[Progisus]

I guess if a scaler goes further than 1M.. the DAC must be able to take more then 768K samplerate.. meaning todays generation will be stuck on its 1M setting, current Dave included.

Successors will always keep comin.. as is the drain on our wallets

Should still ve OK as you can have more taps but stay at 768k. With PGGB taps can be 150million (or more) with freg of 768.

 

[Reactcore]

Should still ve OK as you can have more taps but stay at 768k. With PGGB taps can be 150million (or more) with freg of 768.

That would be as looking at a 4k picture on a full HD screen.. i think with the passing of 1M taps the available OPSR on such device will go up too.

 

[ecwl]

Haha… Since we are all having fun speculating, based on what Rob Watts has accidentally said, this is my wild, wild speculation:
1) Choral style DAVE M-scaler that’ll actually still use the M-Scaler chip at 1 million taps. There may or may not be a slight tweak of the WTA1 and noise shaping algorithm
2) But the Choral DAVE M-Scaler will have a function to update the DAVE’s FPGA
3) So DAVE will turn into a dedicated receiver of 705.6kHz or 768kHz signal from Choral DAVE M-Scaler
4) With the new update, the only function of the DAVE would be a much more advanced WTA2 filter to upscale the 16fs signal to 256fs signal and maybe an even better noise shaper for the 20-element Pulse Array DAC
5) The Choral style DAVE M-Scaler has an option to convert the DAVE FPGA back to the original firmware so that if you want to sell It, normal people without Choral DAVE M-Scaler can use it.

Like I said before, because Rob Watts has been working on power pulse array amplifiers and ADC for a very long time, until I see a new product announced from Chord, I don’t really try to speculate too much as to when it’s coming out or what features it’ll have. On the bright side, because the quality of Rob Watt’s Chord products are so high, I usually just tell my dealer to pre-order ASAP when they’re announced as I don’t need to audition the product, unlike most other manufacturers.

 

[alxw0w]

1) Choral style DAVE M-scaler that’ll actually still use the M-Scaler chip at 1 million taps. There may or may not be a slight tweak of the WTA1 and noise shaping algorithm

I'm not a dac designer, my knowledge is limited so I might be wrong.
But there is no noise shaping when upsampling in M scaler.
You have to noise shape the data just before converting digital to analog, when you drop from 16bit to 5 bits (chord dacs being 5bits I believe)
that's when you have to dither (noise shape) the signal to avoid straight forward distortions - as you have to do it in any other sigma delta dacs.

I strongly recommend this channel with its digital audio fundamentals series (in this case: part 7, Dithering Explained). It helps a lot in understanding how the digital audio works - especially for non technical guys.

 

[Ronnie54]

Hello_ I thought I'd update my findings with HUGO TT2 and M-Scaler from my previous post when I though M-Scaler wasn't doing much at all.

As I have (2) Amps playing through 3 speaker systems simultaneously I shut down (2) speaker systems and concentrated on the best Imaging Speakers I currently have.

TT2 AND M-SCALER SOUND FANTASTIC !!

I won't go in too deep but in a nutshell my findings were the same as others with M-Scaler: Better bass-extension, increased fidelity (cymbals, etc)
sound-staging, realistic decay, forward vocals, musicality and a smoothness rivalling/equaling my Turntable setup.


I was then able to add in the other (2) speaker systems and by simply controlling the volume of each, dial it all in again until the balance I wanted was achieved.

To repeat what has been said before: It really IS hard to go back and listen to TT2 by itself... It is of course a great DAC but it's "greater" using M-Scaler.

I had one song that should have been well recorded: "Oh Holy Night" by Celine Dion. Regardless of the file type, it has always been a rather "Glary" pain in the ass to listen to (even with headphones). That issue was somehow resolved with TT2/M-Scaler. I'm not sure what was "repaired" and I guess it doesn't really matter but it finally sounds great.

A couple of side notes: I thought the ARYA headphones were great before but it's nice to know they also responded exceptionally well to the TT2/M-Scaler combo.

QUTEST with M-SCALER on MAXIMUM (White) sounds close to TT2 on it's best setting on my system Green (or Low).

Hans B suggested a reduction in time smearing may be reducing the chance for the room to resonate in a negative way (paraphrasing here) but I can listen now with QUTEST/M-SCALER or TT2/M-SCALER at LOUD levels without thinking I should turn it down. IOW, something has definitely changed for the better.


Ron

 

[Progisus]

That would be as looking at a 4k picture on a full HD screen.. i think with the passing of 1M taps the available OPSR on such device will go up too.

From what I understand taps refer to the length of the filter not the upsampling. A dac with 32fs should sound better theoretically. Listening to 48khz, 24b upscaled to 768khz 164million taps on TT2.

 

[Reactcore]

From what I understand taps refer to the length of the filter not the upsampling. A dac with 32fs should sound better theoretically. Listening to 48khz, 24b upscaled to 768khz 164million taps on TT2.

I go with that too.

But to benefit from an increased filter length.. the time between the samples can be shorten further.. coming closer to the exact transients of the original signal.

Why else do you think going up the filter increases the SR together on current HMS?

