Sort of.
They point out that FIR linear phase filters cause pre-ringing and latency, and that these effects can be minimised or eliminated by using IIR filters (which are not linear phase) instead. The missing piece of the puzzle is that pre-ringing and latency can also be minimised or eliminated by using FIR filters that are not linear phase, such as the minimum phase option of the Wavedream (or by going back to good old analog fiters, which aren't linear phase either).
Bottom line: Linear phase filters cause latency and pre-ringing, but there are various ways to make non linear phase filters that don't have pre-ringing and latency.
Hold on, I'll ask directly in the Chord Hugo thread now, then we'll get the direct answer without speculation.
you can design FIR filters that aren't symmetric, don't have pre-ringing
and emulating a continuous domain filter by the Impulse Matching method does require long filter lengths in FIR
I have read a few comments from Chord's engineer - some sound like "puffery" - perhaps "accurate" in some irrelevant sense
and emulating a continuous domain filter by the Impulse Matching method does require long filter lengths in FIR
I have read a few comments from Chord's engineer - some sound like "puffery" - perhaps "accurate" in some irrelevant sense
+1 I'd like to know this too. I haven't seen him talking about pre-echo at all. Kastor have you turned up the Julian Dunn reference yet?
Actually I read the paper, the first one in the Nanophon link.
It had some interesting information, err, it arrives at a similar conclusion as the Cirrus Logic paper, namely that with 96 kHz the filter can achieve higher accuracy.
Err or that 16-bit / 96 kHz music is more accurate in the 20 - 20 area due to time / phase related effects, which are asserted as less accurate in 48 kHz media.
It all seems to revolve around time / phase.
Now, how does this all relate to the fourier uncertainty principle, I am not aware.
______
The story for the Chord device is simple.
He asserted that 1 million taps in 16-bit audio would result in a perfect mirror-image of reality.
He vaguely wrote somewhere that there is no pre-echo, after all, if there's pre-echo it's not perfect is it?
I'd rather not say where he wrote that, but the 1 million tap length and perfect replication comments are in the Hugo thread in this sub-forum.
I already asked in that thread so now he has the chance to revise.
It had some interesting information, err, it arrives at a similar conclusion as the Cirrus Logic paper, namely that with 96 kHz the filter can achieve higher accuracy.
Err or that 16-bit / 96 kHz music is more accurate in the 20 - 20 area due to time / phase related effects, which are asserted as less accurate in 48 kHz media.
It all seems to revolve around time / phase.
Now, how does this all relate to the fourier uncertainty principle, I am not aware.
______
The story for the Chord device is simple.
He asserted that 1 million taps in 16-bit audio would result in a perfect mirror-image of reality.
He vaguely wrote somewhere that there is no pre-echo, after all, if there's pre-echo it's not perfect is it?
I'd rather not say where he wrote that, but the 1 million tap length and perfect replication comments are in the Hugo thread in this sub-forum.
I already asked in that thread so now he has the chance to revise.
Note, this 1 million tap length comment is actually posing the wild assertion that all digital audio is in fact imperfect, especially Nos.
He wrote he has a prototype six-digit tap length filter, which supposedly sounds even better.
Nos is non-interpolation and this is super interpolation.
I'm neutral, but wild assertions shouldn't be whisked away, especially when they're selling like hot muffins in winter!
He wrote he has a prototype six-digit tap length filter, which supposedly sounds even better.
Nos is non-interpolation and this is super interpolation.
I'm neutral, but wild assertions shouldn't be whisked away, especially when they're selling like hot muffins in winter!
I re-checked the source.
It says, with infinity taps there is no ringing.
Either way, longer tap length is resulting in purer sines, that's the assertion right?
A pure sine is 0.0000000 THD.
I can't see anything about harmonic distortion in here......
Reconstruction filter - Wikipedia, the free encyclopedia
It says, with infinity taps there is no ringing.
Either way, longer tap length is resulting in purer sines, that's the assertion right?
A pure sine is 0.0000000 THD.
I can't see anything about harmonic distortion in here......
Reconstruction filter - Wikipedia, the free encyclopedia
I sent him a private message a few days ago but he is still offline since the 31st of July.
I suspect he'll sign in and say it himself sooner or later.
I suspect he'll sign in and say it himself sooner or later.
Last edited:
In the Chord Hugo thread it already says infinity taps are needed, or 1 million taps.
I'm mostly interested in if the maths and psychoacoustics for this is correct or not.
Don't link me to books if the answer isn't in there.
I'm mostly interested in if the maths and psychoacoustics for this is correct or not.
Don't link me to books if the answer isn't in there.
Seriously, just half an hour with Google or Wikipedia should be enough to learn enough of the basics to avoid embarrrasing youself by taking random internet comments out of context and thinking you've learnt something.
^
All of your comments are vague and empty with an air of comfort-zone in them.
You can consider eating a positivity apple and reading this
The Idealist Attitude
See you
All of your comments are vague and empty with an air of comfort-zone in them.
You can consider eating a positivity apple and reading this
The Idealist Attitude
See you
OK, don't do any homework, I'll spell it out for you.
a) Linear phase filters cause latency and pre-ringing.
b) The Chord Hugo uses a linear phase filter.
c) Therefore the Chord Hugo has latency and pre-ringing.
d) The same type of filter with 1 million taps would also have pre-ringing.
e) The pre-ringing with a million tap filter would not be any less than with a 26000 tap filter or a 2000 tap filter.
a) Linear phase filters cause latency and pre-ringing.
b) The Chord Hugo uses a linear phase filter.
c) Therefore the Chord Hugo has latency and pre-ringing.
d) The same type of filter with 1 million taps would also have pre-ringing.
e) The pre-ringing with a million tap filter would not be any less than with a 26000 tap filter or a 2000 tap filter.
No he did not.
Yes, that's what he actually said.
No, he did not.
This seems to be the core of your misunderstanding. A theoretically perfect reconstruction filter will have pre-ringing.
Last edited:
You are in error.
Perfection, is when it is as if no sampling is taking place at all.
If you look at the impulse of a violin note recorded with a microphone, then you look at that impulse in the DAC, if they differ, then they are different.
Perfect, is the void of difference.
100 ms latency is an imperfection as well.
The first concern is Zepto level perfection, the second concern is perceptivity.
You need to clarify which one it is you are referring to.
I'm starting to think the Chord = huge phail.
Perfection, is when it is as if no sampling is taking place at all.
If you look at the impulse of a violin note recorded with a microphone, then you look at that impulse in the DAC, if they differ, then they are different.
Perfect, is the void of difference.
100 ms latency is an imperfection as well.
The first concern is Zepto level perfection, the second concern is perceptivity.
You need to clarify which one it is you are referring to.
I'm starting to think the Chord = huge phail.
Last edited:
Agreed. Or, put another way, perfect reproduction means the output waveform is absolutely identical to the original input waveform.
That raises the question of what the original waveform actually looked like. Bear in mind that perfect reproduction is only possible if the waveform which is sampled contains no frequency components higher than the Nyquist frequency.
Now, you're worried about pre-ringing on waveforms with sharp corners or steps like a squarewave. But any such waveform has frequency components out to infinity. Conversely, any signal that has no frequency components above a given limit has no sharp corners - it's smooth curves all the way.
Bottom line: If the output waveform looks like a step, but with pre-ringing (and post-ringing), then there's only two possibilities:
A) The original input signal had pre-ringing and post-ringing, and the output is a perfect reprodution of it.
B) The original input signal had frequency components above the Nyquist limit, and it's therefore impossible to reproduce it accurately.
- Status
- Not open for further replies.