Denafrips Pontus 2 (12th Anniversary Edition) Measurements

Denafrips Pontus 2 (12th Anniversary Edition): https://www.vinshineaudio.com/product-page/denafrips-pontus-r2r-dac

Content and analysis made possible thanks to support from https://headphones.com, and https://patreon.com/goldensound supporters.

This unit was loaned to me by a friend to test

ChitChat:

The Denafrips Pontus 2 is an R2R DAC sitting in the middle of Denafrips’ product lineup. Coming in at just over $1850 USD at the time of writing, it’s a revision to the existing Pontus 2 that offers some upgrades primarily to the power supply and filtering areas, but also to the digital signal processing itself as we will see shortly.

One of the changes is that the 12th Anniversary editions of the Pontus 2 and other products now offer a ‘True NOS’ feature.
Previously I’d demonstrated that the Denafrips DACs were in fact not ‘NOS’ (Non-Oversampling) capable even though it was advertised as such, I additionally showed that the NOS mode on the Denafrips DACs did not provide a resulting output similar to an actual NOS DAC.
If you’re interested you can have a look at my measurements of the Terminator Plus, toward the end of the post:
https://goldensound.audio/2021/10/07/denafrips-terminator-plus-with-gaia-measurements/

Are the new 12th Anniversary Editions actually NOS? Well….they’re closer to it, but technically still not NOS….more on this later…..

Design:

The Pontus 2 has a more than substantial linear power supply, accepting all regional input voltages, and keeping the PSU encapsulated within a shielded compartment within the DAC.
You can see in the above image the massive bank of capacitors, providing an extremely low impedance, low noise supply to the DAC.
In terms of why there are so many capacitors, it is done this way due to the fact that many smaller caps in parallel provides lower equivalent series resistance than a larger cap(s) in series.

Upon looking at the main DAC itself, we can see even MORE capacitors in a large bank, with voltage regulators following.
The DSP board itself which handles processing and digital inputs, is galvanically isolated to prevent noise from the digital source having any impact on resulting performance or sound quality.

This then feeds two FPGAs, one for each channel, which in turn control the actual R2R ladder converters. Each channel is fully balanced.
It is worth noting though that the Denafrips DACs do not have an output buffer stage, and instead the ladders drive the outputs directly. This could be argued to be more ‘transparent’ as there is less circuitry in the signal path, but does result in an extremely high output impedance of 1250Ω on XLR and 625Ω on RCA, as opposed to a more typical 50Ω or so as you’d see from most DACs.
This is unlikely to be an issue, as most preamps and amps will have an input impedance in the realm of 50,000Ω or more, but there are some designs where this is not the case.

The Topping A90 for example is only 2000Ω input impedance, so do check your amplifier to make sure you won’t have any issues here.

Measurements:

Test Setup:

– Audio Precision APx555 B-Series Analyzer with 200kOhm input impedance set unless otherwise specified
– USB Source: Intel PC via intona 7055-C isolator
– All measurements shown are with the DAC connected via SPDIF unless otherwise specified
– Measurement setup and device under test are running on regulated 230V power from a Furman SPR-16-Ei
– Audioquest Mackenzie XLR and RCA interconnects
– Intona Reference Impedance Characterized USB Cable
– Audioquest Carbon SPDIF and AES cables (1.5m)
– Exact analyzer/filter configurations for each measurement are detailed in the full reports
– Measurements shown are with the DAC in ‘OS’ mode unless otherwise stated

Full Measurement Reports:

Reports available here:

Dynamic Range (AES17): 118.1dB

SNR: 123.7dB

IMD SMPTE: -79.6dB

Noise Level RMS: 2.7uVrms (20hz-20khz) / 5.1uVrms (20hz-90khz) – 12.1uVrms (20hz-1Mhz)

DC Offset: 3.1mV active, 3.5mV idle

Delay: 12.36ms (Nice! Latency used to be an issue on Denafrips DACs but it seems the issue is resolved)

1khz Sine, 0dBfs, XLR Out (OS):

1khz Sine, -3dBfs, XLR Out (OS):

Performance does increase a bit when not at max output (this is quite common and is why I do -3dBfs tests).

1khz Sine, 0dBfs, XLR Out (NOS):

Performance in NOS in terms of THD+N is seemingly identical to OS.

1khz Sine, 0dBfs, RCA Out (OS):

We do lose a notable bit of performance when using RCA, but given as the increase is predominantly in the 2nd order harmonics, this might actually have a subjectively pleasant effect and you may find the RCA outs to sound slightly ‘warmer’. Try both and see which you prefer.

