Making DSD down sample playback work from Roon Rock till HQPlayer 4.19.1 Embedded

Hi Gents,

Got some various DSD files on my NAS and am currently evaluating the latest HQPlayer Embedded trial as a sample upgrade till my Roon Rock. Playback of any PCM, even high res is working like a charm, as also any file from Tidal.
Now, it is not a giant leap of resampling, just for my older JBL studio monitors capable of 24/96 in on AES3 input.
The successful upsample of 16/44.1 have revealed the superior signal integrity of HQPlayer, compared to upsample in Roon Rock. The monitors really sing, with the HQPlayer handling the mathematics.
Now, trying to playback DSD, any DSD format, will simply not work. I cannot get the grip what the heck I am doing wrong. I have managed to download a HQPLayer Desktop Pro manual, but the embedded version of (web) UI is rather different. I understand it all boils down to the box of DSD playback.
Would anyone be so kind to show me how to get DSD playback settings to work?
I have Roon Rock on one NUC, the HQPlayer on another one feeding the USB to EAS3 converter including the XMOS U208 chip. Sounds wonderful … except dead silent DSD’s :slight_smile:
This is a potential show stopper for me buying the HQP and I really want it to play everything. Is there any software that will be able to downsample DSD’s to my monitors modest 24/96 with a good quality it is HQP, I think.

Can you provide a screenshot of your current settings? Also please check that you are asking for PCM output and not for example “Auto” mode. And with such converter, remember to set “DAC Bits” to 24!

Thanks for your prompt answer and kind attention :slight_smile:

Included herein I’ve pasted my actual configuration screen shoot. Feel free to suggest any changes, including also the working PCM playback, if you think there are additional benefits with alternative config.

Sorry for the small screen, I had to reduce since the UI page did not fit my TV … :blush:

Look OK otherwise, but try setting “Output mode” to PCM just in case.

I would get started with “standard” as noise filter and “poly-ext2” as SDM-PCM conversion. And then change only one of the items at once if you like to. Makes it easier to notice what kind of effect each change has.

Kiitos, Jussi :grinning:

Don’t know if you speak Swedish? But it works now, thanks a lot … Having trouble with computer performance. Shouldn’t have been so stingy and beleive that a 4 core Celeron should manage these maths. It have been so very flexible as a HTPC and also running Roon with substantial sample and EQ going. But I will not run the more extreme filter algorithms in your software. On the other hand, that was also good to find out It is especially the more advanced DSD to PCM calculations that it didn’t like. Steep PCM filter conversion is working good also with Shaped Dithering. Very silent and with the ploy-sinc-long-pm the dynamics and attack is brutal and very live-like. You’ll have an order within short.

1 Like

Now I have got hold of a not so very new but reasonably potent I7 4-core and 8 threads. Doesn’t manage all settings but a lot more. I have stuck with a poly-sinc-hb, said to be very steep filter and silent. It is together with the filter Shaped creating av very live-like playback, with accentuated transients, drum beats are a dream, maybe not as silky in high mids, but recalibrating my EQ, the pieces came together. According to your expertese, is this a filter, with a tight accuracy to the source, leaving a very small own footprint? My studio monitors says it is, but it would be nice to have an expert opinion. Live-like sound is my benchmark. Transients must be lightning fast and with enough amplitude to connect the body and soul. Is it also equally true to the source? Pre- or post-echo?
Just love your work, the last few days with this more potent PC, made a great difference in my rig.

Kind regards, uudelleen kuulemisesta


Yes, since it is non-apodizing it generally doesn’t leave much trace. Due to this, the results vary more depending on the source content quality. On high quality content without inherent problems it can be good. For not so high quality content (like ones recorded with modern ADCs) it will also reproduce the faults.

Apodizing filters can fix many of the faults on recordings made with modern ADCs or some mastering tools. They do this by replacing “footprint” of the ADC with a better one and cleaning up some problems while doing so.

Other filter similar to poly-sinc-hb, but even steeper you could try is poly-sinc-xtr or sinc-L.

Apodizing filters you could try are poly-sinc-ext2, poly-sinc-short (-mp) and sinc-S.

