• 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
OP
amirm

amirm

Founder/Admin
Staff Member
CFO (Chief Fun Officer)
Joined
Feb 13, 2016
Messages
44,376
Likes
234,538
Location
Seattle Area
People have compared filter-less NOS dacs with delta-sigma dacs, reported audible differences, and even stated some advantages. How I imagine one might reply to this on this site is by saying that they tricked themselves or that their brain did, that the distortion gave to a sound effect or that they're enjoying the distortion. The reason I am not quick to dismiss these findings for those reasons is because a double blind test would have to be conducted comparing both.
So? Did you conduct such a test? If not, what on earth are you talking about?
 

Blumlein 88

Grand Contributor
Forum Donor
Joined
Feb 23, 2016
Messages
20,524
Likes
37,057
Broken by design how?

I already stated that 100khz isn't necessary but that there is room for improvement.

I am not talking about spatial discernment, someone else brought that up, I am talking about percussive discernment at the very least.
solderdude has already answered. No reconstruction filter. Means it will have drooping response in the top 2 octaves, and will have imaging artifacts some of which may intermodulate to add components of sound back down into the audible region. The FR alone will make it sound different. Transients with high frequency content are more likely to intermodulate too. Not surprising it might sound different. I've not listened to any myself. If I did and it really sounded different I would expect the above reasons to be why.
 

HP9000

Member
Joined
Aug 20, 2022
Messages
17
Likes
6
No, that waveform will have infinite bandwidth now due to discontinuities at the edges of half cycles. Rest of your paragraph seems like some kind of automated text generation.
I'm not sure what you mean by the edges of the half cycles of the waveform I described, unless you mean that there would be a concave area if graphed, in this case it would just be reversed so that it is convex.
I can say that whatever about it that you're saying requires infinite bandwidth can be made quasi-infinite and it still stands that this shape is not the one you said demonstrates inaudibility (1khz +/- 100khz).

Did you conduct such a test?
I don't really have the resources now.
what on earth are you talking about?
In large part, the audibility of impulse response ringing and rise and fall time.
 

Geert

Major Contributor
Joined
Mar 20, 2020
Messages
1,938
Likes
3,526
I can say that whatever about it that you're saying requires infinite bandwidth can be made quasi-infinite and it still stands that this shape is not the one you said demonstrates inaudibility (1khz +/- 100khz)

If we only could imagine what your waveform looks like...
 

HP9000

Member
Joined
Aug 20, 2022
Messages
17
Likes
6
If we only could imagine what your waveform looks like...
I don't know why you say this when it's not far off from a typical quasi square wave with a 'noisy' sustain, I was just going into detail by using the numbers amirm used which were 1khz and 100khz.
 

Blumlein 88

Grand Contributor
Forum Donor
Joined
Feb 23, 2016
Messages
20,524
Likes
37,057
I don't know why you say this when it's not far off from a typical quasi square wave with a 'noisy' sustain, I was just going into detail by using the numbers amirm used which were 1khz and 100khz.
You are talking about chopping up a square wave I think. So you are effectively converting it into an even higher frequency. One which our hearing cannot and will not respond to in the first place. I cannot understand what you think you are showing by this.
 

HP9000

Member
Joined
Aug 20, 2022
Messages
17
Likes
6
You are talking about chopping up a square wave I think. So you are effectively converting it into an even higher frequency. One which our hearing cannot and will not respond to in the first place. I cannot understand what you think you are showing by this.
No. I'll try to draw a picture of what I was saying. My next post will be my last on this thread.
 

solderdude

Grand Contributor
Joined
Jul 21, 2018
Messages
15,891
Likes
35,912
Location
The Neitherlands
No. I'll try to draw a picture of what I was saying. My next post will be my last on this thread.
So no more discussion about your theories ?

Can you post (zoomed in) real world music recordings showing the need for a high bandwidth/rise/fall times ?
 

Blumlein 88

Grand Contributor
Forum Donor
Joined
Feb 23, 2016
Messages
20,524
Likes
37,057
No. I'll try to draw a picture of what I was saying. My next post will be my last on this thread.
I'd ask if you have some examples of the type of sounds that require this extended response (if that is what you are referring too as I'm still foggy on what you have in mind)?

