• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

CHORD M-Scaler Review (Upsampler)

Rate this product:

  • 1. Poor (headless panther)

    Votes: 358 88.2%
  • 2. Not terrible (postman panther)

    Votes: 13 3.2%
  • 3. Fine (happy panther

    Votes: 7 1.7%
  • 4. Great (golfing panther)

    Votes: 28 6.9%

  • Total voters
    406

Blumlein 88

Grand Contributor
Forum Donor
Joined
Feb 23, 2016
Messages
20,521
Likes
37,050
Alas, no I think he got what he came for . I doubt he will be back.

It seems he wants to leave it as (paraphrased)

They didnt measure the right things . nothing they have disproves that it does what it says it does. I prefer Headfi. I must be objective because I didnt prefer the DAVE.


As of around 10pm last night UK time:

"Mike, you need to take the measurements and, more importantly the conclusions drawn, on ASR with a bucket of salt. My attempts to discuss the M-Scaler review on ASR showed an unwillingness to discuss the relevance of the measurements and conclusions drawn. Instead the emphasis was on the argument that the M-Scaler can't do anything because it can't do anything. I've got the DAVE here now (going home today) and it sounds great. If it were so flawed, it would not perform as it does in listening tests. It's not my favourite DAC and I have no desire to buy one (even if I could afford it), but ASR are doing a disservice with their butchering of these products that it seems they don't understand beyond "digital in, analog out". I'd recommend visiting the Head-Fi threads on DAVE and M-Scaler for far more balanced info with solid technical evidence being shared in both directions."
Show less
This is the same head fi where you cannot discuss measurements or proper listening tests except in a special subforum which is also heavily censored? If you do you get banned. Oh yeah real balanced. If balanced means only subjective results that help sponsors because it balances the accounts of head fi.
 

Jimbob54

Grand Contributor
Forum Donor
Joined
Oct 25, 2019
Messages
11,066
Likes
14,697
Strange that someone simply cannot understand that a device that makes things worse -and even in theory can't make things better- is not worth the money. I was pretty impressed with the patience shown here. And the wonderful explanations. Not sure what else we can do but proceed where we left. There is much more interesting stuff to do and test.
To his mind his listening tests are definitive. He hears a difference /benefit therefore there is one.

His youtube channel like so many others relies on everything making a difference and add ons /upgrades almost always making a positive one.

I can't imagine a lot of existing m scaler users wanting to sit down and devise and perform some arduous controlled tests that might undermine their established belief in their £5k magic box. This review and the exchanges here with Lachlan likely won't change that.
 

earlevel

Addicted to Fun and Learning
Joined
Nov 18, 2020
Messages
545
Likes
776
Thank you - this is super-clear, and while I feel pretty dumb right now for making that error (the source of which was, I think, considering the concept of interpolation only in one sense and not the sense in which it actually gets applied in this process), I am glad to have gotten a clear understanding from you (and @danadam , as usual :) ).

After I was first notified of my mistake, I did some searching online and while I came across some good explanations, yours is the clearest I have encountered on the specific point of how ultrasonic frequencies get generated after the A/D step, requiring the low pass filter at the other end too.

Just to make sure I’m not missing anything, the effect of the 4x oversampling in your example is to increase the frequency of the unwanted ultrasonics, increasing the frequency spread between them and the desired signal/audible range, yes? My understanding has been that this is the only effect/purpose of such oversampling (easing/simplifying design of the low pass filter). Is there any additional effect besides that?

Thanks again!
You're welcome, I'm glad that was helpful. And sample repetition is assumed by a lot of people, don't feel dumb. It doesn't make intuitive sense to use zeros and make the sine wave (that most people imagine the process with) look discontinuous. If you want a deeper, but relatively intuitive view of sampling, check out my series here:

Sampling theory, the best explanation you’ve ever heard

As far as increasing the sample rate, it doesn't change any frequencies. But since the "Fs", the sampling frequency was raised, sidebands that were a result of the sampling process are now in the new, wider audio band. (By "audio band", I mean everything below half the sample rate.) So, we need to clear everything but the original signal with a lowpass filter.

