CHORD M-Scaler Review (Upsampler)

[kemmler3D]

I know of at least one who has completely converted, reads here regularly, but has never posted. Sold all his high end cables and purchased some ASR-approved kit with the proceeds.

Bingo. In a debate, there are two people on stage who do all the talking, but almost never change their minds. And then there are 100 people in the audience, waiting to be convinced. Those 100 people are worth the effort of rehashing the same old crap again.

 

[Galliardist]

Thank you for watching over us.
Please, next time, remind those members, that it takes two to tango!
Not engaging possible trolls, stops them at their tracks, doesn't it?
No need to ban anyone. scream at anyone . . . .

The problems:
Some are not trolls here to antagonise the forum. They simply arrive to discuss audio. You can't easily tell a true believer from a troll, and they are likely to act in exactly the same way.
If we ignore such people, then others may arrive and start conversations, creating entire threads with no evidence, recommending snake oil products to each other, making comments that then appear to be approved, and so on. That's how the rot sets in.

 

[solid12345]

Upscaling is a mathematical computation with zero insight into what it is doing. It is simply creating more samples with the same information that was fed to it. If you think enlarging a picture makes it more detailed, then heaven help you.

I haven’t heard the mscaler but I see no reason to believe the future of AI technology won’t be able to take scratchy 1930s-60s recordings and mathematically recreate them as if they were recorded today on modern equipment just as you can now take a low-res jpeg and upscale it with stability diffusion and other similar tools to produce a crisp clean sharp fascimile of it. Already I’ve heard tools to remove the voices from songs for instance.

This is going to give objectivists versus subjectivists fun to debate for years to come because the question will become what is more real, the crude recordings on a tin can or the reconstruction that sounds like you were there in the studio that day.

 

[kemmler3D]

I haven’t heard the mscaler but I see no reason to believe the future of AI technology won’t be able to take scratchy 1930s-60s recordings and mathematically recreate them as if they were recorded today on modern equipment just as you can now take a low-res jpeg and upscale it with stability diffusion and other similar tools to produce a crisp clean sharp fascimile of it. Already I’ve heard tools to remove the voices from songs for instance.

This is going to give objectivists versus subjectivists fun to debate for years to come because the question will become what is more real, the crude recordings on a tin can or the reconstruction that sounds like you were there in the studio that day.

You're certainly right about this, but it's still not recovering information that was in the recording, it's inventing new information.

Upscaling is pretty much what it sounds like - basically just representing the same data with bigger numbers. It's like writing your bank balance as $43,201.25000000 instead of $43,201.25. Usually pointless.

When it comes to AI reconstruction or interpolation... it involves creating new information where none was before, based on rules that may end up sounding good, but also don't necessarily correspond to what actually *would have been* on the recording.

I agree that it would be a matter of debate. For example, take Robert Johnson's works. He's one of the best guitarists to ever play, but all of his recordings sound like they were made in a random shack somewhere with substandard equipment even in the 1930s, which they were. I've often wondered if these AI tools could give us a "clean" version of these songs... but you'd always have to wonder, is that REALLY what he sounded like, or is it just a convincing, but ultimately wrong simulation?

 

[ahofer]

but you'd always have to wonder, is that REALLY what he sounded like, or is it just a convincing, but ultimately wrong simulation?

Remember this?

A Few Thoughts on the Authenticity of Peter Jackson’s “They Shall Not Grow Old”

The filmmaker’s process of modernizing footage from the First World War does make it effectively contemporary, but, by altering the mirror of the past, we fail to see the subjects staring into it quite accurately.

www.newyorker.com

How Peter Jackson’s team made World War I footage look new

The way we normally see that footage — black and white, scratchy and silent — is "like a barrier between us and the actual people that were being filmed," Jackson tells Recode’s Kara Swisher.

www.vox.com

 

[Ken Tajalli]

Yes, dumb upscaling (the type Amirm refers to) is . . . well dumb!
But mScaler tries to be not-that.
although 44.1k digital audio can absolutely manage 20kHz, but to get a clean high frequency, requires clever filtering. mScaler uses Rob Watts proprietary algorithm to cleverly guess extra inner-samples to intelligently upscale 44.1k many times. that is its claim to fame, claim being the operative word here!
More samples are good, more bits, not so much. mScaler just upsamples. 16 bits, remain 16.
As such, measuring a device such as this, only shines a light on any quality issues in its manufacture. It has no analog out. one can measure the accuracy of the clock, electrical noise, etc. but one can not realistically measure its sound.
I have heard the mScaler/Dave combo.
I used my own materials, some of them hires. I used headphones, a DCA Stealth, an Audeze LCD-XC and a Hifiman Edition X.
Honestly, on some good Classical music tracks, it seemed to vlean up the sound a bit. On pop and rock music, I couldn't tell.
I certainly wouldn't invest in one, since I can't hear its effect.

Regarding AI upsampling, check this thread:

Real Video Upscaling using AI!