Also taking a guess, maybe high energy cymbals? I have some short recordings of cymbals strikes done at high sample rates. They are interesting in what they contain versus what is imagined.
 
Last edited:

danadam

Addicted to Fun and Learning
Joined
Jan 20, 2017
Messages
957
Likes
1,497
I am not talking about spatial discernment, someone else brought that up
I wonder who:
Our tonal sensitivity is limited to around 20khz, and our sensitivity to amplitude changes in time - of a given audible (fundamental) frequency and a given audible difference in amplitude - is limited to an equivalent of 100khz, which is 5 times as high.
Reference?
"Reference?"
At 13:05 in the video by amirm that is linked he says that time differential is a main factor in localization, and we can ask: what amount of time?
 

Geert

Major Contributor
Joined
Mar 20, 2020
Messages
1,938
Likes
3,526
I'd ask if you have some examples of the type of sounds that require this extended response

Please have a listen:

48px-Ic_play_circle_outline_48px.svg.png
 

DSJR

Major Contributor
Joined
Jan 27, 2020
Messages
3,314
Likes
4,427
Location
Suffolk Coastal, UK
??
Surely all that’s required is to switch the M-scaler in and out in a properly conducted (level matched etc) DBT
But it seems that's not what 'they' do... There appears to be a LEVEL difference usually when lay people and Chord's staff do it and that's what is heard as an 'improvement.'
 

sergeauckland

Major Contributor
Forum Donor
Joined
Mar 16, 2016
Messages
3,440
Likes
9,100
Location
Suffolk UK
But it seems that's not what 'they' do... There appears to be a LEVEL difference usually when lay people and Chord's staff do it and that's what is heard as an 'improvement.'
Here's a great wheeze, create a box that does nothing other than increase the output level by 1dB. Create all sorts of mystique for it with claims of millions of taps (or faucets as you may know them as), package it in a solid aluminium - forged in the white heat of technology, then market the shit out of it.

Genius?
S.
 

Purité Audio

Master Contributor
Industry Insider
Barrowmaster
Forum Donor
Joined
Feb 29, 2016
Messages
9,051
Likes
12,150
Location
London
I see the Mscaler as akin to this,
Keith
 

voodooless

Grand Contributor
Forum Donor
Joined
Jun 16, 2020
Messages
10,229
Likes
17,811
Location
Netherlands
A time value like 10us and 100khz mean the exact same thing, the only translation being made are the abbreviations which are synonymous. This is because frequency denotes to an interval/time window, which is just what the list of seconds, milliseconds etc. denote to. Same denotation different connotations, and not to an extent to where the Fourier transfer need be mentioned.
Okay, so here we are again, this time with a serious answer. Looks like you've never bothered reading one of the previous posts linking to an article about how timing actually works, so let's just do if with a few graphics, that I kindly bastardized from Archimago ;)

So here is a 10 kHz sine wave, one NOS, one oversampled. Obviously, this is a crappy NOS DAC, so worse than a bit more modern one, but that doesn't really matter to make the point. Below is a zoomed and enhanced version (CSI style :facepalm:;) ) that I overlayed. Already from the top image it's clear that the NOS DAC cannot reproduce the 10 kHz sine.
NOS.png

Here you see The actual samples and the staircase that the NOS DAC creates for it. Claiming that the NOS DAC has better timing accuracy is ludicrous. In the first red circle, the NOS DAC is way too slow to catch up with the sine, while in the second one, it's too fast! The triangle in the middle shows the rough error the NOS DAC creates. That is very significant.

But, you say: I want a much higher sample rate. Well yes, as we've already stated, that just shifts these issues up in frequency, but they are still there. NOS is broken from the start, it's just plain wrong. It's easy to see why:

- The NOS DAC holds the value samples at one instance in time until the next sample
- The sample value is however only valid for the moment of the sampling, nothing is known from the rest of the time between samples.
- It's not even correct on average, because for that to be true, the ADC process would need to average the value over the sample time, to begin with.

