Chord Electronics Hugo M-Scaler
Status
Not open for further replies.
Jul 25, 2018 at 5:53 AM Post #18 of 34

Rob Watts

Member of the Trade: Chord Electronics
Joined
Apr 1, 2014
Posts
2,597
Likes
7,673
Will the presentation slides be up on here for the hugo TT 2 for the ppl who couldn't go the show?

Yes as soon as the official thread goes live. Chord are saying later today.

Am sold Rob, I got a deal earlier today and have bought both, TT2 and MScaler. I have no idea when they will ship, but the trigger has been pulled.

Both will be used for headphone listening, one question though, in the TT2 box or MScaler box, is there cables to attach to one another or do I have to source them myself ?

Cheers

I presume so; I will contact Chord to see what will be there.

With the scaler does that now negate the need for music above 44.1/96kHz? Qobuz hires downloads of no advantage now and 192kHz music?

Also Tony i've heard our PM1's will be collector items in the future now oppo have discontinued their audio manufacturing.

The referral to master tape has made me see the scaler in a totally different light tonite. Wow.

All my listening has been with SRC (sample rate conversion) in the way - so a 44.1 file has been damaged by the SRC from 96. And a perfect interpolation filter can't repair the damage (added distortion and noise), so the 96k will always be better.

But what about 88.2 down to 44.1? Well that's currently has problems due to aliasing, and insufficient bandwidth limiting too. But if you had a perfect aliasing performance, then that would be a very interesting test, and it's one I plan to do with the Davian project.

Rob
 
Jul 25, 2018 at 6:58 AM Post #19 of 34

paul2qute

Headphoneus Supremus
Joined
Feb 1, 2017
Posts
2,477
Likes
1,107
I've only just realised I'm on the M scaler thread now lol, be great to see the slides of the hugo TT 2 though, it's great the M scaler thread is up and running now, exciting times
 
Jul 25, 2018 at 7:04 AM Post #20 of 34

bidn

100+ Head-Fier
Joined
Aug 14, 2016
Posts
390
Likes
547
Location
Kingdom of the Netherlands
Yes as soon as the official thread goes live. Chord are saying later today.



I presume so; I will contact Chord to see what will be there.



All my listening has been with SRC (sample rate conversion) in the way - so a 44.1 file has been damaged by the SRC from 96. And a perfect interpolation filter can't repair the damage (added distortion and noise), so the 96k will always be better.

But what about 88.2 down to 44.1? Well that's currently has problems due to aliasing, and insufficient bandwidth limiting too. But if you had a perfect aliasing performance, then that would be a very interesting test, and it's one I plan to do with the Davian project.

Rob

Dear Rob,

Thank you for creating the M - Scaler with a practical form factor :) !

I have a question:
What difference, and why, in SQ would there be between the following audio chains:

- PC → USB → digital upscaling with M-Scaler → digital signal → DAC

- digital upscaling with PC → digital signal → DAC ?

Because the M- Scaler upscales and outputs a digital signal, I don't understand how the output upscaled digital signal would differ from the same produced by software running on a fast PC.

Thank you for your excellent work,
bidn
 
Last edited:
Jul 25, 2018 at 7:14 AM Post #21 of 34

Triode User

Member of the Trade: WAVE High Fidelity
Joined
Feb 5, 2017
Posts
3,458
Likes
2,949
Location
Leics, UK
All my listening has been with SRC (sample rate conversion) in the way - so a 44.1 file has been damaged by the SRC from 96. And a perfect interpolation filter can't repair the damage (added distortion and noise), so the 96k will always be better.

But what about 88.2 down to 44.1? Well that's currently has problems due to aliasing, and insufficient bandwidth limiting too. But if you had a perfect aliasing performance, then that would be a very interesting test, and it's one I plan to do with the Davian project.
Rob

Rob, You may (or probably not - you are or should be busy with other projects!) but I have posted my own experience of preferring a 48 file to the 44.1 version (assuming they are from the same source file) so I always look out for that or even the 96 or 192 versions if available.
 
Jul 25, 2018 at 8:16 AM Post #22 of 34

Rob Watts

Member of the Trade: Chord Electronics
Joined
Apr 1, 2014
Posts
2,597
Likes
7,673
Can someone explain how this works.. how do you "add-in" new information into a 16bit 44.1khz signal from a CD transport? - I love Chord Electronics and love my Qutest - but what exactly will this add to the signal path?

Interesting post - I am actually not adding in any more information - the information is all contained in the original bandwidth limited sampled signal. but what we have is sampled data, and what we need is a continuous un-sampled signal - and we need to do a huge amount of processing to do this without error. I will give you an example. Imagine a sine wave. You can state it's a sine wave; give its frequency and its amplitude. So the information content is fixed; but if you want a waveform of infinite length, and precision, then you would need an infinite amount of processing to create the infinite number of points. And with the sampled data, we can convert it to a continuous signal - with exactly the same information content - and recover the original bandwidth limited continuous signal - if and only if you do an infinite amount of processing and use an ideal sinc function interpolation filter. So I am not trying to create new information - actually we are converting from a sampled bandwidth limited signal to a continuous with exactly the same information. The problem with conventional filters is they are the ones that are adding extra information, as the interpolated signal is different from the original. What I am trying to do is merely reduce these errors, which are audible as it degrades the timing of transients - something which is essential from human psychoacoustics.

Dear Rob,

Thank you for creating the M - Scaler with a practical form factor :) !

I have a question:
What difference, and why, in SQ would there be between the following audio chains:

- PC → USB → digital upscaling with M-Scaler → digital signal → DAC

- digital upscaling with PC → digital signal → DAC ?

Because the M- Scaler upscales and outputs a digital signal, I don't understand how the output upscaled digital signal would differ from the same produced by software running on a fast PC.

