Page 36 of 61 FirstFirst ... 2632333435363738394046 ... LastLast
Results 876 to 900 of 1506
  1. #876
    Senior Member
    Join Date
    Apr 2010
    Location
    Finland
    Posts
    4,408
    Blog Entries
    7
    Quote Originally Posted by Superdad View Post
    I have looked at your graphs for the Teac, and previously for many other DSD and PMC DACs illustrating comparisons of noise and aliasing. I guess what I am most misunderstanding in the discussion/debate regarding the two formats is why all the concern over noise and aliases that are so far above not just out hearing, but well beyond the bandwidth of all speakers and surely most amplifiers. My preamp/amp/speakers, while very good, are not passing DC-to-light. Graphs showing noise 85dB down at 500kHz would not really seem to be of much concern.

    I'm rather agnostic in the debate between DSD and PCM, but it sure seems to me that what happens from 100kHz on down is more important that ultrasonics that won't be passed the rest of the system.
    One reason is because people tend to make such a fuss about it, like how much DSD puts out noise vs PCM. Another reason is that since I work so much on digital filters and modulators, seeing also the "big picture" is important. And it gives some perspective, the kind of measurements rarely published.

    You can measure quite a bunch of DACs using 100 kHz bandwidth and they look pretty much the same. If you look at 20 kHz band, even more so. But when you look at wider bandwidth they don't. I want to avoid tunnel vision. I can already recognize DAC chip manufacturers based on how the wideband spectrum looks like.

    There are few reasons why properties of the noise matters. One is that intermodulation products with non-correlated noise are noise. While intermodulation products with correlated signal are a distinct signal. Wideband look also immediately reveals quite a lot about overall signal reconstruction performance.
    Signalyst - http://www.signalyst.com
    Developer of HQPlayer

  2. #877
    Senior Member
    Join Date
    Apr 2010
    Location
    Finland
    Posts
    4,408
    Blog Entries
    7
    Since some people had worries about DSD performance for multitone high/low level signal combinations or with transients, I created couple of more plots with one of my test signals.

    Test signal consists of two constantly changing signals:
    1) 1 kHz sine at -6 dBFS frequency modulated +- 500 Hz at rate of 1 Hz.
    2) 7.5 kHz sine at -60 dBFS amplitude modulated +-10 dB at rate of 3.7 Hz

    In addition every second there's -10 dBFS dirac pulse.

    This is generated into 44.1/32 PCM file.

    Here is output after upsampling to 88.2/24 PCM:
    cplx-pcm.png

    And here is output after upsampling to DSD256, then converting that to 705.6/32 PCM and that again to 88.2/32 PCM (to make comparison fair - same analysis bandwidth and FFT for both cases):
    cplx-sdm.png

    No problems visible in either one. So this one at least has constantly changing signal. Sometimes I also add third signal at 15 kHz that is both frequency and amplitude modulated at the same time.
    Signalyst - http://www.signalyst.com
    Developer of HQPlayer

  3. #878
    Senior Member Superdad's Avatar
    Join Date
    Apr 2008
    Location
    Mariposa, California
    Posts
    1,925
    Thank you for the clear answer Miska! (BTW, do you stay up all night long on the internet? It is almost 3:00 a.m. there! Get some sleep.)

    I guess the area I am more interested in learning about--and which I know you are an expert in--are the various D-S modulation schemes used in all the sigma-delta based PCM/DSD DACs. It would seem that some of the modulators (and the stages that come right after) would be better designed than others. Do you think measurements will show this or is there a characteristic sound to the different topologies?

    Secondly, from what I have been told about how today's popular DSD-capable DAC chips output DSD, it sure seems that the method of creating the filters they use (semi-analog FIRs with chip-internal caps and resistors?) might be a big sonic compromise. I'd love to here your (and any else's) thoughts on this matter as well.

    Thanks and goodnight,
    ALEX

  4. #879
    Freshman Member
    Join Date
    Jan 2011
    Location
    Madrid, Spain
    Posts
    35
    Quote Originally Posted by jzahr View Post
    ... The range from 20 khz to 48 khz is shown, is this range considered out of band? ...
    Quote Originally Posted by Miska View Post
    Problem in the DSD vs PCM arguments is typically that for PCM in-band is clearly everything within Nyquist band (fs/2). While SDM (DSD) is non-Nyquist sampling system and thus doesn't have clear distinction between in-band and out-of-band.

