That is interesting...so you believe the fundamental pieces of digital are now here at our disposal (well yours actually) to beat analog playback...and the key is implementing it properly to achieve the inherent potential of the existing fundamental pieces. And the key focus is on eliminating the problems that digital ADDS to playback. So if intermodulation noise, maybe on the margin phase shifting, can be solved we are largely there?
Digital that sounds like analog
- Thread starter opus111
- Start date
That is interesting...so you believe the fundamental pieces of digital are now here at our disposal (well yours actually) to beat analog playback...and the key is implementing it properly to achieve the inherent potential of the existing fundamental pieces. And the key focus is on eliminating the problems that digital ADDS to playback. So if intermodulation noise, maybe on the margin phase shifting, can be solved we are largely there?
Yes - that's my 'grand hypothesis' that I'm setting out to test (or alternatively falsify). That digital can beat analog its just a matter of implementing right. I've had that belief from before I heard the first CD. I'm not totally sure that all the relevant issues can be solved at 16bits and 44k1 but probably 20bits and 88k2 are plenty good enough to knock analog into a cocked hat 
Beat it or emulate it?
I remember a debate I had with my brother at the advent of commercial digital video. I said one day digital will have the resolution of celluloid. With 4k and now 8k we are seeing that. The question remains however what is better looking to the content creators and the viewers. Would The Godfather be as good a movie without the look chosen for it? The cinematography is an essential part of how the story was told. Bring that to music, would Born in the USA recorded on 4 track be a better album had it been recorded differently? Sargent Pepper? KOB?
Heavy stuff. I like where Opus is at though. I'm format agnostic. I like analog because at this point in time I think its head start has given it an implementation edge as far as keying in to the aesthetic sensibilities of a large segment of the audiophile market. As such I laud any and all efforts at improving implementation.
I don't participate in analog vs digital debates because I've seen one wipe the floor with the other and it's always been due to software and device implementation not because of the format per se. In A/Bs it's always a case to case scenario.
I remember a debate I had with my brother at the advent of commercial digital video. I said one day digital will have the resolution of celluloid. With 4k and now 8k we are seeing that. The question remains however what is better looking to the content creators and the viewers. Would The Godfather be as good a movie without the look chosen for it? The cinematography is an essential part of how the story was told. Bring that to music, would Born in the USA recorded on 4 track be a better album had it been recorded differently? Sargent Pepper? KOB?
Heavy stuff. I like where Opus is at though. I'm format agnostic. I like analog because at this point in time I think its head start has given it an implementation edge as far as keying in to the aesthetic sensibilities of a large segment of the audiophile market. As such I laud any and all efforts at improving implementation.
I don't participate in analog vs digital debates because I've seen one wipe the floor with the other and it's always been due to software and device implementation not because of the format per se. In A/Bs it's always a case to case scenario.
Emulating its good points and beating it by reducing its primary flaw (noise modulation) without adding additional foibles.
Hoepfully any analog vs digital debates will go on in places other than here - I recognize that in implementation terms digital's not quite there yet but I'd like to set the bar high or there's no challenge in it. Digital holds out the promise of getting a truly satisfying sound without the stratospheric spending that decent analog requires. As such its really a 'people's format' rather than just for the few.
100% agree...its what i as a digital only guy seek. as a non-techie consumer, i seek emulating analog's good points...while maintaining quiet/low noise, detail retrieval, and relative ease of use of digital.
opus111: A 100-ohm load is very low for a typical preamp; you are probably going to have to use an input buffer as well, or change to different source and load impedances more in line with "normal" parts.
I have had very mixed performance using RF parts for audio. One of the things to watch is low-frequency phase/flicker noise; some wideband devices have very high noise near DC, making them poor choices for audio. OTOH, with 100-ohm source impedance, that may knock a lot of it down...
FYI, I use delta-sigma (DS) based on comments from Gabor Temes, a teacher and friend years ago who was one of the first to work with them (along with John Candy and others). He pointed out that the difference (delta) block comes before the summing (sigma) block in circuit, then went on with the insider's knowledge that one of the founding papers was mis-translated, swapping the terms, and so SD took root.
I am not at all sure the rising noise is due to "noise modulation" but have seen the term defined different ways. I started to babble on about architectures and designs and such but realized that is way too deep for here, you probably know it all already, and I am not a DS expert though have designed a few (at much higher frequencies).
I believe the problems with early filters have been well-documented and are certainly one of the reasons for oversampling so reasonable filters can be used. What a lot of people may not realize is that the problem with many early recordings happened on the input (recording, ADC) side, caused by the anti-alias filter, that is as much or more so a problem than the DAC's image filter. ADC aliasing puts stuff back in band where it shouldn't be; DAC images appear at higher frequencies and are also undesirable but are typically beyond the range of our hearing. That is, IMO; I am neglecting the ultrasonic argument for now.
Enough from me, sorry for hi-jacking the thread - Don
I have had very mixed performance using RF parts for audio. One of the things to watch is low-frequency phase/flicker noise; some wideband devices have very high noise near DC, making them poor choices for audio. OTOH, with 100-ohm source impedance, that may knock a lot of it down...
FYI, I use delta-sigma (DS) based on comments from Gabor Temes, a teacher and friend years ago who was one of the first to work with them (along with John Candy and others). He pointed out that the difference (delta) block comes before the summing (sigma) block in circuit, then went on with the insider's knowledge that one of the founding papers was mis-translated, swapping the terms, and so SD took root.
