– Audio Precision APx555 B-Series Analyzer with 200kOhm input impedance
– Audioquest Mackenzie interconnects
– USB Source: Intel PC via ifi iGalvanic 3.0 and iUSB 3.0
– Analyzer galvanically isolated via Intona 7055-D USB 3.0 isolator
– Audioquest Carbon USB and SPDIF cables
– All measurements shown in this post are taken with USB input, XLR output, 44.1khz sample rate unless otherwise specified
– DAC was warmed up for a full 24h before measuring. (Another set of measurements will be taken a week from now, to test if any of the Yggdrasil’s infamously long warmup time has an impact on this model)
– Full reports containing additional data and test configurations are attached
– Exact analyzer/filter configurations for each measurement are detailed in the full reports
Schiit is a company whose approach to product development I really appreciate. Many companies are focused on marketing ‘upgrades’ and ‘the next best thing’ each year. Whereas Schiit is quite comfortable being honest about their products, pricing things sensibly, giving modular upgrades when feasible instead of forcing people to replace an entire product, not releasing new iterations just for the sake of having something new, and in this particular case, giving the customer various options to fit their preferences, rather than a straight tiered approach.
What is even more fantastic is that if you decide you would like more than one version of the Yggdrasil to try or ‘dac roll’, you can do that!
The individual cards can be purchased for much less than the cost of an entire yggdrasil, and so you can swap dacs for a reasonable cost.
HUGE props to Schiit for doing this whereas most manufacturers would have made you buy an entire second dac. You have the choice of swapping these cards yourself, or sending your unit back to Schiit for them to install it.
The Yggdrasil OG used Analog Devices chips, which whilst not strictly speaking full R2R, were still full multibit converters, meaning they could convert 16 bit PCM at 16 bits, without the need for delta sigma modulation to a lower bit-depth.
It was a well loved dac, and still has quite a good reputation, BUT, one common issue that people pointed out was that it, along with many other non-delta-sigma dacs, did not do so well from an objective standpoint.
Rather than outright replacing the Yggdrasil OG, Schiit is now offering two additional versions. Firstly, we have the ‘More is Less’ edition. This uses 4 x TI DAC11001 20-bit chips, and is intended to have some of the lowest THD+N of any ‘multibit’ DAC available today.
BUT, beta testers and Schiit staff say that this was usually the least preferred option, and that the other new arrival, the ‘Less is More’ edition, may measure worse, but sounds better.
The ‘Less is More’ uses 4 x TI DAC8812 16-bit DAC chips. Additionally, both the MIL and LIM use opamps for the output stages, whereas the OG Yggdrasil has a discrete output stage.
So we have:
– Yggdrasil ‘Less is More’ – The best sounding of the three according to many at Schiit
– Yggdrasil OG – The existing and well loved Yggdrasil, for those who may prefer it.
– Yggdrasil ‘More is Less’ – Not as good sounding according to schiit, but ideal for those who want nothing but the best measured performance.
Schiit is using their ‘Thunderdome’ approach whereby several products or versions of a product are released, and only the most successful will stay on. So which versions of the Yggdrasil remain in the future we will have to wait and see. But it’s certainly neat that there is an interesting choice of designs and objective performance available.
My approach has always been that measurements can tell you a lot and guide you in the right direction, but at the end of the day, measurements can explain what we might hear subjectively, but they cannot predict with certainty whether something will actually sound good. There are simply too many variables to measure and and too many things we do not know and cannot predict about our auditory system and personal preferences. So whilst measurements are important, full evaluation of a product must incorporate both objective and subjective discussion.
This unit was sent to me by Schiit. I will be doing a full video review as well as measurements of the ‘Less is More’ edition, and hopefully the OG yggdrasil too, once I’m able to get ahold of them. So for now, I’ll be keeping subjective thoughts to myself until I can compare the Yggdrasil versions together.
But, let’s have a look at the measurements of their (objectively) best performing model!
