-
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
- Thread starter amirm
- Start date
Alcophone
Member
AFAIK, powering audio equipment from balanced power supply (isolation transformer) might help in getting better measurements. I'm not sure Schiit is using regular mains or isolation transformers or not, but if they do...then this could represent a bit of improvement on some measurements.
However, Amir is zooming-in a lot the vertical scale so we can be able to see the measurements better. Some of us may not realize that, but sometimes same measurements with a narrowed or zoomed vertical scale could lead us into the wrong conclusions.
BTW, in regard with DAC's linearity, I personally consider a max. deviation of +/-0.5dB as being acceptable, but I value a lot Amir's detailed graphs and measurements.
Now this is..."challenging" a lot. 
Schiit's linearity test (-0.5dB seems to be around -117dBFS and -1dB seems to be around -122dBFS)
vs.
Amir's linearity test
Like I said before, I'm optimistic and a believer that less that a 0.5dB off vertical axis deviation might not be heard by my >40 years old ears, so please take the above with a grain of salt:
Schiit's test: -0.5dB seems to be around -117dBFS (19.5 bits) and -1dB seems to be around -122dBFS
Amir's test: -0.5dB seems to be around -110dBFS (18.33 bits) and -1dB seems to be around -112dBFS or -113dBFS
So about 7dBFS worse that Schiit's test if looking to a deviation from the horizontal axis of only+/-0.5dB and about 10dBFS worse is looking for a deviation of only +/-1dB. I really think that a firmware upgrade for the DSP can indeed improve the linearity.
Schiit's linearity test (-0.5dB seems to be around -117dBFS and -1dB seems to be around -122dBFS)
vs.
Amir's linearity test
Like I said before, I'm optimistic and a believer that less that a 0.5dB off vertical axis deviation might not be heard by my >40 years old ears, so please take the above with a grain of salt:
Schiit's test: -0.5dB seems to be around -117dBFS (19.5 bits) and -1dB seems to be around -122dBFS
Amir's test: -0.5dB seems to be around -110dBFS (18.33 bits) and -1dB seems to be around -112dBFS or -113dBFS
So about 7dBFS worse that Schiit's test if looking to a deviation from the horizontal axis of only+/-0.5dB and about 10dBFS worse is looking for a deviation of only +/-1dB. I really think that a firmware upgrade for the DSP can indeed improve the linearity.
derp1n
Senior Member
- Joined
- May 28, 2018
- Messages
- 479
- Likes
- 629
Second harmonic is way down (>15dBV) in Schiit's, so there are other differences.
I'm surprised Amir hasn't commented on these measurements yet.
OP
- Thread Starter
- #592
Am I that predictable???
Yes, response will be coming. I am having to juggle a ton of stuff right now.
OP
- Thread Starter
- #593
OK, let me address some low hanging fruit on this with more to come. Let's review what the owner of the unit posted after he sent his unit to Schiit:
All along, I have been saying two things about my measurements versus others:
1. That many of the measurements are actually quite consistent between my data and others. It is just that other reviewers paper over them and pretend that all the data is "excellent" when in really that is not the case at all. See my article on that here: https://www.audiosciencereview.com/...ts-of-schiit-yggdrasil-dac-inconsistent.3812/
2. That where there is distinct and verifiable difference in measurements, the only explanation is that they tested a different revision of the unit than what the 3 customer units were that I tested. This latest turn of events helps with this argument.
As we see above, the owner states that Schiit upgraded the "ROM" in the Schiit Yggdrasil DSP board. But that they had no way of knowing what was there before.
First, no one uses ROMs anymore. ROMs were write-once devices that we used before 1990s before advent of flash memory. Today everyone uses flash that can be rewritten many times allowing for upgrade of units in the field as anyone who has upgraded the BIOS in their PC or countless other electronic devices knows.
As it turns out, I had done a teardown of one of the first two Yggdrasil DACs which I have not yet published. In there, I have a clear shot of the DSP board:
We see the analog devices DSP (signal processor) and next to it, the flash EEPROM. It is made by the company Macronix ("MXIC"). The part number is MX25L4006E (http://www.macronix.com/Lists/Datasheet/Attachments/7393/MX25L4006E, 3V, 4Mb, v1.6.pdf).
