Poly-sinc-gauss filters

@jussi_laako, could you, please, tell us a bit about the new gaussian filters?

What is there to be expected from them in terms of SQ?

I’m currently using the xla version and it sounds very good (Frequency and time response).

Hi, while waiting for Jussi to reply, did you check the updated manual for the description?

1 Like

Of course.

1 Like

One specific question I have for Jussi, is what is “optimal time-frequency response”.

Is this the ultimate best of both worlds, optimal time domain and frequency domain performance?

We’ve reached the holy grail of digital filtering ? :smiley:


That right there is the reason why I posted this thread. :relaxed:


Yes, sort of. I’ve had prototype of this filter since 2015, but never ended up including it for one reason or the other. But now I’ve finished work (so far) on this filter family and I’ve been pleased with technical performance (about 50 bits worth of reconstruction accuracy) and it sounds nice too.

I will likely add one more variant at some point just to see how far I can stretch GPUs for computing this. The current ones are doable on modern CPUs too.

Hard to say, there are many other good filters too, but this can be compared to ext2/ext3 or xtr.

I’m pretty happy with the result. Time will show which one I end up preferring more, ext2/ext3 or gauss.


Hi Jussi , with Poly Sinc Gauss up sampling to 24/192 n a devialet expert would you recommended Gauss or NS9 or another one ? Thanks

Gauss1 or TPDF dither is fine. If you use HQPlayer volume control more than just headroom, you can also try NS9.

Thanks Jussi , I normally have min and max set to -3 and then control volume with the Devialet remote .

Sorry if this is a bit of a basic question but what is it that differentiates these gaussian filters from ext2 and ext3? What makes them gaussian?

They have different time/frequency response. Put a bit simplified, shape of the roll-off curve is different.

This is a bit similar to comparing different types of analog filters.

1 Like