Now back to time resolution: this is not a function of the sample rate. And why should be very clear:

- We have a fixed sample interval. There is no wiggle room there, we cannot deviate from this!
- The actual sample value is quantized based on the bit depth.
- The error of this quantization will become lower if the bit depth becomes higher:
1661266821677.png

This is clearly seen here. The higher the bit depth, the lower the error from the actual sample value. But since we're talking about a continuous signal, that quantized value will probably be valid at some point in time just before or just after the sample moment. And that is where the time resolution kicks in. This "some time in the past or future" is lower with a higher bit depth:
1661267390972.png

Here we see the actual signal in red, the sample values in yellow, and the time errors are the green arrows. Obviously, with NOS, none of this works because it's totally broken technology. Your timing error is basically always increasing towards one whole sample time. This actually means that the time resolution of a NOS DAC is lower than a properly working oversampling DAC ;)

If you still think all of this is wrong, please provide an example with actual band-limited sound content (so no square waves) where a NOS DAC gives an actual timing advantage.

200w.gif
 
Last edited:

HP9000

Member
Joined
Aug 20, 2022
Messages
17
Likes
6
Final comment:

@voodooless You can't say that something is broken by design because it is lacking something if there's nothing stopping you from adding to it. Something is broken by design if it does not allow you to add to or subtract from it. The latter is the issue with delta-sigma dacs being that oversampling is innate to its function. With a filter-less NOS dac you get to decide what kind of reconstruction filter is used; IIR, FIR, high order, low order, etc. More so, you get to decide how much you affect the impulse response. For chip dacs, there are filters named 'NOS', but we know that it is still oversampling. It only serves so much convenience to have onboard oversampling for filtering's sake when it will be inconvenient to be left with irreversible impulse ringing as a result of that oversampling even in the case of playing high bit-rate, high sample-rate, correct ADC material into it.
These problems are really what is wrong with class d amps because all of the slew rate of the transistor is spent on switching. It is not the switching itself that is the problem, but that because it is switching it cannot be directly controlled by the input, it requires the input be probed to activate the switch. This probing frequency results in quantization error, even in the case of 'digital amplifiers'.

@danadam All your comment shows is that amirm was first to bring it up with his reply linking his video on phase, because it occurs in the video and I didn't skip it. Then @Geert said that absolute time resolution isn't indicated by the reference I provided, that it pertains strictly to spatially. I made a comment about how the number hasn't really changed even after further investigation.
In terms of temporal resolution (not regarding difference between two ears/spatiality) there are two numbers; essentially milliseconds, but with a lot of range (0.1, 0.5, 1, 2, 3), and tens of microseconds (22.7, 25). I subscribe to the latter.
Milliseconds:
3 ms at best ('Temporal Resolution in Normal-Hearing and Hearing-Impaired Listeners Using Frequency-Modulated Stimuli', John P. Madden, Lawrence Feth, 1992)
"A quick lit search gives me 1.2ms (Irwin & Purdy, 1982). Also, people were shown to be able to recognize voice when stimuli were only 2ms long (Suied, et al., 2013)!" From the reddit link below.
Microseconds: "I would cut your 1/20,000 in half actually to 1/40,000. Just because that...", "Well I am not really aware of such studies. I can direct you to this file...", "If you create a sound file at 44100 Hz with only silence in it, and set exactly one sample...", "This sounds like a guess to me. What biological processes..." etc.
From: https://sound.stackexchange.com/que...e-shortest-sound-perceptible-to-the-human-ear, https://www.reddit.com/r/askscience/comments/2cbogd
I may edit this in the future to add a picture of the waveform I described earlier.

Let's say - simply going by the tonal limit of our hearing - that 20khz, or 1/20,000th of a second, is the ceiling in terms of our ability to sense a rise time difference for a frequency like 1khz. One might say that this is the limit, so being served a sound containing elements above it would be inaudible, while I'm saying that something like 100khz can be sensed as a difference in rise time. The problem with the idea that "this is the limit, we needn't serve the ear a sound of a more sudden nature because it won't be audible" is that it's a half truth because served a sound that just meets this limit you will actually be hearing something below it given that the ear acts as an impedance to the sound energy. You might not actually be hearing that 1/20,000th sec rise time until you have a sound with 5x that to do the work.*
I have read people saying things like "sounds with a more sudden nature are perceived as louder", and I can't help but feel that in situations like this the words "sounds like" or "perceived" are implying that it comes down to psychoacoustics, but it's not that sophisticated. It is due to that waveform actually having more energy, so it becomes an issue of the kinetic transfer of energy.

