[MLGrado]

@jlaako

I know their are a couple of 'pseudo multi-bit' modulators as you put it in the instruction manual. I have assumed these are modulators that use multiple 1 bit modulators in parallel with unary coded quantizer and filter loop, Like a 3 level system that we might call 1.5 bit binary.

pseudo multi-bit because actually binary introduces its set of issues like glitch issues and increased noise levels due to more switches moving at once. These things are already problematic even in a fully digital system like HQPlayer on a PC, long before any analog conversion and dynamic element matching.

What about the AHM5EC5L and 7EC5L? Are these 'pseudo multi-bit' and does the 5L refer to 5 levels?

Thanks

Andrew

 

[suno23]

Not sure if this is the wrong approach but after some google searches I came across a post that shows how to make a bat. file to start hqplayer with its CPU priority set to high. Now I can upsample everything to 48k base DSD1024 no hiccups with most of the filters. Without setting priority to high I can only do 44.1k DSD1024.
With DAC Correction I can only do 48k DSD512 with all filters.
Would it be better to use DAC Correction at DSD512 or jump to DSD1024?

 

[plumpudding2]

Not sure if this is the wrong approach but after some google searches I came across a post that shows how to make a bat. file to start hqplayer with its CPU priority set to high. Now I can upsample everything to 48k base DSD1024 no hiccups with most of the filters. Without setting priority to high I can only do 44.1k DSD1024.
With DAC Correction I can only do 48k DSD512 with all filters.
Would it be better to use DAC Correction at DSD512 or jump

Can you share those resources? Last time Jussi said anything it was that HQPlayer sets its own thread priorities and you'll only mess that up if you change these things

 

[BShaw]

I'd appreciate some input about my settings, I'm pushing my M2 Mac Mini to it's limits, and was cruising along fine when streaming 44.1k files to DSD, but as soon as I get a Qobuz file above 44 I get stuttering. I checked the activity monitor, and with 44k files I'm running about 48%CPU, and 346% capacity. When I stream a 96k file, it jumps to 74% cpu and 600% capacity! DAC correction is on for my Cyan 2, and I don't want to give it up. My God the sound is unreal, when playing 44k tracks the sound was actually wrapping around me, I've never been able to achieve such alive 3 dimensional sound, it's very beautiful. Below are some pics of my settings, I wonder if I'm off somewhere? Thank you.

 

[genefruit]

I'd appreciate some input about my settings, I'm pushing my M2 Mac Mini to it's limits, and was cruising along fine when streaming 44.1k files to DSD, but as soon as I get a Qobuz file above 44 I get stuttering. I checked the activity monitor, and with 44k files I'm running about 48%CPU, and 346% capacity. When I stream a 96k file, it jumps to 74% cpu and 600% capacity! DAC correction is on for my Cyan 2, and I don't want to give it up. My God the sound is unreal, when playing 44k tracks the sound was actually wrapping around me, I've never been able to achieve such alive 3 dimensional sound, it's very beautiful. Below are some pics of my settings, I wonder if I'm off somewhere? Thank you.

same results if you move away from xla to long for Nx?

 

[MLGrado]

@jlaako

I know their are a couple of 'pseudo multi-bit' modulators as you put it in the instruction manual. I have assumed these are modulators that use multiple 1 bit modulators in parallel with unary coded quantizer and filter loop, Like a 3 level system that we might call 1.5 bit binary.

pseudo multi-bit because actually binary introduces its set of issues like glitch issues and increased noise levels due to more switches moving at once. These things are already problematic even in a fully digital system like HQPlayer on a PC, long before any analog conversion and dynamic element matching.

What about the AHM5EC5L and 7EC5L? Are these 'pseudo multi-bit' and does the 5L refer to 5 levels?

Thanks

Andrew

@jlaako

I think I refined my own understanding. A pseudo multi-bit dsm according to my research (good or bad), simply has multiple 1 bit quantizers working in parallel. They all receive the same signal, process separately, then the (3, 4, 5 etc) 1 bit streams are accumulated into one, presumably with higher linearity and less quantization noise than a single modulator can achieve. This can be further linearized by Dynamic Element matching. It is called 'pseudo' because while potentially better than just DSM with a single 1 bit quantizer, it can't match the linearity and lower noise of an actual multi-level quantizer, (in thermometer code, preferably in my book). A true multi level DSM.

Then there are DSM with binary quantizers. The issue with this is obvious to me in hardware ADC conversion, since their are seemingly unavoidable problems with the DAC in the feedback loop and the introduction of R2R non-linearities, power issues when multiple bits flip, and no easy possibility for DEM.

And as I mentioned before even in all digital systems like HQPlayer, there can still be timing and glitch issues, causing their own noise issues. Nothing is totally immune.

Wondering if I am on the right track, and would greatly appreciate guidance for my thoughts in the correct direction.

Oh, and about those other two DSM with 5L endings? Anything non proprietary intellectual property that can be safely shared?

 

[BShaw]

same results if you move away from xla to long for Nx?

I’ll give it a shot that will probably help, but I’m also just making sure my settings are optimal, I’m pretty new to hqp and still trying to understand all the options and how things work

 

[genefruit]

I’ll give it a shot that will probably help, but I’m also just making sure my settings are optimal, I’m pretty new to hqp and still trying to understand all the options and how things work

optimal for your DAC would be 48kX256 versus the 44.1 you have set. I also use full checkmark on adaptive rate. I don't think XLA at NX brings as much value as having it set at the 1x level. Try poly-sinc-gauss-hires-lp at Nx or even use the same as you have at 1x now (PSGL).