Ending the Windows Audio Quality Debate

[bennetng]

I did not yet. I tried mpv, Spotify and IrfanView, since i have no other apps which read mp3.

The real point though is, neither Windows volume nor EAPO preamp works right.

Both the format decoder (decode into the music player's audio pipeline) and audio output (to Windows mixer) require full floating point support, or the decoder needs to decode to a lower volume. If either of these steps fail, distortion will occur.

 

[Nozza]

Tried it. Cannot hear any difference whatsoever.

Sure this is actually true?

 

[Ron Texas]

Horsefeathers!

 

[bennetng]

Tried it. Cannot hear any difference whatsoever.

Sure this is actually true?

Since you did not quote any post, I don't know if you were replying to me or not. However, if you meant these files:

https://www.audiosciencereview.com/...y-of-software-volume-control.5922/post-172865

If you read the subsequent replies from many other members, up to the last page, things should be very clear.

If you were not replying my posts, then forget about it.

 

[Nozza]

Sorry, I mean the original post and the suggested settings.

 

[DDF]

Sorry, I mean the original post and the suggested settings.

Horsefeathers!

There seems to be some lingering misunderstanding about what the OP is communicating.

• These are repeatable objective test results. They're real (no Henry Blakes required).
• Windows has specific edge conditions where it can unexpectedly distort. These conditions are identified, objectively proven, and free mechanisms provided if you feel they're risks worth avoiding

Making these changes doesn't mean that you will suddenly hear a change in Windows audio quality. To even hear these problems, you need to see the conditions shown. The hidden APO's is especially a crap shoot, there's no way to know if a particular machine has this problem or not without trying it.

I'm coming off like a broken record now, so I'll bow out but I hope this helps clear things up.

 

[Tangband]

Or...or... Simply don't use Windows for audio.

+1.
The biggest issue with windows is not that it CAN be good enough, problem is that you will never know when or why.
A new upgrade comes and you might have trouble, or not.... thats very fatal for creativity when recording or listening and you really dont want to sit and fix things that shouldnt be there in the first place.
Ever since Windows Vista they have shown us all that they really dont care about hi fidelity.

Steve Jobs was an audiophile and that spirit is in all apple products.

 

[Ron Texas]

Let me understand this whole thing. I'm using Foobar with shared WASAPI and Equalizer APO (preamp set to -.1 db in excess of what is needed to get the filters to zero) one machine. Volume is adjusted using the Foobar volume control. This never gets higher than -6 db. Windows output is set to 48 khz 24 bit so that videos played back on VLC don't get resampled.

Do I have to do anything differently? Is -.2 db going to kill the problem off entirely or do I not even need that? I'm ready to take something for my blood pressure.

My other machine is using ASIO. Maybe I should install ASIO on both and just use Equalizer APO for everything other than music.

 

[companyja]

Awesome post! Sorry if I'm derailing anything, just wanted to know two things for absolute peace of mind while also using DirectSound:

1) By applying the negative preamp in EQAPO, is it fine to now put the level in windows at 100%? CAudioLimiter will never be engaged and most intersample peaks will have enough headroom (not that the latter is a DirectSound issue)?
2) You recommend setting the bit depth to 24bit in windows - is there any difference between setting it to 24bit or 32bit? I thought I might as well keep it at 32bit since most DACs advertise 32bit nowadays.

And yeah, installing EQAPO will automatically re-enable Enhancements in the Sound Properties menu - disabling them will disable EQAPO as well.

 

[DDF]

Awesome post! Sorry if I'm derailing anything, just wanted to know two things for absolute peace of mind while also using DirectSound:

1) By applying the negative preamp in EQAPO, is it fine to now put the level in windows at 100%? CAudioLimiter will never be engaged and most intersample peaks will have enough headroom (not that the latter is a DirectSound issue)?
2) You recommend setting the bit depth to 24bit in windows - is there any difference between setting it to 24bit or 32bit? I thought I might as well keep it at 32bit since most DACs advertise 32bit nowadays.

And yeah, installing EQAPO will automatically re-enable Enhancements in the Sound Properties menu - disabling them will disable EQAPO as well.

Thanks, hope it helps.
1) Yes, that was tested and proven to avoid CAudiolimiter in the OP. The 4 dB is recommended by Benchmark and meets RME guidelines as well
2) I haven't tested 32 bit but expect realistically no difference from 24 bit. It'll just zero pad further and use a lower level of dither, but 24 bit dither won't be audible as noise.

 

[Mnyb]

Stupid follow up Q The Topping D70 DAC i just bougth supports 32bit (probably floating point data) should i use that instead of 24bit will the DAC ?