Just because a sine wave reaches its peak and extends in time enough to amount to an audible frequency like 1khz doesn't mean it has filled in that time with as much energy as possible as if to do any more it would require a higher peak or a larger window of time (lower frequency).

A frequency (let's stick with 1khz) with the same peak level but a more sudden start and end in reaching and coming down from that peak packs more energy in the same unit of time than the same frequency with the same peak level having a less sudden start and stop. This is also why the fall time contributes to percussive quality. Think of it like a battering ram but in time not just space. This is why a square wave will sound more percussive than a sawtooth wave of the high attack kind, because it's not just about the attack.

This can also explain why square waves at higher frequencies still audible as tones matter less, because the difference 'filling in' the time window with energy makes becomes more and more marginal.

A high frequency like 100khz alone repeated at a high level won't have a tonal register to our ears because the energy is dispersed by being fractioned in time, there is lots of negative space, there is a lot of time where nothing is happening/there is no energy. This can explain in part why we can't hear these frequencies as tones. In the above, an ultrasonic frequency equivalent rise/fall time has the support of a tonally audible frequency, which is another way of saying it has more energy.

*This shouldn't be read as if there should be compensation for waveforms, it's to say that if you take the time distortion (affected impulse, rise & fall time) away it will sound as it should. It also shouldn't be read as if it only applies to such sudden rise times, this sudden rise time is being used to discuss the limits.

Since this got some backlash:
If we Google search period vs frequency we are told they are opposite. Really, the period is not different for being 'centered between the cycle' because the frequency energy is only occurring over that center. A helpful visual is where there is shading of the wave, which correctly indicates that there is energy and where it is. It's not the lines used to draw the symbols that are the indicators.
"Frequency refers to how often something happens. Period refers to the time it takes something to happen. Frequency is a rate quantity. Period is a time quantity."
While this understanding may be relevant in other contexts, the problem here is obviously that the frequency is being abstracted, when we know that frequency can regard causality (i.e. energy), which is best understood as temporal.

To go back to the Rehdeko speaker, it isn't really known for its soundstage, extension, or ability to play busy tracks. It even has a resonating cabinet which I would say allows us to locate the speaker and detracts from spatiality. Despite this there is still good system Q/damping factor.

It's not just about spatial separation, it's about timbral separation. When I'm talking about single sounds sound summing into one sound wave I'm referring to sound awareness theory (I forget exactly what it's called but it has to do with this). Part of timbre is ADSR, and part of what's going to contribute to our sense of separation of sounds in space is timbral difference.
 
Last edited:

Sir Sanders Zingmore

Addicted to Fun and Learning
Forum Donor
Joined
May 20, 2018
Messages
948
Likes
1,902
Location
Melbourne, Australia
Final comment:

@voodooless You can't say that something is broken by design because it is lacking something if there's nothing stopping you from adding to it. Something is broken by design if it does not allow you to add to or subtract from it. The latter is the issue with delta-sigma dacs being that oversampling is innate to its function. With a filter-less NOS dac you get to decide what kind of reconstruction filter is used; IIR, FIR, high order, low order, etc. More so, you get to decide how much you affect the impulse response. For chip dacs, there are filters named 'NOS', but we know that it is still oversampling. It only serves so much convenience to have onboard oversampling for filtering's sake when it will be inconvenient to be left with irreversible impulse ringing as a result of that oversampling even in the case of playing high bit-rate, high sample-rate, correct ADC material into it.
These problems are really what is wrong with class d amps because all of the slew rate of the transistor is spent on switching. It is not the switching itself that is the problem, but that because it is switching it cannot be directly controlled by the input, it requires the input be probed to activate the switch. This probing frequency results in quantization error, even in the case of 'digital amplifiers'.

@danadam All your comment shows is that amirm was first to bring it up with his reply linking his video on phase, because it occurs in the video and I didn't skip it. Then @Geert said that absolute time resolution isn't indicated by the reference I provided, that it pertains strictly to spatially. I made a comment about how the number hasn't really changed even after further investigation.
In terms of temporal resolution (not regarding difference between two ears/spatiality) there are two numbers; essentially milliseconds, but with lots of range (0.1, 0.5, 1, 2, 3), and tens of microseconds (22.7, 25). I subscribe to the latter.
Milliseconds:


"A quick lit search gives me 1.2ms (Irwin & Purdy, 1982). Also, people were shown to be able to recognize voice when stimuli were only 2ms long (Suied, et al., 2013)!" From the reddit link below.
Microseconds: "I would cut your 1/20,000 in half actually to 1/40,000. Just because that...", "Well I am not really aware of such studies. I can direct you to this file...", "If you create a sound file at 44100 Hz with only silence in it, and set exactly one sample...", "This sounds like a guess to me. What biological processes..." etc.
From: https://www.reddit.com/r/askscience/comments/2cbogd
I may edit this in the future to add a picture of the waveform I described earlier.

