• 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!

Ton of questions (Linearity, Resolution, Bit Depths, Windows Sampler, DACs, DA Filters)

Tks

Major Contributor
Joined
Apr 1, 2019
Messages
3,221
Likes
5,493
Just a few quick questions for a few things I can't get my head around. I know it's a lot, but tackle whatever you see fit if you can some time. It's just been bothering me not knowing, and not knowing how to seek proper answers to these things. I'm sorry if I am asking questions that I should better be served learning prerequisites for (like someone asking about the Big Bang's actual state, without never having taken a physics course for example).

First off, how does one precisely read linearity graphs? I recall in the old measurements there were the linearity graphs as they are now, but would also accompany by telling us how man "bits of resolution". Is there a way to extrapolate this data somehow by looking at the modern linearity graphs, and also.. bits of resolution? Brings me to my next question..

Does that have anything to do with bit depth of an audio file? Or is that the hard-limit of what a device is able to reproduce? Like if some older devices I seen measured a while ago for instance had 20-bits of resolution, does that mean the noise floor is effectively 120db (or something?). And if so, does that mean any device that doesn't have the capability of reproducing something like 16-bit.. you should never bother seeking out higher bit-depth content (not that I do, nor does anyone, but I am asking theoretically does it make sense for a device maker to bother with something like 24 or 32 bit support in-hardware.. aside from just marketing?). Also, does linearity have an overall effect on the final SINAD calculation directly or indirecty?

Another thing I am wondering, that I recently ran into was when playing around with DA filters on my DAC, and using the Non-oversampling filter due to seeing someone say in passing "it basically would produce the best impulse response". The problem being is supposedly hardware has to work harder(think MC from RME mentioned this, I'm not sure though) and the high frequencies begin to fall off (not too bad considering I can't hear 19K or higher). But what I did notice when a handful of songs when I had the volume cranked up for "fun listening".. is there was like some sort of wobbling ringing effect. No idea what that's all about.. Was wondering if anyone has anything about these sorts of topics for a laymen to grasp?

Lastly.. I always hear of DAC's doing their own thing with respect to sampling (like they do on-the-fly conversions and reconversions to reconstruct the signal or something because supposedly having a DAC process an "upsampled" 192kHz file is far easier for the hardware, than having it "natively" process something like 44.1 (which is why I assume no modern DACs support some 32kHz sample rate I see in settings page that asks me to check-off what rates my DAC supports). Also "upsampling" or "downsampling" ..How do I know it's even occurring? This only multiplies my confusion when we take Windows sampler into the equation. By this ordeal, would it make sense just to go into the driver page and set my DAC to 96kHz or 88.2kHz so whenever I'm not using ASIO or WASAPI (like when using Firefox, or playing a game since you don't have the option of NOT setting a sample rate) the DAC can just upsample to those "easier to process" sample rates for the heck of it and possibly less heat generation or whatever?

What's up with PC motherboard SPDIF Optical-output.. and what's up with "disable all effects" settings in the driver menu? (Ill elaborate now with this question)

When I had Equalizer APO running, and I had the Output of Foobar set to the default Windows Mixer optical out (with no ASIO or WASAPI or anything like that) I would instantly hear some insane distortion (sounds like IMD) compared to when I used ASIO or WASAPI push. I had to uninstall Equalizer APO and enable "Disable all Effects" in the windows driver menu.

So getting back to the whole sampling and SPDIF Optical thing.. I seen Roon has a sort of "chain tree" where you can see exactly what's being processed and how. Barring that, how would I know what is happening with the audio I am playing. What on Earth Windows Mixer is doing (this I hear resamples music to what the default sample rate on the DAC is set to through the driver panel in Windows). Is anything happening to my audio stream being fed from the Optical-out from my motherboard vs USB (I seen old posts online talking about how the audio is processed by the built-in "sound card" (or audio codec supplied with the motherboard for example). If so, what is Windows doing that wouldn't be done if I went the USB route vs Optical since it seems Windows Mixer will resample anything regardless that isn't being controlled by a company driver? Can this "mixer" be replaced with something like "PulseAudio" driver native to Linux (and whatever runs Mac) theoretically? Seeing as how this fellow demonstrates some differences, and interesting conclusions. Especially interesting how he goes into NOS DACs and describes what I propose is the same annoying ringing I heard as well with the NOS DA Filter.

