So, with simple words, ringings caused from the reconstruction filter which is an analog low pass filter.
Indeed, square waves are not related with real world acoustics.
Simply are used in workbench tests to stress the DUT (device under test) so that to reveal some of its inherent shortcomings.
By any way, thank you for the information, it is welcomed.
Actually I must confess I was wrong - I had in mind the messenger running all the way from Marathon to Athens and being killed for the bad news.
I was wrong: you guys actually won the battle at Marathon so I'm sure the messenger wasn't killed, more likely given a lot of money and a few women 😉
Jan
Mister Anagnostou did not show you the whole picture.
Your post says that the first is much closer to reality.
If you look at this picture, you can see the impulse response of the NOS DAC is much closer to reality.
If these pictures are correct or not, I don't know, someone would need to email Metrum Acoustics I think and ask him for information, there isn't enough information in these pictures in my view.
The impulse response on the lower right is the interesting one.
What you said about the lack of upper frequency may be true......
However, you need to consider that downsampled 44.1 kHz music is lacking the upper frequency as well.
In the studio, it starts at 96 kHz or 192 kHz, then the final is 44.1 kHz and the music has been "perverted", according to some.
An externally hosted image should be here but it was not working when we last tested it.
Last edited:
Please provide your thoughts on this Cirrus Logic whitepaper which is contradictory to what you've written.
For example
"In addition to the audibility of signal latency in live sound applications, there is a growing body of research
that indicates that the audibility of pre-echoes is a primary source of the sonic differences between the
lower sample rates, 44.1 and 48 kHz, and the higher sample rates of 96 and 192 kHz."http://www.cirrus.com/en/pubs/whitePaper/DS668WP1.pdf
This ABX test as well
Succesful ABX of 24/96 vs. 16/44.1 - Hydrogenaudio Forums
The idea of audio to most people is to reach a perfect replication of reality, the extension of that is to have more detail than reality and then the third path is coloured and aesthetic sound.
In order to have perfect replication you need a perfect reconstruction filter or you need to aim severely in that direction.
I likened this to a megapixel spec in cameras since there is no true perfection, but at a certain point we can say that the pixels have a severe density so that zooming in quite deep is necessary.
Hmm... Although the main source about the battle of Marathon is Herodotus, the legend of the messenger (messengers actually were well trained runners called "day or daily runners") who anounced the victory to Athenians by saying the famous "ΝΕΝΙΚΗΚΑΜΕΝ" (we won) and immediately expired, actually is comming from subsequent sources and not from he. Herodotus just refers a daily runner (ΗΜΕΡΟΔΡΟΜΟΣ in Greek) named Feidippides who was sent from Athenians to Lacedaemonians (or Spartans for the funs of "300") for asking help. Feidippides covered 440km in 4 days from Athens -> Sparta -> Athens.
Plutarch also refers this incident and that the runner-messenger was the Thersippus the Erchieus or a runner named Eucles. Big confusion, ha?
Dear Jan, the only that connect us the "modern Greeks", as well all people aroung the world, with ancient Athenians and the seven wise men of ancient Greek, it is the classic Hellenic education and nothing else. Unfortunatelly just a 5% of modern Greeks are possessors of this education.
Therefore, we, the modern Greeks we don't gave nothing to Marathon runner. Neither ancient Athenians as the runner expired immediatelly (as he was not doped like the modern runners) in the place of AMPELOKIPI of today Athens.
And the heritage of ancient Athenians to all people around the world, was the Direct Democracy and NOT the modern counterfeit Representative Democracy.
That is NOT contradictory to what he wrote.
Delays in live sound (if not deliberate and set so as to reduce the effect!) can cause comb filtering where sound reaches the listener by multiple paths differing in length by a faction of a wavelength this effect is mainly apparent in the frequency domain but is caused by a time domain problem.
This is well known among the smarter end of the live sound and venue acoustics set, and indeed a little more delay then causes comb filtering (say 20ms or so) can actually be useful as the brain tends to localise on first wavefront so delaying the speakers to arrive 10 - 20ms after the direct sound helps with localisation.
I never cease to be amazed by how picky folk get about playback, especially given I know what goes on when making a recording (especially what goes on when making a 'live' recording...).
You really need to go read up on the basics (Hint, the ability to output a non band limited square wave is NOT a selling point in a DAC), and then spend an evening with some paper and pens or a copy of scilab to see what the maths actually has to say.
Regards, Dan.