You could especially compare poly-sinc-xtr vs poly-sinc-ext2, and sinc-L vs sinc-S. But note that the results will largely depend on source content quality. For lower quality (or newer) source content the difference should be bigger. It also depends on music genre. Piano music will likely have very small differences, while some modern rock or pop will have bigger.

1 Like

After reading your manual for Desptop Pro-version, there was a section describing Nvidia GPU and Cuda as possible performance enhancer if CPU power is not good enough. I have purchased the only possible card that do not include fans (noise) and are low profile with enough memory size according to your recommendation, it contains the GT-1030 chip comprising 384 (only) Cuda Cores and GPU memory RAM GDDR5 at 2GB. Came on the market 2017.

My Intel I7 is 4TH gen 4 cores with 8 threads, still cannot handle some of the filters. The Embedded trial I still use until I get hold of how good performance I can achieve. Does the Embedded Linux software find this GPU enhancement itself? Should I force the use of the Intel GPU onboard Intel CPU above, to completely unload the low numbers of Cuda Cores for the purpose of floating calculations? Perhaps this was a way too low upgrade, but for me the only affordable right now …

My dear friend have a newer and more potent PC and claims very good results with the setting proposal you gave us above. It would be great if this relatively small upgrade allows me to reach poly-sinc-ext or even ext2.

My God, there is a difference like night and day adding HQPlayer to the already quite alright Roon. Hope for a devine intervention :innocent: allowing me to reach the higher level of filter with the relatively cheap graphic chip…

Yes, the Ubuntu Bionic build of HQPlayer Embedded has support for CUDA. Other builds don’t. In addition you’ll need latest Nvidia display drivers installed to enable it. Status about the offload can be seen in the web interface during playback as “C”.

That can help sometimes, if there would be notable graphics tasks on the GPU. It depends a bit on the GPU model if it stays active without connected display. But that can be tested.

ext2 could possibly work, depending on output rate. And more likely something like sinc-S which uses different kind of algorithm.

1 Like

Ok, so I guess the 4.19.1 Embedded I downloaded from the bottom of your page: is built on the Ubuntu Boinic edition, is that a correct assumption?

Sorry for all the questions, I am a noob when it comes to custom built softwares on Linux base.

If the latest Embedded HQPlayer is already OK and support the use of Cuda cores, how do I add/implement the nVidia graphic driver, since there is no “windows like” UI? Or is it such a great luck it will itself ask or even download/install without my hands-on?

Never edited Linux before, do not know how. Do you provide a tuturial anywhere on your page? Do you think it is worth adding this in the above Embedded version from the start, perhaps also with auto update, so that anyone connecting a graphic GPU CUDA will benefit from the boost? Just a suggestion … without knowing anything, I have to add … :blush:

Depends on which sub-folder you download it from. If you download it from the Bionic folder then it does have the support.

The Nvidia driver installer runs on text console, so that is not an issue. You can bring the installer in through various routes, either from another computer, or download it with a text mode browser like lynx or links.

No, Linux tutorials are out of my scope, but there are many available online, so it wouldn’t even make sense for me to try duplicating such.

1 Like

Hi again, make a final question, it takes some computer power to manage the full functionality of your software. Knowing this and decided to disregard from adding graphic card and instead go for new mini computer and make the easy installation of the embedded HQP on a usb and no further fuzz.
Is there any reason why not choosing the new AMD Ryzen CPU? These small barebones, are significantly cheaper than the Intel NUC.

The CPUs I know that can largely manage “full functionality” (depending what is meant by that) are Intel’s i9-9900K(S) and and i9-10900K. There may be others, but these are the ones I’ve tested on.

1 Like

Paid and licensed the 4.19.1 Embedded. Best digital signal improvement ever. Running poly-sinc-ext2 with “shaped” dithering @ upsampling from 16/44.1 to 24/96. Running on a finally built dead silent Intel I7-9700 PC.

Will start disengage any/all hardware circuitry in BIOS not necessary for music playback.

F*ck this is good! All new album library …

Thank you, Jussi

btw. same writer as the comments above. The accounts seems to play tricks with me …

1 Like