Saturday 2 March 2024

Home Audio Fidelity's (HAF) X-talk Shaper DSP. And is crosstalk correction/cancellation (XTC) just an "effect"?

See video and plug-in info at Home Audio Fidelity site.

Let's spend some time talking about X-talk Shaper in this post, a new DSP plug-in that will allow speaker system listeners enjoy crosstalk cancellation.

For those who have not read much about this, perhaps review the post from last year written with STC on crosstalk cancellation (XTC) and Ambiophonics. There's also the reposted article written by Ralph Glasgal that discusses some of the rationale for Ambiophonics you might find interesting and I hope provides good background for the 'hows' and 'whys' of this technique.

My understanding is that for awhile now, Home Audio Fidelity has offered room correction with and without incorporating the crosstalk reduction option into the convolution filtering for their customers. This new plug-in then will only utilize the crosstalk reduction portion itself with the ability to customize the sound based on your speaker and head geometry as per the settings:

As discussed recently, I've been using a fanless Intel i3-N305 MiniPC in the audio room and as part of this, have incorporated crosstalk cancellation DSP playback as one of the outputs available to Roon. I've been doing this through the DSP Studio facility in JRiver Media Center and above we see a screenshot from that set-up.

The interface is simple, we have the "Listening Angle" knob which corresponds to the angle between speakers and the listening position. The "Interaural dist" reflects the width of your head between ears - 15cm is the default although like hat sizes, double check for yourself. The "Effect" knob determines how strong the cancellation effect is. "Volume" is self-explanatory, just whether the DSP applies some attenuation to the signal.

In JRiver, turn on "Process independently of internal volume".
Allow DSP to have access to full signal level for processing.

Obviously, have a listen, you'll be able to hear the significant difference crosstalk cancellation makes. Just take "Bypass" on and off. You don't need objective measurements to know if you need or want this.😉

Having said this, since this blog is about looking a little further under the hood at how things work, we can objectively capture/demonstrate the effect... Despite the beliefs of some subjective-only audiophiles, with the resolution and capabilities of modern computers and ADCs, we can always demonstrate a difference if in fact there is something to be heard! It's a matter of knowing what we're testing for. Whether you personally like/enjoy the demonstrable differences is of course the truly subjective part. The only things in the audiophile hobby that cannot be measured or demonstrated as having an objective effect are snake oil products which play on expectation bias and the purely subjective testimonies typically primed by purveyors of such products and echoed by the faithful (for example this laughable recent Stereophile cable hype article).

As demonstrated with the uBACCH DSP, we can examine the digital filter's effect on the impulse response as it is modified going through the processing. Here's what it looks with X-talk Shaper in "Bypass" (no modification) mode, then the default linear phase setting, and finally minimum phase selected:

NOTE: Impulse responses shown to examine morphology.
Linear phase processing will add latency, not shown.

Notice that I'm using a linear phase low-pass upsampling filter hence the symmetrical pre-/post-ringing in DSP Bypass. The intent of this diagram is obviously just to show the morphological change when we apply the crosstalk cancellation. When we use the Minimum Phase setting, like with minimum phase low-pass filtering, there is less pre-impulse effect. The upshot might be as sense of better "immediacy" with transients, at the expense of potentially less effective crosstalk cancellation plus the usual high frequency phase shift found with non-linear filtering (likely not audible).

Obviously changing the speaker angle will have a significant impact on the impulse response as well - here's an example of what happens when the speaker angle is set 30° vs. 60°:

We can surmise the effect is noticeably less strong when speaker angle higher.
(Default linear phase setting used for both.)

The "Effect" knob changes how strongly the crosstalk cancellation signal is sent to the opposite channel. We can easily see this with a 5kHz left-channel-only sine wave sent through the DSP and examine the amount of cancellation content showing up in the right channel:

The "Effect" knob clearly changes the strength of the cancellation signal sent to the other speaker. Notice the phase/time shift between the left and right channels calculated based on the geometry of the speaker angle + interaural distance (space between the ears). While performed differently, this is the same principle as active noise-cancellation headphones sending out a signal that attenuates the noise from the outside reaching your ears; in this case, the signal is intended to reduce the sound heard by the other ear just inches away. Another analogy might be that of a "reverse crossfeed"; crossfeed being the intentional channel mixing that some headphone listeners will use to simulate what happens when listening with speakers. The big difference between this and simulations with basic crossfeed however is that when listening to speakers with XTC, the sounds are still emanating from external space and we perceive it as such using our personal HRTF (head shape, pinnae), unlike headphones and the typical "inside the head" lateral soundstage.