Recently I saw on a website, details about a home software upscaling of StarTrek Deep Space 9, by a fan, and the results were very good. https://www.extremetech.com/extreme/333150-i-can-restore-star-trek-voyager-and-deep-space-nine-to-hd-so-why-cant-paramount That got me intrigued! I did not...

www.audiosciencereview.com

 

[Guermantes]

Remember this?

A Few Thoughts on the Authenticity of Peter Jackson’s “They Shall Not Grow Old”

The filmmaker’s process of modernizing footage from the First World War does make it effectively contemporary, but, by altering the mirror of the past, we fail to see the subjects staring into it quite accurately.

www.newyorker.com

How Peter Jackson’s team made World War I footage look new

The way we normally see that footage — black and white, scratchy and silent — is "like a barrier between us and the actual people that were being filmed," Jackson tells Recode’s Kara Swisher.

www.vox.com

And this?

Is It Live ... or Yamaha? Channeling Glenn Gould (Published 2007)

Glenn Gould is pulled back into the realm of public performance with a “reperformance” of his “Goldbergs” on a specially prepared Yamaha Disklavier.

www.nytimes.com

 

[Jimster480]

Yes, dumb upscaling (the type Amirm refers to) is . . . well dumb!
But mScaler tries to be not-that.
although 44.1k digital audio can absolutely manage 20kHz, but to get a clean high frequency, requires clever filtering. mScaler uses Rob Watts proprietary algorithm to cleverly guess extra inner-samples to intelligently upscale 44.1k many times. that is its claim to fame, claim being the operative word here!
More samples are good, more bits, not so much. mScaler just upsamples. 16 bits, remain 16.
As such, measuring a device such as this, only shines a light on any quality issues in its manufacture. It has no analog out. one can measure the accuracy of the clock, electrical noise, etc. but one can not realistically measure its sound.
I have heard the mScaler/Dave combo.
I used my own materials, some of them hires. I used headphones, a DCA Stealth, an Audeze LCD-XC and a Hifiman Edition X.
Honestly, on some good Classical music tracks, it seemed to vlean up the sound a bit. On pop and rock music, I couldn't tell.
I certainly wouldn't invest in one, since I can't hear its effect.

Regarding AI upsampling, check this thread:

Real Video Upscaling using AI!

Recently I saw on a website, details about a home software upscaling of StarTrek Deep Space 9, by a fan, and the results were very good. https://www.extremetech.com/extreme/333150-i-can-restore-star-trek-voyager-and-deep-space-nine-to-hd-so-why-cant-paramount That got me intrigued! I did not...

www.audiosciencereview.com

AI Upscaling would be a different thing though vs what the mScaler does. Even if he came up with an algorithm that could "clean up" certain tracks specifically; it wouldn't be applicable for all music as it isn't some kind of trained AI model but rather a set of parameters that he set that he thought sounds good. As far as the performance goes; well it doesn't increase the objective performance of the audio equipment as it claims to do.

 

[mccririck]

Can anyone explain why 24 bit doesnt give more fidelity than 16 bit? Why doesnt it give smaller jumps between levels?

 

[gvl]

Can anyone explain why 24 bit doesnt give more fidelity than 16 bit? Why doesnt it give smaller jumps between levels?

24 bits gives you lower noise floor given the rest of your chain doesn’t pollute it with its own. The difficulty is in hearing this improvement as 16 bits is already good enough for most humans and listening conditions.

 

[staticV3]

Can anyone explain why 24 bit doesnt give more fidelity than 16 bit?

24 bit increases the potential dynamic range of an audio file.
However, that potential cannot really be realized in music, as even with 16 bit audio, the bottleneck is not the audio format, but the recording environment and equipment, as well as the playback environment and equipment.

 

[tmtomh]

Can anyone explain why 24 bit doesnt give more fidelity than 16 bit? Why doesnt it give smaller jumps between levels?

More knowledgeable folks than me can give you a more mathematically literature answer, but basically the reason is that a bit is a bit: 1 bit PCM = about 6dB of s/n no matter what. Moving up from 16 to 24 bits does. not reduce the size of the volume increase from one bit to another. Rather, it simply lowers the noise floor. So 24 bit gives you higher dynamic range - a larger volume spread from the noise floor up to max volume aka digital 0.0. It does not create finer gradations of volume steps within the same audible range.

EDIT: Two other folks wrote responses as the same time as I was writing mind. So, what they said.

 

[kemmler3D]

Yes, dumb upscaling (the type Amirm refers to) is . . . well dumb!
But mScaler tries to be not-that.
although 44.1k digital audio can absolutely manage 20kHz, but to get a clean high frequency, requires clever filtering. mScaler uses Rob Watts proprietary algorithm to cleverly guess extra inner-samples to intelligently upscale 44.1k many times. that is its claim to fame, claim being the operative word here!
More samples are good, more bits, not so much. mScaler just upsamples. 16 bits, remain 16.
As such, measuring a device such as this, only shines a light on any quality issues in its manufacture. It has no analog out. one can measure the accuracy of the clock, electrical noise, etc. but one can not realistically measure its sound.
I have heard the mScaler/Dave combo.
I used my own materials, some of them hires. I used headphones, a DCA Stealth, an Audeze LCD-XC and a Hifiman Edition X.
Honestly, on some good Classical music tracks, it seemed to vlean up the sound a bit. On pop and rock music, I couldn't tell.
I certainly wouldn't invest in one, since I can't hear its effect.