Bit-perfect.. What is the point of this outside of studio use for whatever reason? What does bit imperfection even look like or sound like?

Windows/System volume VS AMP or DAC volume. (I hear this isn't even a question anymore in Windows 10 or something for some reason. but lets say for the sake of the argument we're in the Vista era), but yeah, what's worse? Lowering the volume on my AMP/DAC or Windows System (seeing as I can't run both at full blast). I would guess leaving the AMP or DAC at the volume where there is the least THD is the best route, and then lowering software volume as you see fit?

Seriously, what's the point of DSD? (PCM I can somewhat grasp my head around with respect to what finally goes on a CD and such). I've read two things about this ordeal, here is the first and the second one I can't find, but argues the opposite basically in some fashion I can't recollect (I know this question is almost ridiculous but don't need a true answer outside of just people's opinion based on understanding how each is handled).

Lastly, can you forgive the insanely long post :-} ?
 
OP
Tks

Tks

Major Contributor
Joined
Apr 1, 2019
Messages
3,221
Likes
5,493
Oh and, what's up with streaming services like Spotify? No idea of the sorts of file types, or size/quality they use. Also is it possible without some API call with virtualization to set an output device for streaming services on desktop?
 

nemesisrobot

Member
Forum Donor
Joined
Jun 20, 2019
Messages
38
Likes
38
Location
Seattle, WA
As far as I understand it, the linearity test plays a fixed frequency across the range of decibel levels starting from 0 dB and going lower, and the idea is that the measured output from the DAC should be match. Once it deviates from the target, that's where it "fails". The rule of thumb is that 1 bit provides about 6 dB of dynamic range. So, if it's flat until -96 dB, then that's 16 bits.
 

amirm

Founder/Admin
Staff Member
CFO (Chief Fun Officer)
Joined
Feb 13, 2016
Messages
44,303
Likes
233,688
Location
Seattle Area
First off, how does one precisely read linearity graphs? I recall in the old measurements there were the linearity graphs as they are now, but would also accompany by telling us how man "bits of resolution". Is there a way to extrapolate this data somehow by looking at the modern linearity graphs, and also.. bits of resolution? Brings me to my next question..
Any dB value can be converted to bits by dividing by 6. So 120 dB of perfect linearity translates into 20 bits.

The issue is that linearity gradually gets worse. I used to use a very strict value to cut that off and declare the number of bits. People complained and I did not want to deal with it anymore so I stopped. But you can do that yourself.

My new relaxed standard is that I like to see less than 0.5 dB of error or less by -120 dB. That way you essentially have 20 bits of linearity.
 

amirm

Founder/Admin
Staff Member
CFO (Chief Fun Officer)
Joined
Feb 13, 2016
Messages
44,303
Likes
233,688
Location
Seattle Area
Does that have anything to do with bit depth of an audio file? Or is that the hard-limit of what a device is able to reproduce? Like if some older devices I seen measured a while ago for instance had 20-bits of resolution, does that mean the noise floor is effectively 120db (or something?).
The limits of audio reproduction today is 20 bits/120 dB or so. It can be shown that 20 bits/120 dB essentially captures the full dynamic range of a live concert. Beyond that, both technologically and practically, there is no point. So 20 bits is perfection.

Audio standards though jump in 8 bit increments so we go from 16 bits to 24 even though we only need 20 bits. The rest is wasted and holds noise.