This is all calculated by the DSP and applied realtime during playback. With today's modern CPUs it's no problem at all. Here's a look at my CPU load while playing a 24/96 stream from Roon (server computer) → Roon Bridge on the MiniPC → JRiver/X-talk Shaper:

This is the i3-N305 MiniPC with power-limited BIOS setting, processed through 32-bit X-talk Shaper DSP (64-bit version would likely be even more efficient). The computer was able to do this at 10% load running at 1.22GHz - not even the typical base 1.8GHz speed, much less activating turbo to 3.8GHz. As audiophiles, know that audio processing for playback typically isn't a big load for modern CPUs unless you're running a bunch of DSP processes concurrently and possibly in multichannel. More likely to be taxing the computer on the studio production side than consumer playback side.

I hope you found this dive into HAF's X-talk Shaper software and its effects interesting! It's great to see another DSP option available for audiophiles to try out. Here's the link to the X-talk Shaper DSP software which will run in demo mode but intrude every minute with a dip in the volume for a couple of seconds (IMO, a great way to implement software demos). It's available for Windows (32/64-bit versions, VST, VST3, standalone) and Mac (Audio Unit, standalone). Registration is linked to your specific computer hardware and OS based on key ID, and costs €129.

[IMO, this is potentially a much better deal than the unreasonable high cost of uBACCH which also implements crosstalk cancellation through convolution DSP. I'll leave you to subjectively listen and decide for yourself which sounds better in your system.]

The fun of music is in the listening of course, so give it a try yourself with your speaker set-up. Play with the settings; feel free to deviate from your speaker "Listening Angle" in order to achieve most pleasurable effect - I do. If it works well, you should hear neutral tonality yet the soundstage widens with improved depth (better sense of "immersion") as crosstalk distortion is reduced and your ears/brain experience possibly a more accurate interaural time and level differences (ITD + ILD) embedded in the recordings we perhaps never knew were in there!

Tip: to test that you're not hearing weird distortions or tonality, try using mono music with distinctive vocals. For example, the gravelly, detailed vocals of Louis Armstrong like the track "A Kiss to Build A Dream On" on Satchmo Serenades should sound well centered, without artificial glare when played through your DSP of choice.


But is crosstalk cancellation just an "effect"?

I've heard variants of this question asked online and in videos over the last few months. I think the underlying belief is that perhaps there is something pejorative about adding "effects" to the sound we hear. As if this could damage or even destroy some notion of an idealistic "absolute sound" that we're after.

Well, of course crosstalk cancellation has a very significant effect on the sound, so yes, it is and has an "effect"! Putting a mattress or wall between the speakers as a hardware solution to separate the sounds reaching the contralateral ear will also have a noticeable effect. ðŸ˜ [Furthermore, in my mind, choosing to introduce vacuum tubes into the playback chain and selecting vinyl playback are also no less "effects".]

However, the question is really whether that effect is good/beneficial... Whether it serves the purpose of allowing the listener to hear what's actually in the recording - even hearing what's truly in the mix that the artist(s) and studio engineer(s) did not know were there! This idea should not be foreign to us because I think we already know that most musicians and pro audio folks are not obsessive audiophiles.

Crosstalk cancellation (and general Ambiophonics playback) has the potential to remove at least some of the crosstalk distortions from speaker systems which we have all lived with, and experienced as "normal" since the beginning of home 2-channel playback even though we've also all heard the benefits of crosstalk reduction when we wear headphones. Even if you're sitting nearfield listening to excellent speakers (less room effects), notice how much easier it is to use headphones for A/B listening tests to detect subtle changes when we don't hear the distortion from signals intended for the other ear. But other than those who have tried physical partitions between the ears, or DSP like AmbiophonicDSP [note: apparently not being sold anymore], (u)BACCH, X-talk Shaper, etc., probably few music lovers, audiophiles, or even studio production engineers would have heard the expanded immersive soundstage through just 2 speakers on many recordings.

For example audiophiles, despite how often you might have heard a track like "Keith Don't Go" used in audio show demos, I would argue that you have not really heard the best from Nils Lofgren's Acoustic Live (1997, DR11) until you've heard it though good crosstalk cancellation dialed in for your 2-channel speaker system.

Since XTC is not used routinely in music production studios, albums would not have been "checked" for how they sound which means what is heard might not be "as the artist intended".

