Setting up and RC'ing an 5.1 audio system using a home theater processor.

[scintilla]

Hi, I recently added XO to my stable of DSP tool to enable a multichannel correction in HQPlayer for my 5.1 channel ripped music files through a Monoprice Monolith HTP-1. Since that unit does not allow direcly addressing multiple subwoofer channel inputs and I will be using two discrete subs, I am wondering the best approach to building filters. Should I employ the calculated delays and subwoofer management in the HTP-1 but leave Dirac off, use Dirac DLBC below say 100 Hz to manage just the subwoofers or let Audiolense level the bass and set the delays. The subs do have different sensitivities and one is located front right and the other rear left. I am sure there is a best approach here to get the best response. I expect XO is going to yield a lot better definition than Dirac for full-range correction above the bass-steering thresholds. My goal was to improve the mutichannel correction and provide 5.1 input to the HTP-1 at 24/192kHz since most of my multichanel music files are DSD64 and 24/96 PCM and I would like to retain the highest resolution without having to decimate fot the HTP-1 DSP which is limited to 24/48 for processing.

 

[juicehifi]

That was a complex setup and I am not sure I understand if there's a question there.

 

[scintilla]

Just a little followup on my testing:

Here are the sets I measured and built filter sets for:

1. Dirac off, 5.1 channel, front sub only with rear sub disabled. The purest.
2. Dirac bypass using distances from a Dirac measurement and stock bass management for 2 subs, front and rear. The purest with bass management.
3. DLBC on with a flat-bass Harman curve with a turnover at 1kHz to -6db at 21kHz. Double DSP.
4. DLBC on with a 6.7db bass boost bump at ~30Hz and then flat to 1kHz and to -6db at 21kHz. Double DSP with boom.

Measurement sweeps were run through the HTP-1 from Audiolense with the above enabled or disabled and with all other DSP functions disabled (wide synth, dialog enhance, etc.)

The corresponding measurement sets were used to build 4 filter sets in Audiolense, with the 10db LFE boost option disabled in the speaker configurator of Audiolense. I wanted baselines and I can add per-channel gain in HQPlayers matrix.

I played with frequency-dependent windowing to improve the group delay of each set until I was generally happy. I did them en-group rather than individually which might yield better results. Audiolense is a lot more approachable than Acourate and automates most functions. Since I have a lot of experience with Acourate, I found the transition to Audiolense workflow seamless and easy to learn and implement.

There may be some hand-wringing over applying two correction curves to input data, but Audiolense is capable of time domain correction whereas Dirac is limited to a mixed approach using FIR and IIR filters. Therefore, using an Audiolense filter in front of the DLBC filters should provide better time domain behavior.

All filters were doubled-length 131k taps, for no reason other than I could. Filters are 64 bit float at 96kHz. I use HQPlayer for convolution and dithering to output 192kHz/20-bit for the Dirac-off and Dirac-bypass sets and to provide 24bit/48kHz for the DLBC filters, since the HTP-1 decimates all inputs to 48kHz/24 for DSP. The HQPlayer remodulation is much, much better than that offered in hardware on the HTP-1, particularly for DSD source data. Some channel re-mapping was necessary to accommodate the HTP-1, since the sub and center channel are reversed based on my computer's Audio Midi interface.

So, which one is the winner: 3. DLBC flat-bass. So far. In initial listening, to the same content across the filters, I was struck by the quality of bass delineation and definition. DLBC does a great job with bass, cleaning up a lot of mud and providing more bass than without. When listening to the pure AL filter with Dirac off, I expected the best result, especially because I limited that to one sub up front. The time domain corrections should be as ideal as possible in that scenario. However, after switching between the filter sets, I preferred the Dirac bypass more and the DLBC flat the most. The combination of DLBC and Audiolense time domain correction is the secret sauce for truly amazing clarity and definition. One cut from the Al di Meola recording Flesh on Flesh, is positioned in an enormous sound field that is swirling with reverb and places you in a cavern. The immersiveness is palpable. I had hoped that stacking the DSPs would provide a benefit. It does. Dirac is a very good RC system and has a great user interface, but it is not actually state of the art. For that you need the true time domain correction afforded by Acourate or Audiolense. And Audiolense provides an easy entry to multichannel correction, where Acourate does not. If you have to be on the bleeding edge, this is the way. It will hopefully be more-so if we get ART. But for me, I have done enough and will live with these filters, changing periodically to test my perceptions. For my large collection of nearly 600 audio surround recordings, this gets me as close as I can get to ideal with this hardware.

 

[Ofer]

Very interesting. Just so my simple brain could understand. In the signal flow. First you do AL TTD on all channels and subs and then you employ DLBC on all channel and subs?
Did you do independent measurement with REW of the before and after? Was this configuration also better for video?