1. This site uses cookies to help personalise content, tailor your experience and to keep you logged in if you register.
    By continuing to use this site, you are consenting to our use of cookies.

    Dismiss Notice


Discussion in 'High-end Audio Forum' started by magiccabbage, May 14, 2015.
147 148 149 150 151 152 153 154 155 156
158 159 160 161 162 163 164 165 166 167
  1. romaz

    Thanks for the detailed explanation, Rob. Just to make sure I understand, are you saying that with the DAVE, I can attenuate down to -385 dB (inaudible levels?) with no compromise to the waveform amplitude meaning no compromise in dynamics, even with a 24 bit file?

    Based on your opening statement, I am also inferring that it is not possible to create a preamp that would go after the DAVE that could ever achieve this level of performance, is that correct?
  2. Rob Watts
    Hmm - not quite, as it would be noisy at -385 dB! The max attenuation is -75 dB. So a 24 bit ideal signal would be -144 plus -75 so it would now be -219 dB. The truncation noise is at -350 dB (about) so its still better than this by 131 dB.
    The problem we have with this is that if we use a noise shaper within the truncation (and that's the best way of ensuring no extra noise), then the noise shaper degrades the amplitude of small signals - indeed, signals that are smaller than the resolution of the noise shaper (that's the noise shaper noise floor) is lost forever. So as a small signal gets closer to the noise shaper resolution, then small signals get progressively attenuated. This is the source of the small signal amplitude non-linearity, and mathematically it is identical to the non linearity that happens with metal/metal interfaces - and it sounds exactly the same, poor depth reproduction.
    The really interesting thing about the Dave project is that I now have a number for how well a noise shaper needs to perform - and that's 350 dB - and that's actually a reflection of how well I can do with noise shaping. Its possible the human limits on depth perception is even smaller than this. And this is the truly amazing thing about this problem - the implication is that there is no limit to how accurate small signals need to be. And frankly, I have great difficulty with this idea, that the brain can be so sensitive, that zero error is what we need. But I can only report what I can hear, and there is definitely something very weird about depth perception.
    Chord Electronics Stay updated on Chord Electronics at their sponsor page on Head-Fi.
    https://www.facebook.com/chordelectronics https://twitter.com/chordaudio http://www.chordelectronics.co.uk/
  3. romaz