Note that real linearity of devices is far worse than what I show. My linearity tests filter everything out but the test tone. If I don't do that filtering (i.e. what you see in your system), they would perform a lot worse.
 

amirm

Founder/Admin
Staff Member
CFO (Chief Fun Officer)
Joined
Feb 13, 2016
Messages
44,303
Likes
233,688
Location
Seattle Area
What's up with PC motherboard SPDIF Optical-output.. and what's up with "disable all effects" settings in the driver menu? (Ill elaborate now with this question)
Years ago, when my team was revamping the Windows audio stack for Vista, we put in plug-in features to add effects, everything from room eq to 5.1 to stereo virtualization. Alas, once I left Microsoft, my team was disbanded, the new people didn't know or care about those features and they are not really functional anymore. When they were, checking disable all effects disabled the processing I just explained.

I just noticed in the latest update of Windows the sound control panel has disappeared! And with it, the above option. It is a pain now to figure out what the heck is going on.
 

amirm

Founder/Admin
Staff Member
CFO (Chief Fun Officer)
Joined
Feb 13, 2016
Messages
44,303
Likes
233,688
Location
Seattle Area
So getting back to the whole sampling and SPDIF Optical thing.. I seen Roon has a sort of "chain tree" where you can see exactly what's being processed and how. Barring that, how would I know what is happening with the audio I am playing. What on Earth Windows Mixer is doing (this I hear resamples music to what the default sample rate on the DAC is set to through the driver panel in Windows). Is anything happening to my audio stream being fed from the Optical-out from my motherboard vs USB (I seen old posts online talking about how the audio is processed by the built-in "sound card" (or audio codec supplied with the motherboard for example). If so, what is Windows doing that wouldn't be done if I went the USB route vs Optical since it seems Windows Mixer will resample anything regardless that isn't being controlled by a company driver? Can this "mixer" be replaced with something like "PulseAudio" driver native to Linux (and whatever runs Mac) theoretically? Seeing as how this fellow demonstrates some differences, and interesting conclusions. Especially interesting how he goes into NOS DACs and describes what I propose is the same annoying ringing I heard as well with the NOS DA Filter.
There is a ton here to unpack and I am going to sleep in a minute so this will be short.

Internally DAC chips "upsample" content to higher rates. Don't worry about this. It is up to DAC designers to figure this out.

Some DAC units (as opposed to chips), have upsampling too. Berkeley Labs, Benchmark DAC1, etc. are examples of this.

You can also use software to upsample by using Windows to do it (not very high quality) or use the one in your player (e.g. Roon). My advice: don't bother. Play content at its native rate by using ASIO or WASAPI.
 

amirm

Founder/Admin
Staff Member
CFO (Chief Fun Officer)
Joined
Feb 13, 2016
Messages
44,303
Likes
233,688
Location
Seattle Area
Windows/System volume VS AMP or DAC volume. (I hear this isn't even a question anymore in Windows 10 or something for some reason. but lets say for the sake of the argument we're in the Vista era), but yeah, what's worse? Lowering the volume on my AMP/DAC or Windows System (seeing as I can't run both at full blast). I would guess leaving the AMP or DAC at the volume where there is the least THD is the best route, and then lowering software volume as you see fit?
There are lots of volume controls. I plan to test them all and measure at some point.
 

amirm

Founder/Admin
Staff Member
CFO (Chief Fun Officer)
Joined
Feb 13, 2016
Messages
44,303
Likes
233,688
Location
Seattle Area
Seriously, what's the point of DSD? (PCM I can somewhat grasp my head around with respect to what finally goes on a CD and such). I've read two things about this ordeal, here is the first and the second one I can't find, but argues the opposite basically in some fashion I can't recollect (I know this question is almost ridiculous but don't need a true answer outside of just people's opinion based on understanding how each is handled).
Some people like it and record music in it. If so, I prefer to play the DSD version as opposed to whatever they did to convert it to PCM.
 

amirm

Founder/Admin
Staff Member
CFO (Chief Fun Officer)
Joined
Feb 13, 2016
Messages
44,303
Likes
233,688
Location
Seattle Area
Bit-perfect.. What is the point of this outside of studio use for whatever reason? What does bit imperfection even look like or sound like?
All modern operating systems mess with your audio stream in order to mix the sound from all applications into one audio device. For best fidelity, you want to bypass this pipeline. This is called bit-exact and is what you get with WASAPI Exclusive mode or ASIO. It is a good thing.
 

