CHORD ELECTRONICS DAVE
Oct 19, 2016 at 10:49 PM Post #5,281 of 25,754
FWIW the MQA algorithms can reportedly easily determine whether or not Davina was part of the original recording chain. If MQA can do it...


How do you think MQA could determine this? Two completely different companies and approaches.
 
Oct 20, 2016 at 1:09 AM Post #5,282 of 25,754
 
  I don't feel we should get unduly depressed about the state of today's ADC limitations.
 
Much the same ADC solutions can result in fantastic SQ or terrible SQ - depending on the recording/mixing/mastering techniques employed. Meaning that the latter processes have a hugely bigger impact on the end result. So even with Davina in place (eventually), it will be wasted if the recording/mixing/mastering continue to mess it up big time.
 
So to follow on from ubs28's point, I think the term "recordings...are sorely compromised because of the limitations of all current ADCs" is somewhat overstating the case.
Any further increase in transparency in any part of the recording chain is, of course, a good thing, but I'm not holding my breath on this particular improvement

 
Sadly you also have to convince the engineers that Davina would be a lot better than the gear they already think is very good, i dont think they will all be rushing out to do more conversions from tape again. Then with mordern stuff how would one know if its been put through a Chord ADC... 
 
I wonder what PCM sample rate Rob Watts would recommend engineers to use on Davina, simple 24/96 or extreme 32/768 thats quite interesting.

 
I often wear two hat's simultaneously, which sometimes creates schizophrenia at worst, or a headache at best.
 
For sure 768 kHz or 705.6 kHz 24 bit is the best. You don't need 32 bits, as the noise shaper running at 768k is so effective guaranteeing > 350 dB performance. I have been listening to the Blu mk 2 prototype, and whilst coding for it I wanted to add some tests required actually for Davina. This had 768 24 bit output, and it was either noise shaped, or triangular or Gaussian dithered.
 
The noise shaped is technically the best, as I have proven that it can reproduce -301 dB 6 kHz signal perfectly. The other schemes rely on randomization to encode small signals and I wanted to see if there were any losses in employing dither rather than noise shaping. The triangular had the worst sound quality - poorer depth, and detail resolution. Gaussian was better, but the noise shaper was easily the best sounding. Now that's just the issue of how we handle n bits (in this case 54 bits) down to 24 bits, we also have the losses of converting to 44.1/48 K then WTA reconstruction back to 706/768 kHz.
 
So I could simply say lets run at 768 kHz 24 bits and to hell with it. This is where my second hat comes on, and that's me as a music lover. I have 4TB portable hard drive, and its full. It is almost all redbook, and unless we get peta portable hard drives I need 16 bit 44.1 to work acceptably; so I absolutely need small file sizes to work as well as possible, and a lot of the focus on the Davina project will be to find out how much of a SQ loss we actually get, and how to minimize it. 
 
As too recording engineers - well the ones I have met have been passionate about getting the best sound possible. Perhaps its a skewed sample, as recording engineers who wish to talk with me might be more passionate about SQ anyway... But in one sense whether Davina tech takes off big time or not in the pro world is not what motivates me; I can see glaringly obvious problems with modern ADC's and I simply just want to fix those problems. 
 
Rob
 
Oct 20, 2016 at 4:46 AM Post #5,283 of 25,754
I can see how a lot of what you learn from the Davina project wil benefit Blu and Dave 2, not to mention their siblings Rob. It clearly validates the project to my mind.

Fwiw I think your average mastering engineer cares passionately about sound quality. That's why very few ADC brands are present in the worlds top mastering houses. Their reputation relies on using only the best.
 
Oct 20, 2016 at 5:01 AM Post #5,284 of 25,754
Hey, congrats! a little bit OT, but right now I am powering my DAVE with the WireWorld Silver Electra 7, which I find beats the snot out of every PC I've ever tried, including the High Fidelity Cables stuff, all for the ultra low price of $500. I was totally shocked at how well it brought out the imaging, depth and resolution of my DAVE. It was the first PC that I had ever heard that did exactly what it said it would (and this is after being on The Cable Company loaner program for a while now). So, if you know, I'm wondering how the Platinum compares, since that's quite a large jump in price from the Silver to Platinum in price.

Hi, I'm looking for some worthwhile power cable upgrad options for my Dave (still using stock) so am interested in your recommendation of the Wireworld cable.

I've not listened to many cables so far, just a Shunyata Alpha analogue (good punchy base, and clean and dynamic overall) and an AudioQuest NRG-10 (warm attractive bass but takes too much away from the top end); HFC pc cables are not available in the UK to demo. I looked up one review of the Silver Electra (http://www.6moons.com/audioreviews/wireworld2/wireworld_3.html), which seemed to confirm the points you made about imaging, depth and resolution but it did suggest that some might feel it was lacking some "weight and bottom-end grunt". Would you agree with that? Thanks.
 
