Has anyone tried using this crossfeed setup?

[curiouscat]

I've been using this excellent set of equalization/convolution files given here: https://github.com/ShanonPearce/ASH-Listening-Set for the past few days and it's been the best headphone experience bar none across the multiple sets that I own.

For my PC setup it's already as simple as it gets but I'm wondering if there isn't the possibility for a portable version somehow where I think this would really shine. The processing ability required isn't that much (a fraction of a core on even my old PC). A modern phone could blow it away if there was an app that could do it, but some kind of DAC/amp that does it directly would probably be best.

The one I use is the small broadcast room which is somehow the most realistic thing I've ever heard through a headphone. This beats Hesuvi and the old OOYH plugin easily.

 

[dasdoing]

I have worked with (modified) Shanon's old set of IRs before:

didn't know he realeased e new set, will try them out.
as mentioned in the video, I did like the WDR Control Room 1 the most as a starting point.

myself I never considered my work as finished, but I have interacted with him. not sure how much I influenced him

 

[curiouscat]

I have worked with (modified) Shanon's old set of IRs before:

didn't know he realeased e new set, will try them out.
as mentioned in the video, I did like the WDR Control Room 1 the most as a starting point.

myself I never considered my work as finished, but I have interacted with him. not sure how much I influenced him

Yes, I agree with you the Neumann KU100 room recordings are the best. There was a study I saw on this forum that supported that as well, with KEMAR as a close second despite the large difference in response, albeit KEMAR produces a natural Harman-like curve. Normal uncorrected stereo seems to be far inferior at least in comparison for seeming real.

The small broadcast room happened to be coincident with whatever my ear is used to hearing besides, I only had to equalize a small portion to correct some imbalance.

As it is some company could easily produce something like this in a portable format along with a headphone that is close to diffuse equalized and some software algorithms for personal equalization and offer a superlative product. I haven't seen that combination of matched software and hardware yet unless it's gaming-oriented (which usually isn't great for listening) or something with an extraneous ecosystem like Airpods. I think PSB made a new headphone that could be tuned with an app to cover up some hearing deficiencies but that doesn't present the same idea. Maybe not profitable however unless they charge thousands unfortunately.

 

[dasdoing]

The small broadcast room happened to be coincident with whatever my ear is used to hearing

now I realise we are actually talking about the same room. It is the cleanest IR I ever heard, which is probably the best starting point for a room simulation focused on authenticy.

As it is some company could easily produce something like this in a portable format along with a headphone that is close to diffuse equalized and some software algorithms for personal equalization and offer a superlative product.

I wonder if you somehow use convolution on an Android device.
anyways, this stuff has never gained much love from headphone enthusiasts. They actually like the "in your head" sound I found.

 

[charleski]

I've tried various 'virtualisation' plugins over the years and was never really convinced. But these are really impressive! After spending some time playing with them on EqAPO I'm going to convert them to CamillaDSP for my main listening setup.

These IR's are a good ad for high-performance DACs with a lot of dynamic range. A gain reduction of ~18dB on top of replaygain and whatever preamp you have in your headphone EQ adds up to needing a lot of headroom. Luckily modern DACs are so good this isn't really an issue. I've spent a lot of time tweaking my headphone EQ to match my own taste, and found that it seems I've been subconsciously dialing in elements of the tonal colouration produced by these IR's. I've reduced the bass shelf and some of the 'presence band' lifts that I've been using (which happily helps with headroom). If your headphone is in the list for which compensation IR's are provided then you should probably just use that, but any headphone EQ that's been tweaked for standard non-virtualised use may need some modification when you add on the IR's.

They actually like the "in your head" sound I found.

I think you're right. These provide a very different perspective on the performance. But while the spatialisation is very impressive, I find a part of myself still craves the far more intimate presentation given by the classic 'in your head'phone sound. Part of that is clearly just familiarity, though

 

[charleski]

I note they state that "The BRIRs have been equalised to remove undesired spectral colouration and to make the BRIRs compatible with diffuse-field equalised circumaural headphones." Is the exact process for this described anywhere?

I've been comparing the BRIR wavs provided in this collection with what I assume are the source data (eg I'm assuming that rooms 3, 4 and 11 correspond to CR1, CR7 and SBS respectively from the data here) and it's evident that quite a lot of processing has happened just from eyeballing the wavs in an editor. Looking at the matlab code provided at the TH-Koeln site shows filtering for headphone and microphone compensation as well as some windowing and truncation.

 

[curiouscat]

I note they state that "The BRIRs have been equalised to remove undesired spectral colouration and to make the BRIRs compatible with diffuse-field equalised circumaural headphones." Is the exact process for this described anywhere?

