• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required as is 20 years of participation in forums (not all true). Come here to have fun, be ready to be teased and not take online life too seriously. We now measure and review equipment for free! Click here for details.

Review and Measurements of Soekris dac1421 Multibit DAC

amirm

Founder/Admin
Staff Member
CFO (Chief Fun Officer)
Joined
Feb 13, 2016
Messages
13,990
Likes
6,337
Location
Seattle Area
#1
This is a Review and Measurements of Soekris dac1421 Multibit DAC. It is on a kind loan from the designer and forum member, Søren Kristensen. In case you are not familiar with this class of DACs, instead of using an off-the-shelf dac chip, this unit uses a custom implemented DAC using a series of resistors and a custom FPGA (digital logic) controlling them. Technical term for their approach is "Sign Magnitude Discrete R-2R." The dac1421 is unbalanced only and has a retail price of around $1,000 (sold in Euros however).

The unit itself has a simple "design language" and fits in the class as far as fit and finish. Here you see it below SMSL SU-8 which I also tested against it:

Soekris dac1421 vs Schiit Yggdrasil review and measurement.psd.jpg

There is a useful clipping indicator as the volume control can go over reference. For this review, I did not measure the headphone amplifier. Will do that in a future installment.

There is interest in comparing the dac1421 against the Schiit Yggdrasil. For a short period I had both in hand and the measurements below reflect that. A request was also made to compare it to SMSL SU-8 which retails for $250 from what I recall. See its review here: https://www.audiosciencereview.com/...eview-and-measurements-of-smsl-su-8-dac.3778/. And Yggdrasil DAC: https://audiosciencereview.com/foru...measurements-of-schiit-yggdrasil-v2-dac.3607/

I imagine you all are anxious how it did on the bench so let's get into that.

Measurements
All tests are done using USB input and naturally using unbalanced output since that is all the dac1421 has.

Let's start as usual with our dashboard view of 1 kHz, full amplitude signal at 44.1 kHz:

Soekris dac1421 SINAD Dashboard measurement.png


THD+N is 0.006% in one channel at 0.005% in the other. These match what Søren measured prior to sending the unit to me. Both are better than the advertised 0.008% spec. Søren says this is just a unit out of the assembly pipeline as otherwise he would have sent me one that had matching distortion. I believe him.

The dac1421 has an internal switching power supply fed by a two-pronged cord. The case as such is "floating." Teardown will come in a future article. The issue and one that is not unique to the dac1421 is that its case can start to float up causing ground differentials. This shows up in the spike that you see at 60 Hz. There is a 50 ohm resistor between the chassis and signal ground. When I bypassed this with a jumper, it made a significant reduction in mains leakage. Søren says he is not seeing any of this in his testing. Unfortunately the vagaries of unbalanced audio interconnects and grounding in consumer audio is that it is perfectly possible for the two scenarios to exist. Søren is considering lowering the value of that resistor.

The power supply harmonics travel fair bit in frequencies if we look at the FFT spectrum in top right (they get jammed together more and more). Beside that we have a regular train of harmonic distortion. We can see this better if we zoom in and overlay the results on top of Schiit Yggdrasil:

Soekris dac1421 vs Schiit Yggdrasil 1 kHz DAC distortion measurement.png


Schiit Yggdrasil has large number of inharmonic spikes going the full audio bandwidth. Soekris dac1421 doesn't have much of this but then has higher harmonic distortion so overall, it winds up with similar SINAD numbers (difference between signal and sum of the distortion and noise products):

Soekris dac1421 SINAD distortion measurement.png


Looking at intermodulation distortion we get a mixed picture:

Soekris dac1421 vs Schiit Yggdrasil 1 kHz DAC intermodulation distortion measurement.png


At the highest amplitudes the dac1421 does better. At the lowest, it is noise dominated and similar to Yggdrasil. IN the middle range it loses to Yggdrasil by fair bit. Had one of the channels not been higher than the other when it came to distortion, it would have ranked better. As it is, I will give the nod to Yggdrasil in this test.

