[Zenvota]

Just to chime in on my thesis that doing A vs B very quickly on HRIR isn't a great way to test and that you need multiple HRIR's for multiple physical rooms/situations so you get used to the sound of the room. Sean Olive (researcher at Harman) cites some research that may be of interest that seems so support this: https://www.stereophile.com/content/slow-listening

I definitely find there to be a longer acclimation phase using hrir/prir. When I switch between rooms I always pause playback and then try to ignore initial observations and let my brain adjust. Then after a short while I feel I can better analyze the audio.

 

[castleofargh]

Been loving my HRIR for sometime now. I'm spending another month away from home in an airbnb. Still can't get over how with a 15.6" 4k OLED and Impulcifer I can be transported back to my theater room - had a great experience watching Midway on the laptop. Surround track on that is excellent!

Just to chime in on my thesis that doing A vs B very quickly on HRIR isn't a great way to test and that you need multiple HRIR's for multiple physical rooms/situations so you get used to the sound of the room. Sean Olive (researcher at Harman) cites some research that may be of interest that seems so support this: https://www.stereophile.com/content/slow-listening

The actual research is good, the article...
and of course room convolution is something we need to settle in, slowly. Just considering reverb, switching rapidly is like entering the shower rapidly while singing and realizing how much reverb there is. But after a while the brain will identify the reverb components and remove or at least attenuate them, despite still making use of them for some secondary spatial cues.
That need of getting used to sound so our brain can do what it does is very common and an old story. Simply playing around with an EQ makes it painfully obvious. Listen to a song, boost the bass a lot, turn that boost OFF, and for a while the perfectly fine and usual amount of bass will feel weak. That's because instead of comparing a sound to our general habits, we're now comparing it to that other new reference. When we consider the deep power of priming in psychology, it's not hard to imagine the influence that rapid A/B switching can have on our impressions.
Rapid switching is strictly for the purpose of detecting differences between 2 samples. That's what it is for and the only qualitative judgement we should ever get from rapid switching is "do they sound different?". The silly guy who wrote the stereophile article clearly doesn't know that, otherwise he would be ashamed of sentences like this one: "Scientific testing methodologies such as ABX, which require quick and conscious evaluation of a change in the sound, have long struck many of us as insufficient, seeming to miss much that affects our enjoyment of music." ABX to evaluate enjoyment? Did he read anything about ABX before writing that? Thinking himself smart, he's really telling the world that clever him and his buddies have long noticed how a hammer makes for a poor TV remote. He had no business bringing ABX in that article and anytime he does, he makes a fool of himself.

ps: Not sure what Olive has to do with this? Did I miss something?

 

[venkatesh63]

@jaakkopasanen @johnn29 @musicreo can you share your recordings of HRIR.i don't want demo recordings from Impulcifer page. i want your created copies of different speaker's and Equalization's(eq.csv) of headphone's compensation files. i want for Movies and Music to listen copies of hesuvi.wav or hrir.wav you guys created with your gears.

Please anybody give me the files from your created gears i want for Movies and Music.

 

[lowdown]

Please anybody give me the files from your created gears i want for Movies and Music.

Impulcifer is used to create a personalized HRTF specific to an individual's ears and headphones. Your request is kind of like asking for other people's eyeglass prescriptions. Not likely to help you.

 

[venkatesh63]

@lowdown i just asked the files lol !!!. i'm not asking your assets.This group is for sharing and caring. And i asked the above mentioned guys not you @jaakkopasanen @johnn29 @musicreo. why i can't try other's HRTF if i make a mistake i will learn how they did i will analyze and rectify myself HRIR.

 

[johnn29]

ps: Not sure what Olive has to do with this? Did I miss something?

He posted it on his twitter, I assumed he knew the researchers.

More fun with Impuclifer this week - I've been using the AutoEQ transform function to use my existing HRIR with new headphones without having to re-measure to great effect. Also DTS Headphone xv2.0 was updated to support many more headphones - for me it's the best synthetic HRTF I've tried and has the benefit of virtual height channels. But compared to my LS50 Impucifer HRIR it pales.

Only thing I'd like is the ability to trim the reverb to only the bass frequencies. I'm hoping that gets rid of the channel bleed but preserves the pleasing reverb.

 

[musicreo]

One thing I want to share is my actual favourite microphone fixing. I have three pairs of Pui 5024HD capsules and two pairs off Primo EM258 microphones. After testing different options I found that OHROPAX® Silicon is for me the best way for holding the capsules in place during measurements.



