I realise there is a thread for HAF filters already but it’s mainly centred around the measurement and input into Roon. I’d be interested in peoples musical experiences with the filters. Things like;
Do you use the crosstalk reduction or standard filter?
Do you alternate between them?
Do some genres suit the crosstalk filters more?
What type of changes or tweaks did you ask Thierry to do with the first iteration of filters to arrive at where you are now?
Thanks,
.sjb
Why I ask is I’ve just spent this week with filters that Thierry sent me.
First off bass is so much better, more controlled and only there if it’s on the recording. Previously my sub seemed to add bass to everything and so I alternated between a gain of 60 and 70. After Thierry noticed a lack of bass with my first measurement we agreed it was best to measure at the 70 rather than the 60 gain - and boy am I glad I did.
So now I have 2 filters - both sound better than unfiltered but sometimes the crosstalk reduction filter sounds amazing (particularly on Jazz and some classical - although I don’t have much classical) but on some rock/pop can sound more like an “effect” - music coming from way to the left or right of the speaker. Certainly scuzzy rock sounds better (to me) with the standard filter.
So I’m wondering if others experienced this and how they proceeded. Did you ask Thierry to modify the crosstalk filter so that the “effect” lessens but the magic is still there on Jazz etc. Or do you alternate between filters where one just suits a certain album / group/ genre?
I’m also interested in what other changes Thierry may have done for you. For example if you asked for drums to sound more realistic and you couldn’t believe that this could be done so effectively - I’d like to get a feel for what can be done, what parameters can be changed. Can one ask for female vocals to sound better? What trade-offs have you experienced?
Roll up Roll up, step right this way…
.sjb
I’m a big fan. Much better than I could do myself using REW and a UMIK-1. I didn’t originally opt for the crosstalk filters but I recently changed out my speakers and completely reconfigured my room so I’m paying for an update and also added Crosstalk this time.
The filters made a huge difference for me with my precious nearfield setup. My new setup is a pair of Martin Logan Purity electrostats and no longer near field. I setup one seat in my home office for listening from a comfy chair (pic is from the comfy chair) and I “assume” crosstalk will be more noticeable with the setup config and the fact that electrostats are effectively bi-poles.
Near field setups interest me but I admit I have doubted the benefits of this sort of manipulation with them.
Hi Sloop. This is a very good idea for a topic. It’s not easy to describe so forgive the long post.
I started out with Dirac on Windows but wanted room correction integrated into Roon for stability reasons. When Roon released the convolution engine I got Thierry to make normal and X-talk filters for my 2.1 setup. I used the X-talk as standard because that was where I heard the biggest change. I really liked it.
Imaging was better than with Dirac. I didn’t make Thierry change anything about the filters. I noticed that some bass had disappeared but I convinced myself that this was the way it was supposed to sound. I got used to it. Certainly the sounds in the bass region was more defined. You could actually hear the notes being played on a bass instrument.
Over time I came to miss some of the punch in the bass and I also noticed that on rock music and most non acoustic music the X-talk filter had the tendency to sound like an effect the way you describe. Also something about the voice being decoupled from the rest of the sound image. But on something like Peter, Paul and Mary: “All my Trials” the X-talk filters clearly sound better to me – by the way the title “All my trails” rings true trying room correction.
Then I got new speakers this time with Genelec GLM room correction. Their room correction was very good and can be stored in the speakers so they behave very well from the get go without a computer attached. I still wanted to see if the HAF filters could improve it. So I measured my room with the Genelec filters enabled and got Thierry to make new filters. The X-talk filters sounded worse than in my previous setup so I focused on the normal filter instead. I was not able to hear a reliable difference between the Genelec filters and Thierrys normal filter.
But my strategy had changed in the meantime from what I had learned the first time around. I no longer wanted just an accurate response. I also wanted a response I would enjoy – yes I know, I know I’m pretty cheeky also wanting to enjoy this.