    Understood, thanks. This is quite amazing. Even with something like a 32 bit signal, I could maximally attenuate the DAVE down to -75dB and still have zero signal degradation! I can see why it would make no sense at all to add a preamp after the DAVE.
  4. Christer
    While Dave remains the best  consumer Dac  I have heard  I must  again  confess that I am also  looking at all other cheaper options before dishing out the fortune Dave would cost me. 
    Over at Computer Audiophile there seems to be a lot of talk going on around the new german-made  T&A DAC 8 DSD.
    Which is going to be demoed at Munich Highend Show in May.
    But very little, next to nothing at all about DAVE?
    I posted my first DAVE impressions there too and got basically no response at all.
    All the rave over there seems to be about the T&A especially its DSD which seems to have a  native ie a true 1bit converter that works all the way up to DSD 512 and can also be used to upconvert pcm to DSD 512 whatever that implies for SQ?
    The HI FI News review says that DAVE also plays DSD natively. But does it really?
     One of the things that DAVE made even clearer to me via both HD800 and HE1000, than before was that Acoucense's Mahler 5 in 24/192 sounded even better than Channel Classic's DSD 64 version via DAVE than via  both my Hugo and Mojo.
    Is there anyone here who has already heard the T&A DAC8 DSD?
    I remember T&A mades some quite good SACD players.But I haven't heard any of their recent DACs.
    In Europe it seems to sell for around 2700€ which makes it a lot more affordable than DAVE!
    But is it even anywhere near Dave SQ wise?
    According to some at Computer Audiophile its DSD 512 is quote: " entirely in another ballpark".
    I don't pretend to understand all the technical info Rob Watts is providing regarding DAVE nor all that is provided in this pdf :
    But for those who might understand more than me and draw more  conclusions than I am capable of here it is.
    My final verdict is always through listening how close any product gets to the real thing acoustic music as I am used to hearing it live.
    And so far DAVE is the closest digital  approach I have heard.
    But tomorrow morning I will once again  re-calibrate against the only true  reference from early morning until late night,with lunch and a few hours rest in between.
    And no romaz not to worry, it won't be AC/DC.
    According to some news they cancelled a concert  recently because a doctor, you romaz? warned that at least one member of the group risked turning completely deaf if he did one more concert!
    No such risks  with Szymanovsky's 4th  and Tchaikovsky's 5th.
    Well possibly if you stand right in front of  the brass which I did once at some sessions,luckily during some mezzoforte passages but still.
    But I am expecting plenty of real dynamic range and depth perception even from good midhall positions.
  5. ecwl
    Well, I think the T+A brochure says it all. For DAC8, PCM music is played through 4 Burr-brown DAC chips. DSD is played in a true native fashion with a selectable analog filter of 60kHz (for DSD64) or 120kHz (for DSD512). The implementation would be like Emm Labs/PS Audio Perfectstream/Playback designs IF these DACs just let the DSD64/128/256/512 signals pass through without any FPGA conversion.
    As to whether something truly plays DSD "natively", I like this article's definition:
    So according to that artilce, T+A DAC8 would play DSD files in a a true Direct Native DSD fashion whereas Chord DAVE or ESS Sabre DAC chip-based DACs would play DSD files in a non-direct Native DSD fashion. However, there is a difference with DAVE vs Hugo/Mojo/Sabre because the DAVE with the DSD filter does not decimate, meaning the signal probably goes from 1-bit/2.822MHz to 24-bit/2.822MHz for further processing whereas Hugo/Mojo and presumably Sabre DACs decimates meaning the 1-bit/2.822MHz is converted to 24-bit/704kHz (I don't actually know the exact numbers) first before further processing.
    The Computer Audiophile comments are unique in that some of the people on the forum are pretty intense on using their computer to upsample files (e.g. using software like XXXHighend or HQPlayer). So essentially, what they are doing is taking their CD files or DSD64 files and then using their computer as if they are Chord DAVE's FPGA and converting the CD/DSD64 files into DSD512 with their own digital filter/noise-shaping/volume control algorithms. Since you can virtually not find any native DSD512 files available for playback, that's really what those guys are talking about. And then they're sending the DSD512 files into the T+A DAC8 for playback. So if you really want a fair comparison between their DSD512 T+A DAC8 playback with Chord DAVE's DSD playback capabilities, you need to use the same software they're running and convert whatever music into DSD256 and feed it to Chord DAVE via the DSD filter and see what happens. Personally, I think this is pretty insane. And I'm not convinced even in this kind of setup, Chord DAVE would not beat the T+A DAC8 hands down. And I also think if you just feed Chord DAVE the actual CD/DSD64 file, you'll get better sound than tweaking your own computer to upsample it to DSD256 and sending it to Chord DAVE.
    Obviously, actually testing the T+A DAC8 is the best way to evaluate how it sounds in your particular use case. But as Rob Watts said, every DAC technology design has its own fundamental limitations so while designers can optimize their DACs based on Burr-brown chip designs or they can optimize their DACs based on 1-bit true DSD playback at 2.822MHz up to 22.576MHz (DSD512), or based on ladder DACs, you can only optimize the designs to a point and run into the fundamental limits of that particular technology. From what I understand from what Rob Watts is saying, the main advantage of DAVE and in fact of Hugo/Mojo too is that the FPGA/WTA/noise-shaping + Pulse Array DAC designs are not as limited as multi-bit PWM DAC chips/ladder DAC chips or straight DSD/PWM in terms of fundamental performance which is why the sonic reproduction would always be more accurate and as a result, in most DAVE Head-Fi forum commenters' opinion, more sonically rewarding experience.
    JaZZ likes this.
  6. romaz

    I heard this new DAC along with its accompanying headphone amp at the T+A room at CES in Las Vegas in January. It was connected to an HD800S. Because I had my HE1000 with me, they let me listen to my HE1000 with it also. Because this was my first time listening to the HD800S, I spent a fair amount of time there, more than I would have otherwise. They did have a good orchestral track available in addition to some standard studio vocal tracks. Prior to coming into this room, I had just spent a fair amount of time at the Chord room listening to the DAVE. I can assure you, this DAC is not in the league of the DAVE, not really even close.

    Oversampling to DSD is not new. The Directstream and the Nagra HD both do it. The T+A is closer in presentation to the Directstream, which I know well because I used to own it. The Nagra is in a whole different league. It is smoother and laid back yet more resolving and just more engaging in its presentation compared with the drier presentation of the other two. Neither of these DACs have the holographic abilities of the DAVE, however.

