Sorry, I'm just telling the truth, in terms of architecture, DSD DAC is indeed better than R-2R. Since you are a master who specializes in making R-2R, you must be able to understand from my previous description that the above words are true.
Of course, excellent architecture does not mean good sound, so many people don't realize this. Many people have not noticed that the cheap version of R-2R DAC only uses poor quality SMD resistors, so the actual sound quality may be inferior to 9038 and other chip DACs. I'm sorry, but I told the truth again.
It's probably a differential 32-tap single-bit FIRDAC. When it's non return to zero, making it differential does reduce artefacts due to unequal rise and fall times without the (far-off) jitter sensitivity increase you would get from going for a return to zero DAC, among other things.
R-2R is extremely sensitive to mismatch and single-bit sigma-delta modulators have idle tone issues, lots of ultrasonic noise and just about anything can convert some of their out-of-band quantization noise into the audio band. I wouldn't dare to make any sweeping statements that one is conceptually better than the other.
I now suddenly understand what you want to explain. Do you want to show that the R-2R architecture does not require upscaling when processing PCM, so it can handle PCM better than the DSD architecture that requires upscaling? If so, it would be very interesting. In fact, R-2R still needs upscaling when processing PCM, otherwise, why would it still need a clock signal? Any DAC needs upscaling.
Anyone who has used hardware or software upscaling algorithms will definitely know that the PCM processed by an excellent upscaling algorithm will sound better after upscaling. That’s because the external upscaling algorithm is better than the internal DAC. The relationship that makes the internal DAC no longer work.
Soekris's words are misleading.
In other words, if the cheap version of R-2R DAC uses a low-cost digital processor level, the upscaling will definitely not be done well, and the sound quality will of course be poor. Since they all need to upscaling, DSD DAC must choose to use AK4137 or other excellent upscaling algorithms. This is a good thing, which will make DSD DAC have a certain level in the upscaling part.
If you directly play the original DSD recording signal, no upscaling is needed, and the high-resolution advantages of the DSD DAC architecture can be fully utilized.
In addition, for R-ladder DACs less than US$5,000, which products will use high-quality resistors? (The actual price of A&C DSD DAC is less than US$5,000) For R-ladder DACs less than US$5,000, it is generally used Ceramic Composition Resistor has not been used either, because a huge amount of Ceramic Composition Resistor still has to consume a lot of manpower to weld one by one. Most R-ladders still use very cheap SMD resistors, so you can tell from the materials that most R-ladder DACs sound very bad, which is worse than the well-designed chip DACs.
We are a designer and manufacturer, and there is no marketing cost or distributor cost. At the same time, because it is a new product, we need to lower the price as much as possible to attract people to buy it with their own ears, and then spread it through word of mouth.
Since it seems that most Diyaudio website members have serious misunderstandings about many details, so that they are not interested in the better but cheaper R-ladder DAC, I will not answer any questions. Those who are really interested, please contact me via private message.
Once again, R-ladder DACs are excellent in terms of architecture, otherwise the most expensive DACs will not all be R-ladders. The problem is that a large number of low-cost R-ladders all use cheap SMD resistors and other components, so that you have a wrong understanding of R-ladder DAC.
You can certainly work around the issues I mentioned in post #44 and make good DACs based on single-bit FIRDACs, but you can also pay too little attention to those things and make very mediocre DACs.
So far all I know about your DAC is: it uses an AK4137, so you have no control over the modulator algorithm, like you would have with an FPGA. It's apparently a 32-tap balanced FIRDAC using a CMOS logic-resistor structure with expensive resistors. It's unclear what measures you took to prevent crosstalk of the sigma-delta modulate to the reference or the clock, or intermodulation in the I/V board. You argue that it must be good because of the fancy resistors and because the FIRDAC has more taps than there are bits in a typical R-2R DAC, which doesn't make much sense to me. You don't provide any measured data, controlled listening test results or additional technical information.