    Theoretical out-of-band noise for non band limited PCM is infinitely repeating images and mirror images of the signal spectrum. So for example what ever is present at 0 - 20 kHz sampled at 40 kHz will repeat in inverse spectrum from 20 - 40 kHz and again in exactly same spectrum from 40 - 60 kHz and so on. This is presented in Fig 4 and Fig 5 here:
    Nyquist

    As it refers to brickwall filter - in digital domain perfect brickwall filter would be infinitely long and would thus ring infinitely and require infinite number of samples before it would get even one sample out. Perfect analog brickwall filter would have infinite order. Of course anything saying "infinite" is not technically possible in practice...
    Thanks Miska for your clear explanation. In my ignorance, I was under the idea that "out-of-band" is a synonymous of "out-of-audio-band", understanding "audio-band" as the 20 hz to 20 khz frequency range, which is usually cited as human hearing range.

    Out of interest, I have another question for you. Regarding this comment (bold added by me for context):

    Quote Originally Posted by Miska View Post
    ...
    Now after feeding the same sweep (96 kHz logarithmic sweep) through one of my 1-bit modulators to produce DSD256 at 12.288 MHz, then convert it back to PCM at 768/32 and then again convert that one back to 96/32 we get this:
    ...
    As you can see, that even after multiple conversion cycles the performance still significantly exceeds capabilities of 24-bit PCM.
    It seems that here, when going from DSD256 to 24/96 PCM, you go first to 32/768 PCM as an intermediate stage. Is this multi-stage necessary? Does it result in a better conversion that going straight from DSD256 to 24/96 PCM? Could be useful to use more tha one intermediate stage?

    Similar questions can be formulated for conversions in the opposite direction: if a 16/44.1 PCM signal must be sent to the Megaherz range for conversion in a delta-sigma DAC, are there also here one or more intermediate PCM stages, for example 44.1 PCM to 176.4 PCM to megahertz delta-sigma? If so, are these intermediate stages necessary? Do they improve in some way the performance of the global target of converting 44.1 PCM to megahertz?

    Thanks,

    Jorge

  5. #880
    Quote Originally Posted by Superdad View Post
    Secondly, from what I have been told about how today's popular DSD-capable DAC chips output DSD, it sure seems that the method of creating the filters they use (semi-analog FIRs with chip-internal caps and resistors?) might be a big sonic compromise. I'd love to here your (and any else's) thoughts on this matter as well.
    As far as I know, there must be three or more types of architectures in modern DSD-capable DAC chips.
    1. Applying analog FIR filter for 1 bit delta-sigma modulated input data
    Texas Instruments/Burr Brown DSD/PCM-179x (Current output)
    Niigata Seimitsu FN1242A (Voltage output)
    This type can accept 1 bit delta-sigma modulated data of sampling rate up to 24.576MHz.
    No digital volume control is available.
    2. Converting 1 bit delta-sigma modulated input data to a multilevel delta-sigma modulated internal representation
    ESS Technologies ES9018S ( 64 MOS switches in parallel at output )
    A digital volume control is available.
    No switched capacitors at a DA output stage.
    3. ??
    Asahi Kasei Electronics AK4399, Wolfson Microelectronics WM8741 (Voltage output)
    Switched capacitors at DA output stage?

  6. #881
    Senior Member
    Join Date
    Apr 2010
    Location
    Finland
    Posts
    4,408
    Blog Entries
    7
    Quote Originally Posted by jzahr View Post
    It seems that here, when going from DSD256 to 24/96 PCM, you go first to 32/768 PCM as an intermediate stage. Is this multi-stage necessary? Does it result in a better conversion that going straight from DSD256 to 24/96 PCM? Could be useful to use more tha one intermediate stage?
    It is because my SDM-to-PCM conversion class converts down to 1/16th of the SDM rate and 64-bit floating point. Since the command line utility I have currently outputs 32-bit integer PCM I used it here. I have another command line utility to resample PCM to another rate and used that for the second stage. In normal cases the two stages, when necessary, would have 64-bit floating point presentation between, but with these small utilities now used intermediate WAV file. I didn't bother to start changing those utilities to use 64-bit float WAV instead (would be a trivial change, but I didn't do it) although it would have improved quality of the results. I think the results were good enough for this particular purpose anyway.

