• 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!

Review and Measurements of Schiit Yggdrasil V2 DAC

amirm

Founder/Admin
Staff Member
CFO (Chief Fun Officer)
Joined
Feb 13, 2016
Messages
44,368
Likes
234,386
Location
Seattle Area
This is a detailed review and measurements of Schiit Yggdrasil with upgraded "Gen 5 USB" interface and "Analog 2" DAC+analog board. I see this informally called Yggdrasil V2 even though there is no such model name on Schiit website. The retail price as of this writing is USD $2,399 plus shipping. I think it is the most expensive product Schiit makes (?). It is by far the most expensive DAC they make. In the world of high-end, this is actually "cheap." In the world of personal audio, it is of course quite expensive.

Yes, I know, I already reviewed the Schiit Yggdrasil DAC with same upgrades. As with its lower cost brother, the Schiit BiFrost Multibit DAC, it too did poorly compared to not only high-end DACs but also much cheaper mass market products. As you can see in the long review threads on those DACs, due to high popularity of them and the company, there were a lot of questioning and protests. That mostly died down until a couple of weeks ago when a bunch of new measurements posted in two other forums disputing essentially every flaw I had found in the Schiit Yggdrasil. A specific point was made that the Schiit Yggdrasil was architecturally a balanced DAC (no reason given why) and testing it with unbalanced outputs as I had mostly done was unfair to it.

I thought that point was fair and I immediately offered to re-test the Yggdrasil if someone loans it to me as I had borrowed the other two Schiit Yggdrasils for a few hours of testing and no longer had access to them. Sadly, neither Schiit, nor the people testing and publishing their measurements came forward to loan me theirs.

So imagine my pleasant surprise when a member of one of the other forums sent me a message here a week ago offering his Yggdrasil for testing. I of course jumped at the chance to accept it as I thought it was a fair request to test the balanced output and importantly, with yet another randomly selected purchased DAC. He was kind enough to FedEx the unit (plus some other goodies I will test later), and it got to me yesterday. I immediately powered the unit and here we are with brand new set of results for the Schiit Yggdrasil "V2" DAC.

Given the huge amount of attention and scrutiny my Schiit product reviews get in general, and this unit in the specific, I created a set of industry standard DAC tests, trying where possible to stick to standard configuration of such tests in my new Audio Precision APx555. So replicating them should be easier. Still, I will make it even easier by providing all the project setting files including the measurements for what you see on request. So there is full transparency here.

As usual I like to compare two units so that there is context for their measured performance. Importantly any methodology mistakes will be reflected in both products making it a) easier to find and b) invariant to the conclusions if there are difference in performance. For this testing I picked the Topping DX7s which retails for USD $499 but often goes on sale for less. Here you can see the full stack up used in testing:

Schiit Yggdrasil DAC V2 Review and Measurement.psd.jpg


Measurements
As requested by its owner and per above, all the tests here unless mentioned otherwise are with balanced output of the DAC. Again, per proper request, the input used is USB (my previous tests were mostly with S/PDIF). In almost all cases, sample rate is 44.1 kHz. Other sample rates will be tested in the future.

The unit was powered on continuously since arriving and had a few hours of warm up time. It was warm to the touch. Owner has requested week long warm-up testing which I will perform later. Of note, in a few tests I ran when I first powered it on and after a few hours of warm up time, there was no significant change (i.e. fixing any issues that were there).

The balanced measurement cables were from Audio Precision (OP kindly sent me his but they were too short for this use). I have tested these cables and performance is the same as internal loopback in Audio Precision APx555. So they are definitely transparent enough for our testing.

A single rack-mounted power strip provides power to both APx555 and the two DACs as is usually the case in my testing.

My usual Z-series HP laptop with the same two ports I use for testing was used to connect to both DACs. Audio Precision software can only talk to USB devices over ASIO. Schiit doesn't supply ASIO drivers anymore (I tried installing what they had but did nothing). So I resorted to using ASIO4AALL driver which provides "bit exact" communication to the Schiit DAC. Topping does provide ASIO drivers but to keep the variables low, I choose to also use ASIO4ALL to talk to it.