So I asked Thierry if he could add more bass. I didn’t want a boomy bass but a punchy one. Playing with the PEQ filters in Roon resulted in a darker sound not really more punchy. Thierry delivered a new set of filters and this time I was able to tell the difference from the Genelec filters. There was more bass and it was punchy. I really liked it. It was not a big difference and details in the low region was still apparent. I asked Thierry how he accomplished that and he said it also had to do with the decay times in the room. So in my experience this is where it pays to have an actual human expert designing the filters. I’m very happy with my sound now.
My next mission is to find a way to compensate for equal loudness. I spend more time listening at low levels that I do at standard levels. So I’m wondering if Thierry can somehow compensate for that in some way or if its better left to PEQ filters. You write about measuring at 70 instead of 60. Is that in dB at the listening position?
Great post, very interesting to hear others experiences.
The 60 and 70 mentioned is just the display volume on my sub not a measurement.
.sjb
I think this is already feature request for Roon. If not, it should be. It would be a great addition to the DSP settings.
Yes. There is feature request: Equal loudness contour
I’ve been meaning to post there too when I get the time.
My experience with HAF filters that I’m already using for some years is the following:
When Thierry made the first filters for my Gold Phantoms I recognized quite fast that there was a phase problem in the lows. The low bass had disappeared compared to not using the filters and when I inverted one channel it was back. I reported this to Thierry and received corrected filers right away within 24h. I think it had to do with some time alignment issues of the Phantom woofers retracted on the side with respect to the mid and tweeter. Thierry immediately saw what could have been misinterpreted in my sent measurements and corrected it. In terms of sound signature the filters were then removing all the room modes, preserved the fast and precise character of the Phantoms as unported speakers that go low to 16Hz and sounded linear. Compared to the original a bit weak in the sub bass region. So I was hesitating if I should introduce my own house curve to the initial room correction or ask for a correction from a linear sound signature to one that is closer to the originally intended sound designed by Devialet. By the way it was Thierry that asked me specifically to give him my preference on this as he recognized this from the measured curves.
I’m very happy that I decided to ask Thierry to keep the sound signature from the Gold Phantoms and actually only correct the room modes and the reverbation time.
The difference between filters on and filter off now is not changing the sound signature. X-talk, I use it all the time gives this magic of much better instrument separation, more precision in the image. But, the overall pleasing wide sound spectrum is fully retained. Compared to my previous speakers (B&W 802, first edition from 1980) largely extended in the lows and highs. All in all very subtle differences. However, when looking at the sound energy changes over the entire bandwidth I have to reduce the volume by 12dB to not clip (red headroom indicator) in Roon. Without the filters indeed at the same volume setting all is much louder. But the Devialet Phantoms can play very loud. In the volume normally at 70/100 they are at their loudest performance before dynamically reduce the bass to prevent driver damage. And now this limit is rather reached at 75-80, no problem. This shows me that a lot is actually going on to compensate for the room, even if the difference in sound characteristics is moderate. It is more defined, precise, no boom, deep bass, neutral mids and heights, airy, wide, all I need.
So they definitely helped… but I just got my updated HAF filters today (with and without crosstalk)… and WOW, the difference and improvement it makes on this new configuration is awesome. I would still say they’re helpful for nearfield but I would now say it’s a “must have” for normal setups. The bass and lower frequencies are perfectly balanced, no boominess or exaggerated frequencies… while at the same time, I’m able to hear and pick out each instrument/vocal so much easier now.
The crosstalk filters (first time trying them) are great too… best description I can give is that it reminds me of using the “stereo wide” button I recall from my stereos from the 80s… but not an artificial sound… just opens up the soundstage a lot more… it’s like going from coaxial car stereo speakers to components if that makes any sense.
Similar experience here… I listen to the original filters provided by Thierry to ensure they are well balanced but yes, maybe one who admittedly enjoys a 12" subwoofer in my car… the files are “lacking” in bass. Like you said, everything is there, just not a lot of kick drum, low bass guitar or synthesized bass notes.
So I too play around with PEQ, nothing fancy… in fact, a “one band” PEQ that’s a low-shelf filter to add in a house curve. I set it at 100Hz, Q of 0.5 and listened at different gain levels.