So far all I know about your DAC is: it uses an AK4137, so you have no control over the modulator algorithm, like you would have with an FPGA. It's apparently a 32-tap balanced FIRDAC using a CMOS logic-resistor structure with expensive resistors. It's unclear what measures you took to prevent crosstalk of the sigma-delta modulate to the reference or the clock, or intermodulation in the I/V board. You argue that it must be good because of the fancy resistors and because the FIRDAC has more taps than there are bits in a typical R-2R DAC, which doesn't make much sense to me. You don't provide any measured data, controlled listening test results or additional technical information.
Nothing he says make much sense to me neither, it sounds like all misunderstandings and marketing crap....
And the excellent precision thin film smd resistors are not cheap....
And the excellent precision thin film smd resistors are not cheap....
Last edited:
Not many people understand quantum cryptograph and quantum random number generation either but that was my day job.
DSD itself is a subset of delta sigma modulation, it is a trademark for a pulse density single bit modulation. Thus DSD itself is single bit at a higher rate thus per second you have a higher bit rate.
It is entirely possible to create multi-bit sigma delta and the output is therefore a multi-bit encoding rather than a bit stream (even the feedback error quantisation then is multi-bit). The sample rate remains the same, the number of bits per sample can therefore be more than one bit.
PCM and DSD are completely different encoding and to go between requires transcoding. Comparing capability by number of bits from one and number of bits of the other is invalid.
So you're saying that the R2R sounds good because of the software making up for the inadequacies of the R2R hardware?
MarcelD's DSD512 tube DAC fed DSD512 seems to rather good.. and with a few minor military grade IC changes would survive EMI blast and nuclear war.
So if I read this right.. (storm in a tea cup)..
1. you are developing a PCM to DSD transcoder for the front end
2. you are using a simple one bit DSD output stage upsampled to move the noise up the frequency range
3. you have a reconstruction filter to remove noise.
4. you have a standard amplifier then drives the speakers.
Currently a number of 'audiophile' player applications can actually transcode the PCM into a DCD stream in software. There are PCM-DSD FPGA transcoders but like all realtime transcoders, they are limited in knowing the past and current sample. Software presents the option of processing the music file as a complete sample set.
DSD vs other formats will always have their proponents, so simply making statements without sharing the complete test setup and result for independent peer review will end up being rebutted.
Actually playing DSD on a R-2R DAC is easy, to convert DSD to PCM or analog, is basically just a low pass filter, which can easily be done digitally.
Like in my R-2R DACs, they play DSD nicely....
Like in my R-2R DACs, they play DSD nicely....
Understand your point about software, but there can be FPGA transcoders that effectively look into the very near future. Otherwise we couldn't have things like linear phase filters, pre-echo, and so forth.
Last edited:
True - pipeline buffering the input stream and that's probably very likely with reclocking too.
It's simple
Maybe for DSD/SACD recordings it is.
But not for PCM recordings which the majority are, "R2R" converts them "Bit Perfect" where "Delta Sigma" is just a "facsimile" of it.
(Would you rather look at the real Mona Lisa, or a hi rez copy of it?)
Cheers George
^^^ Mostly incorrect.
The real Mona Lisa exists (at least in terms of human perception) in the analog domain. It doesn't have bits. Now, if we were to take the crappiest digital camera ever made and take a digital picture of the analog painting, then the file produced by the camera would be a bit-perfect version of a digitized Mona Lisa. That it is bit perfect until edited or lossy compressed, something like that, is of very little value because the camera producing the original bit-perfect file is so bad.
Regarding modern analog to digital converters used for audio digitization, what comes out of the digitizer is usually already not bit-perfect. The ADC works in a way that produces what is sometimes called Raw Mode format. It is neither standard PCM nor standard DSD. The ADC can be programmed to internally convert Raw to either PCM or DSD. At that point the bit-perfection has already been lost.
