, and what
> happens there, does not happen linearly, so that intermodulation products and
> the like could be controlled.
There's nothing about Nyquist-Shannon that need involve linear interpolation.
You can invert by running the sample impulse train through the same filter that
was used to band-limit before sampling. This gives you the exact intermediate
values, assuming that the band limiting was done sufficiently to meet the
requirements.
Brian Willoughby
Thank you for these titles. I've already found them in the AES library.
Brian Willoughby
On Jan 9, 2022, at 13:43, vicki melchior wrote:
> As far as measurements of how far “into the noise” we can hear, there aren’t
> a lot of good published numbers that I know of (having reviewed
and still be heard,
so it seems possible that certain types of dither can be heard even if they're
below the noise floor of the DAC, amplifier, or listening room.
Bottom line: I agree that subtractive dither may be almost pointless when you
can easily access 24-bit audio with additive dither. But I don't think it's
necessarily cut & dry. There might be room for some improvement.
Brian Willoughby
be realized. But maybe that's not what
your Q'() is trying to solve.
Brian Willoughby
On Jan 8, 2022, at 14:55, robert bristow-johnson
wrote:
> This is true:
>
> Q(s + d) - d != Q(s)
>
> i think this is also true:
>
> Q'(s + d) - d = Q'(s)
>
>
On Jan 8, 2022, at 01:49, Sampo Syreeni wrote:
> On 2022-01-07, Brian Willoughby wrote:
>> For the playback chain - wireless transmission, CD, other - which of these
>> do not employ quantization, and which of them can use subtractive dither to
>> restore the original sig
On Jan 7, 2022, at 23:58, B.J. Buchalter wrote:
> On Jan 8, 2022, at 2:42 AM, Brian Willoughby wrote:
>> Where there is quantization, subtractive dither does not help overall. Yes,
>> it improves S/N on paper, but the resulting correlated noise is *more*
>> audible
>> On Jan 8, 2022, at 00:04, Sampo Syreeni wrote:
> On 2022-01-07, Brian Willoughby wrote:
>> (while I was looking for the reference, I found Craven & Gerzon "Compatible
>> Improvement of 16-bit Systems Using Subtractive Dither" 93rd Convention 1992
>>
ocessing involved in playback, and how
this is supposed to be better. If it's workable, I should be able to follow the
DSP.
Brian Willoughby
On Jan 7, 2022, at 23:24, Zhiguang Zhang wrote:
> i mean i'm not sure what the point of going into great detail about how these
> 'situa
n, CD, other - which of these do
not employ quantization, and which of them can use subtractive dither to
restore the original signal without suffering from correlated quantization
noise?
Brian Willoughby
On Jan 7, 2022, at 22:53, Zhiguang Zhang wrote:
> i think someone was referrin
On Jan 7, 2022, at 22:23, robert bristow-johnson
wrote:
> On 01/08/2022 12:55 AM Brian Willoughby wrote:
>> (while I was looking for the reference, I found Craven & Gerzon "Compatible
>> Improvement of 16-bit Systems Using Subtractive Dither" 93rd Convention 199
it because I don't
understand how truncation is avoided in the desired application signal flow.
While I understand how subtractive dither can remove dither, I do not see what
process reverses the truncation. Meanwhile, if there's no truncation, then
what's the point of the dither?
Brian Willoughby
even cites graphics file formats as an example. Raw pixel data is
difficult to process without standard formats to encode the data and metadata.
Caveat: The language of patents is often distinct from the terms used in
engineering. So, it's possible that I misinterpreted this patent.
Brian W
't hear any cars above the din."
Imagine children born into the world after the last Internal Combustion Engine
vehicle ends up in the dump: Will these children be forced to listen to the
sounds of a dead technology, all while suffering hearing damage?
Brian Willoughby
On Feb 10, 2021, at 16
utput.
> If you made a splitter cable, you could actually do them all at once
>
>> On Sat, Feb 6, 2021 at 12:31 PM Zhiguang Zhang wrote:
>>>
>>> On Sat, Feb 6, 2021 at 1:20 PM Andy Farnell
>>> wrote:
>>>>
>>>> On Sat, Feb 06, 2021 at 1
intrinsic - just latency.
Brian Willoughby
Careful, there. This thread is getting long, and errors are creeping in. Any
confusion can be clarified by simply re-reading what was already written.
First, on the 2nd, Zh. mentions the speed of sound as 343 m/s.
Then, on the 5th, Robert mentions 5 or 6 meters from the guitar amp (sound
sourc
nd accuracy are critical,
like measurement, I highly recommend buying a finished product for the
converter / interface.
Brian Willoughby
p.s. Please trim the digest when quoting. There was way more context than
necessary.
On Feb 3, 2021, at 05:04, Joren Six wrote:
> Hi,
>
> Just to add an
nverter type cannot be
simply looked up. These systems require manual latency correction (and possibly
manual latency measurement to inform the correction).
Brian Willoughby
On Feb 2, 2021, at 10:39, Andy Farnell wrote:
> Historically that was probably true, and the source of so
> much m
han two states would need to have a multi-bit state
>> storage register. Thus, it may be mathematically impossible to operate on a
>> 1-bit input stream using 1-bit arithmetic units to produce a 1-bit output
>> stream (unless you're willing to limit the operations to +0, -0, x1,
ease share. This may be a philosophically loaded question, since any
algorithm with more than two states would need to have a multi-bit state
storage register. Thus, it may be mathematically impossible to operate on a
1-bit input stream using 1-bit arithmetic units to produce a 1-bit output
stream (unless you're willing to limit the operations to +0, -0, x1, /1).
Brian Willoughby
Yes. NN are great for finding parameters that another algorithm will use to
produce audio (DX7, filters, effects, circuits, et cetera). Once these
parameters are found, the NN is no longer used. Some other form of processing
takes those parameters and then processes the new inputs to create sign
s to
3 or fewer bits will saturate and thus fail to perform properly.
Brian Willoughby
On Jan 7, 2021, at 22:59, Xue Wen wrote:
> One way to see NN is as a computation graph made of multiply, add,
> nonlinearity and tunable parameters. In this sense delta-sigma is already a
> recurrent a
ks, I would add that neural networks
are incredibly inefficient for processing audio, especially in terms of energy
expenditure.
Brian Willoughby
On Jan 6, 2021, at 14:45, Dario Sanfilippo wrote:
> I am new to neural networks and I would like to start some investigation for
> a 1-bit audio
l output in the signal flow graph.
Think of your synthesizer in a larger context, such as a DAW, and then it's
arguable that scaling the output is not the responsibility of your unit.
Brian Willoughby
24 matches
Mail list logo