OK, some might have a little problem with the idea that things go on forever, but remember we're just doing math, where things are pure. The same happens in analog sampling, but pulses are imperfect and the images die out as frequency response droops—it still works because we plan to eliminate them with a lowpass anyway. Perfect impulses have no limit to their frequency content, real impulses do. But by keeping only one reading per sample and discarding the rest of the analog waveform, we simulated a perfect impulse. Just saying this to prep for the fact that PCM is a form of amplitude modulation (AM, like the radio), with a perfect impulse, therefore sideband copies—plus and minus—exist mirrored around multiples of the sample rate. And any changes to the part that's in the audio band changes those images as well. We get rid of the images in converting back to analog. So, if we start with a lowpass filtered audio signal that ends up with this bandwidth:

OS1.png

After sampling, we have this in the frequency domain:

OS2.png


Note that that the original band is "mirrored" around Fs (and 2Fs, and 3Fs...), it's reversed below it, and is the forward but shifted in frequency above it.

If we zero-stuff to up the sample rate, the result is this—we only changed where Fs is:

OS3.png


But note that if we look at the new audio band, that below half Fs, we have those mirrored images. They are ultrasonic, but if we run that through a DAC at the higher rate, those frequencies will be in the analog output.

To get back to the audio we started with, we just lowpass filter (again, this and all processes are mirrored) to keep the audio below half the original rate:

OS4.png


If you wanted to play this out a DAC with 20 kHz audio bandwidth, even a very gentle filter will work great—it just needs to ensure the red part is removed.
 

tmtomh

Major Contributor
Forum Donor
Joined
Aug 14, 2018
Messages
2,635
Likes
7,485
You're welcome, I'm glad that was helpful. And sample repetition is assumed by a lot of people, don't feel dumb. It doesn't make intuitive sense to use zeros and make the sine wave (that most people imagine the process with) look discontinuous. If you want a deeper, but relatively intuitive view of sampling, check out my series here:

Sampling theory, the best explanation you’ve ever heard

As far as increasing the sample rate, it doesn't change any frequencies. But since the "Fs", the sampling frequency was raised, sidebands that were a result of the sampling process are now in the new, wider audio band. (By "audio band", I mean everything below half the sample rate.) So, we need to clear everything but the original signal with a lowpass filter.

OK, some might have a little problem with the idea that things go on forever, but remember we're just doing math, where things are pure. The same happens in analog sampling, but pulses are imperfect and the images die out as frequency response droops—it still works because we plan to eliminate them with a lowpass anyway. Perfect impulses have no limit to their frequency content, real impulses do. But by keeping only one reading per sample and discarding the rest of the analog waveform, we simulated a perfect impulse. Just saying this to prep for the fact that PCM is a form of amplitude modulation (AM, like the radio), with a perfect impulse, therefore sideband copies—plus and minus—exist mirrored around multiples of the sample rate. And any changes to the part that's in the audio band changes those images as well. We get rid of the images in converting back to analog. So, if we start with a lowpass filtered audio signal that ends up with this bandwidth:

View attachment 220634
After sampling, we have this in the frequency domain:

View attachment 220635

Note that that the original band is "mirrored" around Fs (and 2Fs, and 3Fs...), it's reversed below it, and is the forward but shifted in frequency above it.

If we zero-stuff to up the sample rate, the result is this—we only changed where Fs is:

View attachment 220636

But note that if we look at the new audio band, that below half Fs, we have those mirrored images. They are ultrasonic, but if we run that through a DAC at the higher rate, those frequencies will be in the analog output.

To get back to the audio we started with, we just lowpass filter (again, this and all processes are mirrored) to keep the audio below half the original rate:

View attachment 220638

If you wanted to play this out a DAC with 20 kHz audio bandwidth, even a very gentle filter will work great—it just needs to ensure the red part is removed.

Once again, thanks very much! Appreciate the combination of clear explanation and in-depth explanation (RE theoretically/mathematically infinite pulses).

Per frequencies being "created," yes, the way you stated it is what I was trying to say - didn't mean to imply that any existing frequencies were changed, but rather that the "new, wider audio band" that you reference is beneficial because the frequency distance (if it makes sense to use the term "distance," at least metaphorically) between the Nyquist limit of the original signal (or perhaps more practically, 20kHz, depending on how one thinks about it, I suppose) and the lowest unwanted sideband is larger than it would be if the signal had not been oversampled.