I've been comparing the BRIR wavs provided in this collection with what I assume are the source data (eg I'm assuming that rooms 3, 4 and 11 correspond to CR1, CR7 and SBS respectively from the data here) and it's evident that quite a lot of processing has happened just from eyeballing the wavs in an editor. Looking at the matlab code provided at the TH-Koeln site shows filtering for headphone and microphone compensation as well as some windowing and truncation.

Good question. I'm not sure how they did it precisely or where they got their headphone measurements from. The wiki has a section that sets out what they did, but not a step-by-step. The listening-set is a newer version of https://github.com/ShanonPearce/ASH-IR-Dataset which has been improved multiple times.

Binaural Room Impulse Responses

A dataset of filters for headphone correction and binaural synthesis of spatial audio systems on headphones - ShanonPearce/ASH-Listening-Set

github.com

 

[dasdoing]

I note they state that "The BRIRs have been equalised to remove undesired spectral colouration and to make the BRIRs compatible with diffuse-field equalised circumaural headphones." Is the exact process for this described anywhere?

I've been comparing the BRIR wavs provided in this collection with what I assume are the source data (eg I'm assuming that rooms 3, 4 and 11 correspond to CR1, CR7 and SBS respectively from the data here) and it's evident that quite a lot of processing has happened just from eyeballing the wavs in an editor. Looking at the matlab code provided at the TH-Koeln site shows filtering for headphone and microphone compensation as well as some windowing and truncation.

import the original IRs into REW and look at the waterfall. there is a lot of "room boom" in them. that's totally normal for room EQs but is annoying on a headphone. also ideally we want the IR to be neutral FR wise, too.
the best way to make a binaural "space-simulation" for headphones would probably be creating binaural IRs in an anechoic chamber with a perfectly neutral system, but afaik noone ever did this

 

[iGude]

this stuff has never gained much love from headphone enthusiasts. They actually like the "in your head" sound I found.

I’m surprised to read that but it sounds like you have experience. May I ask why you think this way or what is your experience that brings you to this conclusion?

 

[dasdoing]

I’m surprised to read that but it sounds like you have experience. May I ask why you think this way or what is your experience that brings you to this conclusion?

the fact that headphone enthusiasts even exists allready shows it, no? if you are an enthusiasts than you don't hear nothing there to fix.
But also if you read about "binaural room simulation" in hifi forums you find that most wont like it.
Even the much lesser crossfeed solutions are niche.
And both of these are nothing new. they acutally only found partial acception with audio engeniers

 

[iGude]

But also if you read about "binaural room simulation" in hifi forums you find that most wont like it.

I would blame that on the lack of convincing and affordable solutions. The Smyth products have quite some supporters, the concept of having speaker sound via headphones seems to catch, at least those who can afford it.
Also those who successfully tried Impulcifer never go back it seems.

 

[Spyart]

now I realise we are actually talking about the same room. It is the cleanest IR I ever heard, which is probably the best starting point for a room simulation focused on authenticy.



I wonder if you somehow use convolution on an Android device.
anyways, this stuff has never gained much love from headphone enthusiasts. They actually like the "in your head" sound I found.

You should try Rootless Jamesdsp app. It requires 4 channel wav files though

 

[dasdoing]

I would blame that on the lack of convincing and affordable solutions. The Smyth products have quite some supporters, the concept of having speaker sound via headphones seems to catch, at least those who can afford it.
Also those who successfully tried Impulcifer never go back it seems.

well, the solution of this topic is free, and you only have to put a config file into EQApo. but this topic has not gained much intrest here on ASR either

 

[iGude]

well, the solution of this topic is free, and you only have to put a config file into EQApo. but this topic has not gained much intrest here on ASR either

There is some truth in it…

 

[isostasy]

You should try Rootless Jamesdsp app. It requires 4 channel wav files though

Can you use rootless jamesdsp for the convolution files shared above? I've tried jamesdsp before and didn't like the built in crossfeed much. I'd really like to find something that works better than my SXFI amp.

 

[iGude]

Can you use rootless jamesdsp for the convolution files shared above?

From what I see, this is not possible, at least it’s not straight forward: JamesDSP features a single convolution stage, but ASH requires two stages in a row, one for the BRIR and one for the HpCF.
Naturally, you could convolve both BRIR and HpCF offline and feed the result to JamesDSP. But even though this is not rocket science, you would need some knowledge and tool to do so, e.g. EqAPO.

 

[iGude]

ASH requires two stages

There might be a workaround though: the frequency response of each HpCF is given as a frequency chart, too. Hence, you could use the convolution stage of JamesDSP for the BRIR and the parametric EQ stage to approximate the HpCF frequency response.
Since this approximation is not required to be very accurate, this could work without audible loss.
But this theory, I’m not using JamesDSP (yet?).

 

[isostasy]

@iGude thanks, I tried a few stereo BRIR files which pushes the sound forward a bit but not really out my head. Jamesdsp has an 'arbritrary response equalizer