For example Bob Marley's "Jamming" (off Legend: The Best of Bob Marley and the Wailers, 1990 CD first pressing) has percussion parts that sound like they're sitting right against the ears as if one is virtually wearing headphones. This is pretty weird but an interesting experience I suppose; not something I imagine Bob would have heard in the studio. Would he have wanted this effect? (Perhaps not.)

There are some other older recordings that really get a boost from XTC, for example Jon and Vangelis' The Friends of Mr. Cairo (1981, DR15), check out the title track! Car whizzing by, car crashes, gunshots in the distance, voices at different depths, damsel in distress, alien-type ray-gun effects, synths, distorted vocals, etc. A cute ode to gangster movies and the golden age of Hollywood. Did Jon Anderson and Vangelis intend for the soundstage to be this wide, and "objects" this well delineated? Have they ever heard this recording with XTC on? Who knows unless we ask them. More importantly, should we even care!? This is clearly an artificial studio production which can obviously sound highly immersive but does not reflect "realism" in that this recording was never meant to replicate the sound from any actual performance one could ever attend.

Modern pop recordings already incorporate a ton of DSP and I find that they tend to sound fine with XTC despite many inherent issues like poor dynamic range. Christina Aguilera's AGUILERA (2022, DR6 - yuck!) sounds pretty good - have a listen to "Pa Mis Muchachas" and the more traditional "La Reina". And Dua Lipa's latest single "Training Season" also sounds expansive already, and a bit more so with X-talk Shaper. Her track "New Rules" from the Live Acoustic EP (2017) also sounds great with XTC. I could go on...

Whereas crosstalk cancellation is still stereo playback (assuming not further processed with surround speakers for ambiance extraction) and there are significant limitations for those listening outside the sweet-spot, true multichannel does provide better imaging for those sitting off the ideal position, is able to anchor the center image better with good center speaker, and can provide sounds originating from the rear (or overhead as in Atmos).

The first track on the Hans Zimmer Dune: Part Two (2024, DR11 multichannel) soundtrack, "Beginnings Are Such Delicate Times" (well known quote from the Frank Herbert book) with its deep bass and massive impression of spatial volume right from the start, mixed with an intimate central melody when heard in a multichannel/Atmos system is an example of impressive immersion that simply can't quite be captured with the 2-channel version even with XTC. This massive spatial sonic gestalt returns in the track "Eclipse", and the cacophony of sounds/noises in "Water of Life" sound phenomenal in multichannel surround.

Consider incorporating crosstalk cancellation in your arsenal of audiophile experiences. Have fun folks as you explore the joys of Fidelity, Immersion, and Realism in your own home and over the lifetime of audio playback!

As always, I hope you're enjoying the music...

BTW: I'm in Houston next week. Probably won't have much time, but let me know if there's an awesome music/audio store to check out in case I have some down-time.


  1. Hej Arch,
    So much to digest! Love your lengthy forays into subjects I need to know more about. I have a multichannel setup and I encourage all 2 channel dinosaurs to at least give this format a proper listen. It is after all the natural evolution of sound reproduction if our audiophile goal is to get as close as possible to a live concert experience. In my naivety I assumed this format would be lauded and embraced by music lovers. Of course I was wrong. I stumbled across a post in one of Roons forums where someone urged people to try the Bacch4macs 3D software as he was completely floored by the experience. He writes, In 38 years of being an Audiophile I’ve never have experienced such an incredible increase in enjoyment in musical realism and added purity through increased dimensional detail and resolution by using distance based or custom (using binaural in-ear microphones) crosstalk cancellation filters. It’s not a frugal endeavor but I do make it a point to mention to audiophiles my experience with it has been revelatory. The response?... I am content with my 2-channel stereo setup, thank you very much. I am here for the music, not for the latest sound-processing gimmick.
    History teaches us that many groundbreaking inventions were ridiculed when they first appeared. The light bulb was not immediately accepted, The British claimed” good enough for our Transatlantic friends... but unworthy of the attention of practical or scientific men." And then we have vaccines, personal computers, internet and so on. All met with skepticism and distrust.
    This inherent negativity and suspiciousness for anything new and revolutionary is sad and sometimes even harmful. Music will be enjoyed regardless of how it is reproduced. But if there is a way to immerse myself even more, then I am all for it!