Looking at more classical THD+N distortion and noise versus frequency, we have:

Soekris dac1421 vs Schiit Yggdrasil 1 kHz DAC THD+N distortion measurement.png


The Yggdrasil oddly has rising low frequency distortion and noise as I have reported (NB: this is unbalanced -- balanced output doesn't have this this problem). In that regard, we have yet another mixed performance and no winners.

Of course our classically designed sigma-delta converter in SMSL SU-8 way outperforms both of these DACs, providing much better signal transparency.

As an interesting aside, the SMSL has "different color" settings and on Tube 1, it almost matched that of Soekris dac1421!

Let's look at our jitter and noise at 44.1 kHz sampling:

Soekris dac1421 vs Schiit Yggdrasil DAC jitter noise and distortion measurement.png


The Schiit Yggdrasil has a couple of sidebands on each side of our main tone. They are not equal in amplitude so they are likely not jitter but unwanted regardless. The dac1421 doesn't have but the spectrum shows the power supply mains leakage. All of these are at -120 dB and lower so absolutely inaudible.

Finally, everyone's favorite measurement, linearity:
Soekris dac1421 vs Schiit Yggdrasil DAC linearity measurement.png


The Soekris dac1421 produces near ideal response with variation of just 0.5 dB at -120 dBFS (20 bits). The limit of measurement is about 0.2 dB so this is almost as good as it gets. Of course, it is worlds better than Schiit Yggdrasil DAC.

Listening Tests
Søren asked me, well pleaded with me :), to also do some listening tests. Regular readers of the forum know that I don't usually perform listening test comparisons on DACs since differences are too small to show up in my casual AB testing. But since he asked, I went ahead and did that against SMSL SU-8 since the Yggdrasil was returned to its owner.

I used Roon to group both DACs and connected their outputs to the two inputs on my Stax SRM-007t headphone amplifier, driving the Omega pro headphones. I then played a series of reference quality tracks from instrumental to classical, jazz, blues and everything else I could. Indeed I am listening to them as I am typing this. The output is simple: there is no audible difference. There just isn't to my ears.

If I could capture the outputs and then perform ABX testing I may be able to tell them apart. But in controlled AB testing, there is no difference that I can detect. Everything sounds as good as the source material allows. Excellent tracks sound superb through both.

Conclusions
Up until now, I had only tested multibit DACs from Schiit. Unfortunately all seem to have glaring design flaws, especially when it came to linearity. Thankfully, there is no such problem with Soekris dac1421. It has excellent linearity, and near absence of inharmonic distortion that plagues Schiit multibit DACs. This tells me the design is thought through and verified.

Alas, coaxing discrete resistors to be as precise as the fewer precision ones inside monolithic DACs is hard. So when it comes to distortion, the traditional sigma-delta DACs produce far better performance.

In my limited listening tests I cannot detect any "multibit magic." If you think you do, or want it regardless, I can recommend the Soekris dac1421.

-------------

As always, questions, comments, recommendations, etc. are welcome.

If you like this review, please consider donating funds for these types of hardware purchases using Patreon (https://www.patreon.com/audiosciencereview), or upgrading your membership here though Paypal (https://audiosciencereview.com/foru...eview-and-measurements.2164/page-3#post-59054).
 
Last edited:

amirm

Founder/Admin
Staff Member
CFO (Chief Fun Officer)
Joined
Feb 13, 2016
Messages
13,990
Likes
6,337
Location
Seattle Area
#3

restorer-john

Addicted to Fun and Learning
Joined
Mar 1, 2018
Messages
644
Likes
595
Location
Gold Coast, Quuensland, Australia
#5
So, basically we have current model, discrete(?) R2R D/A converter that costs $1000 and is outclassed in virtually every technical respect by the very first CD players released to the market in 1983, and it gets your tepid/qualified recommendation?

Can we see a FR plot if you get a chance?

Also, how can a 'linearity' down to -120dB be of any use whatsoever when we have S/N+D in the mid 80dB range? Clearly for that test it was a 24bit signal instead of a 16bit. 1st generation CD players hit 95dB unweighted and 105dB weighted- on 16 bit dithered/undithered routinely and that was 16 bit data.
 