    The DAVE oversamples also but well beyond DSD512 and for different reasons than most. Unlike MQA, for example, which oversamples to address the ringing artifacts introduced by the ADC, Rob did not feel this was important. Here was his response to me:

    "No I over sample to 2048 FS, or a new filtered sample every 9.6 nS.

    The first WTA stage is 16FS. Then the next WTA stage is at 256 FS. Then a three stage filter then takes it to 2048 FS.

    Its done for a number of reasons - to reduce the timing of transients uncertainty problem, to enable the noise shapers to work at 104 MHz so that the noise shapers can reproduce depth correctly, and finally to allow no measurable noise floor modulation.

    So there are a number of reasons why I oversample to such a high rate.

    ADC ringing artifacts is not one of them, as that is irrelevant."

    At CanJam the other week, I asked John Franks if they spent much time listening to their competition and his response was "No." Apparently, this is not their practice at all. As ecwl stated, Rob, having been in this business a long time, is intimately familiar with all current DAC technologies (R2R, delta sigma, etc.) and he was incapable of overcoming their inherent limitations (i.e. R2R is too slow to effectively oversample) so he felt he had to design his own DACs from scratch via FPGA to create what he wanted. Now, it is up to the objective listener to decide if he has succeeded.
    paulchiu likes this.
  7. paulchiu
    Yes.  In my use of the Nagra HD DAC, the analog sound from its headamp is what I like most.  It is pretty much sticking a headphone straight into a top quality record player.  These days, such a phone setup runs into the 50-100K.
    Now, this year, I am looking for more of that soundstage the Dave promises. 
  8. Sunya
    With digital VCs isn't also the analog noise floor of the DAC important even if the digital noise floor is at -385dB? I mean, as you attenuate digitally the signal gets closer to the analog noise floor of the DAC which remains constant, meaning reduced SNR. The quoted 127dB dynamic range translates to a little over 21 bit analog performance which would give you 5 bits of attenuation or 30dB before starting to lose resolution for 16 bit signals. But what happens with a 24 bit signal? Just those 3 extra bits over the analog resolution of the DAC to throw away for VC before starting to affect the sound quality?
    Articnoise likes this.
  9. TheAttorney
    I've dabbled with HQPlayer converting redbook FLAC files into DSD into my iFi iDSD. Even with that low cost portable player driving my trusty old HD600's, the superiority of HQPlayer over JRiver was very obvious at any upsampling/filter option, but converting to 256DSD and upwards really made this combination sing for me, although some on the iFi thread preferred highly upsampled PCM to any DSD. For me though, DSD just sounded more natural and lifelike.
    The trouble was, some of those HQPlayer options put a real load onto my laptop's processor causing the fan to kick in horribly. And the higher rate DSD conversions were by far the worst for loading the processor. Also, the HQPlayer's user interface is an acquired taste, so it can all be hard work to get the best, but what it did prove to me was that manipulating the original redbook file can work wonders.
    The more adventurous that liked HQPlayer thought that the exotic and obscure BugHeadEmperor was even better. I never tried this one because just reading the summary of the user instructions gave me a headache. Apparently BHE fundamentally reconstructs the original file, rather that merely upsampling and filtering, but I didn't delve deeply enough to really understand what it was all about. It's a strange world out there.  
  10. esimms86
  11. paulchiu
  12. Christer

