[taipan254]

Here you go!
I also included a 512 plot so it can be compared a bit more easily

This is so weird to me. Based on this chart, for DSD 256 upsampling, it makes more sense to use the 512+fs modulator (designed for 512 and above). Why would Jussi not recommend using the 512+fs variant for DSD 256?

 

[GoldenSound]

Thank you so much! So this is for the sweep tone at 1khz. I am wondering how the ASDM7EC-Super 512+fs modulator running at DSD256 behaves when the underlying content is hires in the range from 44khz to 88 khz.

The input sample rate won't really affect things, the noise profile you see here will remain the same.

The sample rate of the source material will mostly affect the reconstruction filter itself not the modulator.

This is so weird to me. Based on this chart, for DSD 256 upsampling, it makes more sense to use the 512+fs modulator (designed for 512 and above). Why would Jussi not recommend using the 512+fs variant for DSD 256?

Not 100%

Keep in mind the X axis here is a logarithmic scale. The 512+fs modulator at 256 has higher noise between 800khz or so and 3Mhz. Upto 20-30dB depending on where you look.
Depending on the particular DAC, how its output filter is designed, and how the preamps/amps you're using upstream respond to higher high frequency content, you may be better off using the non-512+fs modulator.

If you have something like a Holo DAC, then its 1-bit converter actually acts as an analog domain moving average FIR filter and so you can use the 512+fs modulator pretty much without any penalty. DACs with more basic 1-bit stages/weaker filters may be better off sticking with the normal one.

 

[Triplefun]

This is HQP5 SincMG converting PCM@192kHz to DSD512x48
SincMG is a good compromise between adequate filter length and not having excessive loading delays
'bobb

Load trace, RTX3080ti GPU load 17%

Hi
Im looking to do dsd1024 on a Holo may. Would you recommend the 7950x3d or would you change to Intel 14909k. I have a 4090 GPU.

 

[GoldenSound]

Hi
Im looking to do dsd1024 on a Holo may. Would you recommend the 7950x3d or would you change to Intel 14909k. I have a 4090 GPU.

Intel has been the preferred choice previously. Both due to a bit of an issue with AMD USB controllers that made DSD1024/PCM1.536Mhz less stable or impossible (though the Gen 2.1 USB card on the May uses Jeff's own USB receiver design and gets around this). But also because the way in which the inter-core communication on AMD worked, it typically was more limited in what it could do on HQP.

Having an NVIDIA GPU helps a lot though. I would recommend the intel CPU if you can

 

[bogi]

Jussi doesn't recommend running 512+fs modulators below their specified rate, but it seems the 512+fs modulator is already superior to "normal" super at 256x rate in the audible band.

512+ modulators are designed for lowest possible noise floor within 100k wide audio bandwith. Such a bandwith is considered by Jussi as sufficient for every existing audio content, incl. DXD or DSD downloads. But Jussi also noted that for most recording 50k is enough bandwith.

When 512+ modulators are used at DSD256 output rate, that 100k wide area is shrunk to 50k. But 192k source recording can contain audio content up to fs/2 = 96k, which is more than 50k. So it is not optimal for high res audio content which can contain audio information above 50k. Not much such content exist. Therefore don't get it as a strict rule. And like you can see above, graphs don't show so steep rise of modulator noise floor with raising frequency. So 192k content can be still ok with 512+ modulators at DSD256 output rate. Listen and compare, how it sounds for you with your chain and source content of your interest.

See https://audiophilestyle.com/forums/topic/19715-hq-player/?do=findComment&comment=1246430

 

[1laraz]

Here you go!
I also included a 512 plot so it can be compared a bit more easily

Dear GoldenOne, there have been measurements made for DSD256 vs DSD512 where DSD512 appears to be a noisier rate (both for the ASDM7ECv2 modulator):

Could you please elaborate how these reconcile with your recent measurements? As per your measurements DSD512 performs just better than DSD256 for the same modulator.

 

[GoldenSound]

Dear GoldenOne, there have been measurements made for DSD256 vs DSD512 where DSD512 appears to be a noisier rate (both for the ASDM7ECv2 modulator):

Could you please elaborate how these reconcile with your recent measurements? As per your measurements DSD512 performs just better than DSD256 for the same modulator.

Those measurements are at the output of the DAC itself. Depending on the DAC/hardware design, some DACs may perform best at higher rates or specific rates.

 

[1laraz]

Those measurements are at the output of the DAC itself. Depending on the DAC/hardware design, some DACs may perform best at higher rates or specific rates.