I then ask Thierry to send me additional files with this house curve integrated into my filters… so I have the standard filters, then one with house curve at +2.5dB and one at +5.0dB. This makes switching my DSP profile super easy from my iPhone while listening in my ideal seat.
I use xtalk almost exclusively. I simply enjoy it more than regular. Occasionally I go back to regular to check but I often don’t last but a song or two. Preference more than anything I suppose. I enjoy the more spacious sound.
I’ve had Thierry make ~10 different sets now over the past 3 years, including both HAF and XTALK versions. My current set I have 4 house curves to choose from in ~0.5dB increments boosting the low end oh so gradually. I use my ‘flat’ curve probably 90% of the time but you know those '80s albums that bring back memories from the time I like to boost a bit as they seem so lacking in the low end mix.
For the past couple of months I’ve been running Roon Shaper along with convolution in the path. Really incredible improvements reducing room resonance!
Very interesting. How did you communicate with Thierry about the four cuves with increased low end? Do you have a graph you can show?
I simply asked him to create several with increasing bass.
Here are 4 impulse response files overlaid one another for each of my ‘bass’ choices. Nothing fancy just a slight increase or decrease depending on perspective from about 50Hz down.
If you want more or less ‘presence’ in the mid bass or elsewhere just ask. He turns them around in mere moments so I have to believe it’s quick and easy for him to adjust.
For me, I used PEQ in Roon and sent him the settings that I preferred.
I’ve not used Thierry’s service, but I have engaged @Mitch_Barnett’s help (via https://accuratesound.ca/) to discover optional filters for my room. I’m curious to know if any HAF users have tried using REW to measure the response after correction. I found this to be an interesting exercise, and it’s not difficult to do.
REW has a Generator function that you can use to save the measurement sweeps to WAV files. You can later import them into Roon and play them back through the DAC you use for day-to-day listening. It’s important to save a separate file for each channel, include a timing reference, and make sure that the timing reference is for the same channel as the sweep. For example:
Next, import these into Roon. You may want to use an ID3 tag editor to add some metadata that will make them easier to find in your library.
Place your mic in the same place you did for the measurements you sent to Thierry (equidistant to the speakers, same orientation, etc.). Then, go to REW, take a Measurement, and immediately hit “Play” in Roon using the correct file for the channel you are measuring. Make sure that the sweep sampling frequency, start and end frequencies, and length settings in REW all match exactly to the parameters you used to create the WAV files. Also make sure that you have your convolution DSP enabled with the HAF filters loaded.
This works because you included the timing reference (you did remember to do that, right?). REW will wait until it “hears” the timing reference before starting the measurement, so even if you start the sweep playback in Roon a few seconds after you tell REW to take the measurement, timing will be perfect.
You can take before and after measurements and compare them in REW, which is kind-of fun. You can also look at improvements to the step response and other parameters. For example, here’s the amplitude response in my room before correction (with Psychoacoustic smoothing enabled):
And, here’s what it looks like after (I’ve since reduced the bass by about 1 dB to tighten things up):
Lots of solutions, including Acourate and Dirac, provide plots of the predicted response after correction, but it’s pretty cool to be able to “independently verify” the actual response using Roon + REW.
I did that the first time I received the filters from Thierry… now, I just load them in as the difference in sound quality is so apparent. Just got my two additional sets of “house curve” files and loaded them up… every request I’ve submitted has been responded to or fulfilled within 24 hours.
Cool. Are you saving all of these as Roon DSP Presets? I found that to be a handy way to A/B changes. I’ll use “Headroom Management” to do some level matching to make the comparisons easier. I do wish switching between presets was a bit faster, but the delay is not terrible.
Yup, a separate preset for each filter. Not so much for A/B testing… but HC1 adds some bass, HC2 adds more… I guess you could say it’s my “loudness” button for low level night listening. I find that without any house curve added sounds good at ample volume.
That’s really helpful seeing them loaded as presets @euge_lee, thanks for that.
I literally received my HAF filters (standard + xtalk) today and was wondering how I could load them both in to easily switch - obviously with Convolution you can only load one at a time.
Problem solved, cheers!
(Sorry for hijacking the thread…)
