SOrry, Lampi did the update for me at RMAF. I do think it is on the Amanero website, however.
OK Thank you ā¦
Xtr filters donāt sound good to me.Since the 3.18 build when I use the XTR filters I get a constant popping sounds appro 10 secs into a song.Gone back to 3.17.
Been playing around with the minringFIR as suggested on here. I quite like it, super detailed and accurate, opens up the soundstage.
The xtr filters are almst the opposite - āanalogueā sounding and very easy to listen to.
The great thing about all this is the choices available. No one filter is going to sound best with every system, every set of ears or even every piece of music. My preferences constantly change, and its nice to be able to do it so easily.
I have not compared Roon upsampling to HQP for a while. I hear it is equally as good these days. What filters are people liking with Roon upsampling?
I tried PCM XTR filters and produced quite good PCM768 sound on my test Hugo2. No quite as good as the Hugo2 internal filters ā¦but pretty good. And they used a negligible amount of CPU on my lowly Lenovo Yoga.
I have this comment from Rob Watts:
https://www.head-fi.org/threads/watts-up.800264/page-28#post-13799581
Can you provide some clarification if his points have validity ā¦or does he just not know HQP filters or how powerful modern systems are.
Thanks
1 Like
Haha, hilariousā¦ 
Even slow small battery powered single core ARM computer running Linux can do DoP packing at DSD512 without hiccups. I donāt know what he usesā¦
The non-2s poly-sinc filters are single stage, going in one step from 44.1k to for example 22.5792 MHz (DSD512), not cascaded. closed-form with about 8M taps at 22.5792 MHz is a about 360 ms long.
I agree the pseudo numbers about taps are meaningless, but for some strange reason he/Chord wants to brag about number of taps like it would mean anything. Iāve said many times, number of taps are meaningless marketing.
Maybe he should look into modern CPUs and GPUs that HQPlayer can utilize in parallel. Iām not going to repeat the specs, but I doubt his DACs can reach the memory bandwidths of Nvidia GPUs exceeding 480 gigabytes/sec. Or clock speeds of latest Intel CPUs running at over 4 GHz. Latest Nvidia GPUs have 12 billion transistorsā¦
In addition, the DSP system in HQPlayer is not run synchronously with the input data, so it can go back and redo the calculations multiple times as necessary to arrive at better solution, based on running analysis of the output data. And still keep on producing DSD512 (or DSD1024) output.
P.S. About the 30 years comment, algorithms in HQPlayer have taken me soon 20 years to refine. Next year is 20th birthday of HQPlayer.
8 Likes
And letās not forget possibility to run digital room correction filters which can be for example 256000 taps long, in addition to upsampling, all for example for 8 channels of audioā¦
1 Like
@jussi_laako
Thanks for your reply. The Hugo2 accepts PCM768 thru USB and I found that the subtlety of difference between āpoly-sync- xtr-mpā from HQPlayer and using the Hugo2 WTA filters to be small but still noticeable. You can insert all the popular audiophile terms here about how Rob Watts WTA revealed more, had more there there, etc. So, my question(s) to you are:
- Do I need to go to DSD or -2s and get a beefier CPU to extract more SQ? (I chose power-of-two PCM up-sampling multiples to reduce CPU usage and never tried the -2s switch)
- We know that WTA does not restore the original samples in the reconstruction - as it was fine tuned thru listening tests. Can we expect Signalyst to ever setup a small audience test to do something similar?
- RWatts will shortly complete a AD-DA loop on live performances using his new hardware and thus further refine his WTA algorithms (#taps and coefficients) to achieve the perfect reconstruction he seeks. Can you do similar ? But perhaps sourcing a few hundred original masters (PCM768) and doing a software decimation-up-sampling loop to determine (in the aggregate) which XTR filter best reconstructs?
Thanks
Dan
Just noting that the -2s filters are computationally less intensive than the āstraightā variant. They make two passes, but the first is the most significant.
Just use the WTA stuff then if you like it. I donātā¦
Different DAC perhaps as a starterā¦ Something that puts all the money on one thing, converting digital samples from input to analog output as well as possible, without messing around with the data in some DSP engine. Or at minimum has option to bypass all the DSP. Fully enabling you to have all the DSP outside of the box.
Iām not going to reveal details how I do fine tuning.
Iāve done similar already, but it is only tiny part of the entire process.
Been using the minringFIR for a few days now. Quite liking what I hear. As lush sounding as the xtr filters are, there is something to be said about clear separation, definition and impact the vocals/instruments have with the minringFIR.
I think the minringFIR is one of the few non apodizing filters in the HQPlayer arsenal. Interesting.
Yeah; minringFIR, poly-sinc-hb, closed-form and polynomial are non-apodizing ones.
closed-form is somewhat similar to xtr in terms of length, but completely different kind of algorithm. poly-sinc-hb is shorter than xtr, but same algorithm just different filter.
With non-apodizing you can expect more album to album variation depending on what kind of tooling was used for making the album. With apodizing filters you get more consistent performance throughout. Of course thereās still quite a bit of variation due to other aspects like microphones, analog equipment, eq, compression, etc.
1 Like
@jussi_laako
I am about to get a Holo CYAN/DSD with a beefy i7 laptop and jump into dsd512 upsampling. In the few times i have heard a Chord DAVE+BluMkII, iāve been floored by the sound quality. My hope is I can achieve something similar by checking all the boxes in HQPlayer - which I assume can be realized by the XTR filters. I know you are not a fan of the Chord sound signature ā¦but id appreciate if you can help suggest what settings you feel make work best.
Thanks
You can try both the xtr filter (linear phase version) and closed-form. For most computers poly-sinc-xtr to DSD512 is too heavy, but the poly-sinc-xtr-2s variant should run on most. ASDM7 modulator is a good starting point.
Two others to try are poly-sinc-ext and poly-sinc-hb, possibly in -2s flavors depending on your CPU.
Leave the Multicore DSP setting in automatic mode (grayed box).
1 Like
Time to bring the thread back to life.
Closed-form is the new filter dejour, and has been for a few months. Stll going pcm to pcm with dsd direct.
More clarity/focus to instruments and vocals over xtr but yet still a relaxed presentation. Xtr is slightly smeared in comparison. @andybob I donāt hear that sizzle you mention in your first post.
Iāve thrown all kinds of 80s hair metal at closed-form expecting bad results but it always delivers.
I havenāt given closed form the attention it deserves lately because itās not autosync. That wouldnāt be a problem with auto rate family ticked, except that my DAC seems to have a hardware issue with 48k and Iām too lazy and cheap to send it back to China for repair (if possible). So for the moment Iām sticking to 705.6 kHz PCM and filters that support it.
I believe the Schitt guys have a deep preference for closed-form filters and it was my initial filter of choice. I understand it is a relatively long filter in the time domain, but not as long as xtr.
I suspect that my instinctive preference for longer filters is because the post-ringing presents as a very slight chorus effect. Perhaps itās not hi-fi, but I like it.
Changed my settings again to DSD512, poly sinc shrt mp and DSD7 256+fs. Iām liking this combination on my gear as it preserves some attack that I think is lost with the AMSDM7 512+fs modulator.
Got one of the new Analog 2 Yggys to test, Iām running naked. Or closed form I guess. 
Interesting, what is the associated modulator you use with Closed-form ?