It is an understatement to say that my measurements of Schiit products create tons of discussions and arguments across multiple forms. So much so that I think they parallel any reality show on TV! Much of that is driven by passion people have on either side of the fence. What I like to address is the accusation that my measurement techniques are incorrect, or cooked in some ways as to generate unfavorable results for Schiit.
As a way of background, so far I have measured three different samples of Schiit Yggdrasil all from actual customers. You can see the review of the latest one here: https://audiosciencereview.com/foru...measurements-of-schiit-yggdrasil-v2-dac.3607/
As shown there, the performance of Schiit Yggdrasil DAC is simply not competitive at almost any price, let alone at its retail price of $2399. Importantly though, I always measure two devices side-by-side. This has two benefits:
1. Any errors in measurement technique translates into both products and as such, the relative performance most likely remains valid.
2. The second product provide an anchor/reference to judge the first. Without it, these are a bunch of graphs that is hard to internalize.
Given this, the data should be pretty compelling and lead the conclusion I stated. Alas, people still complain and go as far as call my competence or intentions into question. I think this is mainly generated by inability of people to truly understand the measurement data, aggravated by others bombarding the reader with tons of obtuse measurements. They then put a few words around it which people run with instead of trying to read and understand the graphs.
Note that in all the tests performed, no two measurements are made from the same unit. I have the units I have tested which all came from customers. They have the units they have tested with unknown origin -- mostly likely given to them by Schiit directly. So some variations cannot be helped. I asked to measure the devices tested by both Jude and Head-fi and AtomicBob (Bob Smith), neither one of whom has agreed to do so. This is very disappointing as Bob Smith lives in drive distance from me.
So how do we make progress? I thought I do something different which is to try to demonstrate that the faults that I see exist very much in the measurements performed by others, in this case, atomicbob. Let me repeat: I will try to make my case with the data presented by the other side.
To do that, I looked at the published tests by atomicbob and spent good bit of time getting my measurement settings to match his. This turned out to be more complicated than I thought due to very poor documentation provided by Bob Smith. And with terrible presentation of his low-contrast graphs. Still, I managed to get a few of them replicated before I had to return my Schiit Yggdrasil to its owner.
So let's dig in.
Measurements
All measurements are from balanced output and digital input.
Let's start with a simple test of 1 kHz tone at 0 dBFS (meaning full amplitude) with levels matched to Topping DX7s as a reference:
Top top graph with black background is from Bob Smith. Everything other than the main tall spike at 1 kHz is unwanted noise and distortion. If you look at the spectrum of distortion and noise from Schiit Yggdrasil in red, you see that there is tons of them. The performance of Topping DX7s is overlayed on top of it (in blue). As we see, all the distortion spikes that are the same frequency, have lower amplitude in Topping DX7s. What's more, the distortions die off sooner than they do on Schiit Yggdrasil.
Note something interesting though: the sampling rate selected by AtomicBob is 96 kHz. Why? That is not a common rate like 44.1 kHz. Why use 96 kHz? And why not note that clearly on measurement graphs? It is these kinds of ad-hoc choices that make it difficult or misleading to try to compare measurements.
That aside, now you see the value of show two devices simultaneously. If we did not have the Topping DX7s performance to compare, we wouldn't know necessarily how good or bad the Schiit Yggdrasil is. Here, we see as much as 30 dB reduction in second harmonic with Topping DX7s! Even the second harmonic is 15 dB better.
Next, let's look at the same 1 kHz tone, but this time at -60 dBFS (much lower level):
The message is the same although sample rate now is 44.1 kHz. Here, the Yggdrasil has a noise floor advantage but it severely underperforms the Topping DX7s with a spray of both harmonic and inharmonic distortions (NOT multiples of 1 kHz). Between higher noise floor and higher distortion of both types, I take the former.
Next let's look at intermodulation distortion by feeding a DAC a pair of frequencies and see what comes out. Anything other than our two sine waves is unwanted noise and distortion. First, 19 Khz+20 kHz tones:
Once again the measurement is at 96 kHz without any due notice to reader. The Topping DX7s has clearly less non-linearities by the fact that it has far fewer and lower amplitude intermodulation products. The Schiit Yggdrasil in contrast creates a spray of unwanted frequencies before and after our dual tones.
