Digital that sounds like analog

[DonH50]

Even if the DAC's anti-imaging filter has no ringing, ringing has already been baked-in because digital requires band-limiting prior to the ADC. This anti-aliasing filter must have ringing, and its non-negotiable - so ringing still exists irrespective of the interpolation algorithm. Seems daft to me to focus on the total elimination of ringing at the DAC. The THD+N claim is percent or a fraction?

No, band-llimiting is required, but a filter does not "have" to ring. I agree any artifacts added at the ADC are going to be impossible (or at least very difficult to the point of impractical) to remove by a DAC at the other end of the chain.

The THD+N claim is commensurate with a 16-bit DAC (0.0018% = 15.47 bits, 94.89 dB dynamic range).

 

[opus111]

No, band-llimiting is required, but a filter does not "have" to ring.

Filters I agree don't have to ring in general, however in the current context they do have to or they'll alias, which is worse. There's only a narrow transition band with RBCD - if we want FR up to 20kHz and decent attenuation beyond 22kHz. This seems to me to be the biggest advantage of higher sample rates - the ability to have AAFs which don't ring.

The THD+N claim is commensurate with a 16-bit DAC (0.0018% = 15.47 bits, 94.89 dB dynamic range).

Yep - but the '%' sign was missing, hence I asked If xxHighEnd is using time domain interpolation its hard to see how they can get distortion down lower than the DAC's spec. I'm certainly interested to know the details.

btw - atm means 'at the moment'

 

[hvbias]

Specific to what? Feel free to chip in on that thread with any questions you might want to ask about DACs. I'd prefer not to derail this thread by talking about my DAC design (much as its a subject close to my heart!).

I was interested in seeing the schematic

Also from the "why do we prefer vinyl more than digital thread", both links you posted point to this thread. One of the links had something about Lite DAC in the URL, but it points to this thread.

As its tangential, more than happy to do that on that same thread, rather than here. Post up that question on there and I'll expand more.

Please do expand, original posts from the other thread

Technically R2R DACs are an inferior breed so your preference for vinyl makes sense to me. S-D DACs (ES9018 being the best example) in my estimation don't reach the heights of R2R.

Can you expand on this? The two sentences read like contradictory statements to me.

 

[Orb]

Ah on the DAC I cheated by buying a Muse 4 * TDA1543 DAC and modifying it. So the DAC design evolved gradually - very hard to say how many hours went into it. It was spread out over more than a year but probably totally a week of time. By which I mean a normal working week since I waste so much time online too

Opus, any reason you did not go with the TDA1541, or try to get hold of Analog Devices DAC chips?
More curious than anything else.
Thanks
Orb

 

[opus111]

I was interested in seeing the schematic

Well a complete one doesn't exist as yet, I'll be gradually putting it out on the modding thread. In outline its nothing special on the digital front end - just the DAC-AH's input circuitry, with a trafo added for isolation and the power supplies slightly improved for HF rejection. The DAC part is 8 TDA1387s (or TDA1545As, depending on which I have to hand) arranged in balanced (positive and negative phases) forming a 3 tap FIR filter. The main coefficient of the filter being provided by two DACs in parallel. A shift register implements the necessary delays to the minor taps - this provides the droop correction. Then there's the passive (LC) filter whose schematic I have already posted. Followed up by AD605s as post-amplifiers.

Also from the "why do we prefer vinyl more than digital thread", both links you posted point to this thread. One of the links had something about Lite DAC in the URL, but it points to this thread.

OK I'll go back and patch up that cock-up, thanks for letting me know!

Please do expand, original posts from the other thread

Will do in a subsequent post...

 

[opus111]

DAC architectures

OK here's a bash at setting out some kind of hierarchy of the technical merits of DACs.

At the bottom we have the S-Ds which by virtue of using feedback in an attempt to linearise a massively non-linear transfer function (truncation) are saddled with seemingly insurmountable noise modulation. This corrupts timbre and squashes dynamics. ESS has made the best out of a bad job with their Sabre32 implementation. This design, while being the top rank of the genre does demonstrate substantial noise modulation - check out the THD+N vs level plot of the Weiss Medea+ DAC.

