New loudspeakers & room digital correction service

[Jud]

4 minutes ago, fresponse said:

It would be interesting to have measurements of your loudspeakers because I have never seen pure linear phase ones.

 

I might be able to provide that to you for free.

 

What types of measurements with what sort of hardware and software would you be looking for?  (It may be the case I don’t have them available, but I thought I would ask.)

 

4 minutes ago, fresponse said:

Nevertheless if there is no phase issue to correct then the DSP will not change anything and you should not have any bad impact on imaging. By curiosity which type of software have you tested ?

 

iZotope SRC 64-bit included with Audirvana Plus; HQPlayer; Audiophile Inventory; SoX.

 

[pkane2001]

13 hours ago, fresponse said:

Just to be clear when I said "optimal" I didn't mean "flat" : as you said optimal refer more to what will sound natural in your listening environment. Looking forward to your measurements !

  

Understood, that's why I said 'optimal, flat', meaning optimal and flat, as opposed to optimal and (house-)curved, which I may want to try later

 

Just sent you the measurements. Looking forward to trying out the filters!

[fresponse]

@Judto assess the phase behaviour of our loudspeakers you need to make an impulse response measurement of one of them at about 1 meter distance. It is better to move it away from room bondaries to ease the removal of room reflections

 

@pkane2001Files received I'll keep you updated by mail

[sixman]

So has anyone tried this? Has it worked out?

[pkane2001]

On 7/28/2017 at 1:02 AM, sixman said:

So has anyone tried this? Has it worked out?

 

I'm in the process of testing. I'll post my findings soon.

[pkane2001]

Impressions

Over the past month, I had a chance to evaluate the service provided by Thierry of HomeAudioFildelity. Thierry offers to create custom convolution filters tailored to your own system and listening space. Two sets of filters are available, depending on your price preference: one with the ability to reduce cross-talk, and the other without this capability, but still tailored to your system to improve phase and frequency response.

 

https://www.homeaudiofidelity.com/

 

(Please note: any the theory included here is based on my understanding. If anything is misstated, the fault is entirely my own)

 

I was particularly interested in the cross-talk reduction version of the filters. I’ve been a long-time user of REW and manually created and optimized convolution filters that have been in my system for a long time (way before I even had a PC-based system).

 

Cross-Talk

What is the cross-talk reduction and why is it important? Well, the theory is that recorded stereo sound when played back over stereo speakers causes mixing, or cross-talk, from left to right channel and from right to left. This is unnatural, since ideally anything recorded in the left channel should only reach the left ear, while anything recorded in the right channel should only reach the right ear.

 

This clearly does not happen when playing music over speaker systems, since some of the sound from left speaker reaches the right ear, and some of the right channel sound reaches left ear. The theory goes that this produces a less natural sound, with less clarity and a more confused soundstage image. Note that this does not happen when listening to music over headphones! There are much more complex solutions that have been proposed to solve this problem, but most require a very tiny sweet spot, or involve complex head tracking algorithms that are as expensive as they are hard to implement.

 

For this reason, when Thierry offered his service, I jumped on it. While his filters promise some cross-talk reduction,  all of cross-talk cannot be completely eliminated with convolution filters.

 

Conclusions

Thierry was very helpful, patient, and provided great support during my trial. I can highly recommend that you try his service, especially the free evaluation he offers. I do suggest that you pick music that is very well recorded, with minimum amount of processing and good spatial cues, as that seems to highlight the effect of the cross-talk reducing filters the best.

 

The filters worked in my system after a few false starts. The final version of XTALK reducing filters I settled on produced a more natural sound for voices, acoustic, and other well-recorded, minimally miked music. The result was a wider sound stage, and better focused, cleaner instruments and better sense of space. The sound extended further out and away from the speakers than with my regular filters. In fact I almost never heard the sound coming directly from my speakers with these filters.

 

The effect was obvious but not overwhelming, I’d say about 15% improvement compared to what I’m used to with my system in the above areas. By way of comparison, I find that USB cables produce no difference in my system, a better DAC might produce a 10-20% improvement, digital and analog interconnects mostly have no effect. So, a 15% improvement is actually significant in relation to other things I could try to improve

 

Large scale, well-recorded choral performances did benefit quite a bit from these filters: individual voices became easier to distinguish and locate in a large group, where previously they sounded more like a large jumble of voices.

 

Sounds of a large orchestra or synthesized or heavily produced rock recordings did not seem to benefit as much, but the filters never made things worse.

 

The non-cross-talk reducing version of the filters turned out to be very similar to the filters I constructed myself, so there wasn’t much reason for me to use these. They might still be very helpful for those who don’t know how to create your own filters by measuring the room response and optimizing phase.

 

Test System

Windows 10 PC with HQPlayer performing convolution, upsampling everything to DSD512, feeding Singxer SU-1 DDC connected to Gustard X20 Pro using I2S connection.

 

I used REW and a Behringer calibrated mic to create the initial measurements required by Thierry’s service.

 

Testing

Testing and evaluation was done by creating separate filter configurations in HQPlayer, one using my own filters, one using the XTALK-reducing filters, and one without the XTALK reduction. I carefully matched levels between all configurations to ensure a fair comparison.