Here is another with 50 Hz tone combined with 7 kHz:
Once again the Schiit Yggdrasil shows a slightly lower noise floor at lower frequencies but spits out massive spray of unwanted distortion products. So much so that they cluster into a solid mass. These all indicate a DAC that is not linear, i.e. it doesn't output what it is told to do. The Topping DX7s produces far cleaner output (in blue).
Finally let's look at harmonic distortion+noise versus level:
AtomicBob has two pairs of graphs one with just the THD and the other, THD+N. I went after the much more common THD+N and as you see, once again I achieved the same results he has. Then when overlayed with results of Topping DX7s, we see that once again the Schiit Yggdrasil underperforms.
The way THD+N works is that if a device has very little distortion, the THD+N keeps getting lower and lower as we increase the amplitude of our signal. When the graph changes direction and it goes back up, it indicates that distortion is dominant rather than noise. In the case of Topping DX7s, that happens a bit in the middle and then at the end of the scale where the output stage starts to saturate.
The Schiit Yggdrasil however, starts to show all kinds of odd distortion onsets with the graph being so chewed up, going up and down as distortion or noise become more dominant.
The Topping DX7s is actually not as good as state of the art devices. The RME is one such and let's look at its performance:
See how perfect the ADI-2 Pro is (the ADI-2 DAC is even better). There are clear engineering and performance problems in Schiit Yggdrasil DAC.
For this last bit, let's review my measurements of Schiit THD+N at 1 kHz at full amplitude and SINAD (signal above power of noise+distortion):
Here is the spec from Schiit's own web site:
So my measurements are absolutely right on the money. And how do they compare to other DACs? Here is a comparison:
Every DAC from other manufacturers outperforms Schiit Yggdrasil regardless of price! Again, remember that my measurements are identical to what Schiit publishes here (SINAD is inverse of THD+N).
Summary
Direct comparison of my measurements and those performed by Bob Smith (atomicbob) show that the two show the same outcomes. Importantly, the performance as measured by him (and replicated by me) is simply not competitive. The difference is not small at all. We are talking 20 to 30 dB deficiency in the output of Schiit Yggdrasil DAC.
As always, comments, corrections or questions are welcome.
As a way of background, so far I have measured three different samples of Schiit Yggdrasil all from actual customers. You can see the review of the latest one here: https://audiosciencereview.com/foru...measurements-of-schiit-yggdrasil-v2-dac.3607/
As shown there, the performance of Schiit Yggdrasil DAC is simply not competitive at almost any price, let alone at its retail price of $2399. Importantly though, I always measure two devices side-by-side. This has two benefits:
1. Any errors in measurement technique translates into both products and as such, the relative performance most likely remains valid.
2. The second product provide an anchor/reference to judge the first. Without it, these are a bunch of graphs that is hard to internalize.
Given this, the data should be pretty compelling and lead the conclusion I stated. Alas, people still complain and go as far as call my competence or intentions into question. I think this is mainly generated by inability of people to truly understand the measurement data, aggravated by others bombarding the reader with tons of obtuse measurements. They then put a few words around it which people run with instead of trying to read and understand the graphs.
Note that in all the tests performed, no two measurements are made from the same unit. I have the units I have tested which all came from customers. They have the units they have tested with unknown origin -- mostly likely given to them by Schiit directly. So some variations cannot be helped. I asked to measure the devices tested by both Jude and Head-fi and AtomicBob (Bob Smith), neither one of whom has agreed to do so. This is very disappointing as Bob Smith lives in drive distance from me.
So how do we make progress? I thought I do something different which is to try to demonstrate that the faults that I see exist very much in the measurements performed by others, in this case, atomicbob. Let me repeat: I will try to make my case with the data presented by the other side.
To do that, I looked at the published tests by atomicbob and spent good bit of time getting my measurement settings to match his. This turned out to be more complicated than I thought due to very poor documentation provided by Bob Smith. And with terrible presentation of his low-contrast graphs. Still, I managed to get a few of them replicated before I had to return my Schiit Yggdrasil to its owner.
So let's dig in.
Measurements
All measurements are from balanced output and digital input.