R2R are the multibit DAC at the bottom of the pile by virtue of the difficulty in timing the switch element transitions to get low enough glitch. PCM1704 is an example of this - they're attempting to correct the zero crossing (biggest carry propagation, all 24 bits switching at once) by moving it to the 25/75% points. Its still there though (albeit at 23bits coz the Colinear blueprint uses two DACs) just now its only there with higher than -6dB signal levels rather than with any signal at all. This certainly helps the lower level THD figures, its a little bit debatable if it helps the SQ on real music overall. Low level performance will certainly be cleaner. Glitches at reduced levels exist at all the other carry points with R2R too.

My preferred kinds of multibit DAC are resistor string and segmented current source. The former is used in the Metrum Octave, the latter in TDA1541A and family variants. These both have a glitch advantage over R2R which translates to better dynamics and superior instrumental timbres. Of the two I prefer the current source kind because resistor string DACs switch voltages and in practice have on-board opamps which do nothing for the SQ.

 

[opus111]

Opus, any reason you did not go with the TDA1541, or try to get hold of Analog Devices DAC chips?

Its an interesting question - thanks for asking.

The reasons for not going for TDA1541A are many and various. Firstly I'm a late-comer to the party and seems to me like the heavy-weight hitters (Thorsten, Pedja, John Brown) have been on this particular case for over a decade. The chip is hot and I really wanted a design which would translate eventually to mobile (the biggest market). I also have a penchant for liking designs which use many small cheap parts to build something bigger - you might call it a 'cellular' ethos. A design which paralleled many TDA1541A in an array would be cost prohibitive and given the chips aren't in manufacture any longer, few people would be able to (or be able to afford to) build it. Given all that, the diminutive TDA1545 captured my inspiration for its superb low power and very decent specs so I felt it would be more fun to branch out a bit and to go where I hadn't seen others go - into arrays of very cheap CMOS DACs.

Some of these arguments apply to the ADI multibit DACs too - they're just not as cute as 8pin SOICs and they're now out of production and what parts are around are substantially more expensive than TDA1545s and TDA1387s with little to show for that extra money in terms of performance.

Help any?

 

[Orb]

Ah k.
Thanks and yes
Thanks
Orb

 

[DonH50]

Virtually any high-resolution conventional DAC uses a segmented architecture to reduce matching requirements and large-scale glitches. Some RF DACs are fully unary although lower resolution (10 bits or less, unless Klaus has done it for 12 bits).

 

[LL21]

Its an interesting question - thanks for asking.

The reasons for not going for TDA1541A are many and various. Firstly I'm a late-comer to the party and seems to me like the heavy-weight hitters (Thorsten, Pedja, John Brown) have been on this particular case for over a decade. The chip is hot and I really wanted a design which would translate eventually to mobile (the biggest market). I also have a penchant for liking designs which use many small cheap parts to build something bigger - you might call it a 'cellular' ethos. A design which paralleled many TDA1541A in an array would be cost prohibitive and given the chips aren't in manufacture any longer, few people would be able to (or be able to afford to) build it. Given all that, the diminutive TDA1545 captured my inspiration for its superb low power and very decent specs so I felt it would be more fun to branch out a bit and to go where I hadn't seen others go - into arrays of very cheap CMOS DACs.

Some of these arguments apply to the ADI multibit DACs too - they're just not as cute as 8pin SOICs and they're now out of production and what parts are around are substantially more expensive than TDA1545s and TDA1387s with little to show for that extra money in terms of performance.

Help any?

Very interesting read. Thanks!

 

[opus111]

Virtually any high-resolution conventional DAC uses a segmented architecture to reduce matching requirements and large-scale glitches. Some RF DACs are fully unary although lower resolution (10 bits or less, unless Klaus has done it for 12 bits).

The DACs built for RF comms use look to be a very promising way for me to build a DAC for hi-res material (>16bits) because of the exceedingly well specific dynamic performance. In my estimation SFDR and multitone IMD figures give a better prediction as to how a DAC's going to sound than THD+N. I recently acquired a few samples of an old NS design, the DAC14135 - http://www.ti.com/lit/ds/snas115/snas115.pdf. I'm impressed, inter alia by its exceptionally low glitch energy of 1pV.s.