So to return to your earlier example comparing a 48kHz sample rate signal with that same signal oversampled 4x to 192kHz, the lower frequency limit of the sideband of the non-oversampled signal would be 24kHz (48 minus 24, yes?), which is also the upper limit of the content of the original signal/recording. But once it's been 4x oversampled, the lower frequency limit of the sideband would be 72kHz (96 minus 24), while the upper limit of the original signal/recording remains 24kHz - yes? I presume that's the reason you said that the non-oversampled signal would need a lowpass filter just under 24kHz, while the 4x oversampled one would need a lowpass filter at 24kHz - you could set it higher of course, but my point is that you didn't say the 4x oversampled signal would need a lowpass filter at, say 71kHz - the whole point is that you have more wiggle room to set the frequency and slope/characteristics of the lowpass filter - yes?
 

earlevel

Addicted to Fun and Learning
Joined
Nov 18, 2020
Messages
545
Likes
776
Once again, thanks very much! Appreciate the combination of clear explanation and in-depth explanation (RE theoretically/mathematically infinite pulses).

Per frequencies being "created," yes, the way you stated it is what I was trying to say - didn't mean to imply that any existing frequencies were changed, but rather that the "new, wider audio band" that you reference is beneficial because the frequency distance (if it makes sense to use the term "distance," at least metaphorically) between the Nyquist limit of the original signal (or perhaps more practically, 20kHz, depending on how one thinks about it, I suppose) and the lowest unwanted sideband has been increased. So to return to your earlier example of a 48kHz sample rate, and then that same signal upsampled to 192kHz, the lowpass filter for the original needs to be set at, as you said, just under 24kHz, since the lowest sideband is (or approaches) 24kHz, which is also the upper limit of the content of the original signal/recording. But once it's been 4x oversampled, the lowest sideband is/approaches 96kHz, while the upper limit of the original signal/recording remains 24kHz - yes? I presume that's the reason you said that the non-oversampled signal would need a lowpass filter just under 24kHz, while the 4x oversampled one would need a lowpass filter at 24kHz - you could set it higher of course, but my point is that you didn't say the 4x oversampled signal would need a lowpass filter at, say 96kHz - the whole point is that you have more wiggle room to set the frequency and slope/characteristics of the lowpass filter - yes?
Yes, that's right—the new "potential" audio band is now 0-96 kHz. But the audio we recorded is still 0-24 kHz. (It's really more like 0-20 kHz for the recorded audio, of course. We generally try for 20k when sampling at 44.1k, and we try for 20k when sampling at 48k—the latter gives us a slight advantage in filter choices. But I use "below half the sample rate", as the theoretical space.)

We're talking about oversampling for conversion, where the only advantage is that we can do some difficult filtering computationally, in the digital domain, to make less work on the analog analog hardware side, with the DAC. And linear phase filters are only possible when we can buffer to simulate looking into the future (of a delayed signal), which is only practical in the digital domain.

But, oversampling is also commonly used when we need frequency headroom. Non-linear processes, such as simulating an over-driven tube amp, which generates harmonics. We need room for them to die out so that the mirrored images don't extend down into the 20k band with enough amplitude to be noticed. Or pitch shifting, to avoid the possibility the top end will mirror back into the audio band.
 

voodooless

Grand Contributor
Forum Donor
Joined
Jun 16, 2020
Messages
10,223
Likes
17,799
Location
Netherlands
@PassionforSound
My attempts to discuss the M-Scaler review on ASR showed an unwillingness to discuss the relevance of the measurements and conclusions drawn.
What unwillingness is that? There was a discussion going on, you just stopped responding to questions posed to you. If there is somebody unwilling, it’s surely you. We’re still here..
Instead the emphasis was on the argument that the M-Scaler can't do anything because it can't do anything.
No, the emphasis was on your way of conducting listening tests, and why your reasoning that these are valid is totally ridiculous.
I've got the DAVE here now (going home today) and it sounds great. If it were so flawed, it would not perform as it does in listening tests.
Yes it would, and Amir said as much in his video review. The fact that you don’t seem to understand this says a lot.
It's not my favourite DAC and I have no desire to buy one (even if I could afford it), but ASR are doing a disservice with their butchering of these products that it seems they don't understand beyond "digital in, analog out".
It’s either badly engineered, or broken. Either way, it’s bad. Regardless of how audible the issues are, for this kind of money, I’d expect perfection.
 

