That speaks to my future. I run Roon and HQPlayer and use convolution FIR. and have plans to load that on a M1 Mac Mini.
1 Like
In my experience, Roon isnt very intensive at all, at least with more modern CPUs. I was running it on a cheap celeron processor for a time without any issues, and that was with HQPlayer and streaming tidal/qoboz as well. $150 celeron fanless computer… I’m running it now with a 10th gen fanless intel i5 with more ram and stuff because I also run HQPlayer on the same server and upscale to PCM1536 and Sinc-L (4 million taps).
I like having Roon + HQP on a quiet, relatively low power consumption mini PC because I can leave it on all the time and not really worry about anything for multiple streaming devices. If it was on my main computer, then the server would only be on if my computer was on, which isn’t all the time.
4 Likes
I have an i5 6700k. Would this be enought to run HQPlayer’s upscaling sufficiently?
Is it worth it compared to an MScaler? Or is there any other program that does things similar as I don’t need any of the integration or discovery of music by Roon?
HQPlayer has a 30 day trial. I recommend trying it out for your use case(s)!
The filter, noise shaper, and how much you upscale to all factor into it, as well as what else you may have going on at the same time on the computer.
I havent tried HQP and Mscaler side-by-side. I think Torq could answer that best, as I think he’s used both.
As far as HQP integration, it has its own music player, but the UI is pretty basic. I don’t know if it integrates with any other software like it does with Roon.
2 Likes
HQPlayer has not been on my radar. What benefit would it have in a reasonably high-resolution 2 channel speaker system? Or is it best appreciated using headphones?
I think its probably used in speaker world just as much as headphone world actually, if not more so.
It’s basically a software upscaling solution to the Chord MScaler. It uses your PC hardware instead of a dedicated hardware to do oversampling and noise shaping. All DACs (well almost all DACs) today do some sort of oversampling already, but this basically can replace that function of the DAC with your customized oversampling configuration.
It’s pretty complex and a lot of options to change how subtle things behave. You can choose between a number of filters, noise shapers, and oversampling rates and combinations of those to your liking.
I prefer the Sinc-L filter which is probably similar to Chord’s MScaler in some ways with similar order of magnitude number of taps – its between 1-4 million in my use cases.
4 Likes
It has wake my curiosity, but I still don’t understand how it works, if the DAC is doing oversampling already, how HQPlayer improves the sound from the DAC? I mean the signal still is gonna go through the DAC, does it work with just specific DACs? I have Roon, so I am curious to give it a try but my DAC is nothing fancy, it is a Sabaj D5 and I also have a Modi Multibit, I mainly use headphones, should I run HQPlayer on my client machine or it doesn’t matter if HQPlayer is on the Roon Core and I use Roon from my client machine? Sorry if this is out of topic.
Some DACs, like the new R2R DACs have NOS mode (non-over-sampling), so it doesn’t do anyting except play the sample rate it was given. In this case, HQPlayer’s upscaling is pretty straightforward.
In most delta-sigma and even Chord DACs, you have to feed it something with a sample rate of max (or near max, depending on the original sample) to get it to not perform its oversampling (since there is nothing to oversample at this point).
So, when I use my Chord Qutest, I tell HQP to upscale to 768 or 705.6KHz and then send that to the Qutest, since it’s first WTA1 filter does a 16X upscale to 705/768. Now that said, Chord DACs may do an additional upscale as well (WTA2 to 256FS) which I can’t override easily, except by changing the filter button to ignore that second stage.
Anyway, I hope that makes sense.
2 Likes
Oops forgot to answer the 2nd part!
It doesnt matter where you run it to be honest, except that it should be what your DAC is connected to, unless you are using a Roon NAA (like a raspberry pi with NAA installed).
1 Like
Thank you, I get it, I may play around with what I have to see if I notice any improvement, I have a Mojo (to satisfy my Chord taste), I am guessing since my stuff is on the low end it might not work well for me. Thank you for the reply.
I guess I got it working, I guess now is a matter to play with the settings and see (hear) if I notice any improvements. I set the sample rate to 768 since my stuff doesn’t have/support NOS. Thanks again @antdroid
1 Like
I know is too soon but I think I like what I hear.
1 Like
Glad its working. If I recall, Mikasa (his username on various forums), owns a Mojo and uses it with HQP. Of course he has a bunch of other DACs too. I also own a Mojo as well. 
One little strange thing with Mojo and Qutest, and I think other Chord dacs is if your PC can’t handle the upsample to 768 flawlessly all the time, it can cause one channel to go static randomly. Once I upgraded my computer, I didn’t have this issue anymore though.
3 Likes
I do get it as well. For my 2 cents - I have a Qutest (I ran a Mojo for many years) into a Drop 789 (as my Phonitor sits in some boat at the California docks…) into HD880S. My setup can be very revealing of poor recordings, The HQPlayer’s upsampling and FIR filtering just makes things sound better (harshness is clarified into smooth high end, etc, and isn’t that the point with all that we do?
4 Likes
I have a Qutest as well and may have to give this a try
1 Like
If you have a DAC that benefits (not all do so to the same degree), can run HQPlayer at the settings you need to get a tangible, reliable, advantage, and don’t mind having (and maintaining) either a PC or another device (e.g. you can do some levels of HQPlayer upsampling/filtering on a Raspberry Pi 4), then maybe.