Let's say - simply going by the tonal limit of our hearing - that 20khz, or 1/20,000th of a second, is the ceiling in terms of our ability to sense a rise time difference for a frequency like 1khz. One says this is the limit, so being served sound containing elements above it are inaudible, while I'm saying that something like 100khz can be sensed as a difference in rise time. The problem with the idea that "this is the limit, we needn't serve the ear a sound of a more sudden nature because it won't be audible" is that it's a half truth because served a sound that just meets this limit you will actually be hearing something below it given that the ear acts as an impedance to the sound energy. You might not actually be hearing that 1/20,000th sec rise time until you have a sound with 5x that to do the work.*
I have read people saying things like "sounds with a more sudden nature are perceived as louder", and I can't help but feel that in situations like this the words "sounds like" or "perceived" are implying that it comes down to psychoacoustics, but it's not that sophisticated. It is due to that waveform actually having more energy, so it becomes an issue of the kinetic transfer of energy.

Just because a sine wave reaches its peak and extends in time enough to amount to an audible frequency like 1khz doesn't mean it has filled in that time with as much energy as possible as if to do any more it would require a higher peak or a larger window of time (lower frequency).

A frequency (let's stick with 1khz) with the same peak level but a more sudden start and end in reaching and coming down from that peak packs more energy in the same unit of time than the same frequency with the same peak level having a less sudden start and stop. This is also why the fall time contributes to percussive quality. Think of it like a battering ram but in time not just space. This is why a square wave will sound more percussive than a sawtooth wave of the high attack kind, because it's not just about the attack.

This can also explain why square waves at higher frequencies still audible as tones matter less, because the difference 'filling in' the time window with energy makes becomes more and more marginal.

A high frequency like 100khz alone repeated at a high level won't have a tonal register to our ears because the energy is dispersed by being fractioned in time, there is lots of negative space, there is a lot of time where nothing is happening/there is no energy. This can explain in part why we can't hear these frequencies as tones. In the above, an ultrasonic frequency equivalent rise/fall time has the support of a tonally audible frequency, which is another way of saying it has more energy.

*This shouldn't be read as if there should be compensation for waveforms, it's to say that if you take the time distortion (affected impulse, rise & fall time) away it will sound as it should. It also shouldn't be read as if it only applies to such sudden rise times, this sudden rise time is being used to discuss the limits.

Since this got some backlash:
If we Google search period vs frequency we are told they are opposite. Really, the period is not different for being 'centered between the cycle' because the frequency energy is only occurring over that center. A helpful visual is where there is shading of the wave, which correctly indicates that there is energy and where it is. It's not the lines used to draw the symbols that are the indicators.
"Frequency refers to how often something happens. Period refers to the time it takes something to happen. Frequency is a rate quantity. Period is a time quantity."
While this understanding may be relevant in other contexts, the problem here is obviously that the frequency is being abstracted, when we know that frequency can regard causality (i.e. energy), which is best understood as temporal.

To go back to the Rehdeko speaker, it isn't really known for its soundstage, extension, or ability to play busy tracks. It even has a resonating cabinet which I would say allows us to locate the speaker and detracts from spatiality. Despite this there is still good system Q/damping factor.

