HQ Player

[numlog]

Is it normal for the Windows volume control to still function in W10 with both ASIO and WASAPI in HQ player?

[numlog]

What DSF settings and modulator is everyone using for PCM to DSD conversion?
choosing the right ones can have a dramatic impact IME.

 

Bit rate is 44.1x128.

poly-lin or poly-mp for conversion and slow- lin or fast-mp noise filter worked well...

if you are using a minimum phase oversampling filter try switching to the linear phase variant and use poly-mp for conversion. transients were nicer this way I found

 

Integrator was best left IIR, it was the most natural.

 

having DirectSDM enabled is better I feel, the volume control is disabled but the gain is reduced just enough for PCM, about -3dB. the manual suggests it only applies when the source is DSD, is this really the case?

 

The manual give much info on modulators. The default ASDM7 is a safe choice,  it sounds clean and seems like the most resolving one. I prefer to use DSD5v2, it sounds fuller and more robust ... maybe it is designed for DSD128?

DSD7 tended to sound unpleasant...

 

My favourite combination is poly-lin conversion and slow-lin noise filter with poly-sinc and DSD5v2. Smooth and resolving in a natural way

[numlog]

3 hours ago, juanitox said:

i'm not sure but i thought Miska said that the DSD settings filters only matters for DSD to DSD or DSD to PCM conversion and not PCM to DSD ?

it would be surprising if either the conversion method or noise filter setting didnt have an effect but its totally possible, the 'differences' are small enough that it could be psychological.

[numlog]

You are right, CIC is pretty good. It sounds colder than IIR so it was hard to get into at first but it seems like its fixing some issues with DSD and tightnening everything up, similar to DSD5v2 and the colder but more accurate ASDM7. PCM was typically thinner and colder than DSD but more accurate, now they are pretty similar except DSD doesnt have the same subtle negative traits I get with PCM (thin, shouty).

 

I think native DSD might sound better (or different at least) but using DoP allows you to control buffer size. With native DSD it is fixed 10ms for DSD64 and 20ms for DSD128, not sure if that is the same for everyone but this is what the ASIO control panel indicates with my DAC.

Using a low bitrate of 44.1x 64 allows a very low buffer size of 5ms without stuttering, the sound quality of 5ms DSD64 is very nice and beats DSD128 at its minimum of 15ms, DSD128 is still more resolving but the 5ms buffer has smoothness and cohereny that is not worth sacraficing for that resolution.

 

Also using HQ Player to set buffer size instead of ASIO control panel didnt work for DoP DSD, a 10ms buffer in HQ player with DoP DSD64 was indicating something weird like 312ms in the Asio control panel.

[numlog]

Has there been any measurement thats show the difference between no dithering and basic dither when upsampling?

Ive found  comparisons of dither and noise shaping.

also comparisons of upsampled 24 bit data truncated to 16 bit vs dithered to 16 bit... that is still different to upsampling 16bit 44.1khz to 24bit without dither and sending it to a 24bit non-oversampling DAC, right? no truncation in that case?

 

 

 

 

 

 

[numlog]

try normal poly-sinc

[numlog]

How 'lossy' is converting DSD to equivalent high res PCM and what is the nature of the loss? or would it technically 100% lossy because of the conversion taking place?

Obviously the only loss to be concerned about is the loss of musical information but don't think that's gonna be easy to objectively quantify. I could test and listen but I'm more interested in the theory behind it.

 

The main things I've been wondering about is the choice of bit depth and also any effect of maintaining a 'native' conversion rate, based on the data rates.

e.g DSD64 = 2.8224 MHz 1 bit

possible PCM 'native' equivalents = 176.4kHz x 16 bit, 117.6kHz x 24 bit, 141.1kHz x 20 bit

 

While 141.1kHz x 20 bit wont be realistically possible (AFAIK), it would seem to be the most balanced choice compared to 16 bit (possibly too low) or 24bit (completely overkill).

That is assuming there is anything remotely 'native' about this.

 

You could use 176.4kHz x 24 bit, but based on the data rate wouldn't it involve some sort of rounding ?
 

Instead you could use 176.4khz x 32bit, but this would still be sort of 'upsampling' (changing the data rate).

Im already guessing having the same data rate is pretty meaningless and any semblance of the original data is already lost in the conversion process. 

However I am dealing with R2R DACs/SPDIF so 32bit isn't any option.