New loudspeakers & room digital correction service

[Alec_eiffel]

Warning: long post.

So here is my experience with HAF filtering, with a bit of background.

I had always been curious about acoustics and sound reproduction in general, being half engineer and half musician/fond of good music across all genres. After 10 happy years with DIY 2-way VIFA speakers floor standers powered by T-AMPs or alike (total cost 1000 €), I decided I deserved an upgrade for the next 10 years. Value for money is important to me, and I like to think long term when I buy or build expensive things.

I started with speakers last year : still an area where DIY allows to reach good performance at a fraction of the retail cost of a commercial product. I assembled my own version of an Atohm Bookshelf kit EURUS 1.0. I recently added a DIY sub  on the side to cover the 20-40Hz range based on a Dayton RSS210, a sono amp, and a miniDSP 2x4HD for filtering. The total cost of my speakers is approx. 3000 €. incl. the sub. 

Then the amp : Devialet D200 bought new for 4500 € some months ago. I guess the 220 Pro is better but was way out of budget for me. 

For streaming and library management, I started with a Raspberry PI / Digi + / Moode Audio solution (80€) and bought a 100€ Digital NAS. Worked very well but I couldn't benefit of the Air input, and the PI is not powerful enough for convolution. I had convinced myself convolution was the only way to cope with a very reverberant untreated living room (concrete walls, RT at 0.6 across the spectrum...). Treating the room was my first choice but not possible. 

I bought an Intel NUC and was convinced with the rest of the family that Roon was a good choice for performance / ease of use / multi-room. Convolution seemed well implemented.
 
I then got into REW/Rephase and thanks to the information gathered by the fantastic community around it (including SwissBear, merci encore !), I managed to build simple filters to remove the main room modes, and later on adjusting tone and phase with Rephase, achieving a "Dirac-like" correction. It was fun to do but it also turned out to be  time consuming and obsessive. There is always something new to try or optimise. I think what I did is OK but can be further improved. I had no time or motivation to further build my experience and expertise.

Overall I was quite happy with my setup. My only remaining frustration was the soundstage, too small (especially missing depth) and missing resolution. Stupid reverberant / room mode-ish living room.

Then  I came across some posts mentioning the approach of Thierry / HAF. Reading his technical paper and annexed references (esp. the Linkwist paper) was enlightening. There were actually filtering techniques that could treat the defects of the speakers and the room at the level of the reverberant field, making it psycho-acoustically more realistic an therefore improving the "illusion" of the stage originated by the speakers. I couldn't fight reverberation with room treatment, but it could be "normalised" through a set of FIR filters by Thierry. PLUS I could get an expert doing the phase and tone correction in a maybe more efficient way. 

All I had to do was to take some measurements with REW : took 20 minutes. A couple of hours later I add 2 demo songs... The first time I played the modded files there was really a WOW effect. I finally had a sound stage. My decision to buy was taken. Thierry was fast in assembling the files according to my need and I have spent a couple of hours today simply enjoying music. The next time I will change the filters will be when I move!

I highly recommend to consider HAF filters in the case you are not satisfied with the 3-dimensionality of your setup, esp. if your room is very reverberant. Additionnally you'll get a Dirac-Acourate like correction of tone (incl. room mode) and phase. The risk is low as you can get free samples. 

Regarding the cost : I cannot think of ANY other improvement of my setup that could bring the 15-20% subjective improvement I got. 

As an illustration here are the filters defined for me by Thierry for the left channel.
The first graph shows the amplitude and phase correction : 
- room modes esp. a very naughty longitudinal one at 55Hz
- softening of a 7Khz bump of the tweeter 
- phase correction is quite soft as my speakers are already well aligned (tweeter filter has a passive RLC "delay") and filtering slopes are only 12DB/oct. The phase curve goes nuts in some highly non minimal phase areas but this is OK


The second filter is actually applied to right channel and then reinjected into the left channel, and is adding some reverberation at the needed frequencies, creating the illusion of a more "natural" reverb correcting the speakers AND the room "unnatural" behaviour because of the speakers directivity mainly .  What is remarkable is that the highest correction is computed in the 2K-5K band which is exactly the overlap of the tweeter and woofer, as predicted by the theory (cf. Thierry's technical paper).