Low level signal output:

1khz -90.31dBfs undithered sine (16 Bit, OS) – 96khz capture bandwidth
1khz -90.31dBfs undithered sine (16 Bit, NOS) – 96khz capture bandwidth
1khz -90.31dBfs dithered sine (24 Bit, OS) – 96khz capture bandwidth
1khz -90.31dBfs dithered sine (24 Bit, NOS) – 96khz capture bandwidth

Filter Ultrasonic Attenuation:

OS
NOS

Note how there is a dip at exactly 705.6khz in NOS…..
This shouldn’t be there

Idle Noise FFT:

THD+N vs Frequency:

This measurement is taken with a 96khz bandwidth to capture harmonics above 20khz. Do not directly compare THD values to regular dedicated THD/THD+N measurements as noise above 20khz is also factored in here. This is intended to show a more accurate profile of rising distortion if it exists, rather than to provide more accurate absolute values below 7khz.

The sudden jump past 15khz or so is a bit unusual here. Usually we see either a relatively flat response, or a slow rise into the higher frequencies.

IMD:

Linearity:

With bandpass filter (exclude noise and only check level of test tone itself)
Without bandpass filter (include noise in measurement)

It seems Denafrips has fixed the transfer function issue that existed previously, which caused DACs to sometimes output incorrect levels in test conditions.

THD+N vs output level (As a percentage)
SINAD/THD+N vs output level (As decibels)

Crosstalk:

Multitone:

Intersample Overs Test:

When DACs oversample, they can sometimes encounter a situation where the reconstructed/interpolated waveform goes above 0dBfs (the maximum possible digital value).
This is particularly common with poorly mastered music that has been ‘brickwalled’ (To clarify: I mean brick-walled as in the loudness-wars term, not brick-wall nyquist reconstruction filters https://en.wikipedia.org/wiki/Loudness_war ).

DACs ideally should have a few dB of digital headroom to accommodate this and reconstruct properly, and many such as Chord, RME, Benchmark etc do, and will output signals with intersample overs without any issue/distortion.
But many do not do this, either out of lack of awareness of the issue or because doing so usually means sacrificing a few dB of dynamic range and/or THD+N performance. And as a result will not be able to properly reconstruct these signals.

We can test this by creating a signal that induces intersample overs.
See in the image below. All the 44.1khz samples (squares) are below the 0dBfs limit, but the actual waveform itself, is above this limit.

The Pontus 2 is susceptible to intersample overs, and unfortunately, in a particularly bad way.
Usually in the presence of intersample overs, a DAC will simply clip, as seen below:

Crane Song Solaris DAC, +3dB Intersample Overs test

However the Pontus 2 does not clip, but instead when a sample value reaches above the maximum, it ‘wraps around’ to the minimum negative value, causing a huge sudden transient which will be very audible and may appear as crackling/popping.

+3dB Intersample Overs Test

Unfortunately even at just +1dB the issue persists, and you can see where the peaks of the sine wave suddenly jumps/distorts to the bottom, and the bottoms suddenly jump/distort to the top.

+1dB Intersample Overs Test

If you hear crackling/popping from your Pontus 2 now and then, this is likely why. Hopefully Denafrips can resolve this in a firmware update, but failing that, add -3dB of headroom/volume attenuation in your media player to resolve the issue.

Jitter:

44.1khz – USB input
48khz – USB input

Is it NOS?:

The first thing to do to see if the DAC is NOS is to check the impulse response. A NOS DAC should output a rectangular/square impulse response, sometimes with a rounded profile if there is an analog lowpass filter present.

Pontus 2 ‘NOS’ IR

That looks pretty rectangular, BUT, there’s a few things to note.
Firstly, the left and right channel are not aligned. This is because the Pontus 2 has a small inter-channel phase difference, which at 20khz is just under 8 degrees.

Nothing major, and not actually related to the topic of NOS, but still worth mentioning (again, hopefully something Denafrips can resolve in a firmware update).

Now, whilst the impulse response is rectangular as you’d expect, there are some odd aspects including:

  • No overshoot at all (uncommon unless there is an analog lowpass filter present or the device itself has a fairly limited bandwidth)
  • The curve at the beginning of rising/falling edge is quite rounded and nearly symmetrical to the curve at the end. With either a low pass filter or limited slew rate, the actual effect would not be this way, and would look closer to the yellow waveform below (though less drastic), with the start being quite sudden/straight and the end being more curved.

