Why usb via schiit sounds thin

[amit11]

Hi friends,

Recently i used async usb with my schiit dac. The source was ipad and also android phone ( with the app usb audio player pro, so that native usb driver of android which is 48khz is bypassed for bit perfect output of 44.1 ).

In general i noticed that usb playback sounded a bit thin compared to optical.

In some songs usb was sounding better and in some songs i preferred the optical.
Also i felt some difference between usb from ipad and usb from android.




Sent from Note5

 

[amit11]

It has been almost one month where I started experimenting with USB input.

I tried various android apps --> USB audio player pro (paid app) which outputs bit perfect.. Then Neutron app... then the Onkyo HF Player.

I also tried few apps on Apple Ios..

I also purchased the Apple Camera Connection Kit to get asynchronous USB out from my ipad.

I also tried Jriver, Foobar, ITunes for USB output to the Schiit DAC.


Initially I had liked it but gradually I started feeling something is not correct. But I could not find out. I also experimented if any USB packets were getting dropped out.

In the last many days...I started feeling ear fatigue. That was not easily clear if it was ear fatigue.... but just a mild feeling that sound is not as it should be. Sometimes I thought my source is not that great hence the bitperfect USB output is showing out the shortcomings...Some songs sounded good and some not...so whats causing the issue was not clear.

Then I experimented with the different cables I had...all various permutation and combinations but still the ear fatigue was not going. I also sometimes thought if my ears are not proper / infected and I cleaned them...But again after listening the harshness/ear fatigue was still there... I started reading more and more on the USB asynchronous stuff.

Then somewhere I had read but had not taken it seriously that Schiit DAC upsamples and applies filter.... All the time I was considering it as a NOS DAC.

Then suddenly yesterday I thought to write email to Schiit and clarify. ....I got the reply. I am copy-pasting the two questions. Answers are immediately below to the questions, as it is replied by Schiit.

a) Suppose I connect USB to it and play music with bitrate 44.1 kHz. Then does the DAC or the filter inside it oversample it to 176.4?

The DAC upsampled to 176.4 then processes it including filters.


b) Suppose via same USB, I play music having bitrate 176.4, then is it true that the filter does not get activated and the DAC plays as it is?

The filter is still used it just doesn't upsample.


........... So what I conclude that Schiit UPSAMPLES (atleast in USB mode) and applies the filter. It it upsamples then its not bitperfect (44.1), and hence the sound signature is getting changed and hence the cause of ear fatigue....


....... I have again emailed them to check if optical also gets upsampled or not. I don't think it gets upsamples via optical as I am already using it and satisfied with that sound.

 

[sound_cycle]

Then somewhere I had read but had not taken it seriously that Schiit DAC upsamples and applies filter.... All the time I was considering it as a NOS DAC.

Look no further than the product page on the schiit website

Sample Rates and Bit Depths:

Modi 2: 16/44.1 to 24/96, including 24/88.2, without Windows drivers, or up to 24/192 with Windows drivers (or using a Mac or Linux machine)
Modi 2 Uber: 16/44.1 to 24/192 via USB. 16/44.1 to 24/192 via Toslink* and Coax.
Modi 2 Multibit: 16/44.1 to 24/192 via USB. 16/44.1 to 24/192 via Toslink* and Coax, with 24/176 and 24/192 being NOS (non-oversampled)


What is NOS-ed

This is from the bifrost FAQ

Whats this about a non-oversampling (NOS) mode?
For 176.4kHz and 192kHz input sample rates, Bifrost Multibit passes the music right alongno digital filter, no oversampling. Have fun! Upsample lower-rate music with another algorithm, and compare to our comboburrito filter.

ciao
gr

 

[amit11]

Modi 2 Multibit: 16/44.1 to 24/192 via USB. 16/44.1 to 24/192 via Toslink* and Coax, with 24/176 and 24/192 being NOS (non-oversampled)

This would normally mean that the DAC supports bitrates from X to Y. It creates confusion whether it upsamples or not. The fact is it upsamples (atleast via USB), and unfortunately there seems to be no way to switch it off.

"For 176.4kHz and 192kHz input sample rates, Bifrost Multibit passes the music right alongno digital filter, no oversampling. Have fun! Upsample lower-rate music with another algorithm, and compare to our comboburrito filter."

This seems to be contradictory to the answer to question (b) in my previous post. Schiit says that if we pass 176.4 to the DAC, the filter is still used (I am assuming there is only one filter under discussion), its just that upsampling does not happen since the input itself is already 176.4


I may have got my understanding wrong, but nevertheless it is disappointing atleast for me, the upsampling part.