And just in case it makes a difference, it was cloudy yesterday and temps inside were around 75 degrees. :D

Enough intro. Let's look at our dashboard view of a 1 kHz tone and see how the Yggdrasil DAC does:
Schiit Yggdrasil DAC Dashboard Measurement.png


We immediately notice a few issues. On the FFT spectrum to the left of our 1 kHz tone (all others are noise/distortion), we see significant spikes at double the mains AC supply. This is created by the internal power supply and is bleeding into the output of the DAC.

To the right, we see an awful lot of distortion spikes. They translate into poor SINAD of just 86 dB. This means that above the energy of distortions+noise, we only have 86 dB to work with. Let's remember that we need 93 dB for our CDs properly dithered.

I circled the frequency counter showing 1.00008 kHz instead of perfect 1 kHz. We can't get perfection here as there are clock errors but this is 80 parts per million which is quite high. Usually I see just the last dig going +1 to -1 for 10 parts per million. The clock source in Schiit Yggdrasil DAC is definitely running fast. Fortunately our ears are completely immune to such errors so not an audible concern. Just lack of precision which should exist in any such caliber DAC.

Here is the dashboard view of Topping DX7s which is created by simply moving the balanced cables to it and changing the output device in ASIO4ALL to it:
Topping DX7s DAC Dashboard Measurement.png


Now this is more like it. There is almost no power supply contributions to the left of our tone in FFT in top right. To the right we have some noise and distortion but far lower resulting in much better SINAD of 108 dB. That is a whopping 22 dB improvement!!!

To put things in context, my new APx555 is just 6 dB better than my 20 year old AP SY2522. So even small amounts of dB improvements matter and hard to come by. So 22 dB is an infinity in that context.

Note that the frequency counter is just off by one digit as I mentioned at 0.99999 kHz. That is the typical performance.

The Schiit Yggdrasil DAC has 4.2 volt RMS versus 4 volts for Topping DX7s. So if you are ever comparing the two audibly, be very careful of this as it would tilt the odds in favor of Yggy DAC just because it is louder. Fortunately for our testing and looking at numbers on a graph, that difference is too small to worry about, making comparisons easy without level compensation.

In my previous review I noted that the "Analog 2" version of Schiit Yggdrasil DAC has a droop in its unbalanced outputs in low frequencies. Testing that with this loaned unit shows the exact same thing:

Schiit Yggdrasil DAC Frequency Response Measurement.png


The balanced output as tested before is fine. The unbalanced should be identical to it, albeit at lower amplitude. But instead, we have "shelf" starting around 300 Hz and going down. FYI it is fine to have a drop at say, 15 Hz as the balanced output is doing. But not the weird shape that we are seeing and the drop so far up the frequency range.

As noted with the cursors, the drop is 0.52 dB so not huge but nailing the frequency response is what DACs do. There is no reason for it to not have flat response on its unbalanced output.

I am sure you all have been waiting for the linearity results since they are so easy to read and heavily disputed by others at this point. This is the only test that is kind of "custom." In order to get rid of noise and distortion contributions in both the analyzer and DAC, a heavy handed filter is used to narrowly filter the source frequency out of the digital generator (over USB). In later posts I will show data using other methods which agree with this measurement regardless:

Schiit Yggdrasil DAC vs Topping DX7s DAC Linearity Measurement.png


The Schiit Yggdrasil is in red. We see that it starts to lose accuracy at around -90 dB and reaches my threshold of 0.1 dB of error at -97 dB. This is 16 bits which fully agrees with my previous measurement of unbalanced output. From there on, we see an exponential rise on error with lots of error at -120 dBFS.

The Topping DX7s in blue in sharp contrast, goes to the limit of the measurements, only showing tiny up and down errors which are at the limit of what the analyzer can do. So while I also give it is 0.1 dB error value of -111 dBFS, it is really "linear" to 20 bits or -120 dB.