    Similar questions can be formulated for conversions in the opposite direction: if a 16/44.1 PCM signal must be sent to the Megaherz range for conversion in a delta-sigma DAC, are there also here one or more intermediate PCM stages, for example 44.1 PCM to 176.4 PCM to megahertz delta-sigma? If so, are these intermediate stages necessary? Do they improve in some way the performance of the global target of converting 44.1 PCM to megahertz?
    Now you are getting to an important point.

    Yes, in most cases DAC chips do the conversion in multiple stages due to resource constraints. The most typical case is that proper digital filters are used to perform conversion to 352.8/384k rate and then subsequent interpolation it just sample-and-hold (SAH or S/H) or comb (CIC) filter due to lack of computing resources. The first "8x" interpolation is traditionally performed with three cascaded 2x stages where each stage uses half shorter filter than previous. When for example double rate (88.2/96) input is used, first 2x stage is dropped out and input goes straight to second stage. Quad rate again drops away another stage. Since the stages are shorter, higher rates will also start rolling off earlier. The following interpolation stage is one of the biggest problems in many of the current implementations. SAH will produce strong image frequencies around multiples of the 352.8/384k internal rate, visible in wideband output of many DACs. CIC being a poor filter produces less articulated images, but still clearly present.

    So the most typical pipeline looks like:
    [2x] -> [2x] -> [2x] -> [16x S/H] -> [modulator]


    On the other hand, my PCM to SDM conversion algorithms use either single stage oversampling, or alternatively for less CPU load two cascade stages (the new *-2s variants) where first stage is always at least 8x regardless of input rate and second stage is also a proper filter. These also allow conversion for example from 44.1 to 6.144 (48x128) rate.

    So either:
    [polyphase] -> [modulator]

    Or:
    [polyphase] -> [FIR] -> [modulator]
    Signalyst - http://www.signalyst.com
    Developer of HQPlayer

  7. #882
    Senior Member Superdad's Avatar
    Join Date
    Apr 2008
    Location
    Mariposa, California
    Posts
    1,925
    Good day Miska:

    Your latest explanation for jzahr is very clear. Of course the information about typical SD DAC chips I already knew. Avoiding the resource constrained oversampling stages of them is easy--and very good sounding with careful SRC filter adjustment--with software that uses the excellent iZotope engine (Audirvana Plus in my Mac case), or your and other Windows s/w. [In my present case--NOS PCM1704--I have to upsample not to eliminate DAC oversampling stages, but to avoid aliasing artifacts if I don't.]

    What I would really like to understand more about (and which you may have tried to tell me in the past--sorry) is what DAC chips both readily accept your (HQPlayer's) 6.144 or higher stream, and your thoughts on which ones do the best sounding job with it on the output side.
    As Bunpei so kindly responded just above your post, the major DAC chip makers fall into several camps with regards to what they do with high rate SD-modulated signals. (Although finding candid and specific information--from the manufacturer--on that last stage of a DAC chip is not easy.)

    And into which of these chips does your s/w SDM most completely bypass the SDM (or is it DSM, they seem interchangeable) of the DAC?

    And back to the output side of things, it seems to me that since you are generating in s/w such high rates, one can question what the DAC ship is then even used for other than as a shitty switched capacitor or resistor network low pass filter. My friend John Swenson could design something much better for you to feed your lovely DSD512 (24.576MHz) into. Getting it from the computer to the his stage is where more work would be required!

    I support your cause, I just wonder how it can be best fulfilled.

    Best,
    ALEX

  8. #883
    Senior Member
    Join Date
    Apr 2010
    Location
    Finland
    Posts
    4,408
    Blog Entries
    7
    Quote Originally Posted by Superdad View Post
    Avoiding the resource constrained oversampling stages of them is easy--and very good sounding with careful SRC filter adjustment--with software that uses the excellent iZotope engine (Audirvana Plus in my Mac case), or your and other Windows s/w.
    But only up to bypassing the 2/2/2 "8x" interpolator that take it to 352.8/384, but not the more constrained SAH/CIC stages...