So there is a trade-off between the worse Holo DAC perception of DSD512 rate and the superior behavior of 512+ modulator at this rate as compared to DSD256. Based on the measurements above is it possible to recommend for Holo DAC owners the optimal combination: DSD256 rate + ASDM7EC-Super 512+fs or DSD512 rate + ASDM7EC-Super 512+fs?

 

[rkt31]

I am not expert in hqplayer setting. Just wanted to know about the current discussion. Is it about converting pcm or dsd to higher dsd using some modulator and then feeding it to a 1bit compatible native conversion? Does converting everything to 1 bit is better than pcm to pcm upsampling?

 

[rkt31]

Did anybody try sinc L long? I could use sinc L medium only upto 768khz but sinc L long skips.

 

[rkt31]

Are there any graphs for sinc L medium or sinc L long or comparison of all sinc pcm to pcm upsampling?

 

[GoldenSound]

I am not expert in hqplayer setting. Just wanted to know about the current discussion. Is it about converting pcm or dsd to higher dsd using some modulator and then feeding it to a 1bit compatible native conversion? Does converting everything to 1 bit is better than pcm to pcm upsampling?

It's not about which is better but rather which one will provide a better result with your DAC.

When you feed most DACs PCM, they will then modulate to a lower bit depth internally (5 bit 11Mhz is common but depends on the particular DAC).
Using a better external DSD modulator can yield better results because you can use much more advanced modulators than what is running on the DAC internally.

If you have an R2R DAC PCM may be better, on many Delta-Sigma DACs DSD will be better. (Though some delta sigma DACs, ESS for instance, cannot output DSD natively and will convert it back to PCM internally so in that case you're better off feeding them PCM)

 

[bogi]

Is it about converting pcm or dsd to higher dsd using some modulator and then feeding it to a 1bit compatible native conversion?

When the talk is about software delta-sigma modulators, then output is 1 bit DSD stream. Yes, best use of 1bit DSD stream is to feed it to a DAC capable of so called direct DSD path. DSD data are then arriving directly to DAC D/A stage, without any oversampling or modulation performed in the DAC device.

GoldenOne's measurements are pure digital domain measurements without D/A conversion involved.
The measurements posted by 1laraz are from Spring3 output.

Does converting everything to 1 bit is better than pcm to pcm upsampling?

It depends on DAC architecture. Not all DACs are of oversampling delta sigma design. HQPlayer is most suitable for DACs with direct DSD path to D/A stage. But it may bring improvement also with other DAC architectures. Also some PCM DACs of R2R type provide direct D/A conversion if input data is fed at a suitable PCM sample rate.

 

[rkt31]

When the talk is about software delta-sigma modulators, then output is 1 bit DSD stream. Yes, best use of 1bit DSD stream is to feed it to a DAC capable of so called direct DSD path. DSD data are then arriving directly to DAC D/A stage, without any oversampling or modulation performed in the DAC device.

GoldenOne's measurements are pure digital domain measurements without D/A conversion involved.
The measurements posted by 1laraz are from Spring3 output.



It depends on DAC architecture. Not all DACs are of oversampling delta sigma design. HQPlayer is most suitable for DACs with direct DSD path to D/A stage. But it may bring improvement also with other DAC architectures. Also some PCM DACs of R2R type provide direct D/A conversion if input data is fed at a suitable PCM sample rate.

Any example of da taking direct modulated stream without using it's own modualation?

 

[bogi]

ESS for instance, cannot output DSD natively and will convert it back to PCM internally

We need to distinguish if the talk is about DAC chips alone or about DAC devices as such.

ESS chips don't convert "back" to lower PCM rates when they are fed by DSD data. No resampling is applied to DSD data in ESS chips. Data format is adapted to allow optional volume level adjustment and to fit modulator input, but original DSD data resolution is not lost like in the case of downsampling to PCM rates (which involves so called decimation). Adapted input data is then re-modulated in ESS chip and finally it is fed to D/A stage. ESS modulator runs at incoming DSD data rate. For DACs utilizing this ESS chip functionality it is suitable to feed them with DSD data. That brings still less internal processing in DAC chip than with PCM input since complete hardware oversampling (including the 2nd simplified stage) is skipped with DSD input.

Some other DAC chips (AKM, Cirrus) provide both direct DSD path and an alternative path with volume processing. It depends on DAC device designer if direct DSD path option is used or not. There are also chips from TI (BurrBrown) and ROHM which provide only direct DSD path with no possibility of volume level adjustment for DSD data.

There are more DAC devices which convert both incoming PCM and DSD data to a common PCM format and rate and then oversample with their own filters. Then they feed oversampled data to a DAC chip or other type of D/A solution. In such a case it is more suitable to use PCM input. An example are Chord DACs.