It's not just about spatial separation, it's about timbral separation. When I'm talking about single sounds sound summing into one sound wave I'm referring to sound awareness theory (I forget exactly what it's called but it has to do with this). Part of timbre is ADSR, and part of what's going to contribute to our sense of separation of sounds in space is timbral difference.
Why are we continuing to discuss (in great detail) the hypothetical reasons for a non-existent audible difference?
 

tmtomh

Major Contributor
Forum Donor
Joined
Aug 14, 2018
Messages
2,636
Likes
7,497
Final comment:

@voodooless You can't say that something is broken by design because it is lacking something if there's nothing stopping you from adding to it. Something is broken by design if it does not allow you to add to or subtract from it. The latter is the issue with delta-sigma dacs being that oversampling is innate to its function. With a filter-less NOS dac you get to decide what kind of reconstruction filter is used; IIR, FIR, high order, low order, etc. More so, you get to decide how much you affect the impulse response. For chip dacs, there are filters named 'NOS', but we know that it is still oversampling. It only serves so much convenience to have onboard oversampling for filtering's sake when it will be inconvenient to be left with irreversible impulse ringing as a result of that oversampling even in the case of playing high bit-rate, high sample-rate, correct ADC material into it.
These problems are really what is wrong with class d amps because all of the slew rate of the transistor is spent on switching. It is not the switching itself that is the problem, but that because it is switching it cannot be directly controlled by the input, it requires the input be probed to activate the switch. This probing frequency results in quantization error, even in the case of 'digital amplifiers'.

@danadam All your comment shows is that amirm was first to bring it up with his reply linking his video on phase, because it occurs in the video and I didn't skip it. Then @Geert said that absolute time resolution isn't indicated by the reference I provided, that it pertains strictly to spatially. I made a comment about how the number hasn't really changed even after further investigation.
In terms of temporal resolution (not regarding difference between two ears/spatiality) there are two numbers; essentially milliseconds, but with a lot of range (0.1, 0.5, 1, 2, 3), and tens of microseconds (22.7, 25). I subscribe to the latter.
Milliseconds:


"A quick lit search gives me 1.2ms (Irwin & Purdy, 1982). Also, people were shown to be able to recognize voice when stimuli were only 2ms long (Suied, et al., 2013)!" From the reddit link below.
Microseconds: "I would cut your 1/20,000 in half actually to 1/40,000. Just because that...", "Well I am not really aware of such studies. I can direct you to this file...", "If you create a sound file at 44100 Hz with only silence in it, and set exactly one sample...", "This sounds like a guess to me. What biological processes..." etc.
From: https://www.reddit.com/r/askscience/comments/2cbogd
I may edit this in the future to add a picture of the waveform I described earlier.

Let's say - simply going by the tonal limit of our hearing - that 20khz, or 1/20,000th of a second, is the ceiling in terms of our ability to sense a rise time difference for a frequency like 1khz. One says this is the limit, so being served sound containing elements above it are inaudible, while I'm saying that something like 100khz can be sensed as a difference in rise time. The problem with the idea that "this is the limit, we needn't serve the ear a sound of a more sudden nature because it won't be audible" is that it's a half truth because served a sound that just meets this limit you will actually be hearing something below it given that the ear acts as an impedance to the sound energy. You might not actually be hearing that 1/20,000th sec rise time until you have a sound with 5x that to do the work.*
I have read people saying things like "sounds with a more sudden nature are perceived as louder", and I can't help but feel that in situations like this the words "sounds like" or "perceived" are implying that it comes down to psychoacoustics, but it's not that sophisticated. It is due to that waveform actually having more energy, so it becomes an issue of the kinetic transfer of energy.

Just because a sine wave reaches its peak and extends in time enough to amount to an audible frequency like 1khz doesn't mean it has filled in that time with as much energy as possible as if to do any more it would require a higher peak or a larger window of time (lower frequency).

A frequency (let's stick with 1khz) with the same peak level but a more sudden start and end in reaching and coming down from that peak packs more energy in the same unit of time than the same frequency with the same peak level having a less sudden start and stop. This is also why the fall time contributes to percussive quality. Think of it like a battering ram but in time not just space. This is why a square wave will sound more percussive than a sawtooth wave of the high attack kind, because it's not just about the attack.

This can also explain why square waves at higher frequencies still audible as tones matter less, because the difference 'filling in' the time window with energy makes becomes more and more marginal.

A high frequency like 100khz alone repeated at a high level won't have a tonal register to our ears because the energy is dispersed by being fractioned in time, there is lots of negative space, there is a lot of time where nothing is happening/there is no energy. This can explain in part why we can't hear these frequencies as tones. In the above, an ultrasonic frequency equivalent rise/fall time has the support of a tonally audible frequency, which is another way of saying it has more energy.