The performance of the Schiit Yggdrasil is very disappointing in this regard especially for those of you who believe in high-res audio. And again, is fully consistent with my previous measurements of other units with my other Audio Precision analyzer.

Jitter and noise is less eventful:

Schiit Yggdrasil DAC vs Topping DX7s DAC Jitter and Noise Measurement.png


We see some sidebands in red hugging our main 12 kHz tone. That is likely caused by the power supply contributions which we also see on the far left of the graph. Fortunately we are quite deaf in low frequencies and perceptual masking will make sure that those jitter components are not heard.

The Topping DX7s has a few spikes of its own but all at lower level. It has an objectively better implementation here despite it costing less than one quarter of Schiit Yggy DAC.

Intermodulation distortion using SMPTE two-tone signal delivers this:

Schiit Yggdrasil DAC vs Topping DX7s DAC IMD Distortion Measurement.png


Despite the high input impedance of my analyzer (200 k Ohms), i.e. no load at all on the output of the Schiit Yggy, we see its output distortion start to rise at -15 dB indicating that it is starting to clip! That is quite surprising and not good. See how Topping DX7s doesn't hardly clip and when it does a tiny bit, it is 25 dB lower distortion than Schiit Yggdrasil!

As I have shown before, the Topping DX7s has a regression with that belly around -40 to -25 dB where its distortion rises (this was NOT the case with Topping DX7). Had it not been the case, it would have hugely outperformed the Schiit DAC.

Let's look sister measurement of harmonic distortion+noise versus frequency:
Schiit Yggdrasil DAC vs Topping DX7s DAC THD Distortion Measurement.png


Wow, oh just wow The Topping DX7s easily outperforms the Yggdrasil and by huge margin. Of note, both measurements agree with manufacturer specs which are apart by a factor of 10 (0.005% versus 0.0005%). What is causing all of that? We saw evidence of it in our Dashboard FFT. But let's zoom in for a better look:

Schiit Yggdrasil DAC vs Topping DX7s DAC1 khz residual Distortion Measurement.png


We are feeding the DACs a pure, digitally created 1 kHz tone over USB and this is what comes out of them. The Schiit Yggdrasil has a huge spray of distortion spikes. We not only have the typical harmonic distortions (i.e. multiples of 1 kHz) but a ton of others in between. Notice how fewer spikes the Topping DX7s produces in blue. That is what we should be seeing.

Sadly this is exactly what I saw in my test of previous Schiit Yggdrasil DACs. It is often said that non-harmonic distortion is not "natural" and is bad for your ears. If so, this DAC is really bad for your ears. :)

We can skin this cat differently by looking at a low amplitude -90 dB sine wave. Typically I show just the "time domain scope" view of this but I thought it is more revealing to look at the spectrum of it at the same time. So here we go:

Schiit Yggdrasil DAC -90 db Measurement.png


Hard to see this on the left but in a real-time display on my analyzer, it was easy to see the waveform dancing up and down a bit to the tune of mains harmonics. The variations became much larger in unbalanced output (not shown). Looking at the FFT on the right, we see confirmation of it with the power supply contributions a larger portion of our -90 dB 1 kHz tone. Likewise we have our spray of distortions to right as before.

Here is Topping DX7s in contrast:
Topping DX7s DAC -90 db Measurement.png


Isn't the contrast amazing? This is what we should be seeing in a DAC that claims to accuracy. Our main tone at 1 kHz and despite its low level, freedom from power supply noise and distortion spikes.

I have more measurements to share and the owner of the unit has requested good bit of others too. But for now, I wanted to simply show my usual measurements I perform in my reviews so time to stop here.:

Conclusions
This Schiit Yggdrasil DAC was the the third unit I have tested. I have now measured this DAC in three locations and with both balanced and unbalanced. All measurements agree hugely with each other despite the variation in units, location, and even which Audio Precision analyzer is used, giving high confidence to the results.

