Passive pre-filtering like what?
The op-amp is usually used as a low-pass filter, the article makes a case for the use of video op-amp's as LPF and that's very interesting!
Causality = latency is true, yes, but latency can be vastly reduced using IIR, minimum-phase or filterless / Nos.
I'd lean toward that you and others are right that latency is not very important in pure audio, aside from audio/visual.
I just cited the "media players / ASIO / Kernel Streaming sounds different" for completeness.
If you connect a Midi synthesizer, an electric guitar or similar to a PC at home though, then the audio playback needs to be in sync with the physical striking.
That kind of latency is perceivable well under 100 ms, I'd suspect it's more like 10 ms.
I am not personally aware if DSD-wide, DSD128, DSD256, DSD512 or something else I've missed is in conflict with Shannon-Nyquist.
If you are I'm sure your knowledge would be appreciated.
I looked around for a while for the controversy and haven't really seen any.
A waveform is a number of sines, sines / frequency is defined by the wave cycle time.
I don't personally see how this is in conflict with video op-amp's or DSD.
An externally hosted image should be here but it was not working when we last tested it.
O.k., with shaped dither we can move noise away from the 2-4 kHz area.
However can the music signal then vary by 120 dB in the 2-4 kHz area as well, or is it still limited to a 96 dB ceiling?
Last edited:
I have several examples of passive post-I/V filters on my blog. I'm even using a kind of opamp after the LC filter in my current DAC incarnation (AD815).
Its an interesting question - one I'm not sure I know the answer to. Instantaneously the answer would have to be 'no' - the instantaneous SNR would have to be limited to the basic quantization noise. However with a bandpass filter in place (passing only 2-4kHz) then better than 96dB can certainly be achieved (engineering-wise) as the bandwidth is considerably lower than the full 22kHz.
I saw an FFT the other day (over here : Conclusive "Proof" that higher resolution audio sounds different - Page 82 ) where the effect of using shaped dither can clearly be seen in the plot. Note how the noise 'floor' disappears off the plot in the critical 4kHz region (the second FFT in the post, the first is using normal dither).
Last edited:
human hearing does use bandpass filters - critical bands are ~20% of center frequency above 500 Hz http://en.wikipedia.org/wiki/Critical_band is a bare start
lots of poorly stated/confusing things about human hearing and noise, noise thresholds, masking can be better understood if you use critical band theory, recognize that hearing has frequency and bandwidth dependencies that make simpleminded engineering's single flat full bandwidth RMS S/N number less useful
weighting curves and the reasons to use them have been around a while now, since 1936? http://en.wikipedia.org/wiki/A-weighting
lots of poorly stated/confusing things about human hearing and noise, noise thresholds, masking can be better understood if you use critical band theory, recognize that hearing has frequency and bandwidth dependencies that make simpleminded engineering's single flat full bandwidth RMS S/N number less useful
weighting curves and the reasons to use them have been around a while now, since 1936? http://en.wikipedia.org/wiki/A-weighting
Last edited:
With 16-bit music made completely in software without an ADC the softest note to the loudest note is limited to 96 dB isn't it?
I mean...... without quantization noise. =)
Last edited:
You're pretty much disagreeing with the definition of "Nyquist frequency". There's a nice explanation here: Nyquist?Shannon sampling theorem - Wikipedia, the free encyclopedia. Some relevant snips:
Further down it explains the problems caused by aliasing.
The reason I mention ultrasonics is because you said this:
The point is that if the original input signal contains no ultrasonic components (above the Nyquist frequency), then reproduction is perfect.
The DAC only exhibits pre-echo if the original input signal does contain ultrasonic components (above the Nyquist frequency),
I'm not, I just couldn't follow what you were saying. It looked like you were implying something like that as long as we apply the rules of Nyquist-Shannon then the reconstruction is perfect.
Alright, I have a question, if there are precisely zero ultrasonic components above the Nyquist frequency during the recording, are you saying minimum-, linear- and maximum-phase in the DAC will all look/perform exactly identical in a scope, or just that we will hear them as identical?
Last edited:
Linear phase reconstruction will look identical to the original.
Minimum phase reconstruction will look different.
I don't know about maximum phase. Is that ever used with DACs?
http://en.wikipedia.org/wiki/Audio_bit_depth
"For example, 14-bit ADC can produce 16-bit 48 kHz audio if operated at 16x oversampling, or 768 kHz"
A 16-bit / 44.1 kHz file just like any other file except it's purely digitally encoded.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.