 

I did a sighted evaluation for a while, recording notes on each filter set and the effect I thought I heard. I evaluated a large sample of music, from vocal to rock, jazz, and classical, orchestral and chamber, and various live recordings.

 

I then did a random blind evaluation. Between my original filters and XTALK-reducing filters, I could correctly identify which filter was playing about 90% of the time (3 sets of 10 trials each). For each of the 3 sets I focused on a specific characteristic of the filter I found significant, using a very short piece of music that I thought illustrated that difference the best (wider, more natural sound stage, more focused singer’s voice, better voice separation in a choir). I’m satisfied that the differences I found were not imagined.

 

Overall, I’m happy with Thierry’s cross-talk reducing filters, they’ll stay in my system, replacing my own. I recommend you try them yourself, but of course, YMMV!

[Jim Smith]

Does Thierry offer a service wherein he comes to your room to evaluate the final result, so that he confirms that his program is delivering as well as it can?

[Kal Rubinson]

5 hours ago, Jim Smith said:

Does Thierry offer a service wherein he comes to your room to evaluate the final result, so that he confirms that his program is delivering as well as it can?

Of course, I doubt it.  OTOH, it might be interesting if he accepted new post-EQ REW measurement files for assessment.   

[fresponse]

Such in situ service (measurements / calibration / validation) would be possible in theory but certainly at a prohibitive cost to cover travel expenses. Validation with post EQ measurements is of course possible but usually the results are very similar to the simulation and this is why I considered it as optional.

 

[Fyper]

Post EQ measurements would show objectively the effectiveness of the suggested correction, and could be the starting point of an iterative correction process if the customer so desires.

Not that optional for a remote service IMHO.

[STC]

On 6/22/2017 at 5:26 PM, fresponse said:

Those convolvers are not compatible with our technology because there are of "one to one" type : this means that left channel is convolved with a filter and send to left output. Same goes for the right channel. What we need is left channel convolved with 2 different filters with respective output sent to left & right channel. This is why I mentioned the butterfly plugin which offer such flexibility (a filter set = 4 different filters)

 

I am using SIR2 with JRiver with 16 stereo IRs (i.e 32 channels fed to 16 speakers) and it is capable of true stereo convolution. 

[fresponse]

My mistake, I did not understood that "true stereo" was actually meaning a 4 impulses filter capability

[STC]

6 hours ago, fresponse said:

My mistake, I did not understood that "true stereo" was actually meaning a 4 impulses filter capability

 

Yes. You worded it better. I meant to say with SIR2 I use 16 distinct stereo impulse response ( 32 different IR channels) . It is good to see a product with both XTC and convolution. Congratulation!

[SwissBear]

I spent quite a bit of time studying these topics of active correction. I started with a Trinnov Amethyst, tried Dirac, then Acourate and finally REW/RePhase.
Only this last solution gave me total satisfaction because it allows for making very clean corrections, with minimal phase, easily portable to all players ...
May I recall that a well-done active correction allows:

1. A guarantee of perfect tonal rendering, ie a reproduction of the sound matching the recording, by correcting the interactions of the speakers and of the room
2. Faithful reproduction of the sound stage by correcting the phase errors essentially introduced by the speaker filters

Thierry has developed a more refined technology. Compared to a traditional correction, which only corrects the direct field of the speakers, Thierry brings two complementary elements:

1. Correction of the reflected field; This correction ensures a faithful reproduction of the sound which takes into account the reflected field created by the walls, the ceiling, etc. The impressive characteristics of this addition is the homogeneity of the sound reproduction throughout the room
2. A crosstalk canceling that allows a much wider sweet-spot

In total, the corrections produced by Thierry have many advantages:

1. They are just as easy to use as other manufacturer's 'ready to use' corrections.
   Just make the measurements and Thierry takes care of the rest
2. They give, with respect to standard corrections (with two pulses, as produced by Acourate or rePhase), a complement of soul to the music. The perfect three-dimensional reproduction of the sound stage gives the illusion that the orchestra is in the room. It is both magical and strongly addictive ...

Congratulations to Thierry and my best wishes to him for success

[Danold]

I got interested in Tierry's method but couldn't find much of a manual or detailed instructions on the website. Did I miss something?

[fresponse]

Danold, if you're question is about how to make required measurements please find below some instructions.

 

As an introduction to REW software, there is a good and simple tutorial available here  https://www.minidsp.com/applications/acoustic-measurements/umik-1-setup-with-rew

 

Once you have set the level (by using left loudspeaker for instance) you'll have to keep it for all the different sweep measurements.

 

From a practical point of view I suggest to follow this procedure:

- set the microphone at the sweet spot position

- set the level using the left loudpseaker

- perform the sweep measurement for this position & left loudspeaker and rename it L1

- perform measurement for this position but right loudspeaker this time, rename it R1

- move microphone to the next position & perform the measurements for right then left loudspeaker, name it R2 & L2

- repeat the procedure for the others different positions (I recommend sweet spot + 4 others as a minimum)

- save the bundle of measurements as a .mdat file

 

Hope it will help,

Thierry

[Celts88]

Thierry

 

I really like the idea of using your expertise via HAF to produce the required filters, and then it should be simple enough to upload into Roon.