Those results unfortunately paint the same picture as before: that there are fundamental problems in Schiit multibit DACs when it comes to doing what they are told to do: convert digital samples faithfully to analog. Whether we use balanced or unbalanced output, problems remain. And not just in one measurement but practically everywhere we look.

I know this is an uncomfortable message for the owners of these expensive DACs. Personally I was wishing the balanced output would measure well so that I would not worry about their angst and all the follow up arguments that are bound to start. But the data speaks otherwise.

If your ear doesn't hear these problems, consider yourself blessed and enjoy your unit. :) But please don't dispute the measurement data. They provide very high confidence results given the multiple times I have tested the model.

I like to thank the kind owner of this unit for loaning it to me. I think it is helping to bring a lot more clarity to the objective performance of this unit and for that, I am immensely appreciative.

Needless to say I can NOT recommend the Schiit Yggdrasil DAC for purchase. If you want a high-performance high-end DAC, my recommendations would be for RME ADI-2 DAC at less than half the price. And the Benchmark DAC3 at similar price. Both are made in the western world so no issue with "buying Chinese."

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

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).
 
OP
amirm

amirm

Founder/Admin
Staff Member
CFO (Chief Fun Officer)
Joined
Feb 13, 2016
Messages
44,368
Likes
234,386
Location
Seattle Area
The owner of the Schiit Yggdrasil DAC had requested that I test the unit after a week of warm-up to see if the outcome changes. I have provided nearly daily snapshots in the thread but here are the summary of changes from day 1 to day 8 with the unit having stayed on the full time. Measurements are with USB input and balanced analog output.

1531072136234.png


All of these are negligible errors occurring from second to second on their own.

So contrary to what some believe, warm-up makes no difference on this DAC even though it is said to need it.

Please don't waste electricity leaving the unit on. Internet folklore is just that: folklore.

Of course, what is shown to be broken in measurements, is broken regardless of time on. To wit, here are the linearity errors that continue to this day precisely as day 1:
Schiit Yggdrasil DAC Day 8 Linearity Measurement.png


A good DAC will have essentially flat line at 0 dB showing no real measurable error. The Schiit Yggdrasil does not do that: it shows errors which increase as one goes beyond 16 bits of resolution. And passage of time makes zero, let me repeat, zero difference. All the graphs are on top of each other.

If temperature equalization was needed, this could have been implemented in the unit in the form of oven controlled oscillators, temperature controlled subsystems, etc. as is sometimes done in high-precision instruments. The fact that it is not there indicates either the designer didn't think it was needed, or that they took the cheap way out by not including it in such a pricey DAC.

So please don't accept internet chatter on matters that are readily measured. If manufacturer insists on warm-up, break-in, etc., tell them to show you measurements of how they ascertained that. If they say they can't measure it, then look for another DAC vendor.
 
Last edited:

RayDunzl

Grand Contributor
Central Scrutinizer
Joined
Mar 9, 2016
Messages
13,200
Likes
16,982
Location
Riverview FL
A sample size of three out of the assumedly thousands of units produced means what, statistically?
 

Ron Texas

Master Contributor
Joined
Jun 10, 2018
Messages
6,077
Likes
8,913
A sample size of three out of the assumedly thousands of units produced means what, statistically?

3 out of 3 randomly chosen is enough. Sorry, I can't do the math for you.

The measurements are actually shocking considering the many glowing reviews the Yaggi received from paid reviewers and its $2200 price tag. This throws into question the entire audiophile review publishing industry.

Freedom of speech is the freedom to say unpopular things. Amir has been banned elsewhere because he has the numbers and the machines don't lie. Another unfortunate situation and an indication of other's lack of respect for freedom of speech.
 
OP
amirm

amirm

Founder/Admin
Staff Member
CFO (Chief Fun Officer)
Joined
Feb 13, 2016
Messages
44,368
Likes
234,386
Location
Seattle Area
A sample size of three out of the assumedly thousands of units produced means what, statistically?
Hard to compute because the samples here are correlated. That is they are all manufactured supposedly to some specifications. So standard formulas for confidence don't apply since each sample (e.g. in a survey) is not independent from any other.
 