Tony gw

New Member
Joined
Jul 26, 2022
Messages
1
Likes
0
Hello everyone, would like to add my bit to the discussion and thanks to everybody's input. I own the TT2 and the M Scaler but we're bought at different points and before I bought them, was and still using Yamaha AS3000 amplifier and the matching CDS3000 SACD player. Speakers are PMC Fact12 and Audioquest cables with balanced connection. The Cd player is the transport to feed the TT2 via toslink.

There is an old saying, "you don't know what you're had until it's gone!" I learnt this when I bought the M Scaler to go with the TT2 but when connecting it up, the TT2 failed due to faulty board contact via BNCs. So I had to send it back with a three page letter to John Franks who owns the company and let Ripp!

So stuck with the M Scaler and my Yamaha gear. I first set up the Yamaha due and on first listening started to question why I bought this bloody Chord gear! But after a while, l found myself loosing interest and skipping tracks and not enjoying the tunes I was so use too playing. Turned it off after a while. This happened again and again until reality set in. The Chord gear is doing something the Yamaha is not, being this Cd player has the famed ESS 9018 chip set. Has all the details but none of the musics life and soul.

I then connect the M Scaler to the Cd player as it's a up sampling DAC too, and used a Blu ray player as a transport via toslink. It made a small improvement to the sound but nothing that justifies the price of adding it. So there is something to be said in the measurements that the Yamaha,s DAC didn't like it like some DACs do.

However, Mr Franks must of read my letter as I was given a brand new TT2 unit. So both units are new and will require many hours to burn in. But from the off, the TT2 and M Scaler combo was a revelation and the jumping from 2+4+16 times then back to bypass was clearly audible and substantially more rewarding and has only improved over the running in process.

I don't get all the technical background babble spoken here and all the measurements details. But do remember, the Japanese Hifi industry prided it's self many many years ago on measurements and technology and specks, But never had a good reputation for sounding that good, but whatever is going on in these units is something I have not heard before in 35+ years with digital audio.

There is a saying where I am from.

" IT DOES WHAT IT SAYS ON THE TIN"

Believe me, this company and this designer are truly respected. I now know why. Please get your hands on a set and listen ☺️
 

iamsms

Member
Joined
Dec 14, 2020
Messages
35
Likes
126
You're welcome, I'm glad that was helpful. And sample repetition is assumed by a lot of people, don't feel dumb. It doesn't make intuitive sense to use zeros and make the sine wave (that most people imagine the process with) look discontinuous. If you want a deeper, but relatively intuitive view of sampling, check out my series here:

Sampling theory, the best explanation you’ve ever heard

As far as increasing the sample rate, it doesn't change any frequencies. But since the "Fs", the sampling frequency was raised, sidebands that were a result of the sampling process are now in the new, wider audio band. (By "audio band", I mean everything below half the sample rate.) So, we need to clear everything but the original signal with a lowpass filter.

OK, some might have a little problem with the idea that things go on forever, but remember we're just doing math, where things are pure. The same happens in analog sampling, but pulses are imperfect and the images die out as frequency response droops—it still works because we plan to eliminate them with a lowpass anyway. Perfect impulses have no limit to their frequency content, real impulses do. But by keeping only one reading per sample and discarding the rest of the analog waveform, we simulated a perfect impulse. Just saying this to prep for the fact that PCM is a form of amplitude modulation (AM, like the radio), with a perfect impulse, therefore sideband copies—plus and minus—exist mirrored around multiples of the sample rate. And any changes to the part that's in the audio band changes those images as well. We get rid of the images in converting back to analog. So, if we start with a lowpass filtered audio signal that ends up with this bandwidth:

View attachment 220634
After sampling, we have this in the frequency domain:

View attachment 220635

Note that that the original band is "mirrored" around Fs (and 2Fs, and 3Fs...), it's reversed below it, and is the forward but shifted in frequency above it.