Importantly what we see is that it is socketed. Usually these are soldered to the board and a connector is used to "flash" them. Here they have gone old school and put it in a socket. This means they externally program them and just insert them in the units at production.
Here is the key point: every flash programmer has the ability to read the flash device as well as write it. As such, Schiit was absolutely in a position to check which version of firmware was loaded into the flash of the owner's device before reprogramming it. I suspect they did not even reprogram it but simply put in a new pre-programmed flash in there and then performed the measurements.
So the story given to the owner is just that: a story. Either there is a new revision of DSP code that is now standard in every new Schiit Yggdrasil DAC. Or a special one exists for reviewers and owners like the person who sent in this unit.
As kindly posted above, the measured linearity has improved post this change. We have known in our investigation that there was a simple signal processing mistake of not dithering 24-bit data before sending it to 20 bit DAC chip that Yggdrasil uses. It may seem that you can just truncate the extra resolution but that creates "steps" in smoothly varying signals resulting in distortion. It is "DSP 101" type of thing that lay people don't know, but every signal processing person should. How Schiit did not know is very strange.
We were not the first to find this problem. JA at stereophile discovered the same in his measurements: https://www.stereophile.com/content/schiit-audio-yggdrasil-da-processor-measurements
When asked where the linearity measurements and hence resolution computation as I do, this is what JA had to say:
I am confident if JA had posted the linearity graphs, it would have looked just like mine with that distinct zigzag shape indicating increasing "quantization noise" or distortion.
The "fix" for this is very trivial and involves just adding some random noise ("dither") to 24-bit data before sending it to the analog devices DAC inside Yggdrasil. And the place to put it would have been inside the DSP and hence the "ROM" change.
This would have been a golden opportunity for Schiit to show transparency, accept the above mistake and offer a cheap upgrade to anyone who asks. The DIP Flash can be sent to users to put in if they know what they are doing, or provide a $200 upgrade fee including shipping. I suspect many owners would have wanted to get the resolution they paid for and gladly pay for the upgrade. Instead we have a charade where Schiit still thinks it can blame this issue on me and obfuscate the real facts.
As I have expressed to the owner, I am super disappointed in this commentary and actions here. Instead of reaching out to me and get advice when Schiit told him what they did, he goes on an accepts their explanation. As a non-engineer, Schiit took advantage of him and sold him a different version of facts. And with it, none of the other owners can benefit from this upgrade (assuming it is a tested and shipping one).
I am here as a service to the entire community. You have a free consultant to clear the air and give you the data and power you need to deal with manufacturers who operate from a much higher level of knowledge. It is like having a mechanic with you for free anytime you go shopping for a used car. Why turn that offer down and instead trust the salesman???
Personally I have better things to do that deal with this saga. I have tons of other gears to review. I rather spend my time on that than deal with the soap opera here that unfortunately users help to create in addition to company's actions.
Summary
The facts here are very clear. I have the same analyzer as Schiit. I have run the exact same linearity script as them. All one has to do is connect the cables and run the script. That the data I get is different than what they get points to one and only one thing: different devices are being tested. In this case, Schiit upgraded the firmware for the DSP and then took measurements. As a minimum they should have performed tests before and after the upgrade. If they were so inept at not knowing what firmware was in there, that alone would have shown us any performance differences and ended this debate.
The fact the owner continues to doubt my work indicates the passion owners have to believe what they want rather than the truth. That he advises people to "strongly" avoid the data I provide does a huge disservice to other users of Schiit products who could benefit from improvements in the product that measurements force to come about. He could have been part of the solution as when he originally loaned me the unit. But now is helping drive this reality show himself.
All along, I have been saying two things about my measurements versus others:
1. That many of the measurements are actually quite consistent between my data and others. It is just that other reviewers paper over them and pretend that all the data is "excellent" when in really that is not the case at all. See my article on that here: https://www.audiosciencereview.com/...ts-of-schiit-yggdrasil-dac-inconsistent.3812/
2. That where there is distinct and verifiable difference in measurements, the only explanation is that they tested a different revision of the unit than what the 3 customer units were that I tested. This latest turn of events helps with this argument.
As we see above, the owner states that Schiit upgraded the "ROM" in the Schiit Yggdrasil DSP board. But that they had no way of knowing what was there before.