Oct 20, 2016 at 5:03 AM Post #5,285 of 25,754
Dear Rob
 
I have been reading about Dave and Hugo on your various posts, so some key differences I could pick up are
 
- Noise Shaper for Dave is at -350 DB resolution compared to -200 DB for Hugo
- The WTA filter for Hugo is as 16 FS, but for Dave it is 256FS WTA filter. Does Mojo also have 16 FS WTA filter?
- Analog stages for Dave vs Hugo, some differences.
- Pulse Array 20 E for Dave and 4E for Hugo
 
You previously posted on your thread watts up "Now replacing the WTA from 16FS (data every 1,417 nS) to 256 FS (data every 89 nS) - - it is a very subtle difference, but was nonetheless extremely audible."
 
I guess all this means that the Dave is really another league of performance compared to Hugo / Mojo, and it was easily audible to my ears, as you said the Dave has better timing.
 
I'm sure many would be interested to know more such details about what separated Dave from Hugo/Mojo.
 
Recently someone posted, that one cannot really gain an understanding of Dave based on the performance on Hugo, having heard both, yes I agree with this statement.
 
You also posted 
 
"The Dave project allowed me to understand exactly what I had stumbled upon, and in the case of Dave, further maximize it. Now the job of a DAC is to converted sampled data back into a continuous waveform exactly as was in the ADC converter, and I had improved the filters within Hugo that go from 16FS to 2048FS - this meant that I had recreated the analogue waveform in the time domain to a much better accuracy than before, and it was this better accuracy that gave the subjective improvements. I had done this in order to improve jitter sensitivity, reduce RF noise levels, all to reduce noise floor modulation, which makes a DAC sound smoother - but it also had these subjective timing benefits.

 

The filtering was a three stage digital filter, and means I can recreate the analogue waveform accurately to a 9.6 nS resolution. All other DAC's work to a resolution of at most 16FS, which is only 1.4 uS. Moreover, getting to this resolution is not good either, as they have very limited tap lengths so it has gross timing errors too. The fact that I have very long tap length WTA filters, plus the fact that filtering is at 9.6nS resolution, gives Mojo this unique timing performance - and its that, above all else, that gives it its musicality."

 
I guess my question is, I'm about confused about what these numbers mean, the 9.6 nS resolution you mentioned compared to the WTA filter resolution you mentioned for Hugo 1,417 NS and then 89 Ns (this I think means before the 2048 FS upsampling). I think the 9.6 probably means after the 2048 FS upsampling, but some more detail from you would be much appreciated.
 
Does it mean, ultimately, Dave, Hugo and Mojo all have 9.6 Ns resolution?
 
Thanks and Regards
 
Oct 20, 2016 at 6:42 AM Post #5,286 of 25,754
Yes all my DAC's filter up to 9.6nS, but they are different to how its done.
 
So Hugo/Mojo is:
 
WTA 16FS > Linear interpolation filter to 2048 FS > Low pass Filter > Low pass filter > noise shaper.
 
The low pass filter is set to 260 kHz, and is mathematically the same as a simple RC low pass filter.
 
Dave is:
 
WTA 16FS > WTA to 256FS > Linear interpolation filter to 2048 FS > Low pass Filter > Low pass filter > noise shaper
 
The addition of the extra WTA filter adds a surprising amount to SQ - surprising because from a HF error point of view there is not much difference between them; but from a time domain POV there is an important difference - the addition of the WTA to 256FS gets us closer to the ideal sinc impulse response. And the closer it gets to ideal, the closer the DAC gets to reproducing the signal in-between samples. What is interesting is these very small differences are very audible and account for Dave's ability to portray the starting and stopping of notes properly.
 
Rob
 
Oct 20, 2016 at 7:53 AM Post #5,288 of 25,754
In which universe can anyone measure -350dB signals, let alone hear anything like that?
What's the noise floor in a typical home listening room?
What's the dynamic range in which a speaker or headphone operates w/o damaging the listeners ear drums?
What's the dynamic range and noise floor of microphones?
... something has to first capture the analog waves of sounds before and super duper WTA What filter can the -350dB magic.
 
This all sounds a little over the top of marketing soft brown substance, at least to me.
rolleyes.gif

 
Oct 20, 2016 at 8:15 AM Post #5,289 of 25,754
In which universe can anyone measure -350dB signals, let alone hear anything like that?
What's the noise floor in a typical home listening room?
What's the dynamic range in which a speaker or headphone operates w/o damaging the listeners ear drums?
What's the dynamic range and noise floor of microphones?
... something has to first capture the analog waves of sounds before and super duper WTA What filter can the -350dB magic.

