So you don’t understand it then…
-
Welcome to ASR. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!
DSD Performance Criteria
- Thread starter Audicron
- Start date
- Thread Starter
- #62
I understand that I can create a sine-wave with just two sample points....
But your claim seems to be that with 2Hz sampling you can recreate the audible spectrum…
- Thread Starter
- #64
No.. I asked: "Then why have a sample-rate greater than 2Hz?" in response to your statement that sample-density does not matter.
Yes, that implies 2Hz is enough..
I am not sure why you invoke DSD-Wide, which is an 8 bits @ 2.8224 MHz digital data transmission format designed and used exclusively by the former Sony's European pro audio division, Sony Oxford (now Oxford Digital), to interface digital signal processing blocks dedicated to DSD production tools without to have to remodulate back to 1 bit data at the end of each processing stage [1].
The point I made is that, unless I'm mistaken, if DSD were decimated (lowered in sample rate) in ESS chips, it would be impossible to obtain signal at frequencies as high as those shown in the above quoted response graph of the DSD low-pass filter. In case of decimation, the stopband would have to start at a much lower frequency, unless significant aliasing distortion were introduced.
[1] Thorpe, Peter; Bentall, Nathan; Cook, Gary; Gerard, Chris; Sleight, Chris; Smith, Mike; Eastty, Peter: DSD-Wide. A Practical Implementation for Professional Audio, AES Convention Paper 5377, 110th AES Convention, Amsterdam in 2001.
- Thread Starter
- #67
That is your interpretation... I'm saying that any LPCM sample-density is not greater even 2.8MHz DSD sample-density... higher resolution is harmonically meaningful.
- Thread Starter
- #68
Why not, if the Hyperstream output signal frequency is greater than or equal to the 1-bit PDM source sample-rate...?
No, it’s not. Just as 192 kHz sampled audio does not contain more information than 48kHz sampled audio in the spectrum below 20 kHz. That is just how information theory works.
- Thread Starter
- #70
You are trying to tell me that the sample-density of a 48kHz signal is equal to the sample-density of a 192kHz signal?
Information density.. which is what matters.
Why not what? Can you elaborate your question?
Again the point I made is that in case of decimation (lowering of the sample rate), for any type of sampled signals (PCM or DSD or whatnot), it is imperative to filter out any signal above the Nyquist frequency (half the target sampling frequency), because any unfiltered signal above that frequency would fold back (alias) in the pass-band and be irrevocably mixed with the original signal content, which would then be severely distorted. This is especially critical with DSD, because this type of signal contains a huge noise power above the pass-band of interest, which means that any aliasing distortion would be of considerable magnitude.
If DSD at 2.8224 MHz were to be decimated, say, by a factor of 4, i.e. at 705.6 kHz, the decimation low pass filter would have to remove as much content as possible just above 352.8 kHz to avoid aliasing.
As can be seen in the above graph I quoted from the ESS ES9008 datasheet, the frequency response of the IIR low-pass filter that is applied to DSD (I insist: this is the response of ESS's design low-pass filter, not the response of the whole DAC chip), there are still significant energy at very high frequencies up to the Nyquist frequency of the DSD input signal (1.4112 MHz), not nearly low enough in my book to get a proper decimation at a lower sample rate. That is a clue that this low-pass filter does not perform decimation or, in other word, keep the original input sample rate of DSD.
To compare, here are the specifications of the decimation low pass filter Sony designed at the end of the 90s to get PCM masters at 44.1 kHz from DSD to produce CD Audio and that was part of the so-called Super Bit Mapping Direct system [1]:
[1] From the Dutch magazine Audio en Techniek, 1997, issue 59, page 37.
Last edited:
In the audible range - up to 20kHz - yes. It carries no more information in that range.
What 192kHz gives you is inaudible frequencies above 20kHz. (up to 80kHz plus). So what? - that information is useless for human hearing. Even if we ignore the fact that in hi res recordings it typically consists of ultrasonic noise.
- Joined
- Jun 24, 2024
- Messages
- 3,057
- Likes
- 5,547
No one is saying that, they are telling you that it is simply irrelevant. You only need a sample rate of twice the highest frequency you are trying to reproduce. So for 20kHz, the generally accepted upper bound of human hearing (though most people certainly can't even hear that high), you need a sampling rate of 40kHz. Hence the common sampling rates of 44.1kHz and 48kHz (the extra sampling rate beyond 40kHz is for technical reasons). With a 48kHz sampling rate, you can perfectly reproduce the entire frequency range that we care about for audio for humans. Additional samples are simply unnecessary information, unless you have a need to capture frequencies beyond what humans can hear. Look up Shannon-Nyquist if you actually care to learn about this.
This is also a great time to once again link to this video:
This is to convert from PCM to DSD, not the other way around.
sam_adams
Major Contributor
- Joined
- Dec 24, 2019
- Messages
- 1,713
- Likes
- 4,275
Anyway... Thank you all for your responses...Over, and Out...
And that's it for today guys. Chonkers has left the dock.
The tour boat leaves in fifteen minutes. Please, be aboard before we cast off.
Similar threads