If we zero-stuff to up the sample rate, the result is this—we only changed where Fs is:

View attachment 220636

But note that if we look at the new audio band, that below half Fs, we have those mirrored images. They are ultrasonic, but if we run that through a DAC at the higher rate, those frequencies will be in the analog output.

To get back to the audio we started with, we just lowpass filter (again, this and all processes are mirrored) to keep the audio below half the original rate:

View attachment 220638

If you wanted to play this out a DAC with 20 kHz audio bandwidth, even a very gentle filter will work great—it just needs to ensure the red part is removed.
You have an excellent knack for explaining things :). Those diagrams are all one need to know about why upsampling is done in audio. It has been 12+ years since I last played with DSP, but it was neat reading your comments.

Just to confirm, with relatively sharp filters available (especially the ones in CHORD's own dacs), do we really need a lot of frequency headroom?
 

Ken1951

Addicted to Fun and Learning
Joined
Sep 28, 2020
Messages
850
Likes
1,775
Location
Blacksburg, VA
Hello everyone, would like to add my bit to the discussion and thanks to everybody's input. I own the TT2 and the M Scaler but we're bought at different points and before I bought them, was and still using Yamaha AS3000 amplifier and the matching CDS3000 SACD player. Speakers are PMC Fact12 and Audioquest cables with balanced connection. The Cd player is the transport to feed the TT2 via toslink.

There is an old saying, "you don't know what you're had until it's gone!" I learnt this when I bought the M Scaler to go with the TT2 but when connecting it up, the TT2 failed due to faulty board contact via BNCs. So I had to send it back with a three page letter to John Franks who owns the company and let Ripp!

So stuck with the M Scaler and my Yamaha gear. I first set up the Yamaha due and on first listening started to question why I bought this bloody Chord gear! But after a while, l found myself loosing interest and skipping tracks and not enjoying the tunes I was so use too playing. Turned it off after a while. This happened again and again until reality set in. The Chord gear is doing something the Yamaha is not, being this Cd player has the famed ESS 9018 chip set. Has all the details but none of the musics life and soul.

I then connect the M Scaler to the Cd player as it's a up sampling DAC too, and used a Blu ray player as a transport via toslink. It made a small improvement to the sound but nothing that justifies the price of adding it. So there is something to be said in the measurements that the Yamaha,s DAC didn't like it like some DACs do.

However, Mr Franks must of read my letter as I was given a brand new TT2 unit. So both units are new and will require many hours to burn in. But from the off, the TT2 and M Scaler combo was a revelation and the jumping from 2+4+16 times then back to bypass was clearly audible and substantially more rewarding and has only improved over the running in process.

I don't get all the technical background babble spoken here and all the measurements details. But do remember, the Japanese Hifi industry prided it's self many many years ago on measurements and technology and specks, But never had a good reputation for sounding that good, but whatever is going on in these units is something I have not heard before in 35+ years with digital audio.

There is a saying where I am from.

" IT DOES WHAT IT SAYS ON THE TIN"

Believe me, this company and this designer are truly respected. I now know why. Please get your hands on a set and listen ☺️
:facepalm:
 

Jimi Floyd

Active Member
Joined
May 5, 2022
Messages
143
Likes
584
Location
Pisa, Italy
Hello everyone, would like to add my bit to the discussion and thanks to everybody's input. I own the TT2 and the M Scaler but we're bought at different points and before I bought them, was and still using Yamaha AS3000 amplifier and the matching CDS3000 SACD player. Speakers are PMC Fact12 and Audioquest cables with balanced connection. The Cd player is the transport to feed the TT2 via toslink.

There is an old saying, "you don't know what you're had until it's gone!" I learnt this when I bought the M Scaler to go with the TT2 but when connecting it up, the TT2 failed due to faulty board contact via BNCs. So I had to send it back with a three page letter to John Franks who owns the company and let Ripp!

So stuck with the M Scaler and my Yamaha gear. I first set up the Yamaha due and on first listening started to question why I bought this bloody Chord gear! But after a while, l found myself loosing interest and skipping tracks and not enjoying the tunes I was so use too playing. Turned it off after a while. This happened again and again until reality set in. The Chord gear is doing something the Yamaha is not, being this Cd player has the famed ESS 9018 chip set. Has all the details but none of the musics life and soul.

