Home Audio Fidelity (HAF) DSP filters

There is always a risk in starting a thread during a process but it does give it more veracity rather than trumpeting a “fait accompli”.

The fact that Roon has the capability for convolution filters has always piqued my interest but this was quickly extinguished when I looked at the manuals for REW and Accolate etc. I’d seen mention of Home Audio Fidelity and Thierry both here and on the Roon forum. The idea being that you measure your room and send the readings to Thierry and he designs a DSP filter.

And so last Christmas (that’s 2019) among other presents I gave my son a UMIK1 USB mic so he could help me measure my room. And so this Christmas I finally decided I needed to take these steps on my own as not much was happening with my son!

Going to the Home Audio Fidelity website there is a downland of a program that allows one to measure the room along with a download of the sound wave track to measure - which I downloaded and saved to my library so it was available in roon.

It’s a simple enough interface, connect the mic, open the program, calibrate the mic - basically hit calibrate and play the track.


My son lent me his mic stand and I’d recommend buying one if you don’t have access to one.

Windows being windows I did need to change the sound settings twice to use the UMIK1 twice and it worked the second time.

So now I have it set up.

The next step is to take the 9 recommended measurements, the first at the main listening position (pictured) and 8 more 30-100cm around the sweet spot…




Good luck John hope it works as well for you as when I used his services. He does great work and he will tweak it to satisfy your needs and is very patient and tolerant. I am tempted for him to update mine for my new speakers just to see what else it might add. Not felt the need to as my old ones really needed help. Doesn’t cost much for updated ones so worth the punt.


So when the measurements are taken they were stored on my laptop in a folder called HAF.


I saved these to onedrive and sent the link to Thierry in HAF. I also sent a song for Thierry to apply the DSP to so I can see what type of changes the filtering will producue before laying any money down.

I choose Thin Lizzy - The Cowboy Song from Jailbreak as it was suggested to choose a song you know very well.

I received a Wetransfer link to download the modified versions : a_mod version corresponding to the standard correction and the _mod_xtalk which has crosstalk reduction on top (generic model) last night.

I’m not quite sure what I was expecting as there are some who say the life is sucked out of the songs and they sound very different while others say the differences are very positive. The caveat here has to be that I’m quite poor at direct A vs B (vs C) comparisons. I test by living with a sound for a while and then reverting back. I cannot tell the difference blind between high res and standard redbook files. Rarely have I had the seismic events of veils tearing and noise floors lowering or the old chestnut of the jaw dropping. I’m just not that kind of guy.

So the first thing that struck was this was no night and day effect. Also to prevent clipping the modified files are at a lower volume and the crosstalk one lower than the standard one. So it was impossible to be sure the volume the same during the listening. So there was no mad displacement of instruments, no features that I’d never heard on the song before, I couldn’t hear Philips clothes brushing off his bass guitar. After a few run through the original and 2 modified tracks it became clear that there was a nicer tone to the modified tracks. The guitar parts were slightly sweeter without losing any power or attack. If anything I thought there was a bit more of the infamous PRAT (as I understand it).
In retrospect I realised that perhaps I hadn’t chosen the best track to actually test the full capabilities of the process and although I could ask Thierry to modify another track there is enough here that warrants buying the filter and I’m actually quite excited at seeing what impact it will have on myriads of my music. I’m going for the “Excellence head” option : 189€ = filters with & w/o cross talk correction (head model).

So I’d imagine I’ll have the filters by Monday, I believe they are easy enough to input through Roon so that will be my next instalment.




I’m pleased I’m not the only one without ‘full sound analysis recall’.


I never went for crosstalk what difference did it make to the ones without crosstalk?

Best of luck! Thierry is great. I had a similar experience as CrystalGipsy and I’m also thinking about trying again with my new speakers.

1 Like

Yes load up the zip file into convolution engine via a pc, apply headroom management and clipping indicator ask Thierry what he recommends for headroom as he might compensate in the filter itself anyway and you might not needs headroom management.

Thanks for posting this @Sloop_John_B I contacted Thierry last week coincidentally to find out how it all works.
I’m currently in the process of sorting out a mic and will then take it from there.
I’m having a go at this out of curiosity to see what, if any, difference it makes in my room. I have a one sided room open on one side and some obstructions on the other so the measurements between left and right will be of great interest whatever.
I shall be interested to see your progress and how easy it was to load the filter into Roon.

You say that your choice of test track could have been better, what would you have chosen if you could choose again and why?
I’ve been wondering if anything in particular will demonstrate the effect better than something else.

1 Like

Well the filters arrived on a Sunday afternoon! As you say very easily added to Roon and Thierry said there was no gain in my filters so I didn’t need any headroom. After the United v Liverpool and our Sunday Covid quiz with the diaspora I’ve been listening for the past few hours. Initial findings are extremely positive on the regular filter. Much more pronounced effect that on the test track on many of the tracks I’ve listened to from various playlists like cable tests / DSD vs PCM are the ones I’ve had issues with and wanted to improve with cables etc.