 

[ecwl]

I'm not a dac designer, my knowledge is limited so I might be wrong.
But there is no noise shaping when upsampling in M scaler.

https://www.head-fi.org/threads/watts-up.800264/page-114#post-15479799

16fs is noise shaped with 11th order noise shaper instead of dithering to 24-bit from 64-bit before being sent to WTA2. This is one of the things HQPlayer and PGGB imitated after Rob Watts mentioned it in this forum. I guess imitation is the best form of flattery to some software fans.

 

[Jawed]

https://www.head-fi.org/threads/watts-up.800264/page-114#post-15479799

16fs is noise shaped with 11th order noise shaper instead of dithering to 24-bit from 64-bit before being sent to WTA2. This is one of the things HQPlayer and PGGB imitated after Rob Watts mentioned it in this forum. I guess imitation is the best form of flattery to some software fans.

I think the 11th order noise shaper @Rob Watts was referring to was for Davina.

 

[Jawed]

I guess if a scaler goes further than 1M.. the DAC must be able to take more then 768K samplerate.. meaning todays generation will be stuck on its 1M setting, current Dave included.

This is wrong.

In upsampling you can use any number of filter taps to produce, for example, 768KHz samples from 48KHz original samples. That's 16x upsampling, which Rob generally describes as 16FS.

It's like calculating Pi: it's 3.14159. But that's only an approximation. Pi is also 3.141592653589793. But that's only an approximation. Pi is also 3.1415926535897932384626433832795. But that's only an approximation...

This is why Rob talks about using infinitely many taps in the ideal upsampling filter. Like Pi with its infinitely many decimal places, there's no limit to the number of taps you can use when upsampling.

 

[ecwl]

I think the 11th order noise shaper @Rob Watts was referring to was for Davina.

https://www.head-fi.org/threads/hug...official-thread.885042/page-651#post-15368410
Nope… Not DAVINA. HMS.

 

[Malevolent]

Whilst we're on the subject of a new M Scaler, I wonder if they'll implement a simple "On/Off" switch.

Thankfully, the M Scaler doesn't get very warm, but it's still a little worrying that it simply cannot be powered off.

 

[Reactcore]

This is wrong.

In upsampling you can use any number of filter taps to produce, for example, 768KHz samples from 48KHz original samples. That's 16x upsampling, which Rob generally describes as 16FS.

It's like calculating Pi: it's 3.14159. But that's only an approximation. Pi is also 3.141592653589793. But that's only an approximation. Pi is also 3.1415926535897932384626433832795. But that's only an approximation...

This is why Rob talks about using infinitely many taps in the ideal upsampling filter. Like Pi with its infinitely many decimal places, there's no limit to the number of taps you can use when upsampling.

Im not calculating on circles lol

I just mean to say if Rob increases his filter length further he'll want to use higher sample rates to get even more accurate than 1.3us sample distance of 768k.

Quote Rob:
"The 768k recordings I have sound very different to 384k recordings, and have the qualities that the M scaler gives; this suggests that there is something special about hitting the 1.3us accuracy of 768k. And certainly, the WTA 2 filter currently does not sound like the M scaler - it only seems to affect the ability to perceive the starting and stopping of notes, not pitch, timbre and instrument separation, which happens when you get to 768k.The 768k recordings I have sound very different to 384k recordings, and have the qualities that the M scaler gives; this suggests that there is something special about hitting the 1.3us accuracy of 768k. And certainly, the WTA 2 filter currently does not sound like the M scaler - it only seems to affect the ability to perceive the starting and stopping of notes, not pitch, timbre and instrument separation, which happens when you get to 768k."

To get even closer to the actual transients one must increase the number of samples further for the DAC to be able to decode the higher resolution.
I'm looking forward to the next generation exceeding 768k.. probably Dave2 being the first.

 

[Jawed]

Im not calculating on circles lol

I just mean to say if Rob increases his filter length further he'll want to use higher sample rates to get even more accurate than 1.3us sample distance of 768k.

Quote Rob:
"The 768k recordings I have sound very different to 384k recordings, and have the qualities that the M scaler gives; this suggests that there is something special about hitting the 1.3us accuracy of 768k. And certainly, the WTA 2 filter currently does not sound like the M scaler - it only seems to affect the ability to perceive the starting and stopping of notes, not pitch, timbre and instrument separation, which happens when you get to 768k.The 768k recordings I have sound very different to 384k recordings, and have the qualities that the M scaler gives; this suggests that there is something special about hitting the 1.3us accuracy of 768k. And certainly, the WTA 2 filter currently does not sound like the M scaler - it only seems to affect the ability to perceive the starting and stopping of notes, not pitch, timbre and instrument separation, which happens when you get to 768k."

To get even closer to the actual transients one must increase the number of samples further for the DAC to be able to decode the higher resolution.
I'm looking forward to the next generation exceeding 768k.. probably Dave2 being the first.

I've talked about an increased scaling for WTA 1, e.g. to 32FS or 256FS before, because it seems like it would be better. It is definitely a mathematically valid approach. But it is not required in order to get more than 1 million taps.

I'm sure Rob would like to have an FPGA that could do 1 billion taps at 2048FS, but...

WTA 1 at 16FS followed by WTA 2 at 256FS are compromises designed to make good use of the available FPGAs.

 

[musickid]

Photo of mscaler with qutest and anni on the qutest thread if anyone wants a look.