CHANGSTAR: Audiophile Headphone Reviews and Early 90s Style BBS
Lobby => Head Amps, DACs, Sources, Portable Equipment Discussion => Topic started by: paras1te on September 01, 2015, 02:13:44 AM
-
Overlooked this, but better late than never ...
Dacs with S-D chips have been accepted at large and how good are these?
Some are acceptable, but most are sub-bar performing.
How do you know (or can prove) they are "sub-bar performing"? IMPLEMENTATION is much more important -- even if they are, ALL ELSE HELD EQUAL, "sub-bar performing" [sounding], product developers can (and do) get easily overcome that low hurdle: just improve quality of passives, get better opamps, tighten up PS/regulation.
On top of everything let's say S-D has potential, but look what it takes to squeeze it out. 8+chips in parallel per channel in some good sounding units, manufacturing costs perhaps going over some good r-2r retail price for maybe roughly the same performance.
HUH???
If you're thinkin' 'bout that 8-chips in parallel processor, noted earlier, look closely at that post. Those are Philips TDA1387 DACs, Philips last multi DACs from the mid 1990. They are still avail. in large numbers dirt cheap, mostly from Chinese stashers (check eBay or Alibaba). And they are a jelly-bean DAC; hobbyists and DIYes like 'em 'cause of their relatively large SOIC footprint, only eight pins to solder, and low price. IMO, the latest ESS or AKM will give you better sound.
R2Rs have very real problems, including sensitively to manuf. tolerances, Johnson noise (in the ladder Rs), higher V_dd pwr/current demands, thermal/performance sensitivity. Chip manuf R&Ders -- when (initially) opting for Bitstream; later for D-S -- didn't just pull a fast one on to "keep the change". Manuf. technology has come a long way, thx to semiconductor demands by Apple/Samsung/etc. So cost-effective and sample-sample consistent AND high-performance multi R2Rs can be realized -- or NOT, judging by the $$ of the newer AD57x1's.
Oh ... BTW .... psst (http://"http://www.elserw.com.pl/baza/PDFy/SAA7350.pdf") ... as used in a certain $12k French D/A. Interesting how often Thetas, Wadias, Spectrals, et. al. get mentioned, and better Bitstreams are forgotten.
Speaking of Theta, did I mention my closetted Chroma 396. "The 396 incorporates the hybrid Burr-Brown 18-bit/1-bit PCM67 DAC. The PCM67 converts the "upper" 10 MSB with an R2R ladder (multibit) converter and the "lower" eight LSBs with a 1-bit converter." This is an approach similar to modern D-S, which are also hybrid.
-
Alright. I like to think D-S is not as evil as some folks think. But this deserves it's own little space IMO. Carry on.
-
r2r does not automatically equal better then d/s... At all period. Look at purrins dac chart.
-
Is jitter potentially a bigger problem for DS?
Where does audio rate, as a far as speed is concerned?
Could be mistaken here, but I recall reading somewhere that 1-bit DS D/A chips (a la ESS Sabre, etc.) are more sensitive to clock jitter than multibit (R2R, switched capacitor, etc) D/A chips
-
Could be mistaken here, but I recall reading somewhere that 1-bit DS D/A chips (a la ESS Sabre, etc.) are more sensitive to clock jitter than multibit (R2R, switched capacitor, etc) D/A chips
I'm not sure about clock jitter sensitivity. But 1-bit DS can have more issues than multi-bit DS, which could reflect on jitter. I'm not sure this extends to multi-bit DS vs pure R2R.
BTW. Apologies for the inconvenience due to the recent thread split and move. Just felt we should leave the "vintage multibit/R2R DACs" thread true to it's topic. Post modifications were only to the title of the post.
-
I'm not sure about clock jitter sensitivity. But 1-bit DS can have more issues than multi-bit DS, which could reflect on jitter. I'm not sure this extends to multi-bit DS vs pure R2R.
BTW. Apologies for the inconvenience due to the resent thread split and move. Just felt we should leave the "vintage multibit/R2R DACs" thread true to it's topic.
Well any time you work with higher frequency signals, jitter becomes more pervasive. I would assume that working with MHz inside of the whole DS oversampling architecture is where the problem becomes more prominent.
In terms of sensitivity to that jitter, it's fairly well accepted that, for example, that SCF's (switched capacitor) on the end of a D/A chip will reduce sensitivity to clock jitter. I'm guessing that's because such a design will be able to hold its value more steady or make smoother transitions, while a 1-bit converter might act less predictably when clock jitter is encountered.
