Currently testing HQPlayer. What would be recommended setup with a Chord Qutest DAC?
Donât⌠No need to upscale or filter⌠The Qutest does everything internally itselfâŚ
Try upsample everything to 705.6/768, with Sinc-M, LNS15
Send 705.6/768k PCM output from HQPlayer always (also for DSD content), âDefaultâ as DAC bits and LNS15 noise shaper. Choice of filter is up to your preferences.
Point of HQPlayer is to do it better and bypass the WTA1 filter in Qutest to improve itâs performance. So of course there is need to upsample, by doing it better than Qutestâ internal stuff.
Thanks, I just want to have a look.
Will try some of the recommendations below and try to compare but I agree the Qutest sounds already very well.
Iâve been running SINC-M/LNS15 to Chord DACs to a while (using SRC-DX so i can use the DBNC inputs). Compared to anything else HQP, and even the latest SINC-S, there is a certain smoothness and increased depth with SINC-M âŚand the 15th order noise shaper helps a lot as well (thanks @jussi_laako) . Definitely go with SINC-M to match the sonic character of WTA1.
Think of it as adding an M scaler but at a fraction of the price
sinc-S was result of my own listening tests, to me it is the sweet spot. And since is scales it behaves the same at all input/output rates.
Maybe this is the thing, because Iâm not comparing to WTA1. I never listen to it, I just do my own things. I donât see any reason why WTA1 should be used as any kind of reference to compare against. I would just ask you to break free from WTA1.
[Moderated] Hey, its not the Taps, it is the Watts Transient Aligned Filter that makes or breaks small signalsâŚ
[Moderated]
No, itâs the algorithms that are used to design HQPlayer filters that really can accurately reconstruct the original signal, even very small details.
[Moderated] We can discuss filter performance in objective terms of course too, if you like.
Now I would like to hear a technical explanation from you why do you think WTA1 in Qutest is better than HQPlayer filters?
Because a single Field Programmable Gate Array (FPGA) chip can do the million taps and filtering without excessive CPU usage or excessive heat generation⌠Because parallel vector software algorithms are hard on PC hardwareâŚ
FPGAs are very power inefficient way to do processing and generate a lot of heat. You can already check how hot the Chord DACs run. In addition, none of the Chord DACs have a million tap filter, only M-Scaler does.
Sounds like you have not even tested what kind of CPU load you get with a million tap filter to the same rates as Chord does. Here is HQPlayerâs sinc-M million tap filter doing 16x to 705.6k output on my ultralight laptop with Pentium Silver N5000 CPU (essentially four Intel Atom cores) that doesnât even have a cooling fan (6W TDP):
HQPlayer takes half of one coreâs processing capacity worth of CPU time. So this computer could do much more. Now compare this 600 EUR laptop to the M-Scaler doing a million tap filter. I can run the filter and browse the web at the same time.
HQPlayer can do million taps to the 16x rate (705.6/768k) with very light CPU load. HQPlayer can also do million taps to 256x rate or higher. Or 16 million taps for example to 512x rate. And run delta-sigma modulators, digital room correction, headphone processing etc at the same time.
Not for me, Iâve been doing DSP software on PCs for nearly 30 years now. We have SSE4.2, AVX, AVX2 and now also AVX-512, on each CPU core.
In addition, HQPlayer supports using Nvidia GPUs for processing too. These have thousands of DSP cores available. When you combine these with the latest CPUs you have nice processing powerhouses.
At the price of M-Scaler you can get very powerful PC with a powerful Nvidia RTX GPU and in addition to playing music you can simultaneously write to this forum with it. You cannot do that with M-Scaler.
If you look over at the mscaler forum on head-fi most of the talk is about mscaler induced rfi. With HQP playing over to ethernet to an naa⌠rfi gone. Added cost $70 for a pi4. Mscaler solution is $1500 cables. The mscaler does match beautifully with other Chord dacs and is very stable but really is Chord centric.
Yeah for years Rob recommended ferrites on BNC cables with the Blu2, saying they can not harm sound quality, they can only help with RFI.
Now with latest M-Scaler model he says ferrites on BNC cables can harm sound quality.
Confusing.
There is so much FUD in here it just does not deserve a reply or correctionsâŚ
A bit late in the day here but do we collectively accept the mathematical premise put forward by Rob Watts that just over 1M Taps is sufficient to reconstruct the digital signal âcompletelyâ (to I think -96dB accuraccy) as per Nyquist/Shannon?
If we do then how one achieves 1M taps - HQP, Blu2 or whatever - becomes a cost/aesthetics/ease of use argument, not an SQ one
Declaration of interest - I own a Blu2/DAVE combo (with ferrites!!)
I donât know how he comes to such conclusion. But HQPlayer can do -240 dB with much less than 1M taps.
It is far from being that straightforward. Filter itself matters a lot, shape of itâs roll-off, whether it is apodizing or not, minimum- vs linear-phase etc. And once you have settled with filter, then in next stage the used modulator/noise-shaper/dither matters a lot too.
So a question if I may @jussi_laako ?
I use HQP to upsample audio and use specialised filters etc etc etc to produce a bitstream that we can call âoptimisedâ
I pass that bitsteam to NAAD residing on some device and NAAD acts as a presumably transparent bridge to my DAC via USB, S/PDIF whatever
My DAC will have itâs own internal filters which, unless defeatable, act on that bitstream as part of the D to A conversion
Assuming said internal filters are âsub optimalâ relative to HQPâs, do I not now lose the âbenefitâ of that originally âoptimisedâ bitstream?
Apologies if that is a naiive or over simplistic question !!
For a long time I was troubled by exactly the same issue. I was thinking of filters from my analog experience, where filters were optional subtractive (usually) manipulations of a signal and later equipment has no way of telling whether a filter has been applied or not.
But inside a modern DAC the filters are used when upsampling an input signal to a higher sampling rate before being sent to the modulator. The DAC can tell what the sampling rate of the incoming signal is, and if it doesnât need to upsample the signal then that stage is just omitted. So the idea behind HQPlayer is to feed the DAC the highest sampling rate it can take and bypass the internal upsampling filters.
Jussi gave an excellent comparison between HQPlayer and typical DAC here.