A few more tracks before bed then…



I think I would choose a track that I thought was too bright or too muddy i.e. one with some issues as because what I seem to be finding out tonight is that many tracks I thought were poorly mastered were just being put through the wringer in my room. I obviously must have had a bad issue with treble reflections as many jarring high hats seem quite tamed.

No doubt there may be some tracks / albums that my room made better through a random set of interactions but the great thing is I just have to hit “bypass filters” in Roon and my room is back!

I have not done any A v B yet as I mentioned above I go for the more organic type of demoing.



Thanks, thats really helpful, Ill have a think whilst my mic is on its way!

My advice if recording at the listening spot is NOT keep the microphone still even for spot recordings. Ideally you want to record by rotating the mic in a slow circular motion around the plane of your ears… then ensure you average the recordings…
Keeping the microphone still is likely to give a false listening response as your mid and hf frequency response will be incorrect. At these levels you will likely get a comb filter response from your room, and in smaller rooms with acoustically shiny surfaces this will likely vary over just a few cms, … hence the need to average out by recording in a circular motion.

When you rotate slowly try and keep microphone cable noise to a minimum.

I would then take half a dozen or more averaged microphone rotated recordings around your listening position, again more important for small rooms.

I have discovered this really does make a significant difference in the mids, to which our hearing is most critically sensitive to.

Once you get the recording to make the response filter, it is also worth filtering that to smooth it… and I always focus on cuts rather than gains, usually keep cuts to 3dB or less. Typically if not I find the artefacts introduced can outweigh the benefits.


I think it’s worth pointing out in this thread that for the Home Audio Filters the following is Thierry’s recommendation.

Here are some guidelines for performing the measurements

  • Use a mic stand for the microphone and stay away from loudspeakers/mic during the measurement
  • If you have the appropriate calibration file (90° for UMIK) point the mic to the ceiling otherwise point it to the midpoint b/w loudspeakers

I think I’d check with him first before deviating from this.



Yes, just ask him he’s really nice and knowledgable. To Simon’s point that’s exactly what I did when taking measurements. I moved the mic to around to take about 9 measurements around my primary listening area. (it’s all automatic with his newer app).

  • a mic stand made things a lot easier and more accurate IMHO
1 Like

Sure, do what you are happiest with, but a static microphone will likely cause errors in the upper mids / rebel unless you are in a large reasonably well damped room… or perhaps you move the mic on its stand horizontally in say 15 cms intervals and take average measures… again more sensitive if you have a small microphone… I use a cheap small calibrated measurement mic a Umik - 1

You take 9 different measurements in 9 different postions for HAF to cover the area no need to move it about, I have heard others using this for a different measuring approach, Thierry knows what he’s doing so I personally would not deviate from what he recommends. He basically follows a method similar to Dirac’s Chairwhich is a cubiod either square or more oblong if you want a wider sweet spot with one in the middle


48 hours down , probably 9 hours of listening.

It’s certainly different to any other “upgrade” or cable change etc. It’s more like your CDs have been remastered, some slightly, others more so. And it’s not as like one masterer has put their sonic stamp on the music, it could be several different people doing this mastering, as there is no one common impact on the music. Some bright music has been tamed and some dull muddy music been released.

The one relative constant is that instruments are clearer and more defined. Pieces that used to be a bit muddled, that I thought were badly mastered or the amp/speakers lost control of, invariably sound better. Born to Run emerges, the 3 choruses? of Us and Them “Forward he cried from the rear …” and the instrumental break in the album version of Cool for Cats. All these are apparently under more control whereas they are jumbled without the filter.

I made the mistake after a day of being very happy of doing some A vs B (filter vs non filter). A - that’s great, B oh that’s very good too and the vocals are much better, A Oh no, vocal are good here too… (and repeat).

It’s not all totally positive though, on some songs there is a certain (hard to describe) squeezed feeling, some central sounds being a little too emphasised or heavy (but this may simply a corollary of reducing reflections).

This is all with the regular filter, I haven’t tried the crosstalk reduction one yet.

Thierry said to listen to a wide variety of music and then report back and see what tweaking may be needed. So I’ll head back to the hifi…



This is also what I did. He’s really easy to work with.

This is what I felt about it, bass was tighter and less dominant, midrange was revealed and vocals where so up close it felt more they where in the room and it tamed brightness which was a major issue. It revealed so much more that the room was hiding. It took about 3 goes at revising filters to get the treble right for me though, he was very patient with me through out this stage. And as I changed a few components and moved the speakers he will tweak afterwards as well but it will cost a few euros for updated filters. This is what I need to do, it’s peaked my interest again.


Thanks for the update, its great getting your feedback in advance of me giving it a go.
At least I am getting a feel for what I may expect and how it all works.

1 Like