Did some brief Google research. This actually quite interesting. I suggest reading on from page 15:
http://www.ece.tamu.edu/~spalermo/docs/clock_jitter_effects_delta_sigma_modulators_saad_2012.pdf
From pg21:
A commonly used solution to alleviate DAC sensitivity to PWJ is the switched-capacito-resistor (SCR) DAC with exponentially-decaying waveform, shown in Figure16. The exponentially-decaying waveform (Figure8) of the SCR DAC makes the amount of charge transferred to the loop per clock-cycle less dependent on the exact timing of the DAC clock-edges [4, 9].
-
I think the paper tackles problems with high-speed Continuous time D-S designs. I think most audio ones are Discrete time D-S designs which may not have that many problems with jitter.
Here is a paper with some discussion about it and other ADC approaches. (http://yadda.icm.edu.pl/yadda/element/bwmeta1.element.baztech-article-BUS8-0020-0049/c/httpaa_czasopisma_pan_plimagesdataaawydaniano4201110samplingjitterinaudioadconverters.pdf)
I also think, though not fully positive, that some good audio D-S designs are return to zero in the feedback path.
*** Corrected link address
-
How do you know (or can prove) they are "sub-bar performing"?
I might have overflown with my statement 'most'. I know what I have heard. And I have probably not heard as many as most US pirates. Sometimes I try to generalise from my exp and knowledge of the internals, but I have been wrong before in this. That is why I am trusting my ears first and perhaps then build any mind construction of the parts playing inside.
For instance the DIY Soekris dac that I built blew my expectations away, I never thought it could rival quite expensive BMC PureDac (a dac with shit loads of tech inside), instead it destroyed the BMC. Other encounters with R-2R dacs have left me with similar impressions to different extent. So this is where I come.
S-D sound in my experience is not playing on the same level.
I'm open to like a S-D dac, I trust most pirates ears to some extent here and some guys at DIYA who have built Buffalos that sound good to them. I trust people's ears more easily than their ability to derive a conclusion for how the said unit sounds from technical viewpoint.
-
Folks ... because of the ever-increasing power of FPGA/DSP, customized and software-based D/A conversion is likely the future...
...check out PS Audio PerfectWave or the top-end Chord DACs (YouTube has some stuff, too).
-
Folks ... because of the ever-increasing power of FPGA/DSP, customized and software-based D/A conversion is likely the future...
...check out PS Audio PerfectWave or the top-end Chord DACs (YouTube has some stuff, too).
Sorry but at the current state, chord or psaudio has not best what MSB analog or Schiit Yggdrasil or even Audio-gd Master 7 can achieve.
-
Folks ... because of the ever-increasing power of FPGA/DSP, customized and software-based D/A conversion is likely the future...
...check out PS Audio PerfectWave or the top-end Chord DACs (YouTube has some stuff, too).
theory ≠ practice
Technically we should all be using Class D amps as well.
And please do not sip the Kool Aid of PS Audio's DAC gnome. He talks a lot, he writes a lot, but many of us remain unconvinced. A lot of the people writing 1000+ word explanation's/artist statements aren't saying much.
-
The latest PS Audio DirectStream DAC (FPGA) is worse sounding than the older PerfectWave mk2 DACs (upgraded from the mk1 models). It's not anymore resolving than the PWD2, sounds very lean in bass, and has an upper-mid emphasis. However, the PSADSD is smooth in the high-end, but with the unfortunate side effect of softened attacks (akin to software upsampling in JRiver to the PWD2 at a higher sampling rate, but much more pronounced). In three words: bright, smooth, forgiving.
The Hugo is much worse than the DirectStream. In three words: Lean, grainy, and boring.
-
My favorite DAC has FPGA, it's used for controlling R2R ladder switches.
-
Folks ... because of the ever-increasing power of FPGA/DSP, customized and software-based D/A conversion is likely the future...
...check out PS Audio PerfectWave or the top-end Chord DACs (YouTube has some stuff, too).
Lol
-
Folks ... because of the ever-increasing power of FPGA/DSP, customized and software-based D/A conversion is likely the future...
...check out PS Audio PerfectWave or the top-end Chord DACs (YouTube has some stuff, too).
The challenge for a DAC, regardless of FPGA or dedicate IC, is still turning bits into accurate voltage levels quickly. FPGA/DSP may indeed have a future, but I don't think it solves these fundamental issues any better.
-
The challenge for a DAC, regardless of FPGA or dedicate IC, is still turning bits into accurate voltage levels quickly. FPGA/DSP may indeed have a future, but I don't think it solves these fundamental issues any better.
Yep.
-
FPGA allows manufacturers to take up their own approaches to delta-sigma instead of relying on someone else's such as AKM, BB, AD, etc.
If you think about, making your own DS D-A converter on a programmable chip is a really neato thing; but so far, I haven't heard any greater "awesomeness" from these FPGA solutions compared to existing off-the-shelf specialized D-A chips, 28 miilion taps or DSDx666 not withstanding.