    [In my present case--NOS PCM1704--I have to upsample not to eliminate DAC oversampling stages, but to avoid aliasing artifacts if I don't.]
    This is a special case, since it's not a delta-sigma DAC, so if you can run it at 768/24 you get most out of it.

    What I would really like to understand more about (and which you may have tried to tell me in the past--sorry) is what DAC chips both readily accept your (HQPlayer's) 6.144 or higher stream, and your thoughts on which ones do the best sounding job with it on the output side.
    TEAC UD-501 can accept up to 5.6 MHz and runs only analog FIR D/A conversion stage with it, one could actually say that it has four different D/A stage configurations available. I'm quite happy what I get with this reasonably priced DAC. And of course it also allows bypassing the first interpolation stage and feeding 384/32 too as alternative, but I don't use it that way myself, other than testing and comparison purposes.

    Then I have my own DAC with CS4398 in DirectDSD mode and it can also go up to 5.6/6.1 MHz. In this mode DSD goes straight to the SCF D/A conversion stage. Plus whenever I have time to play with hardware, I prototype different kinds of discrete converter architectures (many many different possibilities).

    Mytek DAC can take up to 6.1 MHz for it's Sabre chip. And some other Sabre based DACs like exaSound and others can take DSD256 or higher (I'm not sure if these support 48k-base clocks). But AFAIK, Sabre runs DSD through it's DSP on the way.

    For AKM chips I still have not been able to verify what it exactly does with DSD...

    Wolfson supports Direct DSD, but only up to DSD64.

    Getting it from the computer to the his stage is where more work would be required!
    Just get bunch of those Amanero-boards, it is quite flexible...
    Signalyst - http://www.signalyst.com
    Developer of HQPlayer

  9. #884
    Senior Member
    Join Date
    Apr 2010
    Location
    Finland
    Posts
    4,408
    Blog Entries
    7
    Quote Originally Posted by Bunpei View Post
    Asahi Kasei Electronics AK4399, Wolfson Microelectronics WM8741 (Voltage output)
    Switched capacitors at DA output stage?
    AFAIK:

    - AKM has similar SCF conversion stage as Cirrus Logic.
    - ESS has equally weighted elements, like dCS (64 in ESS vs 24 in dCS)
    - Wolfson has otherwise similar conversion stage as ESS, but they have less switches (14) with weighted elements. Equivalent of 78 equally weighted elements.
    Signalyst - http://www.signalyst.com
    Developer of HQPlayer

  10. #885
    Freshman Member
    Join Date
    Jan 2011
    Location
    Madrid, Spain
    Posts
    35
    Thanks Miska for the very informative response.

    Regards,
    Jorge

  11. #886
    Senior Member Superdad's Avatar
    Join Date
    Apr 2008
    Location
    Mariposa, California
    Posts
    1,925
    Quote Originally Posted by Miska View Post
    But only up to bypassing the 2/2/2 "8x" interpolator that take it to 352.8/384, but not the more constrained SAH/CIC stages...
    Yes, I knew that. That is why I am not a big fan of SD DAC chips. And on the output side (SCF as you put it? What does that stand for?: Serial Conversion Filter?) they seem to compromise DSD as well. DSD was just shoehorned into those chips because it was already similar in some respects to what they do in SD for PCM.

    Quote Originally Posted by Miska View Post
    This is a special case [the NOS PCM1704], since it's not a delta-sigma DAC, so if you can run it at 768/24 you get most out of it.
    Of course. I'll be at 352.8/384 next month when my new version of WaveIO USB>I2S board arrives for my DAC. We (Swenson and I) are also developing an advanced product with XMOS (due to very compatible pre-written drivers and lots of code--and also because of ethernet capabilities), so I'm not sure about being able to push out as far as 768KHz with it over USB. Would need to do that with an FPGA or a different processor.


    Quote Originally Posted by Miska View Post
    TEAC UD-501 can accept up to 5.6 MHz and runs only analog FIR D/A conversion stage with it, one could actually say that it has four different D/A stage configurations available. I'm quite happy what I get with this reasonably priced DAC. And of course it also allows bypassing the first interpolation stage and feeding 384/32 too as alternative, but I don't use it that way myself, other than testing and comparison purposes.
    Yes, the Teac is certainly very nice for the price. The PCM1795 is okay and I have never heard the op amp they are proud about using. But I guess that 5.6MHz (DSD256) is only available under Windows with their ASIO driver; Macs should natively support it with DoP up to DSD128. Is that correct or am I getting my math wrong again?