 

I haven't done any sound measurements before (I do have a copy of REW, and a UMIK-1 mic), but am worried that I may not do the best job of taking the measurements, which would limit the effectiveness of the filters you can produce.

 

Is it easy enough for you to see if the measurements have been carried out satisfactorily, and if not advise to take new one's, without having to go through the expense of having to pay twice for a better set of filters (not saying you wouldn't advise, just a genuine question from a newbie).

 

Dumb Questions:

 

1. Where do I get the Pink Noise file for generating the sound

2. Does it make any difference to taking the measurements if having a 2.1 set-up (2 front mains, with a Subwoofer in the mix)

 

Cheers

John

 

 

[fresponse]

Hello John

 

It is pretty easy to make good enough measurements and it is also easy to check them. Before building a correction I always make a quality check.

Regarding your specific questions

1. It is REW that will generate the test signal for impulse response acquisition  (a sweep signal actually)

2. For a 2.1 system you have to measure left channel + sub and then right channel + sub for each location

[Celts88]

Thierry

 

Appreciate you getting back to me.

 

Next week I'll pick a music sample (unsure which will be best), and send it through for the free trial. Not that I'm doubting the HAF correction filters won't work, but will give me an idea of the potential change in sound from the original to your optimised version.

 

Then I'll carry out the measurements via REW (want to take my time and get these right) and can then send through for the full version (most likely will go for the cross talk correction option, then getting the best of both worlds).

 

Cheers

John

 

PS Dumb Question Alert - do I need to carry out any measurements via REW prior to sending through my music file for the free trial

[fresponse]

Yes John you need to make measurements as I generate the appropriate filters for the free trial.

[Celts88]

8 minutes ago, fresponse said:

Yes John you need to make measurements as I generate the appropriate filters for the free trial.

 

As I said, dumb question alert - will do the measurements next week.

 

Have a good weekend

[Celts88]

Thierry

 

Ok, so I gave it a try doing the measurements in REW - 1L & 1R sweet spot, and another 7 each L & R. Still not sure if I've done them correctly, but for the trial song(s) it will give me a good idea of what your filters can achieve.

 

https://www.dropbox.com/sh/g3az7aigjf8deg0/AADuXkqpr-BBPsauK53N3jC0a?dl=0

 

FYI I normally sit offset from the middle of the speakers and have actually did the measurements with me being offset (1/3 to the left side of centre, i.e. 0.5m from the left speaker axis, and 1.8m from the right speaker). Would you prefer if I did the measurements sitting exactly in the middle? (not where I normally sit)

 

Also could I ask a big favour. I've actually linked 2 songs, therefore would it be possible to do your magic on both of these (they’re a bit different from each other, so gives me an idea with differing types of music). If only 1 song possible then that's ok.

 

Don't worry I'm not going to waste your time, as 100% made my mind up to go ahead and get the full filters package from yourself ("Excellence" option: Filters with & w/o cross talk correction), happy to go ahead with the payment if you need that first. I just wanted you to have a look at my measurements & data, and advise if I’ve carried them out correctly, or do I need to redo them (i.e. sitting in the centre of the speakers, or more accurate with my spacing for each measurement, it was a bit more random for locations for 2 to 8).

 

Regarding the full set of filters, are these generated for all specific sample rates, e.g. 44.1kHz, 48kHz, 88.2kHz, 96kHz, 176.4kHz & 192kHz (do you do DSD)?

 

Thanks for your time, and looking forward to hearing the songs rejigged.

 

Cheers

John

 

PS I've sent you all the info in the comment section of your website.

[Brucemck2]

I tried the service and these filters work really well. They improve imaging, "air" and tonality. I preferred the non-crosstalk version, but only marginally.

 

A nice contribution to the hobby!

[marcusman]

I'm late to the party, but FWIW,  this service is fantastic.  It's helped my challenging room greatly.   I'm very happy with what Thierry has done and is continuing to do.   I now have a proper soundstage which was sorely lacking, but more than that with the corrections I now have the correct tonal balance in my living room

[oilpaint]

I also want to recommend people try the HAF filters. I was about to buy software and attempt creating filters myself but then spotted this options. Thank goodness. Doing the measurements (as outlined by Thierry) were not that difficult. I’m sure I could have gotten results with the software I was considering, but, I know it would have been time-consuming and painful…..

 

Although I knew I had a lot of bass bouncing around my room, I was surprised at the first set of filters HAF developed. I think I have been bombarded with bass for so long that I thought the first filters ‘lacked’ bass. After several discussions and iterations with Thierry he added some bass back in ( but no where near what I had) and I am very happy with the results - the bass has oomph but also definition.

 

In the end I liked the cross-talk correction filters. I definitely now have a wider soundstage and better separation between instruments. For the price point (and saved time on my part) I can’t recommend Thierry’s filters enough. Not to mention the excellent service (and patience in my case) he provides. Just looked and I have 40 emails with him!

 

Dan