- Joined
- Feb 23, 2016
- Messages
- 20,525
- Likes
- 37,058
Ringy Dingy
Adding filters and upsampling still leaves it broken. Nothing is undone. you'll just make it a little bit worse. The fact that ones assumes a sample value is valid for the complete sample period makes it broken. All else is collateral.Using a filterless NOS DAC is broken by design. the world filterless implies this. 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.
Adding filters and upsampling still leaves it broken. Nothing is undone. you'll just make it a little bit worse. The fact that ones assumes a sample value is valid for the complete sample period makes it broken. All else is collateral.
So can we please remove the "filterless"? NOS is broken by design, whatever implementation..
And at those higher rates your speakers and your hearing are the filters.I don't agree, assuming bit steps are correct.
When the DAC in question is a filterless design that can handle 706kHz 16 bits then the quantization errors are very, very small and the post analog slow filter that is always present will do the last bit of smoothing the steps.
So... a Chord upsampler can definitely ensure a proper reconstruction filter is in place and a 'steps free' analog output waveform can be created from a filterless NOS DAC.
In essence the Chord + DAC no longer is filterless NOS but simply a properly filtered OS DAC.
Using a filterless NOS DAC as is, at sample frequencies below 88.2kHz, is silly. regardless if goldeneared (and less goldeneared) folks think it sounds fine or even 'better'.
Above that I don't think filterless NOS is very broken when considering music.
What does that mean,"correct bit steps"?I don't agree, assuming bit steps are correct.
None of this means that it's no longer broken, even though you may get a halfway decent result.When the DAC in question is a filterless design that can handle 706kHz 16 bits then the quantization errors are very, very small and the post analog slow filter that is always present will do the last bit of smoothing the steps.
So... a Chord upsampler can definitely ensure a proper reconstruction filter is in place and a 'steps free' analog output waveform can be created from a filterless NOS DAC.
In essence the Chord + DAC no longer is filterless NOS but simply a properly filtered OS DAC.
Using a filterless NOS DAC as is, at sample frequencies below 88.2kHz, is silly. regardless if goldeneared (and less goldeneared) folks think it sounds fine or even 'better'.
Above that I don't think filterless NOS is very broken when considering music.
What does that mean,"correct bit steps"?
None of this means that it's no longer broken, even though you may get a halfway decent result.
Thanks for the clarificationLinearity, which can be an issue for R2R.
A good outcome doesn't mean it's still fundamentally broken. It's like using a hammer to drive in a screw. If you hit it hard and long enough, it works, and probably it will hold perfectly well. But it's still not a good idea.The upsampler + NOS R2R DAC, providing linearity is good and 16 bit resolution can be had is not just halfway decent it can be as good as perfect.
Nothing even remotely broken about it, certainly when properly dithered.
. It's like using a hammer to drive in a screw.
It certainly wont. It will merely remove the 'broken' part only.It won't give you any timing advantages though
non-existent audible difference
I have hesitated from making an explicit admission of being wrong in referencing a study about spatial hearing to back up my claim of 10us/100khz monaural time sensitivity. This is because I have (now had) been skeptical of why the brain would prove to have a 10us capability for two ears and not keep that capability when hearing with only one ear/the same way through both ears. To spare some words I also excluded pointing out the trend toward finer values that the range of microsecond-unit results take on.We cannot hear or "sense" 100kHz.
As delta-sigma dacs/oversampling will only ever be useful as long as high res music is not played into it. You are attributing the issue of scarcity of higher bit and sample rate material to the dac.The tractor will only ever be useful for field work in the mud.
The picture comparisons show differences in the waveforms. Together with that, the sensitivity of the ear is higher than the visual proportional differences.With actual music material, you won't see any of it.
NOS audio looks like a staircase that doesn't even represent the actual signal very well
Either you are taking issue with PCM/digital audio in general ('staircase'), are saying that the bit/sample rate is not sufficient (quantization error), or that no reconstruction filter is in place - not that one can't be added.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.
Hence my use of parentheses around the words, other people call them that.What? Class D amps are not digital...
My first comment on this thread was about how the square wave test is usually excluded. I'd like to see a class d amp tested in that way.we can have class D amps capable of distortion products below -120 dB, all the way up to 20 kHz:
It is already proven. By default they have an unaffected impulse response.prove that NOS DAC's are in any way superior at the timing bit.