*This shouldn't be read as if there should be compensation for waveforms, it's to say that if you take the time distortion (affected impulse, rise & fall time) away it will sound as it should. It also shouldn't be read as if it only applies to such sudden rise times, this sudden rise time is being used to discuss the limits.

Since this got some backlash:
If we Google search period vs frequency we are told they are opposite. Really, the period is not different for being 'centered between the cycle' because the frequency energy is only occurring over that center. A helpful visual is where there is shading of the wave, which correctly indicates that there is energy and where it is. It's not the lines used to draw the symbols that are the indicators.
"Frequency refers to how often something happens. Period refers to the time it takes something to happen. Frequency is a rate quantity. Period is a time quantity."
While this understanding may be relevant in other contexts, the problem here is obviously that the frequency is being abstracted, when we know that frequency can regard causality (i.e. energy), which is best understood as temporal.

To go back to the Rehdeko speaker, it isn't really known for its soundstage, extension, or ability to play busy tracks. It even has a resonating cabinet which I would say allows us to locate the speaker and detracts from spatiality. Despite this there is still good system Q/damping factor.

It's not just about spatial separation, it's about timbral separation. When I'm talking about single sounds sound summing into one sound wave I'm referring to sound awareness theory (I forget exactly what it's called but it has to do with this). Part of timbre is ADSR, and part of what's going to contribute to our sense of separation of sounds in space is timbral difference.

We cannot hear or "sense" 100kHz. Stop. Just stop.
 

voodooless

Grand Contributor
Forum Donor
Joined
Jun 16, 2020
Messages
10,229
Likes
17,811
Location
Netherlands
@voodooless You can't say that something is broken by design because it is lacking something if there's nothing stopping you from adding to it. Something is broken by design if it does not allow you to add to or subtract from it.
That's an interesting definition...
The latter is the issue with delta-sigma dacs being that oversampling is innate to its function. With a filter-less NOS dac you get to decide what kind of reconstruction filter is used; IIR, FIR, high order, low order, etc. More so, you get to decide how much you affect the impulse response.
it's like saying: here I have a nice tractor with big chunky wheels. You can have any engine you like, this tiny one, or this massive one. None of it matters if you want a vehicle to drive on a highway. The tractor will only ever be useful for field work in the mud.
For chip dacs, there are filters named 'NOS', but we know that it is still oversampling. It only serves so much convenience to have onboard oversampling for filtering's sake when it will be inconvenient to be left with irreversible impulse ringing as a result of that oversampling even in the case of playing high bit-rate, high sample-rate, correct ADC material into it.
Why cares? With actual music material, you won't see any of it.

Have to quote archimago again:


You think ringing is an issue, while all your NOS audio looks like a staircase that doesn't even represent the actual signal very well? That's just crazy. BTW, a NOS DAC will ring on EVERY single sample: its step response will also not be perfect. It will always overshoot. It's clearly visible in the pictures I posted earlier.

These problems are really what is wrong with class d amps because all of the slew rate of the transistor is spent on switching. It is not the switching itself that is the problem, but that because it is switching it cannot be directly controlled by the input, it requires the input be probed to activate the switch. This probing frequency results in quantization error, even in the case of 'digital amplifiers'.
What? Class D amps are not digital... And even with all your objections, we can have class D amps capable of distortion products below -120 dB, all the way up to 20 kHz:
index.php


I leave the rest to others... For any of that to be relevant you'll first need to prove that NOS DAC's are in any way superior at the timing bit. So far you have not refuted anything I actually said. Still waiting for your actual example to prove anything I said wrong.
 
Last edited:

solderdude

Grand Contributor
Joined
Jul 21, 2018
Messages
15,891
Likes
35,912
Location
The Neitherlands
With a filter-less NOS dac you get to decide what kind of reconstruction filter is used; IIR, FIR, high order, low order, etc.

Absolutely not so.
Only when you upsample (so not make it NOS) you can have an influence on the filter response. One can do the exact same using DS (and upsampling) and decide to use different reconstruction filters than the one (or more) that come with it.

You can't say that something is broken by design

A filterless NOS DAC is broken by design, the world filterless implies this a vital component is missing. The fact that you can circumvent the broken part by adding a filter and MUST upsample in order to make the filter undo the 'broken' aspect says it all.


You have a very weird idea of how class-D operates as well as about a lot of other things.
 
Last edited:
Top Bottom