Another thing I can say after a lot of testing is that my Sennheiser HD555 (with 595mod) sounds always better than my AKG701 headphones for binaural playback. I have the feeling that the AKG 701 is much more prone to small channel balance errors compared to the HD555.

 

[dwk]

Hi,
New user coming up to speed.

First, major props (and thanks) for implementing this. I've had this idea in my head ever since reading about the Realizer some years ago and have had my Sound Professionals mics stting around for a few years, but never got off my butt to do anything. This is quite impressive.

I'm interested in 'conventional' use of Impulcifier, but I also have a couple somewhat unconventional ideas that I"m hoping to explore.

The first comes from a long-standing feeling that the headphone guys and ambiophonic guys need to talk to each other. Ambiophonic guys are spending a LOT of time and money to set up a speaker system that eliminates crosstalk whereas the headphone guys seem to be constantly looking for new ways to introduce crosstalk. This led me to a theory that what the headphone guys really want is HRTF and not crosstalk.

So, my first application is to try to create 'ambiophonics over headphones' with Impulcifier - i.e. recreating the front stereo dipole. This should actually be really easy - it's just FC-L and FC-R filters applied to a conventional L/R stereo signal. I tried this last night, and the jury is still out on how well it works. It definitely sounds good and opens things up significantly (on my HD6xx), but doesn't quite recreate the mind-blowing sense of space you can get out of an ambio dipole. I suspect this is at least partially due to my measuring the single speaker in the nearfield, so I'm planning on retrying the measurement farther out in-room (need to figure out cabling to be able to do that). Plus, my bass vanished despite my Kef R3s having reasonable response, so I may need to add room correction or otherwise investigate since it made things a bit hard to completely evaluate.. (I've looked at the code and it looks pretty simple to update the output code to add an ambio filter with just FC-L and FC-R, so I may generate a pull request if you're interested)

The other non-conventional application I"m thinking of is DIY speaker crossover auralization. The idea being that you perform measurements of the individual raw drivers in the cabinet in-room, and then individually filter the input channels with appropriate IIR crossover filters ahead of the HTRF processing. As long as you can capture and preserve the relative timing between drivers (might need a separate step), this should allow accurate 'listening' to the candidate crossover without having to build it. There are a couple packages that do anechoic simulation of on-axis speaker crossover response, but nothing that does in-room as far as I know. With the growing appreciation of how off-axis response influences perception, I'm rather intrigued by this capability. This one will take a bit longer, and I'll definitely have to figure out what to do to protect the tweeters from full-scale signals.

BTW - has anyone looked into convolution playback on iOS? If my ambio-over-headphones idea works, even a simple straight L/R convolution might greatly enhance listening on-the-go.

once again though - great job.
Doug.

 

[lowdown]

So, my first application is to try to create 'ambiophonics over headphones' with Impulcifier - i.e. recreating the front stereo dipole. This should actually be really easy - it's just FC-L and FC-R filters applied to a conventional L/R stereo signal. I tried this last night, and the jury is still out on how well it works. It definitely sounds good and opens things up significantly (on my HD6xx), but doesn't quite recreate the mind-blowing sense of space you can get out of an ambio dipole. I suspect this is at least partially due to my measuring the single speaker in the nearfield, so I'm planning on retrying the measurement farther out in-room (need to figure out cabling to be able to do that). Plus, my bass vanished despite my Kef R3s having reasonable response, so I may need to add room correction or otherwise investigate since it made things a bit hard to completely evaluate.. (I've looked at the code and it looks pretty simple to update the output code to add an ambio filter with just FC-L and FC-R, so I may generate a pull request if you're interested)

I'm only using Impulcifer in stereo mode, with FL/FR and am getting extraordinary results. The imaging, detail, soundstage, and tonal balance are much better than any of the audiophile speakers I've heard or owned. The illusion of listening to exquisite speakers several feet in front of me is stunningly real, without any of the artificial sounding artifacts that are obvious in the other virtual surround products I've tried. For me the sound field and space are not inflated beyond what's in the recording, so no simulated surround imaging in stereo mode, but often the sense of sitting in the recording session is uncanny. My suggestion would be don't come to a conclusion on how good Impulcifer can be too soon. I went through numerous recording sessions, tried many different combinations of command line options, and tweaking the room correction files, before I had the version that I'm listening to now. Also, as others have posted it can take some time for your ears to adjust to a different HRIR. It may take some effort but the end result can be way more than worth it.