I then connect the M Scaler to the Cd player as it's a up sampling DAC too, and used a Blu ray player as a transport via toslink. It made a small improvement to the sound but nothing that justifies the price of adding it. So there is something to be said in the measurements that the Yamaha,s DAC didn't like it like some DACs do.

However, Mr Franks must of read my letter as I was given a brand new TT2 unit. So both units are new and will require many hours to burn in. But from the off, the TT2 and M Scaler combo was a revelation and the jumping from 2+4+16 times then back to bypass was clearly audible and substantially more rewarding and has only improved over the running in process.

I don't get all the technical background babble spoken here and all the measurements details. But do remember, the Japanese Hifi industry prided it's self many many years ago on measurements and technology and specks, But never had a good reputation for sounding that good, but whatever is going on in these units is something I have not heard before in 35+ years with digital audio.

There is a saying where I am from.

" IT DOES WHAT IT SAYS ON THE TIN"

Believe me, this company and this designer are truly respected. I now know why. Please get your hands on a set and listen ☺️
7848500110_ff44408cf2_w.jpg
 

DavidEdwinAston

Addicted to Fun and Learning
Forum Donor
Joined
Nov 18, 2021
Messages
754
Likes
566
Hello everyone, would like to add my bit to the discussion and thanks to everybody's input. I own the TT2 and the M Scaler but we're bought at different points and before I bought them, was and still using Yamaha AS3000 amplifier and the matching CDS3000 SACD player. Speakers are PMC Fact12 and Audioquest cables with balanced connection. The Cd player is the transport to feed the TT2 via toslink.

There is an old saying, "you don't know what you're had until it's gone!" I learnt this when I bought the M Scaler to go with the TT2 but when connecting it up, the TT2 failed due to faulty board contact via BNCs. So I had to send it back with a three page letter to John Franks who owns the company and let Ripp!

So stuck with the M Scaler and my Yamaha gear. I first set up the Yamaha due and on first listening started to question why I bought this bloody Chord gear! But after a while, l found myself loosing interest and skipping tracks and not enjoying the tunes I was so use too playing. Turned it off after a while. This happened again and again until reality set in. The Chord gear is doing something the Yamaha is not, being this Cd player has the famed ESS 9018 chip set. Has all the details but none of the musics life and soul.

I then connect the M Scaler to the Cd player as it's a up sampling DAC too, and used a Blu ray player as a transport via toslink. It made a small improvement to the sound but nothing that justifies the price of adding it. So there is something to be said in the measurements that the Yamaha,s DAC didn't like it like some DACs do.

However, Mr Franks must of read my letter as I was given a brand new TT2 unit. So both units are new and will require many hours to burn in. But from the off, the TT2 and M Scaler combo was a revelation and the jumping from 2+4+16 times then back to bypass was clearly audible and substantially more rewarding and has only improved over the running in process.

I don't get all the technical background babble spoken here and all the measurements details. But do remember, the Japanese Hifi industry prided it's self many many years ago on measurements and technology and specks, But never had a good reputation for sounding that good, but whatever is going on in these units is something I have not heard before in 35+ years with digital audio.

There is a saying where I am from.

" IT DOES WHAT IT SAYS ON THE TIN"

Believe me, this company and this designer are truly respected. I now know why. Please get your hands on a set and listen ☺️
I certainly don't get the technical background babble spoken here either Tony.
However, I have come to understand that unless "you", complete proper blind tests on your equipment., Your opinion is less than valid.
 

Blumlein 88

Grand Contributor
Forum Donor
Joined
Feb 23, 2016
Messages
20,521
Likes
37,050
Hello everyone, would like to add my bit to the discussion and thanks to everybody's input. I own the TT2 and the M Scaler but we're bought at different points and before I bought them, was and still using Yamaha AS3000 amplifier and the matching CDS3000 SACD player. Speakers are PMC Fact12 and Audioquest cables with balanced connection. The Cd player is the transport to feed the TT2 via toslink.

