• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. 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!

Denafrips Ares II not NOS

Kw6

Active Member
Joined
Dec 23, 2019
Messages
143
Likes
44
I saw a great review on YouTube clarifying that Ares II by Denafrips is not NOS. It's really oversampling with linear interpolation.

That's why when John Atkinson did a certain test it had a funny error. The on YouTube the guy borrowed a Audio Precision and proved it out. Very impressive!
 

Mister C Melon

New Member
Joined
May 29, 2021
Messages
2
Likes
0
I saw a great review on YouTube clarifying that Ares II by Denafrips is not NOS. It's really oversampling with linear interpolation.

That's why when John Atkinson did a certain test it had a funny error. The on YouTube the guy borrowed a Audio Precision and proved it out. Very impressive!
We might get a reason for this in the following series of articles.
Alvin at Vinshine sent me this link when I asked about that YouTube review.
https://audiophilestyle.com/ca/bits-and-bytes/the-denafrips-design-philosophy-part-1-r1042/
 

JohnYang1997

Master Contributor
Technical Expert
Audio Company
Joined
Dec 28, 2018
Messages
7,175
Likes
18,292
Location
China
The point is, does it even matter in terms of audibility?
In a true NOS dac the imaging starts at very early right after 22.05khz. This calls for extremely sharp low pass filter which is virtually not possible in analogue domain. This either results in a very leaky filter or a filter hugely affects the frequency spectrum. Both can have some kinds of effect on the down stream device and eventually hearing. Oversampling moves this step to the digital domain in which the filter can do the job much more nicely. There's simply no reason for NOS, if high quality reproduction is the goal. :)
 

Music1969

Major Contributor
Joined
Feb 19, 2018
Messages
4,645
Likes
2,812
There's simply no reason for NOS, if high quality reproduction is the goal. :)

Some people like to move all the DSP to a PC where they can control it all in different ways. Then the NOS DAC is just doing D to A.

For example using the PC to combine digital filtering with convolution for room correction.

But yes, NOS alone is bad for the reasons you mention.
 

JohnYang1997

Master Contributor
Technical Expert
Audio Company
Joined
Dec 28, 2018
Messages
7,175
Likes
18,292
Location
China
Some people like to move all the DSP to a PC where they can control it all in different ways. Then the NOS DAC is just doing D to A.

For example using the PC to combine digital filtering with convolution for room correction.

But yes, NOS alone is bad for the reasons you mention.
This has nothing to do with NOS rather. It's simply upsampling before the DAC and feed the DAC with a higher sampling rate sata stream.
 

Music1969

Major Contributor
Joined
Feb 19, 2018
Messages
4,645
Likes
2,812
This has nothing to do with NOS rather. It's simply upsampling before the DAC and feed the DAC with a higher sampling rate sata stream.

Yeh but like I said, some people want to do ALL the DSP on PC side.

With ESS based DACs you can do the most audible upsampling before DAC but you can't do all DSP (i.e. you can't bypass on chip DSP incl. SDM).

At least with a true NOS DAC you can actually move ALL DSP to before DAC.

For those upsampling enthusiasts that like that kind of thing (I'm not saying its the way to go but some people like it)
 
OP
K

Kw6

Active Member
Joined
Dec 23, 2019
Messages
143
Likes
44
Yes, I agree with all your points! Ares II sounds great in my system regardless! I also own Chord Qutest but prefer using Ares II. My system is not perfectly neutral so I am assuming the Ares II distortions work better for me! Do you guys buy the fact that a good analog output stage accounts for most of the sound difference we can hear?
 

kchap

Addicted to Fun and Learning
Forum Donor
Joined
Jun 10, 2021
Messages
579
Likes
566
Location
Melbourne, Oz
Yeh but like I said, some people want to do ALL the DSP on PC side.

With ESS based DACs you can do the most audible upsampling before DAC but you can't do all DSP (i.e. you can't bypass on chip DSP incl. SDM).

At least with a true NOS DAC you can actually move ALL DSP to before DAC.

For those upsampling enthusiasts that like that kind of thing (I'm not saying its the way to go but some people like it)
Edited.

If a DAC is NOS it means the anti-aliasing filter has to be implemented as a classic analogue design like the good old days. :rolleyes:
Remember inductors, capacitors, resistors, op amp based gyrators. Like my old Sony CDP101; that was good, I was sad to see it's demise. It seems that NOS now means something i.e. no filtering.
 
Last edited:

raif71

Major Contributor
Joined
Sep 7, 2019
Messages
2,333
Likes
2,535
Edited.