This had me scratching my chin, and I suspected that actually what might be occurring is that the DAC is indeed oversampling, but using a ‘zero order hold’ filter at likely 16x or 32x the incoming sample rate. This would in many ways LOOK like NOS, and to be honest, probably there wouldn’t be an audible difference to genuine NOS assuming no other DSP is being done. But still, not technically NOS, and leaves open the question of what other DSP is being done.

To verify, we can just put a signal through the DAC and inspect the output at higher frequencies.
Here’s what it should look like on a genuine NOS DAC:

1khz tone output from Schiit Bifrost 2/64 (NOS)

Note how the components of the square wave descend slowly as you go higher in frequency.
But on the Pontus 2:

1khz tone output from Denafrips Pontus 2 12th Anniversary (NOS)

Notice that big dip toward the right? That’s unexpected, and upon closer inspection it turns out the dip is at exactly….you guessed it…705.6khz, EXACTLY 16x the input 44.1khz sample rate.
And when changing the DAC to 48khz, it moves to exactly 768khz.

There’s also a little bit of a spike there even when the DAC is idle in NOS mode too:

This shows that the DAC is unfortunately NOT actually NOS, but instead oversampling in a way that mimics a NOS output.
Similar things are done on the RME ADI-2’s ‘NOS’ filter for example (though they’re quite transparent in the manual about what that filter is.)

I now have a higher bandwidth oscilloscope and if I get a 12th Anniversary DAC back in I’ll do some further testing.

Does this matter?

It depends.

From an audible standpoint? Likely not. I can’t imagine the difference between genuine NOS and ZOH Oversampling at 768khz or higher is going to be audible so long as other forms of DSP and dithering are absent.

But no matter how inconsequential or unimportant something might be, don’t lie to your customers!
If a car doesn’t have a supercharger, don’t call it ‘supercharged’ even if it’s really fast.
If a DAC can’t do NOS, don’t say it can do NOS even if the oversampling mimics it very closely.

To be honest, I’m more concerned about the phase difference between the channels, and the intersample overs being handled in the way they are. This simply should not happen, and due to the massive, full scale transient that is created, if played through a chain without a low-pass/bandpass filtered amp, it could potentially kill transducers.

But hopefully can be fixed via a firmware update.

14 thoughts on “Denafrips Pontus 2 (12th Anniversary Edition) Measurements”

  1. I am disgusted that Denafrips have lied again about the True NOS. In December 2022 I bought the Pontus 12th Edition for the NOS function. I bought it from Vinshine Audio in Singapore and I haven’t unpacked it yet.

    Yesterday I saw Goldensound’s review and emailed Alvin at Vinshine to say that I want him to collect this DAC and refund me in full.

    This afternoon I received the following email from him. Ludicrous. Another liar. You have been warned.

    “Hi Jim

    Good Sunday.

    Let me start by saying this: The audio industry is having a great asset where people like Goldensound, despite at young age, is contributing a lot to the audio community. He is extremely smart despite not having an engineering background, and is able to articulate his thoughts and get the message across with weight, convincing and measurements. He is amazing and talented. The audiophile community needs more people like Goldensound to keep alive.

    Anyhow, I am not certain if it’s intentional for the post to be published on a Sunday where most people will be off from work and enjoy their family time. I did, i was with my 9yo daughter hiking on a trail in Singapore when I received your email. Let’s give the gentleman Goldensound the benefit of doubt. It could be because he has a day job on the weekdays, where on weekends he can be more focused on audio related matters.

    I paused and read the post and nearly sprained my ankle (seriously). My daughter was worried. She was wondering what diverted my attention. And she did show her unhappiness and asked: “I thought Sunday time is for me? Why are you on the phone?” Well… I told her it’s something very important, I have to read and respond.

    Thankfully, she is now a little more mature and understands adults, who, unlike kids, have many things on their plate.

    I quickly reached out to Zhao @ DENAFRIPS, the chief designer. He was also with his family and could only get back to me in the evening, about 2 hours ago, Singapore time. I had a long call with him to discuss this subject.

    The NOS of DENAFRIPS Pontus II on the DSP FPGA level is NOS. There is no oversampling done to the signals feeding into the DAC in NOS mode. The signal is FIFO buffered and reclocked, thereafter, transmitted to two EPM1270 CPLD for balanced R-2R decoding. Each CPLD handle one channel, i.e. L/R.

    Lets define the path of the signal transmission from the DSP/FPGA to the two CPLD called ‘BUS’, where the BUS carries the NOS digital signal in a much higher bandwidth for the CPLD to process the signal for R-2R ladder network decoding. The data in the BUS is not oversample anywhere, hence, the output response of the DAC in NOS mode, as Goldensound measured, is indeed NOS-like, as he noted.