    Quote Originally Posted by Miska View Post
    Plus whenever I have time to play with hardware, I prototype different kinds of discrete converter architectures (many many different possibilities).
    I bet that is where the magic of DSD happens! Will you release a thermometer DAC soon? ;-0

    Thanks for all the interesting information. Now I just dream for HQPlayer to run on OS X without the NAA requirement…

    Goodnight.
    AJC

  12. #887
    Senior Member
    Join Date
    Apr 2010
    Location
    Finland
    Posts
    4,408
    Blog Entries
    7
    Quote Originally Posted by Superdad View Post
    Yes, I knew that. That is why I am not a big fan of SD DAC chips. And on the output side (SCF as you put it? What does that stand for?: Serial Conversion Filter?) they seem to compromise DSD as well. DSD was just shoehorned into those chips because it was already similar in some respects to what they do in SD for PCM.
    SCF is Switched Capacitor Filter. For example CS4398 has built-in digital DSD processor for doing digital filtering and volume control for DSD without PCM down conversion (similar to ESS), but they also provide a mux switch to directly feed the DSD data straight to the actual D/A conversion stage (DirectDSD mode). Chips providing voltage output tend to use SCF output stages, while the current output ones feed switch array outputs directly to the output pins.

    With chips supporting "DirectDSD" mode, it is again possible to use those new chips in the Good Old Days way, where DAC chip only converts digital to analog and nothing else (like PCM1704 does for PCM).

    I wouldn't say DSD was shoehorned to any of the chips, they all seem to have taken some effort doing it, the approaches how they did it varies a lot, just like their actual D/A conversion stages too. Unfortunately the PCM digital sections seem to follow same pattern in most cases.

    But I guess that 5.6MHz (DSD256) is only available under Windows with their ASIO driver; Macs should natively support it with DoP up to DSD128. Is that correct or am I getting my math wrong again?
    5.6 MHz is DSD128 (44.1x128), so far I've been using it only on Windows with the ASIO driver. I think I received one success report also with Linux. I have a BeagleBone Black based NAA 30 cm away from it, so I could try that out on weekend.

    I bet that is where the magic of DSD happens! Will you release a thermometer DAC soon? ;-0
    That's something I don't know. Commercial hardware not. DIY schematic & gerber under CC A-NC, maybe.

    Now I just dream for HQPlayer to run on OS X without the NAA requirement…
    Since Mavericks is supposed to support integer mode, I'll see if there's something sensible possible there.
    Signalyst - http://www.signalyst.com
    Developer of HQPlayer

  13. #888
    ElectrArt in Japan started a new order entry for his UDA2 board, a USB Audio Interface capable of playing 1 bit delta-sigma modulated signal of 24.576 MHz sampling rate and recording that of 12.288 MHz with his proprietary ASIO driver.
    uda2r.jpg

    He also prepared an ARDA Technologies AT1201 ADC chip based ADC board that is compatible with the UDA2 board.
    at1201c.jpg
    at1201s.jpg

    Prices are 28,500 JPY for UDA2 board and 59,800 JPY for AT1201 ADC board, respectively.
    To our regret, the first batch was sold out within a hour!

  14. #889
    Senior Member Superdad's Avatar
    Join Date
    Apr 2008
    Location
    Mariposa, California
    Posts
    1,925
    Quote Originally Posted by Miska View Post
    SCF is Switched Capacitor Filter. For example CS4398 has built-in digital DSD processor for doing digital filtering and volume control for DSD without PCM down conversion (similar to ESS), but they also provide a mux switch to directly feed the DSD data straight to the actual D/A conversion stage (DirectDSD mode). Chips providing voltage output tend to use SCF output stages, while the current output ones feed switch array outputs directly to the output pins.

    With chips supporting "DirectDSD" mode, it is again possible to use those new chips in the Good Old Days way, where DAC chip only converts digital to analog and nothing else (like PCM1704 does for PCM).

