How Chord M-Scaler works in layman's terms

[FunkyBassMan]

Do a level matched, direct A/B switched, blind test with multiple trials averaged and then we’ll talk. Until then, you’re not a reliable witness.

Note: I’m not talking about obsolete NOS DACs or the impedance of the amp stage here. I’m talking about the DAC.

Someone telling me what I can hear and what I can't is as absurd as it is irrelevant.

You think everything sounds the same. Good luck with that, and have a great life.

 

[FunkyBassMan]

Yes and to ordinary and superior human ears.

We don’t say something isn’t inaudible just because one or more of us can’t hear it. We say it’s inaudible because it BOTH falls below the thresholds of audibility AND there is no reliable evidence that anyone can hear it, despite numerous controlled tests over a couple of decades or so.

G

Good for you. Enjoy not hearing things. And I'll enjoy hearing them.

 

[bigshot]

Someone telling me what I can hear and what I can't is as absurd as it is irrelevant.

Don’t feel bad that you can’t hear a difference between DACs. No one can. The people who claim they can don’t know any more about how DACs work than you do.

 

[bigshot]

Good for you. Enjoy not hearing things. And I'll enjoy hearing them.

I SEE DEAD PEOPLE AND I HEAR VOICES IN MY HEAD!

 

[FunkyBassMan]

Don’t feel bad that you can’t hear a difference between DACs. No one can. The people who claim they can don’t know any more about how DACs work than you do.

Don't feel bad your English is so poor in understanding my posts. I'll go slow for you: I'm the one who CAN hear differences (along with millions of others). You're one of the dozens who can't. But I hope you enjoy your music anyway as much as you enjoy telling others what you think they can and can't hear.

 

[FunkyBassMan]

I SEE DEAD PEOPLE AND I HEAR VOICES IN MY HEAD!

This is will surprise no one: I'd recommend a good psychiatrist. Your psychoses and hearing may possibly be improvable with medication.

 

[gregorio]

Someone telling me what I can hear and what I can't is as absurd as it is irrelevant.

Yes sure, science is absurd and irrelevant, which is why ADCs and DACs are built according to subjective smell preferences and Druidic magic rather than science. You can hear radio frequencies without a radio, hear the inaudible and leap tall buildings in a single bound.

Good for you. Enjoy not hearing things. And I'll enjoy hearing them.

I enjoy hearing things that can be heard, you enjoy “hearing” your delusional beliefs.

I'm the one who CAN hear differences (along with millions of others).

If you and millions of others can hear these differences why haven’t you or a single one of those millions ever been able to demonstrate that you can?

You're one of the dozens who can't.

You do know there are about 8 billion people on this planet and not dozens? If you can’t point to a single one of them who can actually hear these inaudible differences then there are about 8 billion who can’t not dozens.

This is will surprise no one: I'd recommend a good psychiatrist. Your psychoses and hearing may possibly be improvable with medication.

I agree with all this! However he was talking about you (or rather the group of audiophiles you represent), so are you going to take your own “recommendation”?

The trouble with making ridiculous assertions is that it’s impossible not to make ever more ridiculous statements when trying to defend them!

G

 

[GoldenSound]

That’s a fallacious argument (the Perfect-Solution Fallacy to be precise). We do not need a “truly perfect” filter as we do not have “truly perfect”: Ears, listening abilities, amps or speakers/HPs.

I agree we almost certainly don't, but that wasn't what I was saying.
Objectively speaking, to achieve perfect reconstruction we need a perfect filter. Where the audible limit lies and for what factors (there are multiple issues/factors to different reconstruction approaches and tradeoffs) is up for debate and there is very little study on the matter.

Furthermore, we don’t need a filter to be “instant” (and that would be impossible anyway), a latency of say 200ms is literally the blink of an eye, actually a particularly fast one (most eye blinks take 300ms). And, filter attenuation to about -90dB is perfectly acceptable/inaudible under any reasonable listening conditions and most DACs, even relatively cheap ones, achieve attenuation lower than that.

Instant is not referring to delay/latency, it's referring to the filter design itself. Instant meaning that everything below 22.05khz is passed through completely unaltered and everything above that is entirely eliminated. As opposed to partial attenuation or slower rolloffs.

The MScaler doesn’t do any reconstruction, it just upscales/upsamples and outputs a digital signal. So how does no reconstruction at all “achieve better reconstruction”?

The upsampling IS the reconstruction filter. It's the same process as done internally in any oversampling DAC just done externally. The maths/process is the same ignoring filter design/performance differences.

Furthermore, by requiring a “lot more compute power” to implement millions of taps it has particularly poor latency. And if that’s not enough, it’s upsampling filter while very abrupt, doesn’t even attenuate to -80dB (still inaudible in the vast majority of reasonable listening scenarios). It also seems to have particularly poor jitter performance (although again at inaudible levels). - Measurement source ASR.

Latency is an inherent tradeoff to high performance reconstruction filters yes. And in many production environments where latency is key this could be an unacceptable issue, but for straight up listening, generally it doesn't matter if your music starts instantly after hitting play or a second or two later.

As to the -80dB claim, that's not true and is unfortunately a result of Amir using less than ideal testing methodology and not setting up the device correctly. At 192khz output the attenuation is over 120dB at least:

https://goldensound.audio/2022/03/17/chord-hugo-m-scaler-measurements-and-technical-evaluation/

And due to how noise shaper effectiveness is linked to conversion ratio, at 768khz output it's better, though difficult to test due to the dual-BNC output so I'm not sure what the exact figure is when in use with a Chord DAC.
Though regardless the 80dB claim is simply not true and is down to Amir's testing not the device's actual performance.

