No, it's an extraordinary misunderstanding.
-
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!
Delta-sigma vs “Multibit”: what’s the big deal?
- Thread starter THW
- Start date
KSTR
Major Contributor
You didn't care to share a source for that plot -- which doesn't help your credibility, btw.[...]
Multi-stage noise shaping = "all missing codes", non-linear.
SAR ADC and human hearing can detect this errors:
UNDERSTANDING SIGMA–DELTA MODULATION: The Solved and Unsolved Issues
And from the same author: Detection and Removal of Limit Cycles in Sigma Delta Modulators
Limit cycles are real in SD-converters, I've measured their symptom spectra with, for example, the AK4490 DAC chip, but I found that
- their levels are extremely low, way below the rms noise floor, only visible with large FFT size and a heck of averaging,
- they only ever occur when the DAC is fed with a single noise-free zero DC-offset 100% pure sine signal in a specific level range (-40dBFS or so).
IMHO it is highly doubtful that there could be any impact on perceived sound quality from this once the DAC is dealing with real-world music signals.
Esotechnik
Member
R2R don't generate this spikes:
DAC chip AD5791:
- 1-ppm Resolution
- 1-ppm INL
- 7.5nV/√Hz Noise Spectral Density
- 0.19 LSB long term linearity stability
- <0.05 ppm/°C temperature drift
- 1μs Settling Time*
* PCM1704 have 200ns
Deltasigma DACs "for audio" do not look like exact instruments, but are like cheap entertaining toys that guarantee nothing butcolorsine distortion.
Last edited:
solderdude
Grand Contributor
uh-huh
Esotechnik
Member
Only 9dB correlation between input and output music signal for RME babyface DA/AD loop (THD <-100 dB) - unreal?
"RME Babyface (didier.brest)
0.4 dB (L), 0.5 dB (R) Corr Depth: 9,2 dB (L), 10,8 dB (R) Difference*: -29.6 dBFS (L) -30.9 dBFS (R)"
https://www.gearslutz.com/board/showpost.php?p=14288579&postcount=1803
This is "all missing codes" due 20-times noise shaping or 10-times ASRC?
Another example. Green and cyan lines is R-2R.
Any audio-amplifiers has significant IMD on ultrasound frequencies.
RC-filters without shields and feed-through capacitor will help poorly, radio waves propagate through the air.
Last edited:
The reason that it has a low null depth. You can simply cherry picking some measurements to "prove" your claim right?
https://forum.rme-audio.de/viewtopic.php?pid=63112#p63112
Esotechnik
Member
"Dynamic properties present the signal-dependent transition between two states. So, they are affected by not only the static properties but also the successive input codes. In the time domain, DACs’ behavior is measured by the settling time, slewing, glitches, and time skew, etc."
KSTR
Major Contributor
This is also directly visible in the files, time-domain. The phase shift for bass frequencies from the ADC highpass is readily visible. When uncorrected, this spoils correlation and null depth.The reason that it has a low null depth. You can simply cherry picking some measurements to "prove" your claim right?
https://forum.rme-audio.de/viewtopic.php?pid=63112#p63112
---------:---------
IMHO, the test method is inherently flawed. Minor and irrelevant linear response changes must all be fully factored out before subtraction, otherwise they will dominate the residual, masking what we really wanted to see: distortion products, spuriae, correlated noise, processing artifacts of any kind etc.
A way to do this is making a robust(!) full range transfer function measurement of the DA-->AD first, obtaining a precision impulse response. Then convolve the original file with this IR (precision required here, time-domain convolution preferred as the kernel is short).
This is the file we have to compare the loop-back recording with, not the original file.
Additionally, several loop-back recordings of the same setup should be made and compared against each other to establish a baseline measurement, with the null depth only limited by drifts, noise etc.
And of course, the residual must be listenened to and rated on a preceptional basis. For example, an undistorted copy of the original, with little frequency response changes, embedded in uncorrelated white and 1/f noise and maybe some hum or other discrete frequency spuriae is completely benign...
This is a problem with the tool used to do the analysis. As stated before, phase and amplitude need to be corrected and DiffMaker doesn't do that for non-linear differences. DeltaWave does, so here's the result with DeltaWave, showing why it's important to correct for phase (blue line is the phase distortion due to Babyface filter). Difference RMS of -70dBA and correlated null of 68dB. Also a fairly high level of jitter, but that DeltaWave doesn't correct for and is obviously not causing an issue in producing a good null:
Nothing to do with Delta-Sigma or R2R, just a simple phase effect caused by the reconstruction filter.
