As far as "there are big huge differences" - again, I'd advise caution with these kinds of claims (ESPECIALLY when they come from manufacturers) - you can find numerous reviews and articles discussing just how small the differences are; look at Tyll's Big Sound comparison as a recent example (here's one of the wrap-up articles that came out of it: http://www.innerfidelity.com/content/big-sound-2015-wrap-what-i-learned). I'm not saying "there are absolutely no differences whatsoever" but we're really talking about very small changes, that in some (many?) cases are probably going to be very hard to notice.
To add some more --- selectively hand-picked --- data-points on real-life R2R vs DS performance as perceived by users. From this thread:
http://www.head-fi.org/t/785367/bifrost-mb-technical-measurements#post_12014048
I have spent the last 7 days listening to this DAC through my various amplifiers with great satisfaction. I am a happy owner of all three Schiit multibit DACs, could spend my time with the Ygg exclusively. But I would be missing out on the auditory experiences the other two provide. These DACs really should be heard. All the flowery prose in the world falls short of the reality presented by these Schiit multibit DACs. I would rather have 16 really good bits with excellent channel matching, than the usual mediocre 17 to 18 from the claimed 24 and 32 bit Delta-Sigma DACs.
Wait, so in multilevel DS (e.g. ES9018) the converter is using a 6 bit sample? Or 6 samples of 1 bit across the 6 switches? If it's the latter, then DS is discarding even more data than I had previously thought...
Neither. DS uses a comparator to send the signal through multiple times (specifically the quantization error; this is a feedback loop), at something like 64x-256x fs (think DSD rates), and then puts it together with a digital filter at the other end; It is not a single pass like an R2R, and it should not be represented as such (the signal is being oversampled and decimated to produce a series of single-step pulses representing relative difference between two samples, and that then drives a voltage level (say between -1 and 1) which produces a sinusoid). Noise is filtered out at the final output (the reason for the feedback is to shift the noise into very high (inaudible) frequencies, and make it less random, so it can be filtered, leaving the (relatively) low-frequency content we want). Multi-bit DS allows for more than one quantization step, and operates basically as a "block" of parallel DS outputting unary code into an array of (equivalent) voltage elements, which is then summed (its similar to R2R at this state, but with (theoretically comparably) better integral and differential linearity). This allows for more complete decorrelation of the quanization error from the signal; this means higher dynamic range and a lower distortion ("noise") floor (relative to 1-bit DS).
Side note:
Cherry-picking R2R-positive reviews from a thread titled "Why Delta-Sigma sucks" is not really a great way to make, or support, an argument either...this also isn't a "versus match" or "deathmatch." Nobody is arguing for or against any kind of technology.
Side note:
Cherry-picking R2R-positive reviews from a thread titled "Why Delta-Sigma sucks" is not really a great way to make, or support, an argument either...this also isn't a "versus match" or "deathmatch." Nobody is arguing for or against any kind of technology.
This isn't quite what I did. The purrin thread is a review of various DACs, and the opinion I cherry picked came from a user who self-avowedly often disagrees with OP's pronouncements. This latest cherry-pick came from a very technical thread on Bifrost MB measurements, from someone who is very clearly a scientifically-minded person. Lumping it all into "Why Delta-Sigma sucks" thread isn't quite right either.
As far as "there are big huge differences" - again, I'd advise caution with these kinds of claims (ESPECIALLY when they come from manufacturers) - you can find numerous reviews and articles discussing just how small the differences are; look at Tyll's Big Sound comparison as a recent example (here's one of the wrap-up articles that came out of it: http://www.innerfidelity.com/content/big-sound-2015-wrap-what-i-learned). I'm not saying "there are absolutely no differences whatsoever" but we're really talking about very small changes, that in some (many?) cases are probably going to be very hard to notice.
I checked the Big Sound article, and it was a very interesting read. Thanks. It also links to this post by Mike Moffat:
http://www.head-fi.org/t/701900/schiit-happened-the-story-of-the-worlds-most-improbable-start-up/7725#post_11921090
[...]
Intrigued, I built a similar box with passive relays and a passive attenuator. Damn, if he wasn't right. It is really difficult to tell differences in an instantaneous blind A/B test between tube gear that I built versus some commercial gear that I was not particularly fond of. I used to bet John beers that I could tell the difference. Usually, I won at 7 out of 10 picks or so – the best I ever did was 9 out of ten. But it was really hard.
This whole deal made me wonder if I was crazy hearing differences between amps. If what John said was true, and many others have said in the passing 40 years or so, there is no point for an audio hobby involving anything other than transducers. What?