It doesn’t yield exactly the same result as an M-Scaler, though the effects are similar, latency will be the same (with PCM material, that’s about 1.4-1.5 seconds for the “equivalent” upsampling/filtering), and it’s a more complicated setup, but it’s also $250 vs. $5,295.
And since there’s a free trial, it’s worth experimenting with.
I’ve found it to be more useful, and more readily apparent, with good NOS DACs than with anything that over-samples (assuming it is using an appropriate length, linear phase, filter).
You might take a look at PGGB (aka “Remastero”). This does something similar to HQPlayer, or the M-Scaler (again, not identical, in either case), using 16fs upsampling and extremely long filters, but works by pre-processing your files and generating a new file that’s been upscaled and had all the necessary filtering applied.
This means that there’s no playback latency, nor any processing overhead for the machine sending/playing the data. In actual playback the resources required are no different to a regular 32-bit 705.6/768 kHz PCM file.
The trade-off is that your files get a lot larger. At its most intensive settings, for 16/44.1 PCM input, stored as FLAC, then at the highest settings, files grew by 64-70x.
It also has some other functionality beyond just raw upsampling and filtering.
The chap that created it was great to chat with, very friendly, absolutely no-pressure and made evaluating it a very enjoyable experience.
Again, there is (or was at least) a free trial.
A while back (I posted a littler about it in another thread), I took advantage of the fact that PGGB and HQPlayer are purely software solutions. This made it VERY easy to do proper, blind, comparisons between processed and non-processed files using standard tools.
You just generate PGGB and HQPlayer* processed outputs, put those, along with the original .WAV or .FLAC files into something that can do a proper ABX test (e.g. Foobar2000 with the ABX module) and see what you come up with.
It’s very nice being to do a proper test like that, solo, and be confident in the results - it’s almost impossible to do that with hardware comparisons.
My net result was that about 1 in 20 of the PGGB processed files (5% of my test track set) was reliably, discernibly (to me, in my rig) different/better into just DAVE than vs. M-Scaler->DAVE. I got similar results with HQPlayer, with high-correlation as to WHICH tracks benefitted.
Ultimately I decided not to use either HQPlayer nor PGGB in my headphone setup. The reasons for this are as follows:
-
I could only tell differences at all when doing very focused, audition/review-style, listening. I don’t listen to music that way when I’m listening for pleasure. In normal, relaxed, had a glass-of-scotch, listening for pleasure, I wasn’t aware of any differences.
-
When I could tell differences/improvements it was only about 5% of the time.
-
With HQPlayer, I didn’t find those differences/improvements to be worth having another machine to manage nor did I want the additional system complexity.
-
With PGGB, processing my entire library would have required spending about $500,000 in storage to accommodate. And my library is just too big to find the 5% (average) of that which I found benefitted to only process and store those files.
Other caveats:
If I didn’t already have an M-Scaler, my perspectives would probably shift a bit and I would likely either run HQPlayer or, if my library was smaller, I’d do PGGB or HQPlayer Pro (since once processed, it’s no change in the system for replay.)
If I was using a non-Chord DAC, especially one that does less intensive upsampling and filtering, I’d be more interested.
If I was using a NOS DAC, I’d definitely be using one or the other for the times when I didn’t specifically want the NOS presentation (with its latent issues). This is a change from a couple of years ago, where if I wasn’t going to run NOS and keep it NOS I just wouldn’t have gone with a NOS DAC in the first place.
At no point did I hear “night-and-day”, “huge”, or “not-subtle” differences, even against a raw (no-M-Scaler) DAVE setup**.*
TLDR; The benefits were not big, nor frequent, enough for me to do this vs. my existing rig. Your mileage may vary. It’ll only cost you some time to test them for yourself. And if you’re diligent about it, you can do so in a proper blind fashion.
*To do this, you’ll either need the “Pro” version of HQPlayer (which is about $3,000) or do the work to capture the output sample data, as the “Desktop” version does not let you pre-process and save the output.
10 Likes
Thank you for the well laid out and detailed post of your analysis.
I apologize for not responding back when you originally posted as it was not something I was ready to dabble in. Definitely not PGGB.
Well, that day has come. I was able to try it out in Roon and can tell a difference. It is thicker and smoother. Hard to describe, fuller, less jaggy in peaks of sounds? Are these transients or ringing?
I am still debating on whether it is worth $250 to not be annoyed by the 30 minute trial. I am in the position that I could afford another $250 rather than a used $4k HMS though. Plus, my source is already my pc that is capable. Just weighing if it is supposedly technically better than the upsampling being done in my X26 Pro. If it is the superior way, then I know what I must do.
I am contemplating setting up a Roon Core but I am on a very limited budget. I will not need to stream anything apart from Tidal or at least very few FLAC albums. So I guess this does not require a very powerful setup. I was thinking getting a 2nd hand NUC or something similar. But how low can I go in regards to specs?
I would like to stream to a RPi4 connected to my BF2 via either Volumio or Ropieee.
2 Likes
There’s a lot of good information in the Roon forum. This might be helpful: Which Intel NUC to choose? - ROCK - Roon Labs Community
3 Likes
I was looking at that just a few weeks back. Seems like the Nuc 8 with i3 is about the furthest back solution that it’s wise to go. Support is still quite active for that version.
I succumbed to temptation and am running ROON on my new MacBook Pro 14…
4 Likes