Ah, understood. But, for a 2.2 system, Audiolense XO will only produce the bass crossovers if there are 4 DAC channels, right? So, it won’t work as you describe when measuring with just a 2-channel DAC. That’s what I want to solve, and I’m not sure how to do so without setting bass crossovers manually in HQP or ahead of HQP in Roon. JCR
Yep correct, AL needs to “see” those bass driver DAC channels to individually correct them and DSP crossover them with other drivers
You need 4 channel (minimum) DAC .
OK, we are now on the same page. Seems that if I do want bass crossovers but will use only my two-channel DAC and no preamp to do the crossovers within, my choices are to manually set some high pass/ low pass filters in either Roon or HQP. In either of these cases, Audiolense XO will see two full-range speakers for which to derive DSP filters.
So, back to my original question: how would I set bass crossovers manually in Roon or HQP? @jussi_laako, perhaps you can take the HQP side of this question. Thanks, all. JCR
And HQPlayer allows you to use any combination and number of convolution filters and parameteric EQ/IIR filters for any of the pipelines…
You disable Convolution engine in HQPlayer and use the pipeline matrix instead. There for each virtual bus (pipeline) you choose source channel, set of filters you want to apply to that bus and then output channel of that bus.
So you can have four pipelines for this, assuming single sub:
1 → main left channel process → 1
2 → main right channel process → 2
1 → left channel sub process with mix gain: -6 → 3
2 → right channel sub process with mix gain: -6 → 3
Thus in this process you have left and right channel mixed for the single channel sub output. If you have two subs, you use four channels instead without mixing.
Then for multichannel DAC you set number of output channels to 4 in HQPlayer Settings (with ASIO you could technically do also 3 channels, but not all drivers support such).
Just as a thought… Is is possible to use separate DAC’s for the channels. Eg. one topping E30 for high-pass and one topping E30 for low-pass?
Depends on the OS, whether you can create an aggregate device. But the problem is that two devices are both clocked by their own clocks and thus eventually run out of sync with each other. In order to create reliable multichannel, you need to have all channels clocked out of single master clock.
Best solution is to use a multichannel DAC. Options are for example Merging Hapi and exaSound s88 or e68.
Now that looks like a nice piece of kit
Ahh yeah, and it’s the pro-side product. The audiophile sorta equivalent is 8-channel version of Merging NADAC.
I am considering HQplayer with Roon or as a standalone player, with a NAA endpoint that is already in place.
My system is a 3 ways multi amplification with digital crossover + room correction made using Acourate. Today, the convolution is done by Roon with an exaSound 8 outputs Dac.
I read that PCM conversion to DSD 256k requires a strong CPU. What will be the impact of the crossover + room correction implemented in the matrix? Will it require 3 times more CPU (convolution to produce 3 stereo ways and then 3 PCM flows to convert to DSD) or slightly more CPU only (1 PCM flow to convert to DSD and then convolution to produce 3 stereo ways)?
Convolution is performed at the source rate, before PCM-to-DSD conversion. So if the source is for example 44.1k PCM, it is not particularly heavy.
But what is heavy is doing multichannel PCM-to-DSD conversion. On this kind of setup, using a powerful GPU can help by splitting the processing efforts.
It’s hard to know if you need a more powerful CPU without knowing what you have now ?
And what OS is that CPU running on?
You can run the trial of HQPlayer for free
There are some really heavy filter/modulator combinations in HQPlayer and there are lighter ones.
With the free trial I would try these filters:
Best is if you can run HQPlayer OS trial - it boots from a USB stick. Then you need to run Roon on a seperate machine.
Burn this to USB stick using etcher (you need to unzip it first):
Thanks for your two answers. So I understand that convolution is done first (2 PCM voices to 6 PCM voices) and then conversion to DSD (6 PCM voices to 6 DSD voices). Thus I understand that it will be CPU or GPU demanding since 6 mono flows will be processed.
I performed a first test this evening with HQP installed on my Windows 11 NUC 10i7 (small computer, 6 cores with HT but low consumption oriented) with proposed settings. I used HQP in standalone mode (not with Roon). I had some problems to build the library with the 1 300 files stored on this PC (WAV format), I need to work this out. The exaSound Gamma Server+e68 was found properly (NAA endpoint). The pipeline matrix is setup with filters at 354.x kHz as suggested in another discussion.
DSD conversion is OK with 3 or 4 cores loaded at 30 or 40% at 3 GHz with 16 bits - 44.1 kHz files. I will push further in coming days to confirm CPU load and test 24 bits up to 192 kHz files and play with filters to analyse CPU load.
Is that HQPlayer Desktop running on Windows or the HQP OS image I linked?
Yes, it is HQPlayer Desktop running on my Windows 11 PC (trial mode). I had some problems to retrieve and navigate in my library using HQP Client (but I am not very familiar with these 2 new interfaces Desktop and Client) and some metadata are sometimes missing (I.e. track number).
I will move to the OS version if my findings are positive (SQ and DSD processing) but with some hardware to be sized and purchased.
Ok keep us updated.
At least you were able to get music playing without any stuttering.
With the settings I suggested if PCM44.1k worked fine, then PCM192kHz will also work (Nx = poly-sinc-gauss-hires-lp)
The exeprience is doing quite fine on my NUC on Windows 11. I temporarily set it to Maximum perf but without many improvements.
ROON server is running on the exaSound Gamma server. Resulting network traffic is Gamma/Roon – 2 voices → NUC/HQPlayer Desktop – 6 voices–> Gamma/Roon → exaSound e68. This will probably be my target setup if I move to HQP, but with HQPlayer OS.
With proposed filters and modulator, I can play from 44.1 kHz/16 bits up to 96 kHz/24 bits. Average CPU is between 70% to 90% with a variable speed that may be related to a variable processing demand. I tried ASDM5 and ASDM7 modulators, they both are doing well. But I get some cuts with 192 kHz/24 bits files.
ASDM5EC, ASDM5ECv2, ASDM7EC and ASDM7CEv2 modulators raise CPU at 100% with regular cuts as from the beginning of the file.
ADMSDM7 512+fs is OK up to 96 kHz/24 bits but NUC is at 100% above with regular cuts too.
Overall, SQ is improved in terms of space and quality of timber. A more “analog” sound, quite “cool” but very detailed. Interesting experience!
Dropouts with Nx=poly-sinc-gauss-hires-lp like I suggested above?
Yes, using this proposed filters.
But I forgot to mentioned that testings were done with Bit rate at 44.1 x 256 as it is the maximum rate accepted by the exaSound e68. I will performed new tests with 128 or 64 rates.
One simple question: does Hyper Threading provide boost to HQP or should I set bios without HT?
You should leave it enabled. HQPlayer knows what to do with it.