amirm

Founder/Admin
Staff Member
CFO (Chief Fun Officer)
Joined
Feb 13, 2016
Messages
44,303
Likes
233,688
Location
Seattle Area
What's up with PC motherboard SPDIF Optical-output..
It is a good thing to have. You get isolation from PC ground as opposed to using USB or coax S/PDIF.
 

solderdude

Grand Contributor
Joined
Jul 21, 2018
Messages
15,843
Likes
35,809
Location
The Neitherlands
I'll pick one...

Another thing I am wondering, that I recently ran into was when playing around with DA filters on my DAC, and using the Non-oversampling filter due to seeing someone say in passing "it basically would produce the best impulse response". The problem being is supposedly hardware has to work harder(think MC from RME mentioned this, I'm not sure though) and the high frequencies begin to fall off (not too bad considering I can't hear 19K or higher). But what I did notice when a handful of songs when I had the volume cranked up for "fun listening".. is there was like some sort of wobbling ringing effect. No idea what that's all about.. Was wondering if anyone has anything about these sorts of topics for a laymen to grasp?

Using non filtered NOS with 44.1 or 48kHz is a bad idea. For one it does not comply to the sampling theorem.
The roll-off is not like linear roll-off (like what one gets with RC filters or any other low pass filters) because the 'roll-off' is sample point (time) dependent and as music is never in 'sync' with the clock the reproduction of higher frequencies varies.

Below a scope shot of the actual output signal of a metrum DAC. As can be seen the signal is supposed to be a nice constant 10kHz sinewave but instead you get a weird signal with slopes that should not be there (and acc. to Sergei must be audible). Also the amplitude varies.
This will also be the case with all high frequencies in music. The amplitude is 'rougher'.

vCFTLJV.jpg


And no .. impulses aren't reproduced better. What one hears is poor reproduction with a sharpish/rawish 'edge' combined with some 'roll-off'.
NOS aficionados just like/prefer this presentation. Better it is not.
When upsampling is used (say 4 or 8 times) this effect is as good as gone, steps become smaller and higher in frequency.
The obligated post filtering that would be essential for NOS 44.1 files is taken care off by the filter in the upsampling algo.
 

solderdude

Grand Contributor
Joined
Jul 21, 2018
Messages
15,843
Likes
35,809
Location
The Neitherlands
Lastly.. I always hear of DAC's doing their own thing with respect to sampling (like they do on-the-fly conversions and reconversions to reconstruct the signal or something because supposedly having a DAC process an "upsampled" 192kHz file is far easier for the hardware, than having it "natively" process something like 44.1 (which is why I assume no modern DACs support some 32kHz sample rate I see in settings page that asks me to check-off what rates my DAC supports). Also "upsampling" or "downsampling" ..How do I know it's even occurring? This only multiplies my confusion when we take Windows sampler into the equation. By this ordeal, would it make sense just to go into the driver page and set my DAC to 96kHz or 88.2kHz so whenever I'm not using ASIO or WASAPI (like when using Firefox, or playing a game since you don't have the option of NOT setting a sample rate) the DAC can just upsample to those "easier to process" sample rates for the heck of it and possibly less heat generation or whatever?

All DS DACs upsample, its how they work. I don't think it is a problem to do it in a DAC.
In software with a fast processor one can probably achieve slightly better technical results. I'll lay of the audibility of this.
Miska's player can upsample (and does a good job at it but is a horrible player to use). There is a trial period. Play with it.
For R2R DAC's things differ. IMO it is good practice to upsample before sending it to the DAC. Certainly when the DAC's have no or poor post filtering.


