconstruction filter.
Oh it just bypasses the first FIR filter stage. I should have known that it isn’t possible to have NOS in a DS DAC. Misleading!Except that the NOS mode, isn't really NOS:
This is the impulse response. That's certainly not what a NOS impulse should look like!
Learn something new every day. Thanks for keeping me honest!
By the way, how is it that you seem to sniff out those invoking scientific incredibility and drop in exactly when your wisdom is needed? On every thread on this site? Are you carbon-based or AI?
solderdude
Grand Contributor
water based ... mostly ... it seems.
Well, you’re a great conductor.
KSTR
Major Contributor
Do they claim filterless NOS (FNOS)? Most likely they don't.
The IR looks like a true NOS Unit Impulse followed by the typical analog Butterworth low-pass filter.
Very similar to what I see at the output connector of my RME's (with AK4490 or 4493). I also measured directly at the DAC pins (it's voltage output type) and there the edges are sharper and show no ringing (but a lot of D/S shaping noise).
At any rate, I think the shown IR fully qualifies as NOS, just a (slightly) filtered one... for the RME the corner frequency is over 100kHz so that in the end it is perfectly filterless NOS except for bats ;-)
Last edited:
KSTR
Major Contributor
Yes, proper upsampling always involves filtering. For a 4x upsampling you insert 3 zero samples after each sample, and then the original waveform is reconstructed from that "pulse train" by a digital low-pass filter.
First of all, I avoid the use of "reconstruction filter" in my writings (earlevel.com), unless I simply want to relate to the term people have heard before. I usually address it as a lowpass filter necessary to remove the modulated frequency products. So I understand why some here would say the digital filter in SRC is also a reconstruction filter.
I disagree, though. The reconstruction filter is so-called because it reconstructs the analog. All frequencies added by the modulation are removed, and it's analog again.
Why not the digital? It only removes some intermediate frequencies—the digital signal still has modulation products not part of the continuous signal. So the continuous signal is not reconstructed. I don't see any candidates for what might have been reconstructed in the digital domain with SRC, so the term seems to not fit, for me.
Please understand I'm just stating my view plainly. I don't mind if you want to think of this word in a different way. I'm neither enamored with the term "reconstruction filter", nor feel it's important to defend what I think it means. But to me, the only value in calling it a reconstruction filter is that it is the lowpass filter that reconstructs the continuous analog signal that was sampled.
Last edited:
Similar threads