Delays in live sound (if not deliberate and set so as to reduce the effect!) can cause comb filtering where sound reaches the listener by multiple paths differing in length by a faction of a wavelength this effect is mainly apparent in the frequency domain but is caused by a time domain problem.
This is well known among the smarter end of the live sound and venue acoustics set, and indeed a little more delay then causes comb filtering (say 20ms or so) can actually be useful as the brain tends to localise on first wavefront so delaying the speakers to arrive 10 - 20ms after the direct sound helps with localisation.
I never cease to be amazed by how picky folk get about playback, especially given I know what goes on when making a recording (especially what goes on when making a 'live' recording...).
You really need to go read up on the basics (Hint, the ability to output a non band limited square wave is NOT a selling point in a DAC), and then spend an evening with some paper and pens or a copy of scilab to see what the maths actually has to say.
Regards, Dan.
I find this National Semiconductor app note pertinent: http://www.ti.com.cn/cn/lit/an/snoa232/snoa232.pdf
You contradict your self: you ask for a
but you are a fan and do a lot of advertising to a NOS DAC without any reconstruction filter.
If somebody can hear over 22.05KHz, I'm agree but I can not hear over 15.3KHZ (now) and I can bet that neither you or someone else can hear over 18-20KHz. This is the dogs and cats range but my dog does not complained about my DAC.
If you want to discuss about perfect Dirak pulse, first you need to provide an answer to my request from the same post:
I thought that the audio equipment are intended for music and sounds of nature reproduction and not to try reproducing theoretical ideal signals.
I have a question: what speakers reproduce ideal pulse from post 164 and how it will look after passing the speaker? but after passing the ear?
As far as my knowledge on digital sound goes:
According to the theory of the Fourier Transformation you need infinite bandwidth to achieve perfect replication with a digital system.
With an infinite bandwidth you can do without the reconstruction filter. If you don't have infinite bandwidth you need the filter to stop digital remnants from polluting the sound signal.
BTW I have serious doubts on the integrity of the information in this picture: no x-y scale, nor a description of how the data was measured.
An externally hosted image should be here but it was not working when we last tested it.
Last edited:
From the picture, it looks like this has been faked. To wit, the 'impulse response' isn't the system impulse response, rather a digital file with illegal values has been created (hex editor or maybe Audacity) and played out through the filterless NOS DAC.
The mistake you are making is assuming that doing a 'join the dots' with straight lines represents reality.
It doesn't, the sample values are NOT the signal, especially when there are high frequency components present.
For example, 0, 1, 0, -1 <repeats> is a Fs/4 sine wave when correctly reconstructed, put it thru a 'DAC' without a filter however and you get this weird pulsy thing simply because you still have components at 3/4Fs and 5/4Fs and so on.
The oversampling DAC in those pictures is the one working correctly and reproducing what I would expect, the NOS one clearly lacks suitable filtering and has gobs of ultrasonics in its output.
Now you can argue with the filter used in the oversampling DAC, it is clearly linear phase but all such suffer from preringing, and reasonable men can disagree about the best tradeoff here, but it is far closer to correct then the NOS one without the filter is.
A good rule of thumb is that any website or paper that makes a big thing about playing back squarewaves via a sampled system can be safely ignored as the author is clearly hard of thinking (or is trying to gull you, any site using the word 'quantum' when trying to sell something is gauranteed to be trying to gull you).
Look, this stuff is complicated and somewhat non intuitive, and there are no really good handwavy explanations that don't have obvious holes, you just really need to hit the books and get your math on.
Claude and Harrys papers are actually not too bad, avoid Vanderkooy and Lipshitz however, great researchers but they like scary mathematics just way too much, for all that they did publish an interesting paper pointing out a problem with 1 bit converters as used in SACD.
73 Dan.
It doesn't, the sample values are NOT the signal, especially when there are high frequency components present.
For example, 0, 1, 0, -1 <repeats> is a Fs/4 sine wave when correctly reconstructed, put it thru a 'DAC' without a filter however and you get this weird pulsy thing simply because you still have components at 3/4Fs and 5/4Fs and so on.
The oversampling DAC in those pictures is the one working correctly and reproducing what I would expect, the NOS one clearly lacks suitable filtering and has gobs of ultrasonics in its output.
Now you can argue with the filter used in the oversampling DAC, it is clearly linear phase but all such suffer from preringing, and reasonable men can disagree about the best tradeoff here, but it is far closer to correct then the NOS one without the filter is.