    1. Nice discussion and point Mike.

      Like it or not, for some, your use of the reference to "dinosaur" is very apt (heck some of the audiophile reviewers probably walked the earth with dinosauria 🙂). IMO, unless the audiophile hobbyist is completely content with being "fossilized" in standard 2-channel audio, forever tweaking with new speakers, changing amps, exploring euphonic distortions of tubes and vinyls, maybe trying cables (which at a reasonable level more than likely will make zero difference regardless of what some people claim to hear), the path ahead will at some level involve digital signal processing.

      Nice quote from the guy who gave BACCH a listen! Obviously, it's fine if some folks don't want a more immersive experience or try out multichannel, etc. I do think it's a very limited, myopic, view of the hobby and the possibilities to explore though.

      All the best, Mike!

  2. Hi Arch. I looked again at Dune 2 on Amazon and now there is no Atmos track at all! I guess like DSOTM it’s Apple paying for some exclusive « spatial audio » version of a popular release for a time. I could stream Apple Music on my Sonos system but I really don’t like their attitude…I’ll wait.

    Regarding XTC, I found out that simply putting both your hands on your nose to make a short 10-finger wall, and already there is more channel separation. Take your hands away and the two channels get mixed up again!

    Brings back memories of the old three stooges eye poke joke. ;-)

    1. Man that sucks that even the two Dune 2 tracks on Amazon disappeared!

      The battle for exclusivity to get patients to sign on for subscriptions is perhaps in full swing among the big players... We're probably at or close to saturation point for subscriptions after all these years and general acceptance of music streaming.

      I do hope Qobuz gets into the game with multichannel content beyond the few albums I talked about last year. Would be great to have Roon handle compressed EAC3-JOC and Qobuz getting involved even if it's not these new exclusive tracks for those of us who would appreciate the convenience when running Roon.

      Yeah, putting hand-to-nose for XTC might help - depends on the size of the hands :-). Could get a bit tiring after awhile.


    2. Of course I meant this as a simple means of hearing the reality of crosstalk, not as a permanent listening method, as you surely guessed. ;-)

      I was really surprised that it worked so I thought of sharing it.

  3. I see you are getting hooked on XTC, I think you can go a long way trying to make the experience perfect by trying to adjust the cross-talk cancellation to your personal head and torso properties. Regarding the question whether it's an "effect" or not—we should start from the fact that stereo recording is already a somewhat artificial attempt to capture or re-create the reality using a very limited representation, much like like a photography maybe. And then we use various ways to extract that reality using our playback system. This is similar to the fact that you can print the same photograph on a different medium, process it using different color profiles, etc—and it will be perceived differently. It's always half engineering, half art.

    1. [Sorry, the first half of my reply was missing]

      Hi Archimago! Thanks for writing about that Jon and Vangelis album. I didn't have a chance to listen to it before, sounds really good! Have you tried the Interstellar soundtrack—it has an incredible sense of depth and envelopment because the organ parts were recorded in a real cathedral. Should sound very real!

    2. Greetings Mikhail!

      I came across that Jon & Vangelis the other week when rummaging through Roon as a recommendation. It has been decades since I last listened to that album!

      Thanks, I'll check out Interstellar later today.

      I agree. Two-channel stereophony is already a kind of abstraction of "how" the soundstage is to be presented. Mix that with our room effects, the nature of our speakers, there are just too many variables. Ultimately, I do agree with the "subjectivists" and their emphasis on enjoyment. But always making sure we know the difference between things that make a difference and the snake oil some seem to be keen to peddle!

      For me, it's nice to have the option of using the XTC DSP on the desktop and sound room even if I can't achieve the ideal Ambiophonics geometry much of the time. Still sounds more immersive and potentially more enjoyable...

  4. Hi Archimago, what is the duration of the impulse response? Thanks.

    1. Hey ST,
      Away currently, but will have a peek when I get back... I'll have a look at the impulse response and see whether I can get an estimate. Of course maybe Thierry @ HAF might want to let us know!

    2. Hey ST,
      Info from Thierry - 32,768 taps at 44.1 kHz; similar to his room correction filters. (And commensurately increases to ~65k with 96kHz to maintain ~1.5Hz filter resolution.)

    3. Hi Archimago,

      Thanks for the reply to you and Thierry. However, I am not familiar with taps or FIR. Based on my limited understand the duration of IR for room analysis is around 227ms or 0.227ms, I have forgotten the exact value which is too short to be audible at all without reflections or resonance of the speakers. Since XTC involves of what actually perceived by the ears, I am at lost what we are measuring since XTC requires and additional inverted signal for the original signal and typical that signal will be less than 220 Ξs. The colouration happens because of the additional signal or more correctly the inverted sound waves to reach the ear for cancellation.