For the bit perfect and DSD info there are lots of threads here... the search function is your friend.
 

reza

Active Member
Joined
Mar 27, 2019
Messages
110
Likes
131
When I had Equalizer APO running, and I had the Output of Foobar set to the default Windows Mixer optical out (with no ASIO or WASAPI or anything like that) I would instantly hear some insane distortion (sounds like IMD) compared to when I used ASIO or WASAPI push. I had to uninstall Equalizer APO and enable "Disable all Effects" in the windows driver menu.

This may sound obvious, but I am guilty of neglecting it when I first installed EQ APO and getting distorted sound:
In the lower left panel of APO, make sure your peak gain is not above zero (see pic below). If it is, reduce "gain" in "preamplification".

1561882614092.png
 

edechamps

Addicted to Fun and Learning
Forum Donor
Joined
Nov 21, 2018
Messages
910
Likes
3,619
Location
London, United Kingdom
what's up with "disable all effects" settings in the driver menu? (Ill elaborate now with this question)

This option disables all Audio Processing Objects (APOs). APOs can come from Microsoft, from the audio adapter manufacturer, or (much more rarely, and not really supported by Microsoft) from third parties, such as with Equalizer APO. They are sometimes called "Effects" or "Enhancements" in the Windows audio control panel.

When I had Equalizer APO running, and I had the Output of Foobar set to the default Windows Mixer optical out (with no ASIO or WASAPI or anything like that) I would instantly hear some insane distortion (sounds like IMD) compared to when I used ASIO or WASAPI push.

Most likely your Equalizer APO filter is applying too much gain and running into clipping. Try reducing the gain in Equalizer APO. Otherwise it could be an Equalizer APO bug.

What on Earth Windows Mixer is doing (this I hear resamples music to what the default sample rate on the DAC is set to through the driver panel in Windows).

Here's what the Windows Audio Engine does:

- It handles sample type conversions (typically from/to integer formats and 32-bit IEEE Float, the format it uses internally), with dithering.
- It applies APOs (described above).
- It converts the sample rate from the application to the one configured in the control panel. Note that AFAIK this is not a high-quality conversion, so you might want to avoid it (either by matching sample rates, or by bypassing the engine entirely).
- It mixes the audio from multiple applications that might be playing at the same time.
- After all that's done, it hands the audio over to the hardware audio adapter driver.

If all you have is a single application playing, with a sample rate that matches the one in the control panel, and all APOs disabled, then all the Windows Audio Engine might do is convert sample types to/from IEEE Float, which is benign and will not result in any audible difference.

Is anything happening to my audio stream being fed from the Optical-out from my motherboard vs USB (I seen old posts online talking about how the audio is processed by the built-in "sound card" (or audio codec supplied with the motherboard for example). If so, what is Windows doing that wouldn't be done if I went the USB route vs Optical since it seems Windows Mixer will resample anything regardless that isn't being controlled by a company driver?

I'm not sure I understand the question. As long as you're using PCM (i.e. you're not trying to bitstream Dolby Digital/AC3 or something like that), the Windows Audio Engine treats a optical S/PDIF output mostly the same as an USB Audio output. It applies the same sample rate conversion logic in both cases. The hardware drivers, however, are different, so there could be differences there.

Can this "mixer" be replaced with something like "PulseAudio" driver native to Linux (and whatever runs Mac) theoretically?

Well, technically PortAudio is documented to work on Windows, but considering the state of its Windows support, I doubt it works on modern systems.

That's not really what you're asking however - you're asking if it's possible to replace the format adaptation and resampling logic in the Windows Audio Engine. Not to my knowledge, no. Well, okay, I guess it could be possible for third-party software to use hooks or override the WASAPI COM classes to inject itself into the audio APIs, but that would be difficult, ugly, and hard to keep stable and working over time.

Alternatively, to use your own mixer, all your applications could agree to bypass the Windows Audio Engine and then work together to mix their own streams. One potential way of doing that is to use JACK as the mixer and then have all your applications use the JACK ASIO driver to talk to JACK. That assumes all your applications have ASIO support. Also, it wouldn't exactly be the most user-friendly system to set up and use.