A good rule of thumb is that any website or paper that makes a big thing about playing back squarewaves via a sampled system can be safely ignored as the author is clearly hard of thinking (or is trying to gull you, any site using the word 'quantum' when trying to sell something is gauranteed to be trying to gull you).
Look, this stuff is complicated and somewhat non intuitive, and there are no really good handwavy explanations that don't have obvious holes, you just really need to hit the books and get your math on.
Claude and Harrys papers are actually not too bad, avoid Vanderkooy and Lipshitz however, great researchers but they like scary mathematics just way too much, for all that they did publish an interesting paper pointing out a problem with 1 bit converters as used in SACD.
73 Dan.
It's noteworthy how this thread is now, in some way, shining light on the false advertising of the Phasure NOS1 and the Metrum Acoustics lineup.
"You contradict your self: you ask for a
Quote:
perfect reconstruction filter
but you are a fan and do a lot of advertising to a NOS DAC without any reconstruction filter."
My naivety at the start of this thread was bit depth, 4-bit sounds remarkably high resolution and I learnt yesterday that cassette-tape is around 6-bit.
Now I have the clarity to know that resolution is in the 44,100 samples per second, even at 6-bit it can sound excellent.
As Shannon-Nyquist-Kotelnikov states those samples are error-free, if we cut reality into 44,100 slices we transfer those slices accurately into the placeholder, such as digital media.
They are accurate transfers, from air to digital there is not a slice which is incorrect.
The incorrectness is only potentially within the studio equipment such as wires, microphone, ADC, filter, quantization noise......
The catch is that these horizontal samples are actually not sufficient, for highest resolution sound, we must make sense of them with sinc functions and interpolation to try to mirror reality!
The first CD players in the 90's did not interpolate, interpolation was an invention which came later.
The modern Nos DAC's have two viewpoints, the first one is colouration, that they prefer this lower resolution sound, akin to someone preferring VHS tape.
The second is they assert that interpolation is flawed, indirectly I think it's akin to saying that both alternatives are flawed and they prefer the dynasty of non-interpolation.
In my experience, I tend to much prefer the sound of interpolation.
However, I am open-minded to the fact that interpolation succinctly differs in it's accuracy.
"If you want to discuss about perfect Dirak pulse"
Yes, this is quite important because Dirac HD technology assert improvement in their signal processing.
"You contradict your self: you ask for a
Quote:
perfect reconstruction filter
but you are a fan and do a lot of advertising to a NOS DAC without any reconstruction filter."
My naivety at the start of this thread was bit depth, 4-bit sounds remarkably high resolution and I learnt yesterday that cassette-tape is around 6-bit.
Now I have the clarity to know that resolution is in the 44,100 samples per second, even at 6-bit it can sound excellent.
As Shannon-Nyquist-Kotelnikov states those samples are error-free, if we cut reality into 44,100 slices we transfer those slices accurately into the placeholder, such as digital media.
They are accurate transfers, from air to digital there is not a slice which is incorrect.
The incorrectness is only potentially within the studio equipment such as wires, microphone, ADC, filter, quantization noise......
The catch is that these horizontal samples are actually not sufficient, for highest resolution sound, we must make sense of them with sinc functions and interpolation to try to mirror reality!
The first CD players in the 90's did not interpolate, interpolation was an invention which came later.
The modern Nos DAC's have two viewpoints, the first one is colouration, that they prefer this lower resolution sound, akin to someone preferring VHS tape.
The second is they assert that interpolation is flawed, indirectly I think it's akin to saying that both alternatives are flawed and they prefer the dynasty of non-interpolation.
In my experience, I tend to much prefer the sound of interpolation.
However, I am open-minded to the fact that interpolation succinctly differs in it's accuracy.
"If you want to discuss about perfect Dirak pulse"
Yes, this is quite important because Dirac HD technology assert improvement in their signal processing.
Last edited:
We can correct the pre-ringing with IIR / minimum-phase filtering!
The Chord Hugo has a so called "very long FIR filter" which I think they state exceed any other modern reconstruction filter, but after reading the Cirrus Logic paper it dawned on me that the Hugo must have immense latency!
In my view high latency is unacceptable, especially in such a pricy product.
Last edited:
Thanks for the education Fotios!
Hmm. Direct democracy - would that be practical with 50 million people?
But no politics here...
- Status
- Not open for further replies.