 

[jaakkopasanen]

So, my first application is to try to create 'ambiophonics over headphones' with Impulcifier - i.e. recreating the front stereo dipole. This should actually be really easy - it's just FC-L and FC-R filters applied to a conventional L/R stereo signal. I tried this last night, and the jury is still out on how well it works. It definitely sounds good and opens things up significantly (on my HD6xx), but doesn't quite recreate the mind-blowing sense of space you can get out of an ambio dipole. I suspect this is at least partially due to my measuring the single speaker in the nearfield, so I'm planning on retrying the measurement farther out in-room (need to figure out cabling to be able to do that). Plus, my bass vanished despite my Kef R3s having reasonable response, so I may need to add room correction or otherwise investigate since it made things a bit hard to completely evaluate.. (I've looked at the code and it looks pretty simple to update the output code to add an ambio filter with just FC-L and FC-R, so I may generate a pull request if you're interested)

This is very interesting. I'm not very familiar with ampiophonics but from what I quickly read it's basically stereo pair separated by 20 degrees and cross-talk cancellation. If this is all there is to it then this is super easy to do in Impulcifer. Normal stereo measurement and then just muting FL-right and FR-left tracks in the output HRIR file. Pull requests are more than welcome!

 

[castleofargh]

The first comes from a long-standing feeling that the headphone guys and ambiophonic guys need to talk to each other. Ambiophonic guys are spending a LOT of time and money to set up a speaker system that eliminates crosstalk whereas the headphone guys seem to be constantly looking for new ways to introduce crosstalk. This led me to a theory that what the headphone guys really want is HRTF and not crosstalk.

A stereo album is for the time being practically always made for stereo speakers. Trying to correct the inherent default of stereo speakers(having 2 sound sources pretending to be only one at another position), can only become the right way if albums start being made with that correction in mind. Otherwise it's just a fancy DSP that some people happen to like. It has no objective legitimacy if the way albums are made doesn't change.

There is no accurate 3D audio for headphones without HRTF, or at least without HRIR like here to provide some virtual speakers at a given position. so you're clearly right about that.



About dealing with mono sounds, apparently not everybody reacts the same way. The general concept is that with perfect mono, we mostly have a little idea of the elevation of something somewhere in front thanks to the outer ear and the general signature of the music(with very poor sense of distance beside close vs far). And that's about it. It's easy for the brain to find conflicting information between that small cue and sight, our dominant sense. So for a few people, mono will forever be crappy unrealistic placement with headphones(in the head, on the forehead, ...). Sadly impulcifier might not solve the problem as the cause is us.

If you have a clear issue with the elevation of mono sounds, then perhaps instead of trying to force a center channel, you might want to try what's in this video from Mr Griesinger.

it's irrelevant for the headphone's own frequency response when using the tools in this thread, but maybe the mics and how they're placed in the ear cause some frequency variations? Maybe there remains a difference because we're not getting the ear canal resonance? Whatever the reason, it's possible that the frequency response variation is off just enough to mess up with your perception of elevation(this is my guess I don't actually know that for a fact ^_^).

If you're really prioritizing the center, following the tuto in the video is fine. But if you're just trying to make sure that your Head Related Impulse Responses give you a response close enough to your speakers, then I would suggest to run the same experiment in the video but with the usual speaker angles(the way they were when you recorded your impulses). Then if done right, chance are that the center would still improve just from being created by more convincing virtual speakers with the right frequency response at your eardrum.

It's not simple, and it's a long shot just to try and confirm if you're one of the people who really need that extra accuracy to get an immersive experience, so I don't know if I can strongly advice to do it. But I mention it just in case.
The next step if even that doesn't push the center forward, would be to add head tracking somehow. Because then any small head movement will turn mono and too little cues, into sound at an angle with the all list of psychoacoustic helpers including the legendary co stars, ILD and ITD. It's a lot of efforts just to improve the center image, but some people need all that. While others seem to feel like they're watching a band live while listening to the nonsensical stereo of default headphone playback. So finding where you as a listener is on that crazy wild spectrum of humanity can be its own adventure.

The last attempt is to fight fire with fire, we fail to accept that a sound source is at a given place in front of us because we see that it's not in front of us. So just place a possible sound source in front of you and find out if you can fool yourself with that. If I see a speaker right in front of me, or watch TV and see the guy talking, my brain will sooner or later, end up placing the sound on it. Sometimes it's fast, sometimes it's not when things just are too unrealistic, but if I spend enough time in that configuration, progressively my brain will move the sound over there. Sight always wins in the end in my case.