If a DAC is NOS it means the anti-aliasing filter has to be implemented as a classic analogue design like the good old days. :rolleyes:
Remember inductors, capacitors, resistors, op amp based gyrators. Like my old Sony CDP101; that was good, I was sad to see it's demise. It seems that NOS now means something i.e. no filtering.
So to use the filters, the ares II has to be in OS mode ?
 

aj625

Senior Member
Joined
Aug 31, 2021
Messages
325
Likes
226
see denafrips joins the two points by inserting 15 more points through 16x oversampling by a straight line. so after every point there is a sharp transition. other do it with best fit curve with lot more points for averaging instead of just joining two points by straight line and 15 additional points on that straight line . in nos there are no additional points between two points, just joining by horizontal line (sample hold ) so linear 16x interpolation is somewhere in between the two scenario. my question is with no os the steps become so large and how that helps in sq ?
 

kchap

Addicted to Fun and Learning
Forum Donor
Joined
Jun 10, 2021
Messages
579
Likes
566
Location
Melbourne, Oz
So to use the filters, the ares II has to be in OS mode ?
Not quite. Based on Golden Sound video (this has been discussed in another thread but I could not find the link), it appear Denafrips is implementing a First order hold DAC.

Standard R2R and Weighted Resistor DACs are usually Zero order hold DACs. A Zoh DAC generates a sample voltage which is held for the entire duration of the sample period; 22.676 μs for a standard CD. This gives that classic "staircase on a sine wave" look to the explanatory drawings that must be included in all articles about DACs. As a result Zoh DACs have a drooping frequency response. This droop is usually corrected in the post conversion anti-alias filter. The droop starts around 17 kHz, is -3 dB at 20 kHz and -3.9 dB at 22.05 kHz.

Foh DACs interpolate the missing values between 2 adjacent samples by integration. The staircase is replaced with a series of sloping lines. A test signal looks more like sine wave but it's still a bit wonky. Foh DACs do not suffer from droop. Maybe it's less droop; I did not find a definitive answer. Denafrips method of integrating the samples is to over sample so the large step between 2 samples is replaced with a series of smaller steps. Zoh and Foh are valid techniques but both still need an anti-alias filter.

Why does the subjectivist camp which seems to include Golden Sound, (he does do the odd measurement) claim NOS R2R DACs with no anti-alias filter sound superior? Could it be the frequency droop that is the standard behaviour of Zoh DACs.

Finally just in case people have the wrong impression, I am no expert on signal processing. Most of what I said is a rehash of Professor Google. I remember reading all about CDs in the early 80s. I had this vague recollection that many articles mentioned that the anti-alias filter needs to apply a 2-3 dB hf boost prior to rolling off so it was off to the interwebs. The observation about the subjectivist camp; that's mine.

EDIT: Golden Sound not goldenears. Explains why I could not find the link!
 
Last edited:

aj625

Senior Member
Joined
Aug 31, 2021
Messages
325
Likes
226
Not quite. Based on goldenears video (this has been discussed in another thread but I could not find the link), it appear Denafrips is implementing a First order hold DAC.

Standard R2R and Weighted Resistor DACs are usually Zero order hold DACs. A Zoh DAC generates a sample voltage which is held for the entire duration of the sample period; 22.676 μs for a standard CD. This gives that classic "staircase on a sine wave" look to the explanatory drawings that must be included in all articles about DACs. As a result Zoh DACs have a drooping frequency response. This droop is usually corrected in the post conversion anti-alias filter. The droop starts around 17 kHz, is -3 dB at 20 kHz and -3.9 dB at 22.05 kHz.

Foh DACs interpolate the missing values between 2 adjacent samples by integration. The staircase is replaced with a series of sloping lines. A test signal looks more like sine wave but it's still a bit wonky. Foh DACs do not suffer from droop. Maybe it's less droop; I did not find a definitive answer. Denafrips method of integrating the samples is to over sample so the large step between 2 samples is replaced with a series of smaller steps. Zoh and Foh are valid techniques but both still need an anti-alias filter.

Why does the subjectivist camp which seems to include goldenears, (he does do the odd measurement) claim NOS R2R DACs with no anti-alias filter sound superior? Could it be the frequency droop that is the standard behaviour of Zoh DACs.

Finally just in case people have the wrong impression, I am no expert on signal processing. Most of what I said is a rehash of Professor Google. I remember reading all about CDs in the early 80s. I had this vague recollection that many articles mentioned that the anti-alias filter needs to apply a 2-3 dB hf boost prior to rolling off so it was off to the interwebs. The observation about the subjectivist camp; that's mine.
even after 16x os there will be zoh but it will be very small. you can see small horizontal lines even in denafrips 16x 1st order linear interpolation. a best fit curve is exactly like the original band width limited analog wave which is theoretically possible if one uses infinitely long samples for with sinc interpolation. but this is not possible as there will be infinite delay so a windowing function with less number of samples is required. key is to use maximum number of samples practically possible with maximum possible coefficients of sinc function so that reconstructed wave is as close as possible to original wave. obviously 1st order interpolation and nos are very far from original analog wave.
 
OP
K

Kw6

Active Member
Joined
Dec 23, 2019
Messages
143
Likes
44
aj625 thanks for explanation! Can you you provide an example of a inexpensive dac which is close to analog?
 
Top Bottom