-
DSD/Sigma Delta/Class D: What does all these have in common? Use of lossy pulse based coding to represent the original data/waveform/power etc.
Sony likes to invent stupid/useless/expensive/redundant formats: Mini-disc, Memory Stick, UMD, Betamax, and you can add SACD/DSD to the list.
-
Mad, it's not a lossy pulse. There is not compression in the same sense as in information theory.
Believe, I've done this shit on commercial IC. This is not the same. Maybe I'll do some explanation write up of what's going on. But the appoximation concept is not it.
Also, Merv is right. You can do some of this with the FPGA and write your own stuff in an FPGA. The timing of the signals needs to be carefully taken care off though. And one still needs a DAC (though with less bits). It's better to have a dedicated IC that is internally optimized (IMO).
-
If you think about, making your own DS D-A converter on a programmable chip is a really neato thing; but so far, I haven't heard any greater "awesomeness" from these FPGA solutions compared to existing off-the-shelf specialized D-A chips, 28 miilion taps or DSDx666 not withstanding.
Also, Merv is right. You can do some of this with the FPGA and write your own stuff in an FPGA. The timing of the signals needs to be carefully taken care off though. And one still needs a DAC (though with less bits). It's better to have a dedicated IC that is internally optimized (IMO).
Sounds great in theory, but OTOH, a great opportunity for less capable designers who might not know the scope of what they're doing and therein screwing things up. We're talking about very complex systems inside these ASIC chips made by very knowledgeable teams of engineers at very large and resourceful mega electronic manufacturers. The sound of the DS-DAC and Chord stuff make sense when you look at it from this lens.
-
Dsd/class d in theory is not lossy, but in real world application it is.
Dsd is explained here:
http://www.thewelltemperedcomputer.com/Lib/SACD.pdf
Class d:
http://www.extron.com/company/article.aspx?id=ts122001
Class D is band/duty cycle limited by the speed of switching power supply. Also The faster the switching power supply is, the more switching noise is generated. It's paradoxical in a sense. You want faster switching, but you also don't want more noise.
There is a reason why class d amps are typically used only in low frequency subwoofers and why class A and A/B is used for higher frequency. Just check out JBL studio monitors for example.
-
Yes. 1-bit DS has issues, that are substantially mitigated in more advanced multi-bit DS designs.
I'm not going to argue about Class D amplifiers because that's not something I have direct experience with. But I believe Class-D amplifiers are not multi-bit Delta Sigmas.
-
...28 miilion taps or DSDx666 not withstanding.
funny.
-
I'm not going to argue about Class D amplifiers because that's not something I have direct experience with. But I believe Class-D amplifiers are not multi-bit Delta Sigmas.
I thought class T is closer to DS.
Class D is just PWM based on the input signal.
Could be wrong though... I never got an interest in class-D or T designs.
-
Seems like so from some wikipedia articles. I doubt they are very sophisticated DS designs though. Likely 1-bit DS, and who knows what random order and topology. I honestly don't know.
Lately, for DACs, it seems that folks at TI are using hybrid (segment) DACs where the upper bits are sort of R-2R like while the lower bits are multi-bit (5-level, 2.32... bit) 3rd order Delta Sigma. Or so it seems with the PCM1794A (datasheet page 24):
http://www.ti.com/lit/ds/symlink/pcm1794a.pdf
Not sure what the DWA stuff is. And dunno if the PCM1794A sounds awesome.
-
Yes, most likely 1 bit 'DS'.
When Sony came out with their first 1 bit DAC's they did not even mention this.
The datasheet of that player simply mentioned ... 18 bits resolution or something like that.
It was in the time that the bitrace began ... 18 bits... 20 bits etc and guess Sony did not know how to 'defend' going to 1 bit.
Technics, at the same time, came out with MASH converters and made a big deal of those.
Lots of 'technical info' on how it worked was available (I got a 2 day course in this technology when I worked for Technics)
These already were 4 bit 'ladder' DAC's that used these 4 bits in 'DS' for the smaller bits.
Talking about >20 years ago already.
-
I like the sound of my Class T amps. They use a control system similar to ds. I think some class d amps might even now be using ds. Regardless I like the sound of my Bel Canto Tripath chip based EVO4. It replaced another tripath based Bel Canto Evo 2. Running the EVO4- 4 channels in bridged mode gives me lots of power and very nice sound indeed.
My second system also uses a tripath chip in a simple yet very good sounding Virtue Audio integrated amp. The sound of this amp is quite reminiscent of some tube monoblocks I had from Rogue. Even when I had a simple 10 watt per channel Trends Tripath chip based amp it handily beat a 100 watt Emotive power amp. A number of friends in the Denver area with speaker based systems have chucked their class a or ab amps for class d, especially some of the modder types who have taken some Pascal amps and modded them with very good results.