01-Oct-2021, 07:51
There was a recent post from HQPlayer developer Jussi Laako (Miska) on AS that interested me:
When doing any DSP in HQPlayer, like convolution, you will always need to use some dither/noise-shaper. Otherwise you will be adding a distortion generator.
When you upsample in HQPlayer when doing convolution or similar, you gain additional one bit worth of dynamic range headroom every time you double the output sample rate. This will give you some additional resolution when you do such DSP operations.
Currently I am using the Focus Fidelity convolutions in Roon. This suits me well at the moment, as within Roon it is possible to load a number of slightly different convolutions, and switch between them very easily, even remotely via iPad. As an example, I have two convolutions at the moment that I like very much. They are slightly different, one is based on the classic Harman house curve, the other is similar, but using more of a "best fit" curve per the tools in Focus Fidelity. I am still not sure which one I ultimately prefer (or even which one is technically most accurate), but with Roon it is very easy to switch between the two if desired when listening to a particular track or album.
Looking forward, I think that maybe I would get better results sonically by using the convolutions in HQPlayer. Mindful of the above quote, and that currently most of my listening is with 16/44.1 material, currenly 16/44.1 is used for the convolutions in Roon, before being sent to HQPlayer which upsamples to 24/192 etc. If I use the Focus Fidelity convolutions in HQPlayer together with the 24/192 upsampling, I should get some benefit per Jussi's explanation above.
Something to think about and maybe try in the future. I wonder how much difference this might make in practice?
When doing any DSP in HQPlayer, like convolution, you will always need to use some dither/noise-shaper. Otherwise you will be adding a distortion generator.
When you upsample in HQPlayer when doing convolution or similar, you gain additional one bit worth of dynamic range headroom every time you double the output sample rate. This will give you some additional resolution when you do such DSP operations.
Currently I am using the Focus Fidelity convolutions in Roon. This suits me well at the moment, as within Roon it is possible to load a number of slightly different convolutions, and switch between them very easily, even remotely via iPad. As an example, I have two convolutions at the moment that I like very much. They are slightly different, one is based on the classic Harman house curve, the other is similar, but using more of a "best fit" curve per the tools in Focus Fidelity. I am still not sure which one I ultimately prefer (or even which one is technically most accurate), but with Roon it is very easy to switch between the two if desired when listening to a particular track or album.
Looking forward, I think that maybe I would get better results sonically by using the convolutions in HQPlayer. Mindful of the above quote, and that currently most of my listening is with 16/44.1 material, currenly 16/44.1 is used for the convolutions in Roon, before being sent to HQPlayer which upsamples to 24/192 etc. If I use the Focus Fidelity convolutions in HQPlayer together with the 24/192 upsampling, I should get some benefit per Jussi's explanation above.
Something to think about and maybe try in the future. I wonder how much difference this might make in practice?
1000 Pro - KEF Blade - iFi Zen Stream - Mutec REF10 - MC3+USB - Pro-Ject Signature 12