    I wouldn't say DSD was shoehorned to any of the chips, they all seem to have taken some effort doing it, the approaches how they did it varies a lot, just like their actual D/A conversion stages too. Unfortunately the PCM digital sections seem to follow same pattern in most cases.

    5.6 MHz is DSD128 (44.1x128), so far I've been using it only on Windows with the ASIO driver. I think I received one success report also with Linux. I have a BeagleBone Black based NAA 30 cm away from it, so I could try that out on weekend.


    Since Mavericks is supposed to support integer mode, I'll see if there's something sensible possible there [for HQPlayer].
    Thank you again Miska! I learn a great deal from your posts. DAC chips' handling of DSD often seems like a black box since the data sheets and block diagrams don't usually spell out what they are doing it.

  15. #890
    Senior Member Superdad's Avatar
    Join Date
    Apr 2008
    Location
    Mariposa, California
    Posts
    1,925
    Quote Originally Posted by Bunpei View Post
    ElectrArt in Japan started a new order entry for his UDA2 board, a USB Audio Interface capable of playing 1 bit delta-sigma modulated signal of 24.576 MHz sampling rate and recording that of 12.288 MHz with his proprietary ASIO driver.
    Click image for larger version. 

Name:	UDA2R.jpg 
Views:	613 
Size:	101.6 KB 
ID:	8902

    He also prepared an ARDA Technologies AT1201 ADC chip based ADC board that is compatible with the UDA2 board.
    Click image for larger version. 

Name:	AT1201C.jpg 
Views:	567 
Size:	79.5 KB 
ID:	8903
    Click image for larger version. 

Name:	AT1201S.jpg 
Views:	256 
Size:	64.6 KB 
ID:	8904

    Prices are 28,500 JPY for UDA2 board and 59,800 JPY for AT1201 ADC board, respectively.
    To our regret, the first batch was sold out within a hour!
    Bunpei: Thank you for bringing the newest ElectrArt boards to our attention. I have been curious about his boards for about a year. I do wish he had an English site with full details.

    Can you explain the I/O on the UDA2 board? Obviously USB input, but what are the outputs? I assume I2S, and it also looks like an S/PDIF output. What are the power supply voltage requirements?

    Of even greater interest to me is the A/D board. I/O for that one? Looks like balanced XLR only for analog input. What are the outputs, and what data formats does it output? Can it work without the UDA2?

    I guess what I would really want for analog recording would be a USB output.

    I can't afford a Grimm, Horus, or even an Ayre QA-9. So the ElectrArt is of interest.

    Thanks.

  16. #891
    Sophomore Member
    Join Date
    May 2012
    Location
    USA Virginia
    Posts
    305
    Quote Originally Posted by Miska View Post
    AFAIK:
    - AKM has similar SCF conversion stage as Cirrus Logic.
    - ESS has equally weighted elements, like dCS (64 in ESS vs 24 in dCS)
    - Wolfson has otherwise similar conversion stage as ESS, but they have less switches (14) with weighted elements. Equivalent of 78 equally weighted elements.
    TI has 66 levels S-D too in their top-of-the-line PCM179x... 5 levels D-S (lower 18 bit) and 62 levels (upper six bits) are in PCM format.
    I keep saying that there is a reason why the manufacturers departed from pure 1 bit, but I guess some people think it's all a conspiration.

  17. #892
    Senior Member Superdad's Avatar
    Join Date
    Apr 2008
    Location
    Mariposa, California
    Posts
    1,925
    Quote Originally Posted by SoNic67 View Post
    TI has 66 levels S-D too in their top-of-the-line PCM179x... 5 levels D-S (lower 18 bit) and 62 levels (upper six bits) are in PCM format.
    I keep saying that there is a reason why the manufacturers departed from pure 1 bit, but I guess some people think it's all a conspiration.
    Sorry, at first I read your last word as "constipation!"

    For what it's worth, Miska does enjoy the Teac UD-501, and that is a T.I. PCM1795.

  18. #893
    Sophomore Member
    Join Date
    May 2012
    Location
    USA Virginia
    Posts
    305
    Quote Originally Posted by Superdad View Post
    Sorry, at first I read your last word as "constipation!"
    Haha, I almost wrote that too...
    I also enjoy my PCM1792A (pin-compatible with slightly better specs than the 1795, but 'only' 24 bit). OS is performed by Denon Alpha24 DSP technology, I can't say if is better or worse than Miska's.
    I use it mostly for SACD's, the PCM CD formats I like them better trough my 'old' PCM61.