    Thanks for your input again romaz.
    I did not really expect T+A to beat or even be as good as DAVE. And having read your posts and credentials here I trust your opinions almost as my own. Just my  wishful thinking and at least hoping I could save some money and still be in DAVE territory.
    I guess I will just have to dig very deep into my sparse resources and bite the  sweet&sour DAVE apple in the end too.
    Regarding the technical bits you quote from Rob "every 9,6 nS"  should it be read as nanoseconds?
    Regarding your question whether Chord listen to the competition, I would say I wish more  DAC designers than Chord, listened to real live acoustic music instead  of tuning and trying to  cover up faults in  their gear according to expected subjective preferences among their customers.
    Oh, they should have heard this morning's Tchaikovsky 5 rehearsal . Now that was truly in a league well beyond any HIFI experience.
    That'd give them something real  to strive for.
  13. rkt31
    digital volume control is there in mojo and Hugo too. mojo even at very low volumes does not lose resolution as it is being enjoyed by very sensitive iem users too. so Dave with so much more processing power should have no problem
  14. Christer
    Thanks for your clarifications too ecwl.
    Just  another case of wishful thinking and false hope  from me then, I suppose.
    There as here, there are few posters with any real reliable references imho.But what got me interested apart from the price issue.Who wouldn't want DAVE performance for 2700€?, was that some of those who have both real references and access to live and masterfiles of acoustic music seem to be a bit  interested in the T+A DAC8DSD.
    At Computer Audiophile there are some hardcore DSD believers possibly sticking to the last chance? for DSD to prove itself superior to HD PCM. which seems to be with  DSD 512. Although DSD 1024 is already talked about in some quarters.But as you say there is little or no available natively recorded source material even  at DSD 512.
    I am not really interested  in upsampling low res material in my computer. All else equal as I have said and Rob confirmed recently, native HD is the way to imo. But there seem to be quite a number of both DSD 128 and 256 releases of acoustic music available from several download sites. Most notably for me on classical sites like native DSD.com and the new site Spirit of Turtle which sells both some native DSD recordings , where many recordings from Challenge Classics can be downloaded  at all resolutions from DXD native to 16/44.1 and  DSD 64,128 and 256.
    I am  certainly no "hardcore DSD believer" any longer. But I used to be, until I heard DXD raw at sessions.I had already heard DSD64 raw. But I have to say that both DSD 64 and DSD 128 which is what I can play now and have heard via Dave, can  also sound very musical  and sweet.
    And yes I would still like to know if Rob Watts when he says DSD is fundamentally flawed,means  at all present and future DSD rates, not only DSD64 and 128 which if I remember correctly, he still finds too noisy?
    Cheers Chris
  15. ecwl
    I think Rob Watts has answered this question a few times in a few different ways in this forum and in the Chord Hugo forum. But I have to admit I have read his answers multiple times and I'm not 100% sure if I fully understand it. But maybe me trying to explain it in the simple way that I understand the DSD data format issues may help clarify things for Christer.
    As a file format, 1-bit/2.822MHz DSD has two fundamental problems that in Rob Watts's mind is inferior to PCM, even possibly 16/44 PCM. One issue is that if there is a very low level signal, in PCM, even though people say it is a 16-bit file format and theoretically should have 96dB dynamic range, in reality, you can dither to a much lower signal level and we know that Rob Watts, and most DAVE "believers/lovers/admirers" believe that these theoretically inaudible low-level signals are what actually creates additional depth and musicality and accuracy to the music. When you record music in DSD, because you only have 1-bit to work with, you have already dithered the music in 1-bit. which is what he meant that DSD is already noise-shaped. So people calculate that DSD is the equivalent of 20-bit 44kHz PCM except even with the best noise-shaping, you will never get beyond the 20 bits because the noise shaping is baked into the DSD file. You cannot recover what you've lost in the recording. Whereas it's possible that some 24-bit dithered low-level signals are preserved in the 16-bit/44kHz files. Since these are ultra-low-level signal levels, they are not going to manifest as noise, they should according to Rob appear in an accurate DAC as better soundstage depth and better timbral accuracy of voices and instruments.
    The second fundamental problem that I think Rob Watts has with DSD is that if you have a dynamic transient, you can have the 16-bit dynamic range at 44kHz in PCM. Meaning that you can jump from no signal to full signal (65535 = 96dB) in 22.7ns. With DSD, you simply cannot record this jump because you're fluctuating 1-bit at a time so even if you record an increment at 0.3ns intervals, after 22.7ns, you're only going to get to 64 (= 36dB), which is much less than 65535. So you can imagine then DSD would sound soft and the timing and transients would sound poor as a result.
    With the DAVE, I do find that really good 16/44 recordings sound better than really good DSD recordings with respect to soundstage depth, timbral accuracy and particularly timing and transients. Is it because I'm imagining all of this because of what Rob Watts said and I was already biased? Not sure. However, there are just a lot more bad 16/44 recordings than bad DSD recordings so on average, it feels like DSD sounds better when in fact it probably does not as a file format.
    That said, I may be completely misunderstanding what Rob Watts is saying. But this is my layman's understanding and explanation of it. I have been known to be corrected about a lot of these technical aspects. I think sometimes, I truly misunderstood the concepts but other times, my simplification of the concepts becomes too simple that more technical people are a little offended by the inaccurate (or slightly-off) representation of reality. Please feel free to correct me if something I said is incorrect.
147 148 149 150 151 152 153 154 155 156
158 159 160 161 162 163 164 165 166 167

Share This Page