Thank you very much to @George Hincapie for lending me his Pontus II 12th Anniversary Edition to do a bit of testing.
I'll post a full set of measurements on my site soon, but just wanted to quickly share some results regarding whether or not the Denafrips DACs are indeed now 'True NOS' or not.
And just to be 100% clear: I like the Denafrips DACs, none of this is about me saying they sound bad or you shouldn't buy them. They have some fantastic and compelling products, but you can like a product whilst also taking issue with misleading marketing.
Regardless of how good a product otherwise is, or how inconsequential a a particular lie might be, manufacturers should not lie to or mislead their customers. Don't make false claims!
TLDR: The Denafrips 12th anniversary DACs are still not actually NOS (non-oversampling). They are now just using zero-order-hold oversampling instead of linear interpolation oversampling to make it LOOK like NOS.
Long version:
A bit of background:
Denafrips has always had a 'NOS' function on their DACs. However I mad a video review a while ago of the Ares 2 and whilst subjectively I really enjoyed the it (and later the Terminator Plus as well), I found when testing that the DACs were in fact not 'NOS' but using a method called linear interpolation. This has some downsides to NOS such as rolling off treble more than NOS does, but more to the point, the issue was simply that the marketing was not truthful.
Denafrips later responded and claimed that I was wrong, and their DACs WERE indeed NOS, however provided no evidence or explanation.
And despite this, they later released their '12th Anniversary' editions of various products, with one of the big marketing points being that they were now 'True NOS!'
Given the fact that the Denafrips DACs seem to be quite heavily reliant on DSP in general, I was a bit doubtful that this was true, but now I can check.
Firstly, NOS means 'non-oversampling'. The DAC does not add extra samples/digitally interpolate like most DACs do. And so if we do an impulse response test, which is silence, followed by one full scale sample, and then silence again, this is what we get, shown by the blue line:
Sample 1: Sample has a value of 0. DAC continues to hold at 0 output
Sample 2: Sample has a higher value. DAC output moves up to meet this value then holds there until the next sample.
Sample 3: Sample has a value of 0. DAC output moves down to 0 again, and holds.
Sample 4: Sample has a value of 0. DAC output continues to stay at 0.
However notice the triangle waveform overlaid. This is how the Denafrips DACs were ACTUALLY behaving in 'NOS' mode.
Sample 1: Sample has a value of 0, but instead of holding at that sample value, the DAC linearly interpolated (draws a straight line) to the next sample, and moves along that path.
Sample 2: Immediately upon reaching sample 2, the DAC then linearly interpolates directly toward sample 3.
Sample 3: DAC then linearly interpolates toward sample 4, which given as it has a value of 0, means it stays at 0 output.
We can see the behaviour of NOS from a real NOS DAC such as the Phasure NOS1A for example:
However the Denafrips DACs instead looked like this:
In fact if we measured with a higher bandwidth we could even see these steps from the extra interpolated samples:
But now, this is the impulse response of the new Pontus II 12th anniversary edition:
That looks a lot more like NOS!......but is it actually NOS? Well, let's do a little more digging.
The first thing that struck me as odd was that there is no ringing.
"No Ringing? But it's NOS! There shouldn't be any ringing!" you might say.
Well, yes, there shouldn't be any ringing from the DAC. But we should still see some ringing on the measurement itself because if a signal exceeds the bandwidth of the analyzer itself, you'll get ringing.
A 'perfect' square wave (or NOS impulse) would require infinite bandwidth to describe. And so when you try to capture it with a limited bandwidth ADC, you'll get ringing.
If the square wave has extremely fast slew and therefore exceeds the analyzer bandwidth by a large amount, such as from the Rockna Wavedream Signature Balanced DAC for example, you'll see quite a lot of ringing:
Whereas if you measured the above signal using a much higher bandwidth oscilloscope, you don't see any ringing. Because it was induced by the filter/bandwidth limitation of the analyzer not the DAC itself.
If the signal only exceeds the analyzer bandwidth by a little, you'll see a fair bit less, such as with the Schiit Bifrost 2/64 in NOS:
If we look on an FFT at a 1khz Sine in NOS from the bifrost 2 this is what we get:
There's the main 1khz tone, and then all the stuff to the right are the aliased content and components that make up the 44.1khz square wave caused by running a DAC in NOS @ 44.1khz.