  19. #894
    Senior Member
    Join Date
    Apr 2010
    Location
    Finland
    Posts
    4,408
    Blog Entries
    7
    Quote Originally Posted by SoNic67 View Post
    TI has 66 levels S-D too in their top-of-the-line PCM179x... 5 levels D-S (lower 18 bit) and 62 levels (upper six bits) are in PCM format.
    I keep saying that there is a reason why the manufacturers departed from pure 1 bit, but I guess some people think it's all a conspiration.
    I just cannot locate performance reasoning for it, because the chips still give out better performance with DSD when fed by a good modulator.

    Main reason I can think of is that the companies didn't manage putting higher interpolation rates and better than 3rd order modulators in, so they had to try to counter that by adding more bits. But still, if they'd had the resources on chip they could have used higher rate, higher order modulator and more bits.

    Although number of "bits" in delta-sigma DAC is not so clear cut, since in most cases there's no 1:1 relation between output "bits" and number of 1-bit D/A elements. So you can utilize almost any number of output elements in different configurations for converting DSD bitstream, like happens in PCM1795 for example. (something you can't do with R2R ladder, because bits are not equally weighted there)

    Moore's law theoretically allows me to double the performance every two years...
    Signalyst - http://www.signalyst.com
    Developer of HQPlayer

  20. #895
    Quote Originally Posted by Miska View Post
    --- the chips still give out better performance with DSD when fed by a good modulator.
    Miska, is some (good) modulator involved in DSD D/A process (CS4398 or WM8741 for example)? If yes, how Lampizator did totally analog conversion of DSD (w/o modulator)?
    Or modulator is needed only in DSD signal creation stage?
    In Saracon DSD converter we can found three types of modulator: 6th, 8th and 10th order. Unfortunately I don't know about Korg's AudioGate.
    And there is some DSD A/D chips, what order on modulator they used?

  21. #896
    Senior Member
    Join Date
    Apr 2010
    Location
    Finland
    Posts
    4,408
    Blog Entries
    7
    Quote Originally Posted by Maldur View Post
    If yes, how Lampizator did totally analog conversion of DSD (w/o modulator)?
    Or modulator is needed only in DSD signal creation stage?
    Like someone said in some interview, DSD is almost analog already represented in digital form. A bit like class-D amplifiers that you can also implement without any digital circuits. Delta-sigma modulator is used to generate DSD bit stream, either from analog or digital source signal.

    Modulator is only needed when you input PCM into delta-sigma DAC, then it needs to be converted inside through modulator. So for PCM inputs, there's a DSP engine inside DAC chip to handle digital interpolation filters and the modulation. For DSD inputs, the idea is that you don't need any of the DSP...

    In Saracon DSD converter we can found three types of modulator: 6th, 8th and 10th order. Unfortunately I don't know about Korg's AudioGate.
    10th order may get you into trouble with many DACs... Unless they tame it a bit which is also doable.

    AudioGate looks like typical 5th order modulator.

    And there is some DSD A/D chips, what order on modulator they used?
    I have not studied that side extensively yet...
    Signalyst - http://www.signalyst.com
    Developer of HQPlayer

  22. #897
    Quote Originally Posted by Miska View Post
    For DSD inputs, the idea is that you don't need any of the DSP...
    This is a good idea, it's like me very well. Thanks for explanation .

  23. #898
    Quote Originally Posted by Superdad View Post
    Can you explain the I/O on the UDA2 board? Obviously USB input, but what are the outputs? I assume I2S, and it also looks like an S/PDIF output. What are the power supply voltage requirements?
    UDA2 board playing/recording version supports both playing and recording of music via USB.
    UDA2 board has two modes, "Bulk Transfer Mode" and "USB Audio Class 2 Mode". One of these mode can be selected by setting jumper pins at its power-on timing.