A few years ago when I first started to use a bad DIY version of what impulcifier allows to do, it never felt good(it was just slightly better than crossfeed) until I happened to get a pair of monitors(speakers) to put on my desk. Then in a matter of days, I was happy with the perceived position of the convolved sound on my headphones as it started to better match the position of the speakers. Again, my anecdote, I cannot tell if you would get similar experience.

 

[jaakkopasanen]

A stereo album is for the time being practically always made for stereo speakers. Trying to correct the inherent default of stereo speakers(having 2 sound sources pretending to be only one at another position), can only become the right way if albums start being made with that correction in mind. Otherwise it's just a fancy DSP that some people happen to like. It has no objective legitimacy if the way albums are made doesn't change.

There is no accurate 3D audio for headphones without HRTF, or at least without HRIR like here to provide some virtual speakers at a given position. so you're clearly right about that.



About dealing with mono sounds, apparently not everybody reacts the same way. The general concept is that with perfect mono, we mostly have a little idea of the elevation of something somewhere in front thanks to the outer ear and the general signature of the music(with very poor sense of distance beside close vs far). And that's about it. It's easy for the brain to find conflicting information between that small cue and sight, our dominant sense. So for a few people, mono will forever be crappy unrealistic placement with headphones(in the head, on the forehead, ...). Sadly impulcifier might not solve the problem as the cause is us.

If you have a clear issue with the elevation of mono sounds, then perhaps instead of trying to force a center channel, you might want to try what's in this video from Mr Griesinger.

it's irrelevant for the headphone's own frequency response when using the tools in this thread, but maybe the mics and how they're placed in the ear cause some frequency variations? Maybe there remains a difference because we're not getting the ear canal resonance? Whatever the reason, it's possible that the frequency response variation is off just enough to mess up with your perception of elevation(this is my guess I don't actually know that for a fact ^_^).

If you're really prioritizing the center, following the tuto in the video is fine. But if you're just trying to make sure that your Head Related Impulse Responses give you a response close enough to your speakers, then I would suggest to run the same experiment in the video but with the usual speaker angles(the way they were when you recorded your impulses). Then if done right, chance are that the center would still improve just from being created by more convincing virtual speakers with the right frequency response at your eardrum.

It's not simple, and it's a long shot just to try and confirm if you're one of the people who really need that extra accuracy to get an immersive experience, so I don't know if I can strongly advice to do it. But I mention it just in case.
The next step if even that doesn't push the center forward, would be to add head tracking somehow. Because then any small head movement will turn mono and too little cues, into sound at an angle with the all list of psychoacoustic helpers including the legendary co stars, ILD and ITD. It's a lot of efforts just to improve the center image, but some people need all that. While others seem to feel like they're watching a band live while listening to the nonsensical stereo of default headphone playback. So finding where you as a listener is on that crazy wild spectrum of humanity can be its own adventure.

The last attempt is to fight fire with fire, we fail to accept that a sound source is at a given place in front of us because we see that it's not in front of us. So just place a possible sound source in front of you and find out if you can fool yourself with that. If I see a speaker right in front of me, or watch TV and see the guy talking, my brain will sooner or later, end up placing the sound on it. Sometimes it's fast, sometimes it's not when things just are too unrealistic, but if I spend enough time in that configuration, progressively my brain will move the sound over there. Sight always wins in the end in my case.

A few years ago when I first started to use a bad DIY version of what impulcifier allows to do, it never felt good(it was just slightly better than crossfeed) until I happened to get a pair of monitors(speakers) to put on my desk. Then in a matter of days, I was happy with the perceived position of the convolved sound on my headphones as it started to better match the position of the speakers. Again, my anecdote, I cannot tell if you would get similar experience.

This is what I thought too. Albums are made for stereo speakers so "fixing" the limitations of stereo speakers with anything doesn't sound realistic. However the ambiophonics are promising more than just better center image, namely 120 to 150 degree sound stage compared to the normal 60 degree one. I've no idea how this works but it sounds interesting. At least I'm going try it out myself at some point.

 

[dwk]

This is very interesting. I'm not very familiar with ampiophonics but from what I quickly read it's basically stereo pair separated by 20 degrees and cross-talk cancellation. If this is all there is to it then this is super easy to do in Impulcifer. Normal stereo measurement and then just muting FL-right and FR-left tracks in the output HRIR file. Pull requests are more than welcome!