No. There are R2R dacs (an example of a dac architecture that is filter-less and NOS) that have analogue low pass filters, digital filters, and oversampled-prior material played into them. There are various configurations.Only when you upsample (so not make it NOS) you can have an influence on the filter response.
Yet you don't get to determine that it won't oversample in the case that it doesn't need to be doing that. It is in this way inflexible.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.
"Broken by design" has a specific meaning. It is to state that there is something innately wrong versus there being something exteriorly wrong.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.
A triangle wave reads or probes the input signal, a power transistor works by it/is not doing that itself, and is acting as a switch and not a linear amplifier. Explain what is weird about this understanding.You have a very weird idea of how class-D operates
Like what?as well as about a lot of other things.
So they are broken when playing natively high bit and sample rate material into them while having no upsampling/filtering?So in that light... yes all real NOS R2R DACs are broken when used without upsampling/filtering.
This sentence is a non-sequitur. It would read correctly if you said "if you have a ex. a Redbook file, the best you can do is oversample to a high rate and still have an acceptable result."if you would have a filterless NOS DAC, the best you can do is oversample to a high rate and still have an acceptable result.
I honestly don't understand your point. Thing is, I don't need to.have hesitated from making an explicit admission of being wrong in referencing a study about spatial hearing to back up my claim of 10us/100khz monaural time sensitivity. This is because I have (now had) been skeptical of why the brain would prove to have a 10us capability for two ears and not keep that capability when hearing with only one ear/the same way through both ears. To spare some words I also excluded pointing out the trend toward finer values that the range of microsecond-unit results take on.
"The shortest auditory click I was able to find in the literature, and which was used in a psychophysical context (i.e., audible to a human) was 10 microseconds..."
"The upper edge of human hearing is 20 kHz so it is unlikely that we could detect a difference between a 25 us impulse and a 10 us impulse (or even a 1 us impulse), but assuming sufficient energy, all would be detectable.", As in my previous post, where I talked about energy.
From: https://psychology.stackexchange.co...re-the-temporal-limits-of-the-auditory-system
I think it's no coincidence that this number is the same between left/right differential and amplitude/time differential.
I could make a big post refuting all the mis-guided ideas you parrot here. I say parrot because you clearly lack understanding of your own. The easiest one underlying everything else is you have the wrong idea that PCM is a stair-case signal. It isn't. The idea is one of those that "is not even wrong". Ideas so incomplete you cannot even make useful predictions. A filter-less DAC is broken by design so much it isn't even wrong. It is broken. And you don't understand that.I have hesitated from making an explicit admission of being wrong in referencing a study about spatial hearing to back up my claim of 10us/100khz monaural time sensitivity. This is because I have (now had) been skeptical of why the brain would prove to have a 10us capability for two ears and not keep that capability when hearing with only one ear/the same way through both ears. To spare some words I also excluded pointing out the trend toward finer values that the range of microsecond-unit results take on.
"The shortest auditory click I was able to find in the literature, and which was used in a psychophysical context (i.e., audible to a human) was 10 microseconds..."
"The upper edge of human hearing is 20 kHz so it is unlikely that we could detect a difference between a 25 us impulse and a 10 us impulse (or even a 1 us impulse), but assuming sufficient energy, all would be detectable.", As in my previous post, where I talked about energy.
From: https://psychology.stackexchange.co...re-the-temporal-limits-of-the-auditory-system
I think it's no coincidence that this number is the same between left/right differential and amplitude/time differential.
~
As delta-sigma dacs/oversampling will only ever be useful as long as high res music is not played into it. You are attributing the issue of scarcity of higher bit and sample rate material to the dac.
The picture comparisons show differences in the waveforms. Together with that, the sensitivity of the ear is higher than the visual proportional differences.
Either you are taking issue with PCM/digital audio in general ('staircase'), are saying that the bit/sample rate is not sufficient (quantization error), or that no reconstruction filter is in place - not that one can't be added.
Hence my use of parentheses around the words, other people call them that.
My first comment on this thread was about how the square wave test is usually excluded. I'd like to see a class d amp tested in that way.
It is already proven. By default they have an unaffected impulse response.