    ElectrArt's proprietary playing/recording program named "PlayAudio"(Windows version only) supprts the "Bulk Transfer Mode" while every "USB Audio Class 2" campatible players on both Windows, Mac and Linux can output to UDA2 board.
    The screenshot of the "PlayAudio" is like this.
    uda2play.jpg

    Bulk Transfer Mode covers;
    [Stereo play] PCM up to 384kHz/24 bit WAV, DSD up to 22.6MHz (DSD512) DSDIFF
    [Stereo record] PCM up to 384kHz/24 bit WAV, DSD up to 11.3MHz (DSD256) DSDIFF
    [4 Ch play/record] DSD source up to 5.6MHz (DSD128) DSDIFF

    USB Audio Class 2 Mode covers;
    For PCM,
    <<Mac OS X(10.6.5 or higher)without any drivers>> [Stereo] PCM up to 384kHz/24bit
    <<Windows with proprietary ASIO 2.2 compatible driver>> [Stereo] PCM up to 384kHz/24bit

    For DSD, UDA2 supprots both "DSD Audio over PCM Frames Version 1.1" and "ASIO 2.2 Native DSD"
    <<Mac OS X (10.6.5 or higher) with Audivana Plus(1.4.1 or higher)>> up to 5.6MHz
    <<Windows with foobar2000(Ver.1.1.11)+ASIO Output Plug-in>> up to 5.6MHz

    UDA2 board outputs I2S & DSD-raw with a word clock and S/PDIF.
    The board inputs I2S, SDIF-3 and a work clock.
    It accepts an external DC +5V power by setting a jumper.

    Quote Originally Posted by Superdad View Post
    Of even greater interest to me is the A/D board. I/O for that one? Looks like balanced XLR only for analog input. What are the outputs, and what data formats does it output? Can it work without the UDA2?

    I guess what I would really want for analog recording would be a USB output.
    AT1201 ADC board accepts XLR analog input and output SDIF-3 digital and a word clock.
    It can output PCM up to 384kHz/24bit and DSD up to 11.3MHz(DSD256).
    It can work without UDA2 as far as you have any SDIF-3 compatible recorder.

  24. #899
    Quote Originally Posted by Miska View Post
    Quote Originally Posted by Maldur View Post
    And there is some DSD A/D chips, what order on modulator they used?
    I have not studied that side extensively yet...
    About 1 week ago, I visited Professor Yamazaki of Waseda University in Tokyo. He is said to be a person who proposed 1 bit Audio technology to SONY at the beginning. (By the way, he does not like the naming "DSD(Direct Stream Digital)" by SONY. His interpretation is that "DSD" stands for "Delta-Sigma Direct" or something like that.)

    He said Paganini's Violin Solo recording of 11.3 MHz sampling, which is downloadable from 1 Bit Consortium Web page, was recorded by him with his proprietary recorder of 2nd order delta-sigma modulation. Kristof Brarati played the piece with his Stradivarius at the Stradivari Society in Chicago.
    He added that other recordings of 2.8 MHz or 5.6 MHz available on the web site were recorded by recorders with commercial ADC chips of 5th or 6th orders.

    By the way, I will demonstrate a play of the recording of 11.3 MHz sampling at "1 Bit Audio Research Meeting" held at Waseda University on December 20th. Is there any other sources of 11.3 MHz sampling which is available to audiophiles like me?

  25. #900
    Junior Member
    Join Date
    May 2009
    Location
    Middle East/USA
    Posts
    587
    Does recording via DoP make sevse to anyone? This works? PS Audio recording as DoP .wav? Then burning to DVD...

    Recording DSD, playing back over PWT - PS Audio Discussion Forum
    DIGITAL: Windows 7 x64 JRMC19 >Adnaco S3B fiber over USB (battery power)> Auralic Vega > Tortuga LDR custom LPSU > Zu Union Cubes + Deep Hemp Sub

    ANALOG: PTP Audio Solid 9 > Audiomods Series V > Audio Technica Art-7 MC > Allnic H1201 > Tortuga LDR > Zu Union Cubes + Deep Hemp Sub

    ACCESSORIES: PlatterSpeed, BlackCat cables, Antipodes Cables, Huffman Cables, Feickert Protracter, OMA Graphite mat, JRemote

LinkBacks (?)

  1. 07-15-2013, 06:28 AM
  2. 07-07-2013, 09:13 AM
  3. 05-28-2013, 03:26 PM
  4. 05-14-2013, 05:42 PM