Yeah, 20-degree is generally the recommendation. The math for XTC gets ill-conditioned at the separation gets smaller, and the wider you get the more the 'tonal inaccuracies' of 2-channel creep in. My inference is that the ideal is to be centered, but that doesn't work mathematically. I started with a single FC since it offered the potential for a simple single-measurement solution, but also seemed to be in line with the theoretical ideal.

Having made a few passes, my conclusion is that I need a few more . I took a single mono center run in the 'big system' today rather than the desktop system, and in certain respects it's wildly successful. Center-anchored vocals are superb - to the point I immediately understand the comments saying "I can sell my speakers". Sounds that are spread out laterally are less successful, and hard-panned sounds collapse back into the ear - you get a bit of a horseshoe effect. This is interesting to me, since hard-panned sounds in my ambiophonic experiments have been problematic for the opposite reason - they end up "way over there" and actually take explicit focus to try to integrate into the soundstage.

I did take a conventional L/R measurement in the 'big system' as well, but haven't listened to it yet since I'm using JRiver and need to set up a convolution config for it. (Is there a canonical config for the common cases? Not a big deal to set up, but it one exists it saves a step).

Once again though - great work. Pretty straightforward to use, and my $350 headphone rig (Khadas Tone Board, JDS Atom, NAD Viso HP50) sounds spectacular.

 

[dwk]

A stereo album is for the time being practically always made for stereo speakers. Trying to correct the inherent default of stereo speakers(having 2 sound sources pretending to be only one at another position), can only become the right way if albums start being made with that correction in mind. Otherwise it's just a fancy DSP that some people happen to like. It has no objective legitimacy if the way albums are made doesn't change.

Yes, this is the debate that seems to rage any time ambiophonics gets discussed. It's tough to argue with - 'authorial intent' and all that. OTOH, on the right recordings ambiophonics can be absolutely mind-blowing - it was definitely the most jaw-dropping audio experience I've had since the first time I heard Spica TC-50s and realized that spatial presentation of audio was possible at all. It works best on minimally mic'd / processed acoustic recordings where it's more believable that the natural soundfield has passed through. Heavily processed studio recordings frequently don't work well at all (Rodrigo Y Gabriella is the worst I've run into - the hard-panned technique makes things sound pretty comical). You also sacrifice a lot of headroom to the XTC process, making it tricky to preserve full dynamics. (my most successful trial was using Yorkville U15 Unity PA cabinets, so headroom wasn't an issue)

I could never fully commit to an ambio system as a primary system, but that's the beauty of Impulcifier - use it for times/albums where it works, and use a different config in other cases. Following that train of thought - I'm seriously thinking of building some Lx Mini speakers strictly to set up once and measure - I suspect that if augmented with some open-baffle woofers it might make the perfect system for chamber and solo piano music.

 

[jaakkopasanen]

Yeah, 20-degree is generally the recommendation. The math for XTC gets ill-conditioned at the separation gets smaller, and the wider you get the more the 'tonal inaccuracies' of 2-channel creep in. My inference is that the ideal is to be centered, but that doesn't work mathematically. I started with a single FC since it offered the potential for a simple single-measurement solution, but also seemed to be in line with the theoretical ideal.

Having made a few passes, my conclusion is that I need a few more . I took a single mono center run in the 'big system' today rather than the desktop system, and in certain respects it's wildly successful. Center-anchored vocals are superb - to the point I immediately understand the comments saying "I can sell my speakers". Sounds that are spread out laterally are less successful, and hard-panned sounds collapse back into the ear - you get a bit of a horseshoe effect. This is interesting to me, since hard-panned sounds in my ambiophonic experiments have been problematic for the opposite reason - they end up "way over there" and actually take explicit focus to try to integrate into the soundstage.

I did take a conventional L/R measurement in the 'big system' as well, but haven't listened to it yet since I'm using JRiver and need to set up a convolution config for it. (Is there a canonical config for the common cases? Not a big deal to set up, but it one exists it saves a step).

Once again though - great work. Pretty straightforward to use, and my $350 headphone rig (Khadas Tone Board, JDS Atom, NAD Viso HP50) sounds spectacular.

Could you elaborate on what did you do exactly? What I gathered from this is that your measured a mono system and somehow that creates a sound stage. I'm sure this is not the case. How many speakers did you measure? At which positions? Did you do something extra to the output HRIR file that Impulcifer doesn't do? Do you need some kind of special DSP processing to make this work with Impulcifer created HRIR? Did you have some kind of special DSP processing for the speakers during the measurement? I'm baffled.