Let's start with a simple test of 1 kHz tone at 0 dBFS (meaning full amplitude) with levels matched to Topping DX7s as a reference:
Top top graph with black background is from Bob Smith. Everything other than the main tall spike at 1 kHz is unwanted noise and distortion. If you look at the spectrum of distortion and noise from Schiit Yggdrasil in red, you see that there is tons of them. The performance of Topping DX7s is overlayed on top of it (in blue). As we see, all the distortion spikes that are the same frequency, have lower amplitude in Topping DX7s. What's more, the distortions die off sooner than they do on Schiit Yggdrasil.
Note something interesting though: the sampling rate selected by AtomicBob is 96 kHz. Why? That is not a common rate like 44.1 kHz. Why use 96 kHz? And why not note that clearly on measurement graphs? It is these kinds of ad-hoc choices that make it difficult or misleading to try to compare measurements.
That aside, now you see the value of show two devices simultaneously. If we did not have the Topping DX7s performance to compare, we wouldn't know necessarily how good or bad the Schiit Yggdrasil is. Here, we see as much as 30 dB reduction in second harmonic with Topping DX7s! Even the second harmonic is 15 dB better.
Next, let's look at the same 1 kHz tone, but this time at -60 dBFS (much lower level):
The message is the same although sample rate now is 44.1 kHz. Here, the Yggdrasil has a noise floor advantage but it severely underperforms the Topping DX7s with a spray of both harmonic and inharmonic distortions (NOT multiples of 1 kHz). Between higher noise floor and higher distortion of both types, I take the former.
Next let's look at intermodulation distortion by feeding a DAC a pair of frequencies and see what comes out. Anything other than our two sine waves is unwanted noise and distortion. First, 19 Khz+20 kHz tones:
Once again the measurement is at 96 kHz without any due notice to reader. The Topping DX7s has clearly less non-linearities by the fact that it has far fewer and lower amplitude intermodulation products. The Schiit Yggdrasil in contrast creates a spray of unwanted frequencies before and after our dual tones.
Here is another with 50 Hz tone combined with 7 kHz:
Once again the Schiit Yggdrasil shows a slightly lower noise floor at lower frequencies but spits out massive spray of unwanted distortion products. So much so that they cluster into a solid mass. These all indicate a DAC that is not linear, i.e. it doesn't output what it is told to do. The Topping DX7s produces far cleaner output (in blue).
Finally let's look at harmonic distortion+noise versus level:
AtomicBob has two pairs of graphs one with just the THD and the other, THD+N. I went after the much more common THD+N and as you see, once again I achieved the same results he has. Then when overlayed with results of Topping DX7s, we see that once again the Schiit Yggdrasil underperforms.
The way THD+N works is that if a device has very little distortion, the THD+N keeps getting lower and lower as we increase the amplitude of our signal. When the graph changes direction and it goes back up, it indicates that distortion is dominant rather than noise. In the case of Topping DX7s, that happens a bit in the middle and then at the end of the scale where the output stage starts to saturate.
The Schiit Yggdrasil however, starts to show all kinds of odd distortion onsets with the graph being so chewed up, going up and down as distortion or noise become more dominant.
The Topping DX7s is actually not as good as state of the art devices. The RME is one such and let's look at its performance:
See how perfect the ADI-2 Pro is (the ADI-2 DAC is even better). There are clear engineering and performance problems in Schiit Yggdrasil DAC.
For this last bit, let's review my measurements of Schiit THD+N at 1 kHz at full amplitude and SINAD (signal above power of noise+distortion):
Here is the spec from Schiit's own web site:
So my measurements are absolutely right on the money. And how do they compare to other DACs? Here is a comparison:
Every DAC from other manufacturers outperforms Schiit Yggdrasil regardless of price! Again, remember that my measurements are identical to what Schiit publishes here (SINAD is inverse of THD+N).
Summary
Direct comparison of my measurements and those performed by Bob Smith (atomicbob) show that the two show the same outcomes. Importantly, the performance as measured by him (and replicated by me) is simply not competitive. The difference is not small at all. We are talking 20 to 30 dB deficiency in the output of Schiit Yggdrasil DAC.
As always, comments, corrections or questions are welcome.