 

[prem]

Schitt never made NOS dacs. I have had their Yggdrasil. It uses a burrito filter which gives you the best of NOS and upsampling dacs. It's an R2R dac. Many assume R2R means NOS. If you want NOS, buy Kitsune or Audiogd or Metrum

 

[amit11]

Schitt never made NOS dacs. I have had their Yggdrasil. It uses a burrito filter which gives you the best of NOS and upsampling dacs. It's an R2R dac. Many assume R2R means NOS. If you want NOS, buy Kitsune or Audiogd or Metrum

Hmmm..
Is the FIO D03k NOS dac?, or it also upsamples and then gives the output? I had it as my first DAC and still have. After purchasing Schiit I never connected FIO again until this USB thing happened.

What I observed in my Schiit Multibit DAC (Modi), that USB is not at all pleasing. Anything which causes ear fatigue is a total turn-off even if the other characteristics of audiophile sound are present. Comparatively the FIO dac is bearable even if it is inferior.

When I play via the optical (and that is via airport express the oldest generation), it sounds musical even though it may not be the perfect. I am of the thinking that when optical is used the upsampling does not happen, either by design or probably optical is not meant for upsampling on the fly. But who can confirm about this optical upsampling happening or not? I have written to Schiit again and awaiting their reply..

And yes the DACs you mentioned I would like to have but too expensive for me.. especially the Holo DAC.. However I am wondering few people in this forum who bought this DAC have not written their views in details... they have written in short only. Either they are very happy with that dac and not inclined to write much about it or they might be feeling its falling short of expectations and not writing..

PS: I can recall you had sold your Yggdrasil dac.... Even with all this upsampling and filter it might sound better but still not musical/analogish/originalish.

 

[MaSh]

Nice insights and information. I recently got a schiit modi 2, onlu USB version. What's the best setting to run it?

Sent from my Redmi Note 3 using Tapatalk

 

[sound_cycle]

I may have got my understanding wrong, but nevertheless it is disappointing atleast for me, the upsampling part.

Why

The secret sauce in the Schiit m/b DACs is the "Mega Combo Burrito" filter to oversample

Can you turn off the damn filter

No says Mike Moffat (writing as baldr here

There is no way to turn of the mega combo burrito filter in any Schiit Multibit DAC. If you have a Bimby, it is not in place when 176.4K or 192K material is played through, a Gumby or Yggy double that.

If you really prefer NOS 44.1/48K stuff, you are far better off with other DACs. If that is the case, I caution you to make sure said DAC has a true analog Brick Wall filter on board.​

What happens is that

Gungnir/ Yggy oversampled to 384 or 352.8, Bifrost and Modi everything is oversampled to 192 or 176.4


ciao
gr

 

[amit11]

Nice insights and information. I recently got a schiit modi 2, onlu USB version. What's the best setting to run it?

If you are feeding it via a laptop and software (e.g. jriver/foobar), then on the screen somewhere in the settings (audio-playback), we have the option to choose the "driver". You will get options like "Schiit USB driver (WASAPI)" or "ASIO USB Driver", etc. Those sound better compared to any default driver of windows. (I learned it from someone in the forum itself)

Also there might be other settings (either default or we need to change), which will keep the bitrate same as the source. So those settings should also be active. I think these are the 2 main settings which will output bit-perfect.

 

[amit11]

Why

There is no way to turn of the mega combo burrito filter in any Schiit Multibit DAC. If you have a Bimby, it is not in place when 176.4K or 192K material is played through, a Gumby or Yggy double that.
gr​

My understanding based on Schiit's reply is, even if we feed 176.4 K to the DAC, the filter will still get activated. (I am taking example of the modi multibit)

There are two things which happen in sequence.
a) Upsampling
b) Digital Filter which they call as burrito filter


a) Now upsampling is nothing but repeating of the values. e.g.
if the sample is 1, 4,6,3
then upsampling four times (i.e. from 44.1 to 176.4) will result in
1,1,1,1, 4,4,4,4, 6,6,6,6 3,3,3,3

b) After this the burrito filter CHANGES this sequence of samples. Yes, out of those 4 samples for each original sample, it will KEEP 1 of them and CHANGE the remaining 3 based on mathematical computation. That's why they say they do not discard the original sample.

so after the filter is applied, the samples become something like this.

1,x,y,z, 4,a,b,c, 6,l,k,m, 3,p,q,r
(where xyz, lkm, pqr are computed based on their logic / interpolation etc etc)

They do take care of phase and other things and that's why their filter may sound cleaner compared to other DACs.

However it is still not the original output. And my point is the filter is ALWAYS active irrespective of the input rate.​

 

[sound_cycle]

the filter will still get activated
<snip>
) Now upsampling is nothing but repeating of the values. e.g.
<snip>
After this the burrito filter CHANGES this sequence of samples. Yes, out of those 4 samples for each original sample, it will KEEP 1 of them and CHANGE <snip>
However it is still not the original output. And my point is the filter is ALWAYS active irrespective of the input rate.

I've read several monster threads on the topic. I am not enough of a math/physics guy to be explain it correctly (which is not to imply I have understood it all/ correctly). And therefore cannot explain why you are wrong on all counts

The filter is always active. The upsampling is much more complicated that. It is as close to the input as can be. Delta Sigma never was (and uses far fewer bits). The Schiit magic is how they upsample.