First, no one uses ROMs anymore. ROMs were write-once devices that we used before 1990s before advent of flash memory. Today everyone uses flash that can be rewritten many times allowing for upgrade of units in the field as anyone who has upgraded the BIOS in their PC or countless other electronic devices knows.
As it turns out, I had done a teardown of one of the first two Yggdrasil DACs which I have not yet published. In there, I have a clear shot of the DSP board:
We see the analog devices DSP (signal processor) and next to it, the flash EEPROM. It is made by the company Macronix ("MXIC"). The part number is MX25L4006E (http://www.macronix.com/Lists/Datasheet/Attachments/7393/MX25L4006E, 3V, 4Mb, v1.6.pdf).
Importantly what we see is that it is socketed. Usually these are soldered to the board and a connector is used to "flash" them. Here they have gone old school and put it in a socket. This means they externally program them and just insert them in the units at production.
Here is the key point: every flash programmer has the ability to read the flash device as well as write it. As such, Schiit was absolutely in a position to check which version of firmware was loaded into the flash of the owner's device before reprogramming it. I suspect they did not even reprogram it but simply put in a new pre-programmed flash in there and then performed the measurements.
So the story given to the owner is just that: a story. Either there is a new revision of DSP code that is now standard in every new Schiit Yggdrasil DAC. Or a special one exists for reviewers and owners like the person who sent in this unit.
As kindly posted above, the measured linearity has improved post this change. We have known in our investigation that there was a simple signal processing mistake of not dithering 24-bit data before sending it to 20 bit DAC chip that Yggdrasil uses. It may seem that you can just truncate the extra resolution but that creates "steps" in smoothly varying signals resulting in distortion. It is "DSP 101" type of thing that lay people don't know, but every signal processing person should. How Schiit did not know is very strange.
We were not the first to find this problem. JA at stereophile discovered the same in his measurements: https://www.stereophile.com/content/schiit-audio-yggdrasil-da-processor-measurements
When asked where the linearity measurements and hence resolution computation as I do, this is what JA had to say:
I am confident if JA had posted the linearity graphs, it would have looked just like mine with that distinct zigzag shape indicating increasing "quantization noise" or distortion.
The "fix" for this is very trivial and involves just adding some random noise ("dither") to 24-bit data before sending it to the analog devices DAC inside Yggdrasil. And the place to put it would have been inside the DSP and hence the "ROM" change.
This would have been a golden opportunity for Schiit to show transparency, accept the above mistake and offer a cheap upgrade to anyone who asks. The DIP Flash can be sent to users to put in if they know what they are doing, or provide a $200 upgrade fee including shipping. I suspect many owners would have wanted to get the resolution they paid for and gladly pay for the upgrade. Instead we have a charade where Schiit still thinks it can blame this issue on me and obfuscate the real facts.
As I have expressed to the owner, I am super disappointed in this commentary and actions here. Instead of reaching out to me and get advice when Schiit told him what they did, he goes on an accepts their explanation. As a non-engineer, Schiit took advantage of him and sold him a different version of facts. And with it, none of the other owners can benefit from this upgrade (assuming it is a tested and shipping one).
I am here as a service to the entire community. You have a free consultant to clear the air and give you the data and power you need to deal with manufacturers who operate from a much higher level of knowledge. It is like having a mechanic with you for free anytime you go shopping for a used car. Why turn that offer down and instead trust the salesman???
Personally I have better things to do that deal with this saga. I have tons of other gears to review. I rather spend my time on that than deal with the soap opera here that unfortunately users help to create in addition to company's actions.
Summary
The facts here are very clear. I have the same analyzer as Schiit. I have run the exact same linearity script as them. All one has to do is connect the cables and run the script. That the data I get is different than what they get points to one and only one thing: different devices are being tested. In this case, Schiit upgraded the firmware for the DSP and then took measurements. As a minimum they should have performed tests before and after the upgrade. If they were so inept at not knowing what firmware was in there, that alone would have shown us any performance differences and ended this debate.