Regarding AI upsampling, check this thread:

Real Video Upscaling using AI!

Recently I saw on a website, details about a home software upscaling of StarTrek Deep Space 9, by a fan, and the results were very good. https://www.extremetech.com/extreme/333150-i-can-restore-star-trek-voyager-and-deep-space-nine-to-hd-so-why-cant-paramount That got me intrigued! I did not...

www.audiosciencereview.com

I see, it's a resampler, not just a bit-depth-increaser.

In real life it's hard to imagine how this could really help fidelity or sound quality.

The only reason to use a sampling rate higher than 44.1khz is to represent frequencies higher than 22,500hz. Even if you could hear those frequencies, you'd have to be adding content (what kind of content, anyway?) that doesn't exist in the original recording, for the ultimate analog output of the resampler to be any different than the original. Why would you want to do that?

 

[mccririck]

More knowledgeable folks than me can give you a more mathematically literature answer, but basically the reason is that a bit is a bit: 1 bit PCM = about 6dB no matter what. Moving up from 16 to 24 bits does. not reduce the size of the volume increase from one bit to another. Rather, it simply lowers the noise floor. So 24 bit gives you higher dynamic range - a larger volume spread from the noise floor up to max volume aka digital 0.0. It does not create finer gradations of volume steps within the same audible range.

EDIT: Two other folks wrote responses as the same time as I was writing mind. So, what they said.

ok so PCM is stopping the differences in volume level being smaller than it is?

 

[staticV3]

More knowledgeable folks than me can give you a more mathematically literature answer, but basically the reason is that a bit is a bit: 1 bit PCM = about 6dB no matter what. Moving up from 16 to 24 bits does. not reduce the size of the volume increase from one bit to another. Rather, it simply lowers the noise floor. So 24 bit gives you higher dynamic range - a larger volume spread from the noise floor up to max volume aka digital 0.0. It does not create finer gradations of volume steps within the same audible range.

In theory, the same volume of signal can be stored and played back more precisely in 24 bit, which becomes most noticeable as you approach the DR limit of 16 bit:

(Source: https://www.stereophile.com/content/dcs-vivaldi-apex-da-processor-measurements)

In practice, we have dithering, real music signals, and flawed listening environments, negating the advantages of 24 bit audio.

 

[kemmler3D]

ok so PCM is stopping the differences in volume level being smaller than it is?

When you go from 8 to 12 to 16 bit, what you hear is less noise. If you want to try this yourself you can use this plugin in EQAPO and increase/decrease the bit depth on your audio as you please: (MDA Degrade lets you select the bit depth in real time) https://www.kvraudio.com/developer/mda

The noise you hear is the random error between the value encoded and the real signal. So the smaller steps between levels = less error = less noise.

The thing is, 16-bit already has so little noise that most equipment can't even keep up, so moving to 24-bit doesn't sound any better for playback in 99.99% of situations. It's good for recording when the noise floor is a particular concern. It's fine for playback in a "more is more" sense and in that it doesn't hurt anything, but isn't necessary for good sound.

 

[mccririck]

When you go from 8 to 12 to 16 bit, what you hear is less noise. If you want to try this yourself you can use this plugin in EQAPO and increase/decrease the bit depth on your audio as you please: (MDA Degrade lets you select the bit depth in real time) https://www.kvraudio.com/developer/mda

The noise you hear is the random error between the value encoded and the real signal. So the smaller steps between levels = less error = less noise.

The thing is, 16-bit already has so little noise that most equipment can't even keep up, so moving to 24-bit doesn't sound any better for playback in 99.99% of situations. It's good for recording when the noise floor is a particular concern. It's fine for playback in a "more is more" sense and in that it doesn't hurt anything, but isn't necessary for good sound.

ok, so there are actually smaller steps between levels in 24 bit compared to 16 bit?

One thing - does 24 bit use more energy to convert?

 

[ahofer]

I believe that upsampling can facilitate other forms of DSP, and can eliminate some noisy edge cases in using digital volume control. Perhaps more useful in recording/mastering than straightforward playback.

 

[kemmler3D]

ok, so there are actually smaller steps between levels in 24 bit compared to 16 bit?

One thing - does 24 bit use more energy to convert?

Yes, in a sense the steps are smaller.

It may use a bit more energy because it's more CPU intensive, but realistically nothing in digital audio playback is that difficult for today's CPUs to handle, until you reach 384 Khz / 32-bit with multiple FIR filters, or something.

 

[SIY]

In theory, the same volume of signal can be stored and played back more precisely in 24 bit, which becomes most noticeable as you approach the DR limit of 16 bit:
View attachment 297069

Less noise, not more precision. There would be more precision in a noise-free system, which doesn't exist, or if the signal is undithered, which describes 0.0000001% of recordings.