So I tried something new – I still did the A/B tests, matched levels, but allowed long-term listening to each; at least an hour or two with known recordings. Guess what! Suddenly I knew which was what. I tried it out on John B and Mike and Dave and all my other audio buddies. They called it too – tubes vs a bad solid state preamp. Every friggin' time. My enthusiasm had returned. This taught me that the human ear is an integral, NOT differential device.
So much for the blind A/B instantaneous naysayers. All that matters is frequency response, they say. People can't hear anything much above 20KHz in their prime, less later. The ear has a short memory, it is all bias, blah, blah. They should take up a different hobby, say stamp collecting.
My own intuitions on the subject of blind-testing audio equipment / sources go very much along the same lines: instantaneous A/B blind-testing isn't a very useful measuring tool (e.g. you definitely don't want to do an A/B test on a symphony by using short 10sec samples), and as Tyll and Moffat remark it will often result in subjects not being able to easily discern much if anything. Going from there to concluding that "there are NO huge differences" is a big step. However long-term A/B blind-testing is a whole different kettle of fish, and may allow for reliable and objectively measurable differences to be discerned. And since we tend to use our audio gear for long listening sessions, extending from days to years, well....
I have news for all those trumpeting the superiority of R2R designs, but especially those who proclaim that DS decoding "drops bits" while R2R does not.
The great majority of great-sounding digital recordings are done in DS ADCs. The great PMD ladder ADC has been well surpassed by the best of the new devices including those from Merging and Grimm. The output to PCM formats from those ADCs involves far more computation than the DSD output. So you really do love DS codecs, you just can't say it.
Now, I am agnostic about DAC topologies: the Yggy, for example, has outstanding reviews and measured performance that nearly equals the best DS designs. The Phasure DAC does very well too. I have no clue whether my next DAC will be a DS or an R2R design. But on the recording end there isn't a contest...for the moment.
Delta Sigma happens to have attractive properties for audio vs R-2R DAC
differential linearity is important for audio - look at the GedLee Metric (highly weights "zero crossing" nonlinearity, tests well for correlation in listing tests)
of course the differential linearity is "perfect" in single bit DS - but that has the other processing problems arising from the nonlinear saturation and audible patterns, "birdies" in early lower order implementations
Multibit Delta Sigma does have to use clever tech to hide the linearity errors of the low (5-7?) bit count internal DAC - but it is proven and measured to work very well, gives differential linearity deep into the noise floor of practical electronics
so Delta Sigma excel in low level linearity where we have evidence that our hearing cares most
a ESS whitepaper has -60 dB sine fft plot showing spot noise floor ~100 dB further down from the -60 tone, any sum of distortion bins had to be less than -90 dB below the -60dB sine since there are no visible harmonic peaks in the plot, that's -150 dB THD re full scale!
John Atkinson's Stereophile measurements of 6-7 year old DACs in universal players show -90 dB sine with just noise in several reviews
the trade is the high oversampling, complicated digital processing to achieve the noise spreading and filtering
there can be issues around noise modulation - but at the levels in current "flagship" monolithic audio DAC chips the evidence for hearing these is scant
in another plot in the ESS presentation I believe the "bad competitor" DAC audio band noise floor rose ~ 10 dB from -117 dB to -106-7 dB as signal amplitude rose into the top -10 dB to 0 dB full scale of the converter - that should be Loud!
using estimates of recording mic noise, home listening room noise floor, masking curves I simply don't see where that "bad" DAC's noise floor modulation is going to be audible with music played in the top 10 dB of the DAC, not even with 120 dB SPL peak system capability
can you really hear noise modulation within 10 dB of our hearing threshold in quiet at the same time the music is blasting at >100 dB SPL?
full bit depth R-2R DAC is pushing the tech very hard to keep the major bit carry error below desired audio DAC resolution requirements
the differential nonlinearity of full bit depth DACs around zero crossing almost always are going to give correlated distortion
so do you want correlated distortion that is audible or noise that may modulate at levels way below masking thresholds when the output is near full scale - and at a 10x cost differential too
the yggy's AD5791 actually uses a segmented architecture with the top 6 bits being equal weighted instead of R-2R - its easy to see evidence of cycles of ripple in the INL and glitch vs code plots in the AD5791 datasheet
I would say the full bit depth approach is "the hard way" today for "hi rez" digital audio
a post of mine recycled from the "Thoughts..." thread
In the context of R2R, is there any need for "Hi-Res"?