The fact the owner continues to doubt my work indicates the passion owners have to believe what they want rather than the truth. That he advises people to "strongly" avoid the data I provide does a huge disservice to other users of Schiit products who could benefit from improvements in the product that measurements force to come about. He could have been part of the solution as when he originally loaned me the unit. But now is helping drive this reality show himself.
svart-hvitt
Major Contributor
- Joined
- Aug 31, 2017
- Messages
- 2,375
- Likes
- 1,253
OK, let me address some low hanging fruit on this with more to come. Let's review what the owner of the unit posted after he sent his unit to Schiit:
View attachment 14533
All along, I have been saying two things about my measurements versus others:
1. That many of the measurements are actually quite consistent between my data and others. It is just that other reviewers paper over them and pretend that all the data is "excellent" when in really that is not the case at all. See my article on that here: https://www.audiosciencereview.com/...ts-of-schiit-yggdrasil-dac-inconsistent.3812/
2. That where there is distinct and verifiable difference in measurements, the only explanation is that they tested a different revision of the unit than what the 3 customer units were that I tested. This latest turn of events helps with this argument.
As we see above, the owner states that Schiit upgraded the "ROM" in the Schiit Yggdrasil DSP board. But that they had no way of knowing what was there before.
First, no one uses ROMs anymore. ROMs were write-once devices that we used before 1990s before advent of flash memory. Today everyone uses flash that can be rewritten many times allowing for upgrade of units in the field as anyone who has upgraded the BIOS in their PC or countless other electronic devices knows.
As it turns out, I had done a teardown of one of the first two Yggdrasil DACs which I have not yet published. In there, I have a clear shot of the DSP board:
View attachment 14534
We see the analog devices DSP (signal processor) and next to it, the flash EEPROM. It is made by the company Macronix ("MXIC"). The part number is MX25L4006E (http://www.macronix.com/Lists/Datasheet/Attachments/7393/MX25L4006E, 3V, 4Mb, v1.6.pdf).
Importantly what we see is that it is socketed. Usually these are soldered to the board and a connector is used to "flash" them. Here they have gone old school and put it in a socket. This means they externally program them and just insert them in the units at production.
Here is the key point: every flash programmer has the ability to read the flash device as well as write it. As such, Schiit was absolutely in a position to check which version of firmware was loaded into the flash of the owner's device before reprogramming it. I suspect they did not even reprogram it but simply put in a new pre-programmed flash in there and then performed the measurements.
So the story given to the owner is just that: a story. Either there is a new revision of DSP code that is now standard in every new Schiit Yggdrasil DAC. Or a special one exists for reviewers and owners like the person who sent in this unit.
As kindly posted above, the measured linearity has improved post this change. We have known in our investigation that there was a simple signal processing mistake of not dithering 24-bit data before sending it to 20 bit DAC chip that Yggdrasil uses. It may seem that you can just truncate the extra resolution but that creates "steps" in smoothly varying signals resulting in distortion. It is "DSP 101" type of thing that lay people don't know, but every signal processing person should. How Schiit did not know is very strange.
We were not the first to find this problem. JA at stereophile discovered the same in his measurements: https://www.stereophile.com/content/schiit-audio-yggdrasil-da-processor-measurements
View attachment 14535
When asked where the linearity measurements and hence resolution computation as I do, this is what JA had to say:
View attachment 14536
I am confident if JA had posted the linearity graphs, it would have looked just like mine with that distinct zigzag shape indicating increasing "quantization noise" or distortion.
The "fix" for this is very trivial and involves just adding some random noise ("dither") to 24-bit data before sending it to the analog devices DAC inside Yggdrasil. And the place to put it would have been inside the DSP and hence the "ROM" change.
This would have been a golden opportunity for Schiit to show transparency, accept the above mistake and offer a cheap upgrade to anyone who asks. The DIP Flash can be sent to users to put in if they know what they are doing, or provide a $200 upgrade fee including shipping. I suspect many owners would have wanted to get the resolution they paid for and gladly pay for the upgrade. Instead we have a charade where Schiit still thinks it can blame this issue on me and obfuscate the real facts.
As I have expressed to the owner, I am super disappointed in this commentary and actions here. Instead of reaching out to me and get advice when Schiit told him what they did, he goes on an accepts their explanation. As a non-engineer, Schiit took advantage of him and sold him a different version of facts. And with it, none of the other owners can benefit from this upgrade (assuming it is a tested and shipping one).
I am here as a service to the entire community. You have a free consultant to clear the air and give you the data and power you need to deal with manufacturers who operate from a much higher level of knowledge. It is like having a mechanic with you for free anytime you go shopping for a used car. Why turn that offer down and instead trust the salesman???
