The HQPlayer thread

Agreed that HQplayer will work even with a Raspberry Pi 4 if the goal is to stick with PCM say up to 24/384. For multi-bit DACs - i.e. IFI burr brown chips that supports Native DSD support, the higher end hardware can be utilized for on the fly conversion to DSD and apply the noise modulator

 

Okay, DAC and digital boffins, questions...

I've got a Gustard R26 DAC. I'm thinking of trying it in NOS mode with HQP doing upsampling.

--Should I be upsampling to PCM or DSD? I really only stream (Qobuz through Roon) and don't bother with local files. I assume since the R26 is PCM native, it'd be better to upsample to PCM? (DSD Direct seems to go to a separate Sony chip.)

--What sort of sampling rate should I be aiming for? Highest possible?

--Related: I have a spare Lenovo Intel i5 Gen 8 laptop that's not being used for much. Is this a good candidate for Roon Core and HQP? I know it's not particularly powerful, but if I'm just upsampling to PCM, that seems doable on most machines?

--As an HQP endpoint, I plan to try either the built-in LAN streamer in the R26 or my iFi Neo Stream. Both seem to support NAA.

--Anyone have suggestions on what filters to use in HQP?

This stuff sort of makes my head spin, and google searches haven't termed up many obvious answers, so thanks for any suggestions!

 

Can’t comment on DSD, never had a DAC that was DSD capable since I have HQP.

The highest rate makes sense, of course, though some filters only support integer upsampling, i.e. 44.1 can’t be upsampled to 384, only to 352.8.

My Linux i5 desktop isn’t plagued much by upsampling to 192 kHz, I guess it would be able to do 1.5 MHz PCM. Never tried DSD.

That’s the hardest part, IMHO. So many options, and you have to listen closely for differences. Personally, I liked sinc-M best with my Clear Mg, to me it added some weight and emotion compared to the default polysinc filters; I also used poly-sinc-gauss-xla for a while. With the HEDDphone 2, though, I prefer the defaults (poly-sinc-gauss-long for 44.1/48 and poly-sinc-gauss-hires-lp for higher rates), as they produce more detail and better timbre for me, which plays to the strengths of the HEDD.

Also, I’m using NS9 dithering, which sounded best to me and, according to the handbook, is great for the 192 kHz I’m upsampling to.

Definitely start with the defaults, they’re a good option. The PDF handbook of HQP also has a table with all the filters, some hints about what they are optimised for and suggestions which kind of music might sound good with them.

Yes, it’s a hassle, but it’s worth it. To my ears, the advantage provided by HQP is larger than the difference between 320 kbit/s MP3 and CD quality.

 

Appreciate the thorough reply. I'm going to give HQP a try on this laptop some time over the next few days. (Feel a mild cold coming on, so might wait till any congestion passes.) Here are some other settings suggestions gleaned from searching various forums. Surprisingly hard to find this kind of stuff. I'll check out the HQP manual and see if anything more specific there.

For the R26:

768khz is max sample rate for PCM (on the R26 itself, so no sense going higher, and I don't see much need for DSD conversion)

16 or 19 bit depth (Bit of a guess here. I think the 16-bit suggestion comes from @GoldenOne )

Sinc M filter

LNS15 noise shaper

NOS mode on the R26


Happy to be corrected or nudged on any of these. I recognize that the filter and the noise shaper are settings to be played around with, but I find I have minimal patience for this sort of thing, and sort of just end up going with a consensus opinion (if/when there is one).

 

Probably not definitive but maybe a useful table with notes on various filters:

https://audiobacon.net/2021/03/17/hqplayer-better-than-a-5000-upscaler/4/

 

Comments on filter choice for R26:

https://community.roonlabs.com/search?expanded=true&q=R26 #audio-gear-talk:hq-player order:latest

 

I have worked with HQPlayer for many years and have found it to be a game-changer for certain DACs. I have found that DSD-capable NOS (like Holo Spring) or DSD Direct capable DACs (AKM with DSDD enabled) can really benefit from HQPlayer upressing to DSD. It is a different sound than PCM - a bit less slamming, more sense of space and room, much much better imaging, better sense of "room" and "presence". I have found that DSD upressing really allows you to tailor the sound to your chain and preferences.