Last edited:

amirm

Founder/Admin
Staff Member
CFO (Chief Fun Officer)
Joined
Feb 13, 2016
Messages
13,990
Likes
6,337
Location
Seattle Area
#6
So, basically we have current model, discrete(?) R2R D/A converter that costs $1000 and is outclassed in virtually every technical respect by the very first CD players released to the market in 1983, and it gets your tepid/qualified recommendation?
Well, no. The qualification was if someone wanted multibit for whatever reason. Then this is a better option than the other multibits I have tested.
 

amirm

Founder/Admin
Staff Member
CFO (Chief Fun Officer)
Joined
Feb 13, 2016
Messages
13,990
Likes
6,337
Location
Seattle Area
#7
Also, how can a 'linearity' down to -120dB be of any use whatsoever when we have S/N+D in the mid 80dB range?
Linearity test filters out almost all the noise and distortion. So it is only sensitive to actual output of the DAC, sans the other components.
 

derp1n

Active Member
Joined
May 28, 2018
Messages
270
Likes
201
#8
Thanks Amir, have been looking forward to these results after seeing @BE718's dam1021 measurements earlier.
 

gvl

Senior Member
Joined
Mar 16, 2018
Messages
484
Likes
132
Location
SoCal
#9
If the DAC itself has good linearity, where do all those distortions come from? Analog output stage?
 

amirm

Founder/Admin
Staff Member
CFO (Chief Fun Officer)
Joined
Feb 13, 2016
Messages
13,990
Likes
6,337
Location
Seattle Area
#10
The linearity test is a static one. RMS voltage is measured and analyzed. Real content has dynamic waveform and non-linearities in the dynamic range show up there. And of course there is also noise.
 

restorer-john

Addicted to Fun and Learning
Joined
Mar 1, 2018
Messages
644
Likes
595
Location
Gold Coast, Quuensland, Australia
#11
If the DAC itself has good linearity, where do all those distortions come from? Analog output stage?
With the numbers Amir measured, there's roughly 125uV of other noise (residual/distortion/psu harmonics etc) with respect to the rated output voltage.

Whatever way you look at it, that is a terrible result.
 

gvl

Senior Member
Joined
Mar 16, 2018
Messages
484
Likes
132
Location
SoCal
#13
The linearity test is a static one. RMS voltage is measured and analyzed. Real content has dynamic waveform and non-linearities in the dynamic range show up there. And of course there is also noise.
I see, roughly speaking the fact the DAC can linearly track RMSV doesn't necessarily mean it outputs a perfect sine, right?
 

Blumlein 88

Major Contributor
Joined
Feb 23, 2016
Messages
4,049
Likes
2,235
#14
So just guessing, Amir is probably saying to himself, "if I filter out noise leaving only low level linearity people get upset. People get the wrong idea. If I don't filter it out pretty well, other people get upset and get a different wrong idea." :)

It would be interesting to know how much of the distortion is from the analog output itself. Perhaps Mr. Kristensen could explain how his output stage is done and provide insight into that question. If not for that distortion, this would be a rather stellar result for multi-bit. And Kudos to Mr. Kristensen for sending in his unit for such public scrutiny.
 

restorer-john

Addicted to Fun and Learning
Joined
Mar 1, 2018
Messages
644
Likes
595
Location
Gold Coast, Quuensland, Australia
#15
So just guessing, Amir is probably saying to himself, "if I filter out noise leaving only low level linearity people get upset. People get the wrong idea. If I don't filter it out pretty well, other people get upset and get a different wrong idea." :)
Well said. :)

It would be interesting to know how much of the distortion is from the analog output itself.
Does it really matter? Residual noise, non-linearity, THD- who cares. It's all superfluous spuriae and shouldn't be buried.

The limiting parameter here is clearly S/N (residual noise) and that appears to be way back in the analog era numbers. I have cassette decks with better S/N than that. Even an ancient VHS HiFi VCR (with its head switching anomalies) can achieve S/N into the early 90s.