Ron Texas

Master Contributor
Joined
Jun 10, 2018
Messages
6,077
Likes
8,913
I am thinking with even a 1% defect rate (which is high) the chances of getting 3 bad ones in a row are around one in a million.
 
OP
amirm

amirm

Founder/Admin
Staff Member
CFO (Chief Fun Officer)
Joined
Feb 13, 2016
Messages
44,368
Likes
234,386
Location
Seattle Area
Ah, Jude was kind enough to post his linearity script for Audio Precision APx555 which we both have. I took his script which already had his measurements and ran it against my Yggdrasil DAC. I had not seen it mention that he was using balanced AES digital input on Yggdrasil before. So I did the same. Here are the results with full screen of the app visible so that there is no question about what is being run:

Schiit Yggdrasil Jude Script Linearity Measurement.png


As you see, my results are completely different than his. And quite consistent with what I post in the review with my custom filtering.

So at this point we can put aside the food fight on measurement setting differences.

The unknowns left are different units being tested. Hmmm. :)
 

gvl

Major Contributor
Joined
Mar 16, 2018
Messages
3,427
Likes
3,981
Location
SoCal
I hope you can get your hands on another quality oversampling multi-bit DAC, such as those from Soekris, and compare it to the Yggy, or even a vintage one from the 90s based on the 20-bit chips from that era, e.g. PCM63 or UltraAnalog. It would be more of an apples to apples comparison than putting it against a Delta-Sigma unit that's based on the (almost) top of the line ESS chip of current production. That said, there is no excuse for the mains noise.
 
Last edited:

mindbomb

Active Member
Joined
Nov 11, 2017
Messages
284
Likes
175
Sinad and 0dbfs smpte imd being like 25db worse than its contemporaries is pretty bad. And that's not even mentioning the unseemly power supply behavior.
 

gvl

Major Contributor
Joined
Mar 16, 2018
Messages
3,427
Likes
3,981
Location
SoCal
I live not too bad of a drive from and have been wanting to spend some time in the Schiitr to get some subjective impressions in the natural habitat :)
 

Timbo2

Senior Member
Joined
Feb 27, 2018
Messages
497
Likes
396
Location
USA
Ah, Jude was kind enough to post his linearity script for Audio Precision APx555 which we both have. I took his script which already had his measurements and ran it against my Yggdrasil DAC. I had not seen it mention that he was using balanced AES digital input on Yggdrasil before. So I did the same. Here are the results with full screen of the app visible so that there is no question about what is being run:

View attachment 13509

As you see, my results are completely different than his. And quite consistent with what I post in the review with my custom filtering.

So at this point we can put aside the food fight on measurement setting differences.

The unknowns left are different units being tested. Hmmm. :)

Trying to head off any FUD I note this was done at 48kHz, maybe try 44.1kHz while they are both on the bench. Not that I expect a difference...
 

SIY

Grand Contributor
Technical Expert
Joined
Apr 6, 2018
Messages
10,383
Likes
24,749
Location
Alfred, NY
Amir, I'd certainly like a copy of the project files. That's great of you to offer this.
 
OP
amirm

amirm

Founder/Admin
Staff Member
CFO (Chief Fun Officer)
Joined
Feb 13, 2016
Messages
44,368
Likes
234,386
Location
Seattle Area
Amir, I'd certainly like a copy of the project files. That's great of you to offer this.
Please start a conversation with me and we can figure out a way to send you the file.
 
OP
amirm

amirm

Founder/Admin
Staff Member
CFO (Chief Fun Officer)
Joined
Feb 13, 2016
Messages
44,368
Likes
234,386
Location
Seattle Area
Trying to head off any FUD I note this was done at 48kHz, maybe try 44.1kHz while they are both on the bench. Not that I expect a difference...
It makes no difference at all. Here it is:

1530388738443.png


It lands right on top of my previous measurement pixel for pixel. I also have tested with Toslink with the same result.

Edit: that is, it is just as bad as my previous measurements.
 
Last edited:
Top Bottom