      Looking at IRs diagrams not telling much about the accuracy since it is a total decoding of the brain of multiple duplicated signal of the original sound inverted and not inverted separated by the durations in Ξs ( based on speakers location. I tired to get some clarification from BACCH when they produced diagrams and graphs to prove the point ( sadly, I didn’t receive any reply) and I am still at lost how these chart in the post established the tonal accuracy.

  5. When I listen to music with headphones, I use 112dB Redline Monitor VST crossfeed plugin. So, if I want to try this X-talk Shaper VST plugin (for headphone listening), does it make sense to use it before the 112dB Redline Monitor?

    1. Hi fgk,
      No you would not want to use X-talk Shaper for headphone use since the job is just to remove the crosstalk from speaker playback. You actually want that crosstalk since you're presumably preferring to add crossfeed.


    2. My thinking was that recorded music already contains some crosstalk in order not be "hard stereo". So, if we just simply apply crossfeed (which introduces crosstalk), we end up with stacking new crosstalk on top of the existing one. Would it not be beneficial to "clean" first the recorded audio file from the existing crosstalk and only then add new crosstalk by applying crossfeed effect?

      Crossfeed (HRTF processing) = emulation of speakers (or speakers sounding in a room).

      Why would I want to emulate speakers with crosstalk if I can emulate a better version of them (i.e. speakers with crosstalk cancelled)?

      I asked AI this quesition and here's its answer:

    3. The crosstalk in the music is part of the recording and not to be confused with inter-aural acoustic crosstalk. Basically, XTC meant to prevent the sound from one speaker heard by the other ear, just like headphones as you won’t be hearing the sound from one side headphone in the other ear. So when you use XTC with loudspeaker sound you perceive the recording like a binaural recordings that you get with headphones listening of binaural recordings.

      This is more about how human ears work.

    4. Hi Archimago,
      Sorry to have to give out this Caveat Emptor, but I made a purchase of the AmbiophonicDSP XTC program from at the end of February; my order was accepted; I was redirected to PayPal where I paid my $10 then I was sent back to where I was given a numbered invoice and told to wait a couple of days. Three days ago I emailed Howard Moscowitz at asking for the download link. Unfortunately, I have not had a reply. Luckily, my loss is not at the level of uBACCH-pricing, but the experience with leaves me an unhappy potential-Ambiophile.

    5. Please go to the forum and go to the Chat at the bottom of the page and inform mosc about your problem. If I remember correctly, they stopped selling AmbioDSP but unsure why the have not updated the pages.

    6. Thanks for the discussion guys... Been away down in Houston for the last while :-).

      Hey s=klogw, hikes. I heard that there might be some problems around whether they were still selling AmbiophonicDSP. I'll remove the link from the main text and put a note about that.

      Hopefully Moscowitz replies soon or you get resolution at the forum as per ST. Let us know!

    7. Hi Archimago, According Howard he already sent the ZIP file and I hope s=klogw could confirm. Anyone here who are having trouble with the purchase can email me directly and I will assist to get them the DSP. Meanwhile, Howard may make the DSP available free or as Donationware but subject to approvals of co developers.

    8. Hi Archimago and ST, thanks to ST's advice I joined the forum and the guys there were very helpful. They said Howard didn't come to the forum much any more but they would PM him. Howard did then send a ZIP file to me which he said contained everything I needed: "a ZIP file of all I have on this project. It includes a DLL file for a 32 bit VST and some documentation".
      The ZIP file's properties state it contains 2 folders. However unzipping only gets me one file: the freely available AmbiophonicDSP User Guide pdf. I'm not an OS-virtuoso so can any one advise me of how to get at the hidden .dll file, please? I'm on Windows 10. Thank you to the great community on Archimago's peerless blog!

    9. Try to unzip the file with admintrative rights.

      To Archchimago, Google sign in has stopped working with iPad and Iphone since the latest updates and therefore users with Apple devices can no longer post comments. Perhaps, you want to allow other methods to aloow commenting. Thank you.

  6. Hi Archi,
    Do you know Anaglyph (high-definition binaural spatialization engine)?
    I would be interested to hear your impressions about it.

  7. I just tried this out and I can't hear ANY difference whatsoever having this on or bypassed. The volume slider inside the VST seem to be working at least so I guess the rest should be working as well, except that nothing really happens when working the sliders. I think I have a quite good system and my hearing shouldn't be too bad either, so why am I not hearing any difference?