DSD upressing is really taxing on your system, though. A beefy GPU is helpful as is a strong CPU. For example, I am running a 13700K and an RTX 4090, and I can make it choke with HQPlayer settings if I'm not careful.

The nice part of this is that Jussi's NAA (network audio appliance) can be installed on anything and is super light. It just receives the DSD stream over the network. If you're running any VLANs or have multiple NICs running on your main machine, you will have some network troubleshooting to do, but if everything is on the same LAN, the setup is straightforward.

You'll need a reasonably powerful PC running HQPlayer to do the upressing - HQPlayer can run Qobuz directly or you can work with Roon which you can link to HQPlayer. From HQPlayer, you can choose a local DAC or send to your NAA which would be connected to a DAC. I do both.

I personally use a simple Intel Upboard SBC + nice PSU with Jussi's stripped-down NAA Linux image as an endpoint, and it works great...as does direct USB from my workstation to a DSDD DAC.

As for DSD upressing settings, there is a magic starting point. Rate: 44.1k/48k x256 depending on your DAC, poly-sinc-xtr-lp (for both 1x and Nx), Modulator: ADSM5ECv3/ADSM7ECv3 depending on your DAC. If you can get this running, you will get a sense of what DSD upressing can do. From there, you would try other filter combinations (Gauss, and sinc-M variants are very compelling), rates up to 512, and 512 specific modulators.

Upressing to PCM requires much less processing power and produces results that are more subtle than DSD upressing. There are very worthwhile changes that you can make upressing to PCM.

I have used DSD upressing with a Holo Spring, a Denafrips Ares, RME ADI-2 FS AKM, and currently a Geshelli J2 AKM in which Geno had turned on the DSD Direct functionality, In all cases, it is like having many DACs in one.

I have a Yggdrasil LIM, which I love. There are times and materials in which I prefer DSDD upressing.

It IS a rabiit hole/pain getting this all sorted, but if you're inclined, it's pretty fun and productive.

 

I will be honest, hqplayer is a game changer for me. There simply isn’t any dacs out there tbat would not benefit from hqplayer. From nos r2r dacs to multibit dacs with native dsd… heck even chord dacs benefits from pcm upsampling

I have 4 different systems at home and I have had a number of different dacs - from xonar, Cayin RU6, mojo 2, ifi neo, ifi pro idsd, holo cyan 2…

I would say the ifi dacs and the Cyan 2 have the biggest impact with dsd upsampling. The most dramatic improvement was with the Cayin RU6 tho, it literally sounded like trash without hqplayer but sounded like a hidden gem with hqplayer upsampled to 15/384!!!!

I have an Intel 13600k and a 3080ti as my hqplayer desktop pc that does all the upsampling and insert convolution filter specific for each of my 4 systems, all connected by a NAA. I use jplay iOS on my iPhone to manage qobuz and my Dlna server content.

I could not imagine how else I can set up my 4 systems at home without hqplayer really. The flexibility is insane there are simply infinite number of ways you can build your system(s) around it…

with its new NAA v5 that supports simultaneous input and output, you can easily set up for Apple streaming or Amazon hd, Spotify……

 

I agree about the flexibilty and transformation of the sound of some DACs. The RME ADI-2 FS AKM was a meh DAC in PCM. In DSDD with HQPlayer it was wild and exciting. Are you using embedded or desktop? The input/output upressing for streaming is interesting! How are you doing this?

 

I am using desktop. But with the new NAA feature, I have built a “upnp hub” with rpi4 so I can use jplay iOS upnp control app to connect to the upnp hub then sent it to hqplayer desktop.

essentially the “upnp hub” is a rpi4 in bookworm, with 1) upmpdcli and mpd as the upnp renderer, 2) minimserver to a 4tb ssd for local content 3) NAA input that that changes the rpi4 usb c port from host mode to device mode then takes the digital content from upnp renderer, sends the content along with correct sample rate information to hqplayer desktop pc through lan.

the same rpi4 can also be used as a NAA output connected to a dac!!!

For Apple streaming, you can connect an iPad as the source device by usb otg to a rpi4 input NAA and send that to hqplayer pc as well!!! Or for Amazon hd, use a laptop with Amazon windows app connected to a rpi4 NAA …

 

Ouch! This is a level or two up from my abilities/level of focus. I'll follow this and see if there is an entry point for me. I've often thought that it would be great tp use HQPlayer with the like of spotify and amazon.