Thank you for your excellent work,
bidn

It's about how accurate the system does it; the PC is not capable of reconstruction to anything like the same accuracy that the M scaler is capable of.
 
Jul 25, 2018 at 8:33 AM Post #23 of 34

paul2qute

Headphoneus Supremus
Joined
Feb 1, 2017
Posts
2,477
Likes
1,107
And an interesting answer Rob, I avoided up scaling like the plague but if your not adding anything but reducing errors so you hear more of the original recording then it makes sense, I never really took much notice with the blu 2 M scaler, obviously price but didn't realise how much impact an M scaler has to the sound quality.
 
Jul 25, 2018 at 9:57 AM Post #24 of 34

bidn

100+ Head-Fier
Joined
Aug 14, 2016
Posts
390
Likes
547
Location
Kingdom of the Netherlands
Rob Watts said:
It's about how accurate the system does it; the PC is not capable of reconstruction to anything like the same accuracy that the M scaler is capable of.

Thank you,
but I would be interested if you would elaborate on this difference and on the reasons allowing the M-Scaler to produce a higher accuracy than a powerful PC.
 
Jul 25, 2018 at 2:23 PM Post #27 of 34

birdlandbill

Head-Fier
Joined
Apr 24, 2017
Posts
69
Likes
85
Location
MN
And an interesting answer Rob, I avoided up scaling like the plague but if your not adding anything but reducing errors so you hear more of the original recording then it makes sense, I never really took much notice with the blu 2 M scaler, obviously price but didn't realise how much impact an M scaler has to the sound quality.
Interesting post - I am actually not adding in any more information - the information is all contained in the original bandwidth limited sampled signal. but what we have is sampled data, and what we need is a continuous un-sampled signal - and we need to do a huge amount of processing to do this without error. I will give you an example. Imagine a sine wave. You can state it's a sine wave; give its frequency and its amplitude. So the information content is fixed; but if you want a waveform of infinite length, and precision, then you would need an infinite amount of processing to create the infinite number of points. And with the sampled data, we can convert it to a continuous signal - with exactly the same information content - and recover the original bandwidth limited continuous signal - if and only if you do an infinite amount of processing and use an ideal sinc function interpolation filter. So I am not trying to create new information - actually we are converting from a sampled bandwidth limited signal to a continuous with exactly the same information. The problem with conventional filters is they are the ones that are adding extra information, as the interpolated signal is different from the original. What I am trying to do is merely reduce these errors, which are audible as it degrades the timing of transients - something which is essential from human psychoacoustics.



It's about how accurate the system does it; the PC is not capable of reconstruction to anything like the same accuracy that the M scaler is capable of.
Interesting post - I am actually not adding in any more information - the information is all contained in the original bandwidth limited sampled signal. but what we have is sampled data, and what we need is a continuous un-sampled signal - and we need to do a huge amount of processing to do this without error. I will give you an example. Imagine a sine wave. You can state it's a sine wave; give its frequency and its amplitude. So the information content is fixed; but if you want a waveform of infinite length, and precision, then you would need an infinite amount of processing to create the infinite number of points. And with the sampled data, we can convert it to a continuous signal - with exactly the same information content - and recover the original bandwidth limited continuous signal - if and only if you do an infinite amount of processing and use an ideal sinc function interpolation filter. So I am not trying to create new information - actually we are converting from a sampled bandwidth limited signal to a continuous with exactly the same information. The problem with conventional filters is they are the ones that are adding extra information, as the interpolated signal is different from the original. What I am trying to do is merely reduce these errors, which are audible as it degrades the timing of transients - something which is essential from human psychoacoustics.



It's about how accurate the system does it; the PC is not capable of reconstruction to anything like the same accuracy that the M scaler is capable of.

Amazing! The information/signals was/were always there. Accurately decoding (with FPGA generated coefficients and WTA magic) it in a transducer-useful manner is only now reaching a point of sound studio reproduction. The sense of something always missing from digital recordings is finally receding and promising to get better still.
 
Jul 25, 2018 at 5:38 PM Post #29 of 34

OG10

100+ Head-Fier
Joined
Jan 2, 2005
Posts
207
Likes
119
Location
Manchester - UK
Interesting post - I am actually not adding in any more information - the information is all contained in the original bandwidth limited sampled signal. but what we have is sampled data, and what we need is a continuous un-sampled signal - and we need to do a huge amount of processing to do this without error. I will give you an example. Imagine a sine wave. You can state it's a sine wave; give its frequency and its amplitude. So the information content is fixed; but if you want a waveform of infinite length, and precision, then you would need an infinite amount of processing to create the infinite number of points. And with the sampled data, we can convert it to a continuous signal - with exactly the same information content - and recover the original bandwidth limited continuous signal - if and only if you do an infinite amount of processing and use an ideal sinc function interpolation filter. So I am not trying to create new information - actually we are converting from a sampled bandwidth limited signal to a continuous with exactly the same information. The problem with conventional filters is they are the ones that are adding extra information, as the interpolated signal is different from the original. What I am trying to do is merely reduce these errors, which are audible as it degrades the timing of transients - something which is essential from human psychoacoustics.



It's about how accurate the system does it; the PC is not capable of reconstruction to anything like the same accuracy that the M scaler is capable of.

Thanks for taking the time to respond back Rob, so in terms of re-creating the the wave form via PCM through SPDIF how will the M Scaler handle any clock variations. Is the need for an I2S link between source and DAC a bit exaggerated, my reason for asking is if there is jitter from the source transport via SPDIF is there a chance that the jitter could be amplified by the upscaling before it goes to say the Qutest, causing additional distortion?

Or is that too minimal an effect to consider ?

Thanks in advance.
 
Status
Not open for further replies.

Users who are viewing this thread

Top