Bit-perfect.. What is the point of this outside of studio use for whatever reason? What does bit imperfection even look like or sound like?

A system is called "bit-perfect" if it guarantees to pass the digital audio data (i.e. the sample data itself, the "bits") completely untouched, with no conversions of any kind taking place - i.e. the audio buffers are copied as-is (they might be split or merged to fit buffer size constraints, but that doesn't affect the samples themselves). It's a useful concept if you're paranoid and don't trust software developers or hardware manufacturers to process the audio correctly.

"Bit imperfection" therefore just means some software or hardware somewhere altered the data in transit; for example, it did a sample rate conversion, or a sample format conversion. When done correctly this wouldn't normally result in audible differences, but the point of bit-perfectness is to eliminate that risk entirely.

The Windows Audio Engine is not bit-perfect, since it does sample format conversion, sample rate conversion, mixing, and APOs. This particular source of "bit-imperfectness" can be removed by using ASIO, WASAPI Exclusive, or WDM-KS, which all bypass the Windows Audio Engine. Whether or not this would "improve" the audio in audible ways depends on the particulars of your system, but in most cases, I would say no.

Windows/System volume VS AMP or DAC volume. (I hear this isn't even a question anymore in Windows 10 or something for some reason. but lets say for the sake of the argument we're in the Vista era)

Why do you think it's not a question in Windows 10 anymore?

what's worse? Lowering the volume on my AMP/DAC or Windows System (seeing as I can't run both at full blast). I would guess leaving the AMP or DAC at the volume where there is the least THD is the best route, and then lowering software volume as you see fit?

If the DAC has a hardware volume control, it's better to use that rather than a software volume control, otherwise you're reducing your effective signal-to-noise ratio and are operating closer to the limits of linearity.

Seriously, what's the point of DSD?

You can put that in the same bucket as "high definition audio": that is, you can safely dismiss it as a marketing gimmick as it will not make any audible difference. DSD has the additional twist of being significantly harder to set up and use (since it's a completely different format from PCM, it's not just a higher sample rate/bit depth). It's not worth it.
 
Last edited:

edechamps

Addicted to Fun and Learning
Forum Donor
Joined
Nov 21, 2018
Messages
910
Likes
3,619
Location
London, United Kingdom
People tend to worry about trivial LSB errors on digital volume controls while ignoring audible clipping distortion on the MSB.

Right. What I forgot to add was that it's often a good idea to set the software volume control to slightly below 0 dB. For example -3 dB. This mitigates the potential for intersample peaks which most DACs do not handle well.
 

bennetng

Major Contributor
Joined
Nov 15, 2017
Messages
1,634
Likes
1,691
Right. What I forgot to add was that it's often a good idea to set the software volume control to slightly below 0 dB. For example -3 dB. This mitigates the potential for intersample peaks which most DACs do not handle well.
Not only intersample clipping (which can be solved by using hardware volume control on DACs), there are also codec clipping which can only be avoided by using software volume control/management:
https://izotope-rx.livejournal.com/5760.html
 

Crane

Active Member
Forum Donor
Joined
Dec 16, 2018
Messages
145
Likes
163
Years ago, when my team was revamping the Windows audio stack for Vista, we put in plug-in features to add effects, everything from room eq to 5.1 to stereo virtualization. Alas, once I left Microsoft, my team was disbanded, the new people didn't know or care about those features and they are not really functional anymore. When they were, checking disable all effects disabled the processing I just explained.

I just noticed in the latest update of Windows the sound control panel has disappeared! And with it, the above option. It is a pain now to figure out what the heck is going on.

They do like to complicate things, if you open sound settings it'll be on right side sound control panel
https://www.addictivetips.com/windows-tips/access-control-panel-sound-settings-windows-10-1903/

Alternatively you can download an App from the windows store called EarTrumpet which gives you more option with sound control and makes it easier to access things.
 

Attachments

  • EarTrumpet.png
    EarTrumpet.png
    90.9 KB · Views: 284
Top Bottom