I don’t see the point of that? Of course some filters will be audibly different, especially ones designed to be audibly different, say NOS emulation filters or any other filter with a roll-off easily within the audible band. If you take a decent filter though, say the typical filters used for the last 25+ years or so, a linear phase filter with a roll-off starting around 18kHz-19kHz and a transition band of around 2kHz, good luck ABX’ing these.

I'm comparing filters that are higher performance than those inbuilt into most DACs. Not 'slow' or broken ones.

No there’s not, there’s numerous devices and there has been for years. Of course, that depends on exactly what you mean by “very high performance filters”. I take it to mean any filter that does it’s job without audible artefacts.

By high performance I'm meaning those using at least a few thousand filter coefficients. Most DACs use 128-1024 taps which is very low.
As to 'without audible artefacts', my testing has shown that the limit is surprisingly high. Video coming soon

 

[GoldenSound]

Was he talking about filters? When I asked for clarification, he got all dodgy.

I didn't 'get all dodgy'. I just pointed out that I'd answered this question to you directly previously and am not really wanting to answer the same questions again and again when I've already given answers.

https://www.head-fi.org/threads/why...t-bad-for-music.716822/page-205#post-17296200

https://www.head-fi.org/threads/why...t-bad-for-music.716822/page-210#post-17312291

I am MORE than happy and would be quite grateful for constructive input to make this more robust, but most of the replies thus far have just been akin to 'nope you're wrong' with no backing to those statements, or just referring to Amir's previous statements. Neither of which are constructive

 

[VNandor]

The upsampling IS the reconstruction filter. It's the same process as done internally in any oversampling DAC just done externally. The maths/process is the same ignoring filter design/performance differences.

This is not what people usually mean by the reconstruction filter. The reconstruction filter is the analog filter at the end of the signal chain, often times also called the anti-imaging filter. DACs typically do oversampling but the reconstruction filter is not used during that step. The m-scaler doesn't have an analog output so I don't think it has a reconstruction filter either.

 

[GoldenSound]

This is not what people usually mean by the reconstruction filter. The reconstruction filter is the analog filter at the end of the signal chain, often times also called the anti-imaging filter. DACs typically do oversampling but the reconstruction filter is not used during that step. The m-scaler doesn't have an analog output so I don't think it has a reconstruction filter either.

It's entirely contextual but these digital filters are reconstruction filters and both are actually the same thing. A reconstruction/anti-imaging filter can be implemented in analog or done digitally. Which of the two makes sense depends on the situation. Class D amps for example you need to go analog. DACs you cannot feasibly make a sufficient filter in the analog domain so it's done digitally (often with a secondary analog filter depending on the DAC design). But both are fulfilling the same task and are practically speaking the same thing.

https://en.wikipedia.org/wiki/Reconstruction_filter

 

[VNandor]

From the article you linked:

"In a mixed-signal system (analog and digital), a reconstruction filter, sometimes called an anti-imaging filter, is used to construct a smooth analog signal from a digital input, as in the case of a digital to analog converter (DAC) or other sampled data output device. "

Note that the output of the reconstruction filter is an analog signal according to the wiki.

A reconstruction/anti-imaging filter can be implemented in analog or done digitally.

My knowledge is shaky when it comes to implementation details, but as far as I know, a DAC will put out a zero-order hold signal at some point which ideally have to be filtered by an analog low pass filter before actually outputting the signal. The digital signal that's used to create the zero-order hold signal is usually low passed by a digital filter before but the zero-order hold signal will still include images unless it's being filtered away. I think it's the analog filter specifically that's being called the reconstruction filter.

 

[GoldenSound]

From the article you linked:

"In a mixed-signal system (analog and digital), a reconstruction filter, sometimes called an anti-imaging filter, is used to construct a smooth analog signal from a digital input, as in the case of a digital to analog converter (DAC) or other sampled data output device. "

Note that the output of the reconstruction filter is an analog signal according to the wiki.

Yes, and by that definition the term applies.

You can take the sampled data and use a digital reconstruction filter to provide a reconstructed/smooth waveform vs a sample and hold output. Whether or not further analog filtering is necessary depends on the DAC topology and design goals.

Regardless, this is semantics, if you'd prefer to specifically refer to it as a 'Nyquist Reconstruction Filter' rather than just 'Reconstruction Filter' you're free to do so

 

[VNandor]

You can take the sampled data and use a digital reconstruction filter to provide a reconstructed/smooth waveform vs a sample and hold output. Whether or not further analog filtering is necessary depends on the DAC topology and design goals.

From the article you linked: "Alternatively, a system may have no reconstruction filter and simply tolerate some energy being wasted reproducing higher frequency images of the primary signal spectrum."
Are you talking about the above case? Note that the wiki specifically mentions no reconstruction filter for this case.

I think using the word correctly have some practical implications. If a DAC has bad reconstruction filters, the m-scaler won't solve that problem because it doesn't do reconstruction, it only does upsampling, which are different things in a lot of cases.

 

[GoldenSound]

From the article you linked: "Alternatively, a system may have no reconstruction filter and simply tolerate some energy being wasted reproducing higher frequency images of the primary signal spectrum."
Are you talking about the above case? Note that the wiki specifically mentions no reconstruction filter for this case.

I think using the word correctly have some practical implications. If a DAC has bad reconstruction filters, the m-scaler won't solve that problem because it doesn't do reconstruction, it only does upsampling, which are different things in a lot of cases.

No, that quote is effectively describing a NOS DAC which is not what I'm referring to
Again, digital upsampling IS reconstruction. It's not just adding extra samples for no reason