Why should DAC's upsample here is a linky (with a link to a more detailed article inside)

ciao
gr

 

[prem]

Amit11, Yggdrasil is a very nice dac. It's just that most of the music I listen to is mastered better on vinyl.

Between a good NOS dac like Metrum and Kitsune as compared to Yggdrasil will be that the tonal density of Yggdrasil will be superior whereas the other two will sound more spacious and liquid. So you need to go with what floats your boat.

 

[MaSh]

If you are feeding it via a laptop and software (e.g. jriver/foobar), then on the screen somewhere in the settings (audio-playback), we have the option to choose the "driver". You will get options like "Schiit USB driver (WASAPI)" or "ASIO USB Driver", etc. Those sound better compared to any default driver of windows. (I learned it from someone in the forum itself)

Also there might be other settings (either default or we need to change), which will keep the bitrate same as the source. So those settings should also be active. I think these are the 2 main settings which will output bit-perfect.

I use schiit through foobar only and the output is Wasapi. I do have the schiit driver installed too in the system. The only other thing I changed was to set the bit depth to 24 bit 192 in playback device settings.

MaSh

Sent from my Redmi Note 3 using Tapatalk

 

[dheerajin]

Schitt never made NOS dacs. I have had their Yggdrasil. It uses a burrito filter which gives you the best of NOS and upsampling dacs. It's an R2R dac. Many assume R2R means NOS. If you want NOS, buy Kitsune or Audiogd or Metrum

Yggdrasil is multibit DAC not R2R ladder DAC..just to clear the things

 

[dheerajin]

And yes the DACs you mentioned I would like to have but too expensive for me.. especially the Holo DAC.. However I am wondering few people in this forum who bought this DAC have not written their views in details... they have written in short only. Either they are very happy with that dac and not inclined to write much about it or they might be feeling its falling short of expectations and not writing..

Come over...not only Holo spring DAC..you will get to listen fantastic source with linear power supply...

 

[prem]

Dheerajin

The Analog Device chip used in Yggdrasil is a R2R chip. You can read the review below

Schiit Audio Yggdrasil DAC | The Absolute Sound

I guess this is akin to the Burr Brown 1704 chip

 

[dheerajin]

Dheerajin

The Analog Device chip used in Yggdrasil is a R2R chip. You can read the review below

Schiit Audio Yggdrasil DAC | The Absolute Sound

I guess this is akin to the Burr Brown 1704 chip

R2R means resistor to resistor ladder network, if chip is being used, it is considered as multibit. even in this review it is not mentioned as R2R. Yes Burr Brown 1704 chip is multi bit chip.

 

[jls001]

a) Now upsampling is nothing but repeating of the values. e.g.
if the sample is 1, 4,6,3
then upsampling four times (i.e. from 44.1 to 176.4) will result in
1,1,1,1, 4,4,4,4, 6,6,6,6 3,3,3,3

b) After this the burrito filter CHANGES this sequence of samples. Yes, out of those 4 samples for each original sample, it will KEEP 1 of them and CHANGE the remaining 3 based on mathematical computation. That's why they say they do not discard the original sample.

so after the filter is applied, the samples become something like this.

1,x,y,z, 4,a,b,c, 6,l,k,m, 3,p,q,r
(where xyz, lkm, pqr are computed based on their logic / interpolation etc etc)

They do take care of phase and other things and that's why their filter may sound cleaner compared to other DACs.

However it is still not the original output. And my point is the filter is ALWAYS active irrespective of the input rate.

Is the above really how upsampling happens?:indifferent14:

But more importantly ask yourself why upsampling is done.

 

[prem]

Moffatt selected an unusual DACa 20-bit R/2R ladder device from Analog Devices (the AD5791) that is used in applications where bit-perfect conversion is essential, such as in instrumentation, weapons systems, and magnetic-resonance imaging (MRI) machines. No one else uses these DACs for audio.


Dheerajin, above is quote from the Absolute Sound review

 

[amit11]

Is the above really how upsampling happens?:indifferent14:

But more importantly ask yourself why upsampling is done.

Well... in laymen terms that is what I understood about upsampling. The actual process and the steps and the calculations can be done in N number of ways. But the important thing happening is for each original sample data point, 3 extra are getting generated. (for 44.1 to 176.4)

Now those 3 extra which are generated for each and every sample, that is done to enable create the original wave form. How good/better/best it is done - there can be various ways since it is finally the mathematics being done. Either joining the DOTS via a straight line, or joining the dots with a curve by looking at the immediate next DOT or looking even further to get a more accurate representation.

My understanding is it may initially sound good....but not in the long run. For me its ok if 44.1 is less than vinyl, but those extra 3 dots can never 100% represent the original howmuch ever we try... This brings in artificialness which may be less or more.