 

[opus111]

I'm gradually making my way through the boxed set of Karajan's 1960s recordings (82 CDs in total). None yet have disappointed in any way, but this one in particular is a stand out for the intimacy of the vocal performances and (naturally enough) the recording dynamics. Its only marred by premature fade-outs at the end of a number of items - the natural die-away sounds truncated often.

http://www.amazon.com/Haydn-Schopfung-Creation-Gundula-Janowitz/dp/B000001GXN/

<edit> About the fade outs - these may be a bug in my player, I need to investigate further...

 

[opus111]

Yes the fade-outs are the el-cheapo Chinese DVD player marking the track changes, probably nothing to do with the disc.

Here's my current favourite version of the Beethoven Diabelli Variations - the most intense reading I've come across : http://www.amazon.com/Variations-Diabelli-Op-120-Sokolov-Grigory/dp/B0000CNTJ1/

Its a live recording, there are a few audience coughs, nothing too distracting from the supremely focussed playing of Mr Sokolov.

 

[opus111]

As my appetite for Russian music seems insatiable, here's what I'm listening to right now - http://www.amazon.com/Ballets-Russes-Russian-Dances/dp/B00011RULK/

Excellent foot-tapping dynamics and some really intense, biting brass sounds - don't get this if you don't love a good blast

 

[LL21]

have you tried this one?

Yes the fade-outs are the el-cheapo Chinese DVD player marking the track changes, probably nothing to do with the disc.

Here's my current favourite version of the Beethoven Diabelli Variations - the most intense reading I've come across : http://www.amazon.com/Variations-Diabelli-Op-120-Sokolov-Grigory/dp/B0000CNTJ1/

Its a live recording, there are a few audience coughs, nothing too distracting from the supremely focussed playing of Mr Sokolov.

 

[opus111]

No but thanks for the heads up on this - seems the lead reviewer at Amazon thinks the one I just mentioned is equally fine as Stephen Bishop's ( I still remember him by his old name which was the name he had when I was first introduced to him by my father).

 

[egidius]

diabelli

Yes the fade-outs are the el-cheapo Chinese DVD player marking the track changes, probably nothing to do with the disc.

Here's my current favourite version of the Beethoven Diabelli Variations - the most intense reading I've come across : http://www.amazon.com/Variations-Diabelli-Op-120-Sokolov-Grigory/dp/B0000CNTJ1/

Its a live recording, there are a few audience coughs, nothing too distracting from the supremely focussed playing of Mr Sokolov.

I used to be a complete crazehead for the Diabellis - funnily enough I like that reading along with one of the three Richter ones as well. Had a time with Ugorski, but in the end tend to come back to Brendel: he does it all, without ado.. much the opposite of Ugorski who not only does it all but keeps saying "look here I am doing it", a big no no in acting!!

 

[LL21]

I used to be a complete crazehead for the Diabellis - funnily enough I like that reading along with one of the three Richter ones as well. Had a time with Ugorski, but in the end tend to come back to Brendel: he does it all, without ado.. much the opposite of Ugorski who not only does it all but keeps saying "look here I am doing it", a big no no in acting!!

I have Richter in Prague which was very difficult to find and excellent (imho)...i also have Stephen Kovacevich...any thoughts on his Diabelli?

 

[egidius]

diabelli

I have Richter in Prague which was very difficult to find and excellent (imho)...i also have Stephen Kovacevich...any thoughts on his Diabelli?

YES! forgot that one. Indeed, the diabellis are so outrageous, that any pianist going for it, is to be hatoffed
There is very good sounding Konstantin Scherbakoff version on Naxos, and of course a mediocre sounding CD version of Serkin, who is actually one of my favourite pianists (I did a project on Adolf Busch, and will do a project on Serkin:the website www.adolfbusch.ch will have more details somewhere in April..) - I still hope to find sometime a better transfer of the CBS tapes.

 

[opus111]

Time for another de-recommendation, this time rather a surprising one. Normally I associate the EMI label with decent dynamics, here's an exception which sounds like it has been mastered with DSD. The date though suggests not (1999) - John Dunkerley is the recording engineer who normally is with Decca and has impeccable taste. Argerich and Perlman are of course top notch artists so I'm wondering how much the sucky dynamics has influenced the Amazon reviews which are decidedly mixed : http://www.amazon.com/Beethoven-Violin-Sonata-Kreutzer-Franck/dp/B00002CF0S/

@egidius - what label is Brendel's Diabelli on?