[SoupRKnowva]

This is another interesting interview with Rob



Some interesting tidbits. Ultima Dac will be dual chassis, separate power supply, but it will still be SMPS. but with more advanced RF filtering than before.

The advancements in the new choral scaler isn't just more taps. Rob had lots of time during covid to dig deeper into why the 5 variables in his windowing function (the WTA filter) were so sensitive. So with more work, he had made a breakthrough in, to me what seems like, potentially a new windowing function, less sensitive to the accuracy of the values? Thus allowing him to get even better performance without needing 10^17 taps? I sure hope so. a more fundamental breakthrough instead of more and more taps. He said the new scaler will have several large FPGAs. Larger probably means a step up from the Artix series to the Kintex line of FPGAs. this means a great deal more available logic cells and DSP blocks to work with.

All of this has me even more excited about the future choral scaler, I hope it's a 2023 product with the Ultima dac coming out later in 2024 or 2025. One thing I hope makes it into the choral scaler is to not have the garbled sound when switching sample rates on anything but USB input, that keeps me from using optical with mine.


The one thing I am still wondering is if the Ultima dac will require the new scaler, or if it will have the new scaling tech built in, for the cost it will be, I kinda hope it is the later.

 

[Rob Watts]

Carrying on the theme of "Balanced", I wonder if I may ask another question of Rob please?

- You mentioned a while back that, from the DAVE, a single-ended (i.e., RCA) output is superior to the balanced (XLR).

I understand this, but having just purchased a rather splendid Chord Ultima-5 power amp, I'm told by various people that this is a "balanced" design and would benefit from balanced input.

- What would be your recommendation for connecting DAVE to Ultima-5: RCA or XLR?

Using the balanced output will result in a small loss of transparency, as an extra amp stage has been added. But there are benefits in connecting via balanced in terms of common mode noise suppression. I have heard amps sounding much better when using balanced. Whether balanced is better depends upon the noise levels, earthing, siting of gear, the interconnects used, and the length of interconnects. For short connections, you are almost certainly better off using the RCA phono. For long lengths of cables, the balanced XLR may be better. Only way of telling is to try with an AB listening test - but remember that balanced has +6dB more gain, so you will need to turn down the volume by -6dB when using balanced.

 

[Reactcore]

This is another interesting interview with Rob

Nice summarizing. Thx for posting this!!
Makes me all the more curious to whats to come

 

[dcp10]

Using the balanced output will result in a small loss of transparency, as an extra amp stage has been added. But there are benefits in connecting via balanced in terms of common mode noise suppression. I have heard amps sounding much better when using balanced. Whether balanced is better depends upon the noise levels, earthing, siting of gear, the interconnects used, and the length of interconnects. For short connections, you are almost certainly better off using the RCA phono. For long lengths of cables, the balanced XLR may be better. Only way of telling is to try with an AB listening test - but remember that balanced has +6dB more gain, so you will need to turn down the volume by -6dB when using balanced.

Thanks, Rob. I'll try some listening tests, although I don't have the same type of cables for XLR and RCA.

I guess that a 1.5 m cable length would count as short (=> RCA?)

 

[Aslshark]

I prefer using Dave with RCA to my Gryphon Diablo 300. I have tested with very good XLR cables and the Gryphon is marketed for its balanced design as well. However the sound through RCA is super crystal clear. Love it.

I use a Gryphon Scorpio CD S as transport. The next big buy will be an m-scaler or something similar.

 

[Rob Watts]

Thanks, Rob. I'll try some listening tests, although I don't have the same type of cables for XLR and RCA.

I guess that a 1.5 m cable length would count as short (=> RCA?)

Yes. The time when I heard that balanced was a lot better was with cables that were greater than 5m long.

 

[dcp10]

Yes. The time when I heard that balanced was a lot better was with cables that were greater than 5m long.

Many thanks, Rob.

 

[Jawed]

@Rob Watts I’m very interested in your upcoming Choral Super Scaler and Ultima DAC. It sounds like the CSS is in the final stages of preparation for production. I guess that means the UDAC design is still in progress.

As far as I know, you plan to make it a two-box system: a power supply and a head unit.

I have some suggestions for the UDAC design:

• The power supply should have all the user interface features, such as a screen, knobs and buttons.
• The power supply should have all the digital inputs, such as optical, AES, BNC and USB.
• The head unit should have only a power connector and a digital audio connector from the power supply.
• The head unit should have only analogue outputs and a headphone jack.
• The head unit should have minimal openings on its surface to reduce RF noise interference from outside.

And here’s an idea that might surprise you:

• The power supply should include a Super Scaler, the same one as in the CSS.

What do you think of this idea?

I think that having a Super Scaler in the UDAC would make it a much more attractive product.

I also think that this would avoid a “product stack” problem that might arise after you launch both the CSS and the UDAC. When you develop DAVE 2, there is a chance that CSS + DAVE 2 will outperform UDAC if it doesn’t have a Super Scaler. People who follow your comments about future products might see this problem and hesitate to buy UDAC, knowing that they could get a better performance from CSS + DAVE 2 for a similar price.

Of course, UDAC with a built-in Super Scaler would cost more than UDAC without it, but it would be a worthy long-term product that would stay at the top of the market for 10+ years.

 

[adyc]

Please use USB chip XMOS for the Ultima DAC.

 

[GuiltyRocker]

Hmmmmmmmm, I wonder if Rob will take into account any of these suggestions.

 

[miketlse]

Deleted

 

[GuiltyRocker]

Arguably could already be too late.
If it is due to ship in say Q3, then the pre-production design could already be completed, and approx 12 months ago Rob was pointing out that the lead time for many components was of the order of 12 months. Once you factor in the time for Robs rigorous testing of those prototypes, and there is probably little time available for late design changes.
Even so Rob is known to pull the odd rabbit out of the hat.

Nice, can't wait to see and hear what he's cooking. I'm very happy with his current stuff, so it'll be interesting to hear the improvements.

 

[Jawed]

If it is due to ship in say Q3,

2024, not this year. Choral Super Scaler might launch this year, first, before Ultima DAC.

 

[miketlse]

Rob sometimes posts that he feels that FPGA development has slowed, and that Moores Law is weakening.

I have just finished reading 'The Intel Trinity' which includes the origins of Moores Law, and Intels success at maintaining it for several decades.
I found it an interesting book, but it was clear that the perpetuation of Moores Law did require consumers to feel the need to upgrade their PC/laptops/phones/tablets etc every two years. I can understand why this may not align nowadays with some consumers, concerned about the implicit exponentially increasing use of raw materials to fabricate ever more powerful chips, when the legacy i3 family of chips may be powerful enough for their needs.

It would be interesting to hear the reasons why Rob feels that Moores law is weakening.

 

[Rob Watts]

In one sense Moore's law has already broken - in that it was a doubling of transistors every 2 years or so, for the same die area, and the assumption being that the cost for that die would be constant - hence the cost per transistor (or FET) would half every two years. This rule was broken some time ago in that price per component is rising as the difficulties in going to smaller feature sizes means much more costs - indeed a 3nm mask cost is expected to be an extraordinary 500M to 1.5B USD according to here. Masks are unique for the design of chip that you want - so a new design needs another set of masks, and in this case another 1B USD or so. Scary numbers...

Mask costs for a 150nm was a relatively trivial 0.3M USD when I was working with silicon companies. At the moment 28nm is the sweet spot for silicon, in terms of price per transistor.

And of course being at 3nm we are now talking about 15 atoms of silicon (and before anybody comments node size and actual true feature size is no longer the case due to marketing lies), so Moore's law is approaching destruction anyway.