Reading Monty Python on bit depth, he insists that 16 bit is the absolute maximum that humans will ever need in terms of dynamic range:
http://people.xiph.org/~xiphmont/demo/neil-young.html#toc_tdro1b "16 bit linear PCM has a dynamic range of 96dB according to the most common definition, which calculates dynamic range as (6*bits)dB. Many believe that 16 bit audio cannot represent arbitrary sounds quieter than -96dB. This is incorrect. [...] Our -96dB noise floor figure is effectively wrong; we're using an inappropriate definition of dynamic range. (6*bits)dB gives us the RMS noise of the entire broadband signal, but each hair cell in the ear is sensitive to only a narrow fraction of the total bandwidth. As each hair cell hears only a fraction of the total noise floor energy, the noise floor at that hair cell will be much lower than the broadband figure of -96dB. Thus, 16 bit audio can go considerably deeper than 96dB. With use of shaped dither, which moves quantization noise energy into frequencies where it's harder to hear, the effective dynamic range of 16 bit audio reaches 120dB in practice [13], more than fifteen times deeper than the 96dB claim."
It's clear that 24 bit files are absolutely futile in DS with its native 5-6 bits, and it doesn't seem to be necessary at all in R2R. From what I see, however you look at it 24 bit files are simply a lot of wasted space and bandwidth, especially given that the human dynamic range plateaus at ~130 dB. Looked from this perspective, 16/18/20 bit R2R D/A converters are perfectly sufficient to cover all our dynamic range needs forever and do not require the insane manufacturing precision that 22/24 bits would require...
If we switch to sampling speeds, R2R seems perfectly capable of fully retrieve analog waveforms from slow speed PCM. Monty insists 44.1 kHz is sufficient, with higher sampling speeds like 192 kHz being useful for processing only but not for playback:
http://people.xiph.org/~xiphmont/demo/neil-young.html#toc_o "Sampling rates over 48kHz are irrelevant to high fidelity audio data, but they are internally essential to several modern digital audio techniques. [...] This means we can use low rate 44.1kHz or 48kHz audio with all the fidelity benefits of 192kHz or higher sampling (smooth frequency response, low aliasing) and none of the drawbacks (ultrasonics that cause intermodulation distortion, wasted space)."
Dan Lavry concedes ~60 kHz as an "optimal" sampling rate, which means that 96 kHz fully covers the technological needs of R2R converters:
http://lavryengineering.com/pdfs/lavry-sampling-theory.pdf "While this article offers a general explanation of sampling, the author's motivation is to help
dispel the wide spread misconceptions regarding sampling of audio at a rate of 192KHz. This
misconception, propagated by industry salesmen, is built on false premises, contrary to the
fundamental theories that made digital communication and processing possible.
The notion that more is better may appeal to one's common sense. Presented with analogies
such as more pixels for better video, or faster clock to speed computers, one may be misled to
believe that faster sampling will yield better resolution and detail. The analogies are wrong.
The great value offered by Nyquist's theorem is the realization that we have ALL the
information with 100% of the detail, and no distortions, without the burden of "extra fast"
sampling.
Nyquist pointed out that the sampling rate needs only to exceed twice the signal bandwidth.
What is the audio bandwidth? Research shows that musical instruments may produce energy
above 20 KHz, but there is little sound energy at above 40KHz. Most microphones do not pick
up sound at much over 20KHz. Human hearing rarely exceeds 20KHz, and certainly does not
reach 40KHz. The above suggests that 88.2 or 96KHz would be overkill. In fact all the
objections regarding audio sampling at 44.1KHz, (including the arguments relating to pre
ringing of an FIR filter) are long gone by increasing sampling to about 60KHz. [...] So if going as fast as say 88.2 or 96KHz is already faster than the optimal rate, how can we
explain the need for 192KHz sampling? [...]"
So is there any benefit to playing back files in "higher-res" than 16 bit/96 kHz on R2R D/A converters?
R-2R, or hybrid segmented, "full bit depth" multibit DACs have been developed to a very high standard - so even with my characterization of it being "the hard way" today for Audio there are a few quite good enough for the best we know of home recorded Music Audio playback requirements
better differential nonlinearity is good for lower distortion - Jason correctly points out Schiit's lower cost multibit upgrade DAC has sub lsb DNL even though its only 16 bits
the 20 bit linear DAC in the Yggy should deliver better distortion numbers even when fed with only 16 bits - because each of the 16 bit lsb is correct to ~6% (1/16th)
and Monty of Xiph is just Monty - no snakes appended
1. Listen to a great R-2R DAC, then come back here
2. Buy the one you like the sound of, not the one that reads better in the tech pages or what the manufacturer is telling you is great
3. IMO most R-2R have a different 'take' on the sound, it sounds more organic and real to my ears, maybe not others.
4. R-2R done well is more expensive to do than DS which is literally cheap as chips (pun intended)
5. Many DACs regardless of topology have poor amplifier stages and simple power supplies
6. One big difference between some of the better R-2R is the complete removal of the 'digital filter'
7. The use of I/V transformers in the digital to analogue stage can reap huge rewards
If the R-2R world is bugging you DS guys, get a Yggy on sale/return or better, the TotalDAC. Bet you keep it.....