And Kudos to Mr. Kristensen for sending in his unit for such public scrutiny.
Absolutely. I reckon there are plenty of us here who have toyed with the concept of making a D/A converter and trimming our own R2R arrays at some point. Kudos for actually bringing a product to market.
 

Blumlein 88

Major Contributor
Joined
Feb 23, 2016
Messages
4,049
Likes
2,235
#16
Well said. :)



Does it really matter? Residual noise, non-linearity, THD- who cares. It's all superfluous spuriae and shouldn't be buried.
Well it was a point of curiosity to me. Unlike a certain infamous company most other aspects of this DAC look well put together.
The limiting parameter here is clearly S/N (residual noise) and that appears to be way back in the analog era numbers. I have cassette decks with better S/N than that. Even an ancient VHS HiFi VCR (with its head switching anomalies) can achieve S/N into the early 90s.
Actually in this case it looks like the 3rd harmonic distortion and slightly lower 2nd harmonic distortion are the main problems spoiling the numbers. The noise otherwise while not SOTA isn't too bad. Which means unlike a noise limitation, at lower levels the distortion probably isn't an issue either. And by lower levels I mean just a few db off max. Not trying to sugarcoat it or give a pass on a design problem. But this is a different result than one with simply too much noise or with bad jitter or strange glitches. It is educational to look at Amir's 1 khz noise and distortion graph. You have about the same single number in decibels for two devices. But were I choosing between the two, I'd choose the Soekris with higher THD and not those other spurious spikes in the noise floor. Again neither is exemplary, but they aren't equivalent even with the same spec number.



Absolutely. I reckon there are plenty of us here who have toyed with the concept of making a D/A converter and trimming our own R2R arrays at some point. Kudos for actually bringing a product to market.
 
Joined
Feb 9, 2018
Messages
34
Likes
51
#17
Let me do a couple of comments:

On my test bed I have zero power line harmonics, can't explain my Amir get so much (relatively), although its below the human hearing limits...

The output buffer is a zero feedback design so it add a little 2nd and 3rd harmonics at high levels, lower the level a couple of dB and they don't dominate anymore.... Still, they're way below what you can hear.

The slight different THD and Intermodulation results between the channels will mostly be present due to the slightly varying values of the resistors, there are 200 pcs of very precise ones.... They have a max specification but are typically significant better than the spec.

One reason I prioritize performance at low levels is the perfect digital volume control, my DACs are designed to be connected directly to a power amplifier or active monitors, or to headphones using the builtin headphone amplifier.

If you want to see a Sine down to -120 dB, one of the other forums had the balanced version, the dac1541, under test... The dac1421 is basically the unbalanced version of the dac1541.

And thanks to Amir for taking his time, results are better than what you could have feared here :)
 
Joined
Feb 22, 2018
Messages
59
Likes
18
#18
Not a big fan of comparing a $1000 amp/DAC to a (poorly measuring) $2300 DAC.
It should be compared to other sub-$1000 amp/DACs regardless of the design choices.
 

dc655321

Active Member
Patreon Donor
Joined
Mar 4, 2018
Messages
229
Likes
110
#19
Not a big fan of comparing a $1000 amp/DAC to a (poorly measuring) $2300 DAC.
It should be compared to other sub-$1000 amp/DACs regardless of the design choices.
I've never understood this type of thinking.
By that logic, there should be a strong correlation between price and objective performance.
Yet, there is very little.
 
Joined
Feb 22, 2018
Messages
59
Likes
18
#20
I've never understood this type of thinking.
By that logic, there should be a strong correlation between price and objective performance.
Yet, there is very little.
The line of thought is that the consumer who got a $1000 budget and is looking for an amp/DAC combo unit does not care about $2k+ standalone DACs, but instead if anything better than the dac1421 can be had for those money.
I get that they get compared in this review because they're both multibit, but from a consumer perspective, it's an apples and oranges comparison.
 
Top Bottom