Last edited:
watchnerd
Grand Contributor
And if any differences are below the audible thresholds, it's debatable if it matters at all.
R2R lost to DS DACs for economic reasons.
The chips are a hell of a lot cheaper to make. They have economies of scale that continue to improve.
They're used in cell phones, TVs, smart speakers, and everything else in the modern digital and wireless world where mass market DACs with low cost and low power consumption matter.
For that reason alone, R2R aren't going to make a come back in the real world, so this is all pretty much just mental masturbation.
R2R lost to DS DACs for economic reasons.
The chips are a hell of a lot cheaper to make. They have economies of scale that continue to improve.
They're used in cell phones, TVs, smart speakers, and everything else in the modern digital and wireless world where mass market DACs with low cost and low power consumption matter.
For that reason alone, R2R aren't going to make a come back in the real world, so this is all pretty much just mental masturbation.
Looking around the AD5791 has ~80 dB SFDR and does not really target audio so not a relevant comparison.
As has been said the test methodology in the OP's reference is severely flawed.
Today's delta-sigma converters typically include high-order and cascaded loops using multibit converters inside the loop as well as noise decorrelation (dither) so tones and limit cycles are essentially impossible to find. I suppose some pathological test condition might find them but in a lot of testing of RF versions I do not recall seeing them.
Seems like we're being trolled by someone with an agenda, or a firm and unwavering belief in the results of a questionable report.
As has been said the test methodology in the OP's reference is severely flawed.
Today's delta-sigma converters typically include high-order and cascaded loops using multibit converters inside the loop as well as noise decorrelation (dither) so tones and limit cycles are essentially impossible to find. I suppose some pathological test condition might find them but in a lot of testing of RF versions I do not recall seeing them.
Seems like we're being trolled by someone with an agenda, or a firm and unwavering belief in the results of a questionable report.
R2R don't generate this spikes:
I just wanted to point out that the paper you got that image from is a paper on the design of the WM8740, a multilevel delta-sigma device. As demonstrated in that paper, modern multilevel DS converters address most of the issues inherent in simple DS devices.
The R2R PCM1704 is a really nice part though, it's a shame they don't make it any more.
For frequencies < fs/2 the NOS filterless DAC is perfectly accurate (Ok, there is some attenuation towards fs/2 caused by the sample and hold).
All inaccuracies are in the folded spectra >fs/2. If the rest of the system does not make any new signals out of these in the <fs/2 range, and the ear perfectly ignores all of these folded spectra, a NOS dac could sound ok
watchnerd
Grand Contributor
That's the ticket right there.
Vintage DAC chips are the new NOS tubes.
Hoard the remaining supply right now so that late 21st century audiophiles can wax prosaically about how good 1st generation digital was and build anachronistic DACs built on old topologies and sell them in expensive audio jewelry cases, much like the way tube amp lovers harken back to the old "Golden Age of HiFi' with SETs and tubes.
Get them now while if you still can.
Maybe NOS DAC promotion is the agenda of our recent poster?
I am holding on to my 20+ year old PCM63 DAC! I'll try to measure it one day, once I dig it out from storage
Esotechnik
Member
No. I want to return the SAR ADCs (which still measure signals in digital oscilloscopes) to the practice of audio recording, then you yourself will understand what is the true quality of the DAC.
You can also record and compare sounds through 2 fundamentally different channels of Audio Precision - LF (deltasigma) and HF.
https://www.ap.com/analyzers-accessories/apx-overview/bandwidth/
It can be improved via digital (table) correction.
Last edited:
watchnerd
Grand Contributor
As someone who does a lot of recording, I can tell you that the ADC/DAC quality is far less important than everything else that happens.
Microphone selection and placement alone dwarf anything you're obsessing over.
Wait, do you mean it's not already better than the majority of the modern Sigma-Delta DACs? I thought R2R was perfect
I have a modern R2R DAC as well. It's pretty good, but not better than a good S-D one (note to myself: need to get rid of some of these DACs, have way too many!)
Mr Rick
Member
Similar threads