 

Actually for Spotify and Amazon or tidal it would be pretty easy to do. Just burn a NAA image for rpi4 to a sd card, change the config.txt and toggle on a command line at the bottom of the .txt file.

use a laptop with Spotify or Amazon app installed, connect to the RPI4 usb c port
In hqplayer, set input device as RPI4 usb, type in the url line in hqp client app, “audio:default/0/2” press enter. Your laptop will now show “hqplayer” as the wasapi exclusive driver, which you can choose in the Spotify or Amazon app as the output device and there you go!!

 

Super cool. What toggle would I change in the config.txt file?

 

It’s at the very end. Miska had put a description there that “uncomment this line if you like to enable peripheral mode”
you cant miss it.

 

@dericchan1 Very cool, I had been wondering how to get this done.

Do you know a way without USB, network only, for Spotify? I’m running HQPlayer embedded on my Linux desktop. I’ve tried once for an hour or so to have HQPlayer use ALSA as input and make Spotify play to that ALSA device (and I could then control this Spotify instance from my phone), but it didn’t work, possibly due to Pulseaudio getting in the way or something like that—I’ve never done much with Linux sound, and it’s a mystery to me.

 

Sorry, not too familiar with Linux, but I know at least for Spotify people can do that in windows with “virtual cable” and for Mac with “blackhole”

 

Inspired by this, I just tested closed-form-M with the HEDD2, which is very nice, too. In fact, I had only tried sinc-M and the poly-sinc-gauss-long/hires-lp with it—as I’m on vacation, I hope I’ll find the time to test them all in the next few days.

closed-form-M is much fatter than poly-sinc-gauss-long. Not as clear by far, but delicious.

I also found again that it’s important to find the right synergy for every headphone—and possibly for different DACs/amps, too, if one has several? For once, I feel lucky not to be in that situation.

Anyway, poly-sinc-gauss-long with the Clear Mg is unimpressive and feels uninspired to me, a bit lifeless. sinc-M OTOH is totally wonderful. Since I have the HEDD, I never bothered to switch to sinc-M for the Clear Mg, which was a mistake, to my ears it’s SO MUCH better with sinc-M. A reason to keep it, really. Full of emotion and actually much better timbre than psgl, great voices and detail.

sinc-M with the HEDD2 doesn’t fit, though, IMHO. A bit flat, missing slam and detail, timbre fucked up. psgl in comparison is like lifting a veil, great clarity, everything stands out much more. Also better in general than closed-form-M, though the latter might be nice from time to time to add some darkness and weight. Looking forward to all the options and what I’ll find!

I’m getting the feeling that HQPlayer might be a great synergy building tool—roll filters instead of amps and DACs for different headphones? Of course, I don’t know how far one can get, but my guess is that it at least helps a lot making headphones sound better on existing gear.

 

Great to hear a little experimentation has been fruitful.

Though, ha, now I'm somehow doubly daunted by all of the tweaking possible in HQP. Tbh, I don't always hear the difference between DACs, so I'm wondering how far I can get into this side of digital audio. Given that HQP lets you no-risk trial, however, I don't see a reason not to do a little exploring in the land of impenetrably named filters.

 

For Holo DACs, I've landed on poly-sinc-gauss-xla (1x), poly-sinc-gauss-hires-lp (Nx), ASDM7EC-super, x256 SDM, which work perfectly both on i9-11900 (Ubuntu Server, v5 HQP Embedded) and on Mac Mini M1 (v5 HQP Desktop). According to Jussi, 256 is the sweet spot for Holo (he owns a Spring 2), and in any case I can't go to x512 on this hardware.

That HQP configuration sounds excellent with my current DACs (May KTE and Spring 2 KTE), amps (Stellaris, DSHA-3F, Mjolnir 3) and headphones (ZMF Atrium for Stellaris and 3F, Caldera for MJ3).

 

Jussi also has a Cyan 2 and you are correct that dsd256 is indeed “the sweet spot”. My system is capable of running dsd1024x48 but I ended up settled on dsd256 with the default filter settings gauss long for 1x and hires gauss lp for nx with 7EC super modulator.

Most of my friends who own the Spring 3 and May also set to dsd256 and never look back.