This all sounds a little over the top of marketing soft brown substance, at least to me.:rolleyes:


http://www.head-fi.org/t/766517/chord-electronics-dave/4710#post_12859578
 
Oct 20, 2016 at 8:55 AM Post #5,290 of 25,754
Absolutely this performance is measurable. This is the FFT from data coming out of Dave's noise shaper:
 

 
This confirms that the noise shaper accurately resolves a -301 dB 6 kHz signal.
 
My contention, based on my own listening tests, is that this performance is required to reproduce depth correctly. Now I can see that making claims that the brain needs this level of accuracy as being absurd; how can the ear/brain be so sensitive? Indeed, I was in two minds about whether I should post about these results, as its easy to ridicule them. But I took the view to follow the evidence, no matter how absurd it may appear, and I stand by my claims, based upon repeated listening tests, that the digital signal path must have absolutely no error (no matter how small) in small signal amplitude linearity in order to reproduce depth correctly.
 
Actually some of the listening tests I have planned for Davina actually will be about this issue; I will do depth recordings, and some will be with Davina using 350dB noise shaping, and some will be with less capable noise shaping, as some of the integrators will get turned off. Now if it transpires that nobody can hear the difference then I will eat my proverbial hat.
 
Rob     
 
Oct 20, 2016 at 9:32 AM Post #5,291 of 25,754
Hi, I'm looking for some worthwhile power cable upgrad options for my Dave (still using stock) so am interested in your recommendation of the Wireworld cable.

I've not listened to many cables so far, just a Shunyata Alpha analogue (good punchy base, and clean and dynamic overall) and an AudioQuest NRG-10 (warm attractive bass but takes too much away from the top end); HFC pc cables are not available in the UK to demo. I looked up one review of the Silver Electra (http://www.6moons.com/audioreviews/wireworld2/wireworld_3.html), which seemed to confirm the points you made about imaging, depth and resolution but it did suggest that some might feel it was lacking some "weight and bottom-end grunt". Would you agree with that? Thanks.


I would not agree with that. In fact, quite the opposite as I've been using HD800s, and depending on the recording there is plenty of articulate umph in the lower registers. It's not like these cables are cutting off any of the signal. They are the purest PC I've ever heard, so with the DAVE if you expect pure, the pairing is very much the sonic characteristics of the stock cable, yet expanded to bring out the best in the DAVE.
 
Oct 20, 2016 at 11:20 AM Post #5,292 of 25,754

The point I was trying to make is that each ADC apparently has its own "fingerprint," if you  will, a set of characteristics that MQA is able to determine EVEN IF the ADC was not documented in the engineer/producer/record company's files. If MQA can do this then it means that it is no longer a problem without a solution.
 
This is in response to @Deftone's question: "Then with mordern stuff how would one know if its been put through a Chord ADC... "
 
Oct 20, 2016 at 12:28 PM Post #5,293 of 25,754
 
The point I was trying to make is that each ADC apparently has its own "fingerprint," if you  will, a set of characteristics that MQA is able to determine EVEN IF the ADC was not documented in the engineer/producer/record company's files. If MQA can do this then it means that it is no longer a problem without a solution.
 
This is in response to @Deftone's question: "Then with mordern stuff how would one know if its been put through a Chord ADC... "

 
To be able to do so in anything even approaching a deterministic manner (and anything less is, at best, guessing), it would need a reference to a parallel recording with known ADC characteristics in order to compute the differentials for the undocumented ADC.  You can, in (mathematical) theory, use other proxies for the missing ADC documentation, but all of them require additional, more invovled, information in order to function.
 
Oct 20, 2016 at 1:23 PM Post #5,294 of 25,754
I would not agree with that. In fact, quite the opposite as I've been using HD800s, and depending on the recording there is plenty of articulate umph in the lower registers. It's not like these cables are cutting off any of the signal. They are the purest PC I've ever heard, so with the DAVE if you expect pure, the pairing is very much the sonic characteristics of the stock cable, yet expanded to bring out the best in the DAVE.

Thanks. I think I can get a loan of a Silver Electra so will give it a go.
 
Oct 20, 2016 at 2:30 PM Post #5,295 of 25,754
   
To be able to do so in anything even approaching a deterministic manner (and anything less is, at best, guessing), it would need a reference to a parallel recording with known ADC characteristics in order to compute the differentials for the undocumented ADC.  You can, in (mathematical) theory, use other proxies for the missing ADC documentation, but all of them require additional, more invovled, information in order to function.


There is a fairly short list of ADC's used by engineers in the recording industry. If one, for example, looks at the extensive Neil Young catalog you have a large number of records made with the highly popular Pacific Microsonics ADC. Once Chord releases recordings made with Davina there will similarly be a Davina database to be analyzed.
 

Users who are viewing this thread

Back
Top