However if we look at the same FFT but from the Pontus II 12th in NOS, here's what we see:
Can you spot the difference?
There's a large dip heading right down to...you guessed it...705.6khz. Exactly 16x the input sample rate.
And if we change the input sample rate to 48khz, then that dip moves over to 768khz, which is 16 x 48khz.
This is because the DAC is not actually operating in NOS. It is oversampling at seemingly 32x, for a bandwidth of 768khz or 705.6khz.
The content you see on the FFT above 768/705.6khz is a result of imaging/aliasing due to the oversampling filter being used not filtering out content above the nyquist frequency.
Additionally even when the DAC is idle we can still actually see a bit of residual signal at 768khz:
So no, it is NOT NOS. It is still oversampling, it's just applying an oversampling filter to make the output look like NOS.
Other DACs such as the RME ADI-2 also have options to do this (though they're very clear in the manual about the fact that whilst they label it as 'NOS' in the UI it is not actually NOS and is only labelled that way because it sort of emulates actual NOS).
Is this better than linear interpolation on the previous Denafrips DACs? Yes definitely
Is it audibly different to genuine NOS? Up for debate
Does it matter? Depends on what matters to you. I doubt there's going to be any big audible difference, but the point is that Denafrips' marketing is not truthful, and after being called out for lying about something, instead of either genuinely addressing the problem or just apologising, they've tried to hide the fact that their DACs aren't NOS more convincingly, and also have upped their marketing about 'true NOS'.
If you just care about the resulting sound, go listen to a denafrips DAC and decide if it's right for you. None of this will tell you if you will/will not like it.
But there will be a number of people who specifically want a genuine NOS DAC so that they can listen to their music with no digital processing at all, and the Denafrips DACs whilst claiming to offer this, cannot actually do so. That is misleading, and there's no reason Denafrips should be making this claim.
The fact that they've not only lied about this once but then tried to hide it and then doubled down on it also only throws doubt on what other claims about their products might not be true at all....
Don't lie to your customers.
I'll post a full set of measurements on my site soon, but just wanted to quickly share some results regarding whether or not the Denafrips DACs are indeed now 'True NOS' or not.
And just to be 100% clear: I like the Denafrips DACs, none of this is about me saying they sound bad or you shouldn't buy them. They have some fantastic and compelling products, but you can like a product whilst also taking issue with misleading marketing.
Regardless of how good a product otherwise is, or how inconsequential a a particular lie might be, manufacturers should not lie to or mislead their customers. Don't make false claims!
TLDR: The Denafrips 12th anniversary DACs are still not actually NOS (non-oversampling). They are now just using zero-order-hold oversampling instead of linear interpolation oversampling to make it LOOK like NOS.
Long version:
A bit of background:
Denafrips has always had a 'NOS' function on their DACs. However I mad a video review a while ago of the Ares 2 and whilst subjectively I really enjoyed the it (and later the Terminator Plus as well), I found when testing that the DACs were in fact not 'NOS' but using a method called linear interpolation. This has some downsides to NOS such as rolling off treble more than NOS does, but more to the point, the issue was simply that the marketing was not truthful.
Denafrips later responded and claimed that I was wrong, and their DACs WERE indeed NOS, however provided no evidence or explanation.
And despite this, they later released their '12th Anniversary' editions of various products, with one of the big marketing points being that they were now 'True NOS!'
Given the fact that the Denafrips DACs seem to be quite heavily reliant on DSP in general, I was a bit doubtful that this was true, but now I can check.
Firstly, NOS means 'non-oversampling'. The DAC does not add extra samples/digitally interpolate like most DACs do. And so if we do an impulse response test, which is silence, followed by one full scale sample, and then silence again, this is what we get, shown by the blue line:
Sample 1: Sample has a value of 0. DAC continues to hold at 0 output
Sample 2: Sample has a higher value. DAC output moves up to meet this value then holds there until the next sample.
Sample 3: Sample has a value of 0. DAC output moves down to 0 again, and holds.
Sample 4: Sample has a value of 0. DAC output continues to stay at 0.