Personally I have better things to do that deal with this saga. I have tons of other gears to review. I rather spend my time on that than deal with the soap opera here that unfortunately users help to create in addition to company's actions.
Summary
The facts here are very clear. I have the same analyzer as Schiit. I have run the exact same linearity script as them. All one has to do is connect the cables and run the script. That the data I get is different than what they get points to one and only one thing: different devices are being tested. In this case, Schiit upgraded the firmware for the DSP and then took measurements. As a minimum they should have performed tests before and after the upgrade. If they were so inept at not knowing what firmware was in there, that alone would have shown us any performance differences and ended this debate.
The fact the owner continues to doubt my work indicates the passion owners have to believe what they want rather than the truth. That he advises people to "strongly" avoid the data I provide does a huge disservice to other users of Schiit products who could benefit from improvements in the product that measurements force to come about. He could have been part of the solution as when he originally loaned me the unit. But now is helping drive this reality show himself.
Who needs HBO?
Nobody could possibly have made this up.
Entertainment on a new level!
OP
- Thread Starter
- #595
My feelings exactly. Problem is, unlike the "talent" in HBO shows, I am not getting paid for this!
Disappointing, I’m not surprised though.
It continues to amaze me, you seek to empower people but they seem intent on staying ignorant and end up biting the hand that feeds.
Sometimes I think people are just not worth the bother.
It continues to amaze me, you seek to empower people but they seem intent on staying ignorant and end up biting the hand that feeds.
Sometimes I think people are just not worth the bother.
Just dump that DIP8 EEPROM and let's burn some blanks and get another DAC to test before vs. after.
You could find really cheap compatible blank chis from eBay/Alibaba and the EEPROM writer I was able to DIY-build it within 10-15 minutes, as I didn't want to spend 10 bucks on eBay.
.
OP
- Thread Starter
- #598
Here is the second installment in my response to Schiit's measurements.
When the owner originally contacted me, he was adamant about testing USB input only. He didn't care about any other input which was fine. Let's look at the report from Schiit and see what input they tested:
Yup, all of their tests used Toslink Optical input. Had the owner confided in me I would have told them they did not test USB.
There was another reason that it was important to see USB tests. If you go back to my original review, you see that the USB clock ran too fast:
Since clocks vary, my hope was that this was a signature we could look at to determine if Schiit swapped out that board or not. Alas, since they only tested Toslink, we don't know if this was done or not (my test of AES input showed correct frequency).
That aside, you see why I am so disappointed to read comments from owner to this effect:
Optical input provides full galvanic isolation on top of being different from USB in general. So even if Schiit had run the same tests as I (which they did not), there would be variations in the outcome.
All of this could be avoided if I was contacted during the process, rather than waiting after the fact with such derogatory commentary regarding my work.
When the owner originally contacted me, he was adamant about testing USB input only. He didn't care about any other input which was fine. Let's look at the report from Schiit and see what input they tested:
Yup, all of their tests used Toslink Optical input. Had the owner confided in me I would have told them they did not test USB.
There was another reason that it was important to see USB tests. If you go back to my original review, you see that the USB clock ran too fast:
Since clocks vary, my hope was that this was a signature we could look at to determine if Schiit swapped out that board or not. Alas, since they only tested Toslink, we don't know if this was done or not (my test of AES input showed correct frequency).
That aside, you see why I am so disappointed to read comments from owner to this effect:
Optical input provides full galvanic isolation on top of being different from USB in general. So even if Schiit had run the same tests as I (which they did not), there would be variations in the outcome.
All of this could be avoided if I was contacted during the process, rather than waiting after the fact with such derogatory commentary regarding my work.
Alcophone
Member
Amir, please. Did you honestly expect Schiit to publish measurement results confirming yours? Come on. Did anyone, on either camp, expect a different outcome? Really?
OP
- Thread Starter
- #600
What I honestly expected was that you would ask them to run the same tests I ran, in exact same conditions. That you would run them by me for examination before you went with your public commentary. All of this would have helped us get closer to the truth -- whatever that may have been.
As it is, you are helping cast doubt on validity of my measurements "because Schiit's data doesn't match Amir's." We already had such claims before. The idea of sending your unit to them was to analyze what was going on. Not to help increase the smoke that is covering the facts here.
Similar threads
- Poll