Full Measurement Reports:
Reports available here:
Dynamic Range (AES17): 118.8dB
Noise Level RMS (20-20khz): 3.792uVrms
Noise Level RMS (20-90khz): 8.735uVrms
DC Offset: 6.109mV Active, 5.963mV idle
Absolutely stellar THD+N here! With equivalent SINAD to a Holo May. (Harmonics are lower than the May, but noise is higher. Overall the SINAD value is about the same according to WolfX’s measurements https://www.audiosciencereview.com/forum/index.php?threads/review-and-measurements-of-holo-audio-may-probably-the-best-discrete-r2r-dac.10161/ )
THD+N / Frequency:
What we can see here is that whilst the MIL does indeed have excellent or perhaps even incredible THD+N for an R2R dac at 1khz, the level of distortion and noise rises drastically up into the higher frequencies, with only 0.2% (-73dB) THD+N at 20khz.
Depending on the performance of the other models, this alone could explain why this model was less preferred to the OG and LIM models.
This is an example of why a 1khz test can often miss out important factors about the performance of a product, and further testing than just SINAD and Dynamic Range is absolutely needed.
This behaviour is likely due to the way that the DAC11001A handles ‘de-glitching’ using an integrated sample and hold circuit. For lower frequencies, the voltage between two samples is smaller and so the DAC does not need all too much time to settle after holding.
But for high frequencies, the DAC gets closer and closer to the point where it cannot fully settle before the next sample is handled.
This is unfortunately not something Schiit themselves can do anything about, as it’s just inherent to how this chip works. This chip was not really designed for audio and high frequency applications and was designed for high-end medical applications where DC performance is important.
Because of this factor, higher frequency signals will have more (and different) distortion to lower frequency signals. And additionally, the sample rate of the DAC will have a direct impact on distortion as well.
With a 1khz tone, the pattern of distortion changes when we swap from 44.1khz to 192khz, how much of this is to do with the DAC and how much of it is to do with the Schiit upsampling/filter/DSP I am not sure:
And then if we move to a 17.5khz signal, our distortion is considerably higher, and once again changes if we use a higher sample rate:
But to demonstrate the unpredictable behaviour caused by this chip design quirk, if we go just from a 17.5khz tone to an 18khz tone, our distortion once again changes entirely:
Therefore, this means that with music, which will have content at all frequencies upto 20khz (and above for hires), distortion is going to be very high and unpredictable. This DAC performs very well in strict test conditions, but for actual music, there are some concerns.
UPDATE: There is a discussion ongoing on SBAF looking at this behaviour. It seems that this may be due to schiit’s filter and not the Yggdrasil MIL itself. If you’d like to follow the discussion you can view the thread here: https://www.superbestaudiofriends.org/index.php?threads/schiit-yggdrasil-less-is-more-and-mil-and-og-impressions-measurements.11391/page-13#post-360200
I even encountered some odd situations where the distortion products and noise floor would move/sway slowly even when there was just a static tone playing:
Linearity is excellent!
SINAD/THD+N vs output level:
Now this is quite interesting. Usually with most modern R2R or other ‘full multibit’ designs, you’d see some sawtoothing on this graph as the signal transitions from one converter to another. An example is shown below the Rockna Wavedream.
The fact that we don’t see this suggests that the chip internally is not following this approach.
IMD (SMPTE) vs Output Level:
-90.31dBfs 1khz Sine (Recorded with 192khz ADC Sample Rate):
Filter Ultrasonic Attenuation:
Idle Noise Spectrum upto 1.2Mhz:
Also worth noting, this DAC shows some ultrasonic harmonics of sorts:
44.1khz Jitter (USB):
NOTE: I am using a 192khz ADC sample rate for 44.1khz J-Tests from now on to address a current bug in the AP firmware. Please do not directly compare this to older 44.1khz J-Tests, and note that the noise floor on the FFT will be slightly higher than 48khz J-Tests.
Jitter levels are exceptionally low, Unison USB is clearly doing a great job here!
48khz Jitter (USB):
I also checked 48khz Jitter over SPDIF to ensure that was fine, and it was excellent too.
This will of course be dependent on the quality of your digital source though. Using the APx555B I can actually add jitter to the digital output, and adding 10ns of 1khz jitter caused the J-Test to show as below. The Yggdrasil does have a decent PLL, but quality of digital source should be taken into account (as on the vast majority of dacs).
NOTE: I am using a new, higher FFT length method for this test.