I dont know how the USB interface operates in this regard and if the DAC is using floating piont internally for its calculations ?

I'm fully aware that i could not possibly hear a difference but i just want to set things to whats most "correct" .At i work I'm usually trying to avoid unecessary truncation and dataformat transformations in my programs (but industrial systems are so much worse than audio ).

So if the DAC takes the 24bit integer and transforsm it to float for its internal upsampling and filters i migth just avoid sending integers in the first place ?

 

[mansr]

It is extremely unlikely that the DAC uses floating-point at all. 24-bit or 32-bit integer makes no practical difference since nothing can resolve more than about 21 bits anyway.

 

[Mnyb]

It is extremely unlikely that the DAC uses floating-point at all. 24-bit or 32-bit integer makes no practical difference since nothing can resolve more than about 21 bits anyway.

You are interiely correct , just wondering if there is such a thing as a native input format to the chip .

 

[ninetylol]

You are interiely correct , just wondering if there is such a thing as a native input format to the chip .

Install the XMOS driver for your device and it should Show which bit Format it works with. At least my Loxjie D30 Shows 32 bit regardless of what settings i send.

 

[bennetng]

Install the XMOS driver for your device and it should Show which bit Format it works with. At least my Loxjie D30 Shows 32 bit regardless of what settings i send.

The clipping you experienced in my test mp3 files when using exclusive mode is a proof that there is a conversion stage from floating point to integer before audio data entering the DAC chip.

Stupid follow up Q The Topping D70 DAC i just bougth supports 32bit (probably floating point data) should i use that instead of 24bit will the DAC ?

I think your DAC comes with a driver with ASIO support. Open RMAA, click the ASIO button to check the bit format.

Int32LSB means the driver uses integer. I've seen some virtual ASIO drivers use float though. However, driver is still at software layer, RMAA cannot show what happens later on. For example, if you are outputting to SPDIF, then the format must be capped to 24-bit later on.

 

[ninetylol]

The clipping you experienced in my test mp3 files when using exclusive mode is a proof that there is a conversion stage from floating point to integer before audio data entering the DAC chip.


I think your DAC comes with a driver with ASIO support. Open RMAA, click the ASIO button to check the bit format.
View attachment 111244
Int32LSB means the driver uses integer. I've seen some virtual ASIO drivers use float though. However, driver is still at software layer, RMAA cannot show what happens later on. For example, if you are outputting to SPDIF, then the format must be capped to 24-bit later on.

Clipping appeared with Toslink, I just changed to USB few days ago, and didnt test your Files So far

 

[Mnyb]

The clipping you experienced in my test mp3 files when using exclusive mode is a proof that there is a conversion stage from floating point to integer before audio data entering the DAC chip.


I think your DAC comes with a driver with ASIO support. Open RMAA, click the ASIO button to check the bit format.
View attachment 111244
Int32LSB means the driver uses integer. I've seen some virtual ASIO drivers use float though. However, driver is still at software layer, RMAA cannot show what happens later on. For example, if you are outputting to SPDIF, then the format must be capped to 24-bit later on.

Thankyou another tool to learn

8000 Hz - not supported
11025 Hz - not supported
16000 Hz - not supported
22050 Hz - not supported
32000 Hz - not supported
44100 Hz - supported
48000 Hz - supported
88200 Hz - supported
96000 Hz - supported
176400 Hz - supported
192000 Hz - supported
352800 Hz - supported
384000 Hz - supported
Input channels:
Output channels:
channel: 0 (Analogue 1) - Int32LSB
channel: 1 (Analogue 2) - Int32LSB

 

[bennetng]

Int32LSB is essentially the norm even when the interface only uses 24-bit components like this one (CS4272)
https://prosound.ixbt.com/interfaces/focusrite-scarlett-solo.shtml
24-bit is not as convenient as 32-bit since the program usually needs to read/write a 16-bit data and an 8-bit data, or three 8-bit data.

 

[Katji]

Mine shows 32-bit (IEEE floating point) (RMAA soundcard modes) but I can't find any reference to floating point in the DAC datasheet, just 32-bit.
I'm quite impressed by what I see though.

 

[bennetng]

Mine shows 32-bit (IEEE floating point) (RMAA soundcard modes) but I can't find any reference to floating point in the DAC datasheet, just 32-bit.
I'm quite impressed by what I see though.


View attachment 111564

click the ASIO button

If you click the "Modes" button, even a Realtek will show IEEE float. The "Modes" button shows the formats supported by the Windows Mixer. If a non-virtual, kernel mode ASIO driver supports float, then it means there is a possibility to set the global volume via the product's dedicated mixer to avoid clipping and limiting even when the DAW/playback software is outputting 0dBFS+ data via ASIO.