    Let’s imagine this: NOS is non oversampling, in a simple term, what is in = what is out. No complex oversampling / filtering /and stuff is needed or required in NOS mode. Why would DENAFRIPS want to go through the trouble to make a pseudo-NOS or NOS-like oversampling mode to get us into this unnecessary debate?

    Perhaps, if I may, I would say: We don’t know what we don’t know. This applies to both sides but it is up to us to justify what DENAFRIPS is doing is right or wrong, good or bad. But then again, audio is and has always been subjective and controversial , one man’ meat could be another man’ poison. There is no absolute right or absolute wrong.

    While I need more time to gather more information on this, I do understand that the post by Goldensound may have influenced many of our customers – somehow.

    Thank you for reading this, Jim.

    Cheers to the New Year! May 2023 be an extraordinary one!
    ♩ ♪ ♫ ♬ ♭ ♮ ♯

    Many thanks.

    Yours sincerely,

    Alvin Chee

    VINSHINE AUDIO PTE. LTD. (SINGAPORE)”

    Reply
    • It’s a shame that this has happened.
      What I would say first of all though is that this is NOT Alvin’s fault.

      Alvin is the distributor, but is not the designer for the DACs, and on the first round of NOS issues Denafrips actually was seemingly quite reluctant to provide HIM with info on the topic.

      This doesn’t mean you as a consumer shouldn’t be annoyed, at the end of the day you bought the product for your own reasons according to how it was marketed and if it turns out some of those reasons are not true then you have the right to be angry about it.
      But I do just want to make clear that people shouldn’t assume Alvin himself is at fault here, he isn’t. He’s just unfortunately caught in the middle.

      For transparency, I did not speak to Alvin or Denafrips prior to publishing this article because when I did speak to them during the first NOS debate, Denafrips provided literally nothing in reply other than just saying I was incorrect. Alvin was quite helpful but was limited by the lack of information Denafrips provided.
      In regards to Alvin’s reply, the important part is this paragraph I think:

      “No complex oversampling / filtering /and stuff is needed or required in NOS mode. Why would DENAFRIPS want to go through the trouble to make a pseudo-NOS or NOS-like oversampling mode to get us into this unnecessary debate?”

      Whilst I don’t know the answer to this for certain, the point is that the evidence at the moment DOES show that there is something going on and this doesn’t discount it.

      If that dip was at some random frequency then it could be attributed to an analog lowpass filter and we could double check by probing the board before the LPF components.
      But the fact that it is at EXACTLY 16x the incoming sample rate, AND changes when the incoming sample rate changes, AND there is a small spike of activity there even when the DAC is idle, shows that it is digital domain alteration.
      Exactly what or why, I don’t know, maybe there is some part of the DSP that the DAC relies on for accuracy and switching it off in NOS would result in worse performance.

      In any case, as mentioned in the article, to be honest the NOS issue itself is less of a problem here. It’s an ethical issue about what is ‘strictly’ true vs ‘practically’ true.

      The biggest issue is the intersample overs handling.
      That causes a massive, full scale transient that if played on a chain without a lowpass/bandpass filtered amp, could potentially kill transducers.
      That NEEDS to be fixed.

      Reply
      • Thank you, Cameron. Denafrips have obviously kept Alvin in the dark (again) and I know the NOS problem is not his fault. I’m just pissed off that he won’t accept your findings. Thank you for your wonderful reviews.

        Reply
  2. Denafrips and Alvin know that tests can be done and yet he is still saying that the Pontus 12th is True Nos when the results are there for him to see. I won’t ever buy Denafrips again as I really don’t want to deal with liars like this. Let’s see if Alvin collects my Pontus 12th (I’m in the UK) and refunds me along with my customs fees. I have also started a Section 75 claim with my credit card company who are joint liable. I’m not f***ing about as I cannot be bothered with these rotten companies.

    Reply
  3. Thanks for the comprehensive and well executed tests of this Denafrips DAC. As you point out; very concerning behavior in the intersample overs test and also the inter-channel phase difference. I have not seen this intersample test published for the previous Denafrips DACs you measured or the Holo Audio MAY. Would be very interesting to know if the previous generation behaves as bad and how Holo Audio withstand!

    Reply
  4. So I might be totally misunderstanding something here, but is the intersampling overage still an issue when running in the “NOS” aka ZOH oversampling mode? And I’m curious if any crackling/popping as a result of this could possibly be perceived as a subtle, less-than-great mastering on a revealing system or an obvious/glaring auditory artifact(s).

    Reply

Leave a Reply

%d bloggers like this: