NO!!!!!!!!!!!!!!!!!!!
I think many here on DIYA have not heard the latest digital gear; instead, y'all throw off dissing opinions/generalizations because all you've had exposure to is older and/or low-end devices.
Have you/anyone heard the NAD M2 Direct Digital Amplifier?
(Said another way: even the best Sony/Aiwa cassette Walkmans were not as good as 1st gen Apple iPod/iPhone)
No! Well-implemented digital VCs are not flawed. And further, just like with analog equip., you can't generalize from a few bad apples (if that's all you've tasted).
I think many here on DIYA have not heard the latest digital gear; instead, y'all throw off dissing opinions/generalizations because all you've had exposure to is older and/or low-end devices.
Have you/anyone heard the NAD M2 Direct Digital Amplifier?
(Said another way: even the best Sony/Aiwa cassette Walkmans were not as good as 1st gen Apple iPod/iPhone)
But digital VCs don't suffer from Johnson noise or other trash accumulated in the resistance of that analog pot.abraxalito said:
YES!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
Anyway, Stereophile's May 1998 review of the Wadia 850 CDP noted:
If anyone has that 'White Paper' JA refers to, throw us a link!
I take this back a bit ... decent (but carefully implemented) digital VC's have been around a while--at least since the early 1990s (but likely even earlier, especially gain control in aerospace and industry).
Anyway, Stereophile's May 1998 review of the Wadia 850 CDP noted:
Unless realized with long internal word lengths and optimally dithered, a digital-domain volume control has the potential of reducing resolution with increasing attenuation. If you think about it, reducing the level by 6dB is approximately equivalent to sliding all the bits in a 16-bit digital word one bit to the right and replacing the Most Significant Bit with a "0." What was the Least Significant Bit now is the 17th bit—but only as long as the word length has been increased to more than 16 bits. Otherwise, it just "falls off the end" of a 16-bit word, a process described in engineering terms as "truncation."
Wadia's white paper on its volume control implies that it has been correctly implemented with respect to preserving absolute resolution. I examined its operation by feeding the 850 24-bit data representing a dithered 1kHz tone at -90dBFS and performing a spectral analysis with the volume control set at 0dB, -6dB, -12dB, -18dB, and -24dB. The result, with the reference level normalized for the -90dBFS tone with each successive reduction in level, is shown in fig.4. By comparing this graph with fig.2, it can be seen that the noise floor is at or below the 16-bit level as long as the volume-control operation is between 0dB ("99") and -6dB ("87").
Fig.4 Wadia 850, spectrum of dithered 1kHz tone at -90.31dBFS, with noise and spuriae, with volume control set to 0dB, -6dB, -12dB, -18dB, and -24dB (24-bit data, 1/3-octave analysis, right channel dashed).
If anyone has that 'White Paper' JA refers to, throw us a link!
I remember the top Wadia's having an analogue output preset levels and you needed to adjust this so that it's digital VC at full level was the loudest you ever wanted to listen at on that system. This to me said they didn't want to reduce to far with the digital VC? Why? maybe they knew about "Bit Stripping"
Thought I had it saved on PDF
Cheers George
Thought I had it saved on PDF
Cheers George
I think the Stereophile text and picture in post 62 above may be misleading to some.
You can possibly get the impressin that 0 and -6dB is ok, but that all other setting increase the noise floor. This is not what happens though.
1) Each attenuation step is 6dB.
2) The difference in noise floor is 6dB per step.
3) The files are normalized for the 1kHz compnent.
This means that the noise floor is constant. There's no loss of dynamic range using this specific digital VC.
Yes, the SNR of the signal is worse but that does not matter since the noise floor is the same as when no attenuation is used and the program material contains a signal with the same level.
As long as proper math is used with high enough internal processing (say 32bit or 64bit float) and proper dither before truncation to 24bit, everything is dandy. The analog noise floor of the DAC will still be the limit for dynamic range.
The noise added by the process is at the -140dB level or thereabouts.. which is 15-20dB below the best DAC's that exists... which already have noise floor well below human hearing.
You can possibly get the impressin that 0 and -6dB is ok, but that all other setting increase the noise floor. This is not what happens though.
1) Each attenuation step is 6dB.
2) The difference in noise floor is 6dB per step.
3) The files are normalized for the 1kHz compnent.
This means that the noise floor is constant. There's no loss of dynamic range using this specific digital VC.
Yes, the SNR of the signal is worse but that does not matter since the noise floor is the same as when no attenuation is used and the program material contains a signal with the same level.
As long as proper math is used with high enough internal processing (say 32bit or 64bit float) and proper dither before truncation to 24bit, everything is dandy. The analog noise floor of the DAC will still be the limit for dynamic range.
The noise added by the process is at the -140dB level or thereabouts.. which is 15-20dB below the best DAC's that exists... which already have noise floor well below human hearing.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.