I am not at all sure the rising noise is due to "noise modulation" but have seen the term defined different ways. I started to babble on about architectures and designs and such but realized that is way too deep for here, you probably know it all already, and I am not a DS expert though have designed a few (at much higher frequencies).
I believe the problems with early filters have been well-documented and are certainly one of the reasons for oversampling so reasonable filters can be used. What a lot of people may not realize is that the problem with many early recordings happened on the input (recording, ADC) side, caused by the anti-alias filter, that is as much or more so a problem than the DAC's image filter. ADC aliasing puts stuff back in band where it shouldn't be; DAC images appear at higher frequencies and are also undesirable but are typically beyond the range of our hearing. That is, IMO; I am neglecting the ultrasonic argument for now.
Enough from me, sorry for hi-jacking the thread - Don
Hi Don,
Just curious to learn...if you are not sure intermodulation noise is the challenge, what would you point to as the 'top 3' issues holding digital playback (D/A) at the moment? And what is your view on multibit/NOS vs DS approach to D/A? thanks!
Yes - that's my 'grand hypothesis' that I'm setting out to test (or alternatively falsify). That digital can beat analog its just a matter of implementing right. I've had that belief from before I heard the first CD. I'm not totally sure that all the relevant issues can be solved at 16bits and 44k1 but probably 20bits and 88k2 are plenty good enough to knock analog into a cocked hat
---- That's what Bob Stuart from Meridian says. :b
I am not competent to say as I do not have the systems and ears most on WBF seem to have. Personally, I think perception is a major issue (people think analog is better), poor recordings are a huge issue (digital does not hide flaws as well, and a lot of remasters seem to my ears to be very poor), and technically 44.1 kS/s means high-order filters (both the anti-alias filter before the ADC and image filter after the DAC) that introduce problems well into the audio band.
The noise-shaping in DS converters could certainly be an issue, but I am not sure the impact of noise modulation. I mean exactly that; I do not know the magnitude of the problem in modern converters. Tones were a big problem with early converters, but modern designs using dither and high-order loops (which do their own dithering) have apparently solved that problem.
Oversampling relieves the requirement for high-order filters and thus would be one of my keys to better sound, more so than resolution (others will argue that, of course). There is also the issue of oversampling using a conventional DAC architecture; oversampling and DS are not interchangeable terms technically. There are non-oversampling DS architectures, and you can oversample using a conventional DAC.
Yeah he says that but as far as I know he still uses D-S (or S-D) converters so he won't get there...
I can't understand your list for considerations. You cite a TI part for example in your first line which is D-S and I've already said I'm avoiding them.
Your comment about self-dithering prompts me to ask - how to ensure that the dither is indeed of the correct amplitude and prob. density function in the loop?
You lost me again.
Still confused - selections of what? I'm not going to use upsampling, I'm avoiding a digital filter.
The PCM1704 is too expensive for what it offers and 24bits are unnecessary for redbook. Pointless to be paying for unused bits. PCM1702 would be a better choice but even that suffers from too much glitching compared to the segmented current source DACs.
Nope, I'm clueless on them. As for jitter I'm ignoring it in the main. Until I can hear its effects, its not on the list of things to address.
No worries Bob - the DAC I am using is the NXP (Philips) TDA1387 if that is of interest to you : http://www.datasheetcatalog.org/datasheet/philips/TDA1387T.pdf
This remark has me puzzled. The 100R load is built into the AD603 chip I'm using as the amplification stage post DAC. It has something like 33dB of gain which brings the swing up around the typical level for a digital converter (betwen 1 - 2VRMS). The 100R working impedance also makes the inductors easier to fabricate as higher impedance would necessitate more turns (though somewhat lower copper losses at HF relatively in compensation).
Yes this also was one of my concerns before trying the part out. Many RF designed parts just aren't characterized down to DC as there's no need. So its a bit hit-and-miss whether their noise performance is any good for audio or not! Subjectively so far there isn't a significant problem though the part is only just good enough for redbook noise floor.
I think I recall you mentioning this before - I just have a habit of writing S-D I wonder if its not too hard to break that?
Of course the nomenclature could be function based so while D-S means D before S, we could just as easily say S(D()).Yes it does seem to me that inadequate anti-aliasing filtering is a far worse sin than inadequate anti-image filtering.
Its not a hijack at all, very relevant stuff
The elliptic filter is hooked up now to the LAID DAC (which uses 19 TDA1387s in a transversal filter arrangement) and the difference is total magic. Previously I was using a 4th order LC filter (Butterworth) but the improvement from this 7th order is substantial in terms of two areas - ambience and dynamics. On a really good recording - I'm listening now to a BIS recording of Mozart arranged by Hummel - the bloom of the acoustic has blossomed in all directions behind the speakers. The subjective noise floor seems lower so the dynamic range has improved, even though it was pretty good before. I was only going to test out the filter in this DAC before transplanting it into the simpler design which has no transversal filter, but I can't go back now so I'll have to build another one!
I am with you on the NOS side of things. However, didn't you design a 2x filter as well at some point? If so, why did you opt to go back to NOS instead?
Similar threads
 | Steve Williams Site Founder | Site Owner | Administrator |  | Ron Resnick Site Co-Owner | Administrator |  | Julian (The Fixer) Website Build | Marketing Managersing |