No. There are R2R dacs (an example of a dac architecture that is filter-less and NOS) that have analogue low pass filters, digital filters, and oversampled-prior material played into them. There are various configurations.
Yet you don't get to determine that it won't oversample in the case that it doesn't need to be doing that. It is in this way inflexible.
"Broken by design" has a specific meaning. It is to state that there is something innately wrong versus there being something exteriorly wrong.
In the case of a filter-less NOS dac, you can circumvent the problems posed by circumstances exterior to the architecture before having to swap architectures.
And there is nothing exclusively advantageous about upsampled material vs natively hi-res material. It is not something that must be done.
A triangle wave reads or probes the input signal, a power transistor works by it/is not doing that itself, and is acting as a switch and not a linear amplifier. Explain what is weird about this understanding.
Like what?
So they are broken when playing natively high bit and sample rate material into them while having no upsampling/filtering?
This sentence is a non-sequitur. It would read correctly if you said "if you have a ex. a Redbook file, the best you can do is oversample to a high rate and still have an acceptable result."
3 times by now (excluding my quote)I imagine this ( the most posted video on ASR) has already been linked?
Keith
Yes, there are some, but only where actually the signal is no longer band limited and actually invalid (things like clipping)The picture comparisons show differences in the waveforms.
Obviously not. Visually I can represent any discrepancy, however small it is. No way all of these are audible.Together with that, the sensitivity of the ear is higher than the visual proportional differences.
Staircase != PCM. It's even in the name: Pulse Code Modulation. It's a train of pulses, not a staircase! I'll refer to the Monty video for that. Plus you clearly haven't understood anything of my previous post detailing why NOS is broken by design. It has everything to do with this.Either you are taking issue with PCM/digital audio in general ('staircase')
I asked for actual audio data that proves you are right. I guess you have none?It is already proven. By default they have an unaffected impulse response.
There is no point, is there? You need oversampling to make a delta-sigma DAC work. Flexibility? What for? We get performance that is totally SOTA and beyond what our ears can differentiate.Yet you don't get to determine that it won't oversample in the case that it doesn't need to be doing that. It is in this way inflexible.
Sure, If you already have high-res material, that works as well. The reality is though, that the vast majority of audio I play is not high-res, and I suspect that that is the same for most other people as well.This sentence is a non-sequitur. It would read correctly if you said "if you have a ex. a Redbook file, the best you can do is oversample to a high rate and still have an acceptable result."
I have hesitated from making an explicit admission of being wrong
There are R2R dacs (an example of a dac architecture that is filter-less and NOS) that have analogue low pass filters, digital filters, and oversampled-prior material played into them.
you don't get to determine that it won't oversample in the case that it doesn't need to be doing that. It is in this way inflexible.
A triangle wave reads or probes the input signal, a power transistor works by it/is not doing that itself, and is acting as a switch and not a linear amplifier. Explain what is weird about this understanding.
"Broken by design" has a specific meaning. It is to state that there is something innately wrong versus there being something exteriorly wrong.
In the case of a filter-less NOS dac, you can circumvent the problems posed by circumstances exterior to the architecture before having to swap architectures.
And there is nothing exclusively advantageous about upsampled material vs natively hi-res material. It is not something that must be done.
So they are broken when playing natively high bit and sample rate material into them while having no upsampling/filtering?
Using a filterless NOS DAC as is, at sample frequencies below 88.2kHz, is silly.
I have hesitated from making an explicit admission of being wrong in referencing a study about spatial hearing to back up my claim of 10us/100khz monaural time sensitivity. This is because I have (now had) been skeptical of why the brain would prove to have a 10us capability for two ears and not keep that capability when hearing with only one ear/the same way through both ears. To spare some words I also excluded pointing out the trend toward finer values that the range of microsecond-unit results take on.
"The shortest auditory click I was able to find in the literature, and which was used in a psychophysical context (i.e., audible to a human) was 10 microseconds..."
I have a better experiment: Take a 44.1 kHz file. Do a NOS "upsample" to 705.6 kHz (Zero Order Hold style). Now use a nice brick wall filter at ~21.5 kHz, and then downsample that back to 44.1 kHz. Let's put the difference in DeltawavePlease show me a least 1 example of a NOS DAC that has analog low pass filters that are steep enough to act as a reconstruction filter and shifts frequency depending on sample rate.