Thus, bit-perfect is quite overrated. When format conversions are done, what matters is the quality with which they done. Poor conversion to PCM can be just as bad or worse than poor conversion to DSD. Depends.
The real Mona Lisa exists (at least in terms of human perception) in the analog domain. It doesn't have bits. Now, if we were to take the crappiest digital camera ever made and take a digital picture of the analog painting, then the file produced by the camera would be a bit-perfect version of a digitized Mona Lisa. That it is bit perfect until edited or lossy compressed, something like that, is of very little value because the camera producing the original bit-perfect file is so bad.
Regarding modern analog to digital converters used for audio digitization, what comes out of the digitizer is usually already not bit-perfect. The ADC works in a way that produces what is sometimes called Raw Mode format. It is neither standard PCM nor standard DSD. The ADC can be programmed to internally convert Raw to either PCM or DSD. At that point the bit-perfection has already been lost.
Thus, bit-perfect is quite overrated. When format conversions are done, what matters is the quality with which they done. Poor conversion to PCM can be just as bad or worse than poor conversion to DSD. Depends.
Listen to a direct a/b level match of the two conversion technics with CD 16 or 24bit, and then tell us what you drew you into the music more and wanting more, and which left you a little bored and a little sterilized.
Cheers George
I have listened to both, and I know how they tend to sound. Personally, I prefer a certain type of DSD dac, a very good one. However, I know people who prefer the sound of R2R. This has nothing to do with bit-perfect or not. Its that no dac is audibly perfect, and the quirks of R2R dacs tend to be different from the quirks of DSD dacs.
Regarding DSD in particular, SACD is the worst of it. Its sample rate is referred to as DSD64. DSD doesn't start sounding really good until its up around DSD256 or maybe higher. Also, a really good DSD dac can be pretty costly to make even if its diy.
Regarding DSD in particular, SACD is the worst of it. Its sample rate is referred to as DSD64. DSD doesn't start sounding really good until its up around DSD256 or maybe higher. Also, a really good DSD dac can be pretty costly to make even if its diy.
🧏♂️/🦯forest and trees.
I believe your too hung up on measurements, the objective go hand in hand with the subjective. Why do you think R2R has never gone away, have a listen to good ones.
I have a MSB Discrete R2R, nothing DS has come near it for PCM conversion that can satisfy "musically" as it does.
I know amps that measure insanely good, and the very same amp that's had it's masses of global feedback removed to make it measure so good, and I replaced with just a local feedback becomes far more enjoyable to listen to, but doesn't measure near as good as it did with masses of global feedback.
Applies to the same analogy as my last post with DS v R2R topologies converting PCM
Cheers George
I believe your too hung up on measurements, the objective go hand in hand with the subjective. Why do you think R2R has never gone away, have a listen to good ones.
I have a MSB Discrete R2R, nothing DS has come near it for PCM conversion that can satisfy "musically" as it does.
I know amps that measure insanely good, and the very same amp that's had it's masses of global feedback removed to make it measure so good, and I replaced with just a local feedback becomes far more enjoyable to listen to, but doesn't measure near as good as it did with masses of global feedback.
Applies to the same analogy as my last post with DS v R2R topologies converting PCM
Cheers George
Last edited:
You don't know me very well. I am always talking about problems with spectral analysis (and its human interpretation), and I have never posted an FFT.
Regarding DCS and such, I know of a guy who I can't name, but he is big name guy and consultant in the hi-fi world of his country. He has a full blown $100,000+ dSC dac. But when no visitors are around he prefers a Chord DAVE. He just can't say the DAVE is better, given its much lower cost point. That's because of how audio business politics works where he is.
Last edited:
This is really comical. In almost all DAC threads, there are discussions between Mark and a certain other member, with Mark posting listening impressions and the other member posting measurements, and Mark explaining why he doesn't find the other member's measurements useful and the other member explaining why he doesn't trust Mark's listening tests.