There is an old saying, "you don't know what you're had until it's gone!" I learnt this when I bought the M Scaler to go with the TT2 but when connecting it up, the TT2 failed due to faulty board contact via BNCs. So I had to send it back with a three page letter to John Franks who owns the company and let Ripp!

So stuck with the M Scaler and my Yamaha gear. I first set up the Yamaha due and on first listening started to question why I bought this bloody Chord gear! But after a while, l found myself loosing interest and skipping tracks and not enjoying the tunes I was so use too playing. Turned it off after a while. This happened again and again until reality set in. The Chord gear is doing something the Yamaha is not, being this Cd player has the famed ESS 9018 chip set. Has all the details but none of the musics life and soul.

I then connect the M Scaler to the Cd player as it's a up sampling DAC too, and used a Blu ray player as a transport via toslink. It made a small improvement to the sound but nothing that justifies the price of adding it. So there is something to be said in the measurements that the Yamaha,s DAC didn't like it like some DACs do.

However, Mr Franks must of read my letter as I was given a brand new TT2 unit. So both units are new and will require many hours to burn in. But from the off, the TT2 and M Scaler combo was a revelation and the jumping from 2+4+16 times then back to bypass was clearly audible and substantially more rewarding and has only improved over the running in process.

I don't get all the technical background babble spoken here and all the measurements details. But do remember, the Japanese Hifi industry prided it's self many many years ago on measurements and technology and specks, But never had a good reputation for sounding that good, but whatever is going on in these units is something I have not heard before in 35+ years with digital audio.

There is a saying where I am from.

" IT DOES WHAT IT SAYS ON THE TIN"

Believe me, this company and this designer are truly respected. I now know why. Please get your hands on a set and listen ☺️
Every part of your described experience is psychological and unrelated to the signals you were listening to. Our minds are quite the trickster.
 

danadam

Addicted to Fun and Learning
Joined
Jan 20, 2017
Messages
956
Likes
1,496
Bonus points for illustrating the (less interesting) integer down-sampling case?!?
Just FYI, I already did a similar thing but without the samples graph:

Let's start with a mix of 2 tones:
  • 2000 Hz at -18 dBFS
  • 7500 Hz at -3 dBFS
at 16 kHz sampling rate and 8 bits (all generated files are in the attachment):
Code:
sox -r16k -n -b8 "01_two_tones.8_16.wav" synth 5 sin 2000 sin 7500 remix 1v0.125,2v0.7
01_two_tones.8_16.png


If we do the wrong thing and just pick every second sample, the 7500 Hz tone will get aliased as 500 Hz tone:
Code:
sox "01_two_tones.8_16.wav" -r8k "02_just_downsample.8_8.wav" downsample 2
02_just_downsample.8_8.png


The correct thing to do is to first filter out everything that is above half of the new sampling rate:
Code:
sox "01_two_tones.8_16.wav" "03_low_pass.8_16.wav" sinc -t 400 -3800
03_low_pass.8_16.png


And only then pick every second sample:
Code:
sox "03_low_pass.8_16.wav" -r8k "04_then_downsample.8_8.wav" downsample 2
04_then_downsample.8_8.png
 

Attachments

  • downsampling_examples.zip
    70.6 KB · Views: 39

tmtomh

Major Contributor
Forum Donor
Joined
Aug 14, 2018
Messages
2,635
Likes
7,485
You have an excellent knack for explaining things :). Those diagrams are all one need to know about why upsampling is done in audio. It has been 12+ years since I last played with DSP, but it was neat reading your comments.

Just to confirm, with relatively sharp filters available (especially the ones in CHORD's own dacs), do we really need a lot of frequency headroom?

This was one of my questions too: Chord touts 16x oversampling for "transient restoration." But given that that's not what oversampling does, we're left with @danadam 's and @earlevel 's excellent explanations of what it does do, which is what most of us thought (even if some of us, <cough>me</cough> initially didn't quite grok exactly *how* it does it :) ) - creating more frequency headroom for filtering, as you say.

But @amirm has shown that Chord's filtering is extraordinarily sharp (and on that count, well done Chord!). So it would seem there's no need for an outboard extra-upsampler like the M-Scaler if you are using a DAC (like a Chord DAC) that has such a filter.
 
Last edited:
Top Bottom