Hear! hear! I have (had) several dacs of all sorts, and modified quite a few now myself. The one I use in my main system is just a measly TDA1543x8. I have another dac/cd-player with a perfectly tweaked tube-output. I have a very revealing and natural sounding system and R2R is simply closest to natural sound. The sigma-delta is very revealing and lush, sparkly and bubbly party, just trow whatever adjective you like, it's all that. But the R2R always feels like coming home. It is quiter, easier on the ears, more natural, fast, tight and even though the high do not sparkle (except when it's on the recording, never on it's own) yet, there is more details to be heard. They are cleaner and thus more discernible.
I take mine straight (totally filterless), optimally @24/96. Yes I know 24bits has no usefull purpose other than for digital volume. My amp is always on full blast (the whole 8W!) when playing digital.
here's a cut n paste from what I wrote last week. It's an analogy foor temporal accuracy. I really don't get why all of the discussion in DAC's is almost always focussed on frequency response where the temporal axis is just as important (or more IMHO actually). The pre-ringing is caused by the feedbackloop in sigma-delta, as written above ^^
For an R2R- or ladder-DAC there is a lot less nasty HF artifacts. With 44kHz some filtering (1st order) is useful. At 96kHz it is not required... Strictly speaking and according to the original specsheets it is, but doesn't do any good for actual listening to music. It is the cleanest music in the temporal domain you can get. Highs are natural, seemingly soft but very exact. Placement and soundstage are very lifelike, i.e. like live.
Our human brain is very sensitive for correct phase and not just amplitude. Our brain gets very confused by pre-ringing because it's like time in reverse. It reminds me of the film Minority Report. You see the murder scene in the water with ripples moving along. You think you've seen it before so you discard it. But this is the whole plot of the movie. It isn't. It was something that wasn't, so it could happen again without consequence. Imagine throwing a stone in a pond. In real life you expect ripples expanding and dying out from where the stone drops in the water. But what if when you throw the stone the water would slowly start rippling, getting aplified and converging to the point where the stone is going to hit. And then the same thing in reverse, but now like 3 separate stones hit the water. You would not feel very 'Zen' about that would you? Very disturbing because you cannot tell actually when what stone hit the water where, causing all the ripples and interference patterns. This is not relaxing, causing hearing fatigue.
About the snakes,...
I guess Monty without the python is the full monty.
For an R2R- or ladder-DAC there is a lot less nasty HF artifacts. With 44kHz some filtering (1st order) is useful. At 96kHz it is not required... Strictly speaking and according to the original specsheets it is, but doesn't do any good for actual listening to music.
For FL playback at 24/96, what devices could you recommend? I'd be very curious to try it out... (It looks like Schiit Bifrost MB avoids their filter only at 176.4/192 kHz speeds.)
One thing I don't really get about how a delta sigma works and why 24 bit material would be useless, as it only uses the top 6 bits.
It is said there that you would not be able to hear the difference between a 24 bit and a 16 bit file, with a DS DAC, because it only uses the top 6 bits. But then 16 bits would also be unnecessary. Why not truncate your music files to 8 bits for instance? That would save a lot of harddisk space. But I'm pretty convinced I can hear a difference between a 8 bit and 16 bit file even on my laptop speakers. So I don't get why 24 bit would be unnecessary for a DS DAC.
One thing I don't really get about how a delta sigma works and why 24 bit material would be useless, as it only uses the top 6 bits.
It is said there that you would not be able to hear the difference between a 24 bit and a 16 bit file, with a DS DAC, because it only uses the top 6 bits. But then 16 bits would also be unnecessary. Why not truncate your music files to 8 bits for instance? That would save a lot of harddisk space. But I'm pretty convinced I can hear a difference between a 8 bit and 16 bit file even on my laptop speakers. So I don't get why 24 bit would be unnecessary for a DS DAC.
Schiit has this to say on the topic, and they should know as they produce both DS and R2R DACs:
http://schiit.com/products/bifrost "I can’t get over the fact that Bifrost Multibit is only 16 bits!
You didn’t have any problem with delta-sigma being 2 to 5 bits, did you?
But, 16 bits!
Yeah, and most music is still 16 bits—99.9%+, in fact.
But what happens when I use 24 bit music?
We transform it to 16 bit, and it plays just fine. Just like the 2- to 5-bit delta-sigma DACs do. Except with a lot more bits."
This site uses cookies to help personalise content, tailor your experience and to keep you logged in if you register.
By continuing to use this site, you are consenting to our use of cookies.
--- data-points on real-life R2R vs DS performance as perceived by users. From this thread:
![]eep](https://cdn.head-fi.org/avatars/m/139/139747.jpg?1273084439){kind=avatar}