However notice the triangle waveform overlaid. This is how the Denafrips DACs were ACTUALLY behaving in 'NOS' mode.
Sample 1: Sample has a value of 0, but instead of holding at that sample value, the DAC linearly interpolated (draws a straight line) to the next sample, and moves along that path.
Sample 2: Immediately upon reaching sample 2, the DAC then linearly interpolates directly toward sample 3.
Sample 3: DAC then linearly interpolates toward sample 4, which given as it has a value of 0, means it stays at 0 output.
We can see the behaviour of NOS from a real NOS DAC such as the Phasure NOS1A for example:
However the Denafrips DACs instead looked like this:
In fact if we measured with a higher bandwidth we could even see these steps from the extra interpolated samples:
But now, this is the impulse response of the new Pontus II 12th anniversary edition:
That looks a lot more like NOS!......but is it actually NOS? Well, let's do a little more digging.
The first thing that struck me as odd was that there is no ringing.
"No Ringing? But it's NOS! There shouldn't be any ringing!" you might say.
Well, yes, there shouldn't be any ringing from the DAC. But we should still see some ringing on the measurement itself because if a signal exceeds the bandwidth of the analyzer itself, you'll get ringing.
A 'perfect' square wave (or NOS impulse) would require infinite bandwidth to describe. And so when you try to capture it with a limited bandwidth ADC, you'll get ringing.
If the square wave has extremely fast slew and therefore exceeds the analyzer bandwidth by a large amount, such as from the Rockna Wavedream Signature Balanced DAC for example, you'll see quite a lot of ringing:
Whereas if you measured the above signal using a much higher bandwidth oscilloscope, you don't see any ringing. Because it was induced by the filter/bandwidth limitation of the analyzer not the DAC itself.
If the signal only exceeds the analyzer bandwidth by a little, you'll see a fair bit less, such as with the Schiit Bifrost 2/64 in NOS:
If we look on an FFT at a 1khz Sine in NOS from the bifrost 2 this is what we get:
There's the main 1khz tone, and then all the stuff to the right are the aliased content and components that make up the 44.1khz square wave caused by running a DAC in NOS @ 44.1khz.
However if we look at the same FFT but from the Pontus II 12th in NOS, here's what we see:
Can you spot the difference?
There's a large dip heading right down to...you guessed it...705.6khz. Exactly 16x the input sample rate.
And if we change the input sample rate to 48khz, then that dip moves over to 768khz, which is 16 x 48khz.
This is because the DAC is not actually operating in NOS. It is oversampling at seemingly 32x, for a bandwidth of 768khz or 705.6khz.
The content you see on the FFT above 768/705.6khz is a result of imaging/aliasing due to the oversampling filter being used not filtering out content above the nyquist frequency.
Additionally even when the DAC is idle we can still actually see a bit of residual signal at 768khz:
So no, it is NOT NOS. It is still oversampling, it's just applying an oversampling filter to make the output look like NOS.
Other DACs such as the RME ADI-2 also have options to do this (though they're very clear in the manual about the fact that whilst they label it as 'NOS' in the UI it is not actually NOS and is only labelled that way because it sort of emulates actual NOS).
Is this better than linear interpolation on the previous Denafrips DACs? Yes definitely
Is it audibly different to genuine NOS? Up for debate
Does it matter? Depends on what matters to you. I doubt there's going to be any big audible difference, but the point is that Denafrips' marketing is not truthful, and after being called out for lying about something, instead of either genuinely addressing the problem or just apologising, they've tried to hide the fact that their DACs aren't NOS more convincingly, and also have upped their marketing about 'true NOS'.
If you just care about the resulting sound, go listen to a denafrips DAC and decide if it's right for you. None of this will tell you if you will/will not like it.
But there will be a number of people who specifically want a genuine NOS DAC so that they can listen to their music with no digital processing at all, and the Denafrips DACs whilst claiming to offer this, cannot actually do so. That is misleading, and there's no reason Denafrips should be making this claim.
The fact that they've not only lied about this once but then tried to hide it and then doubled down on it also only throws doubt on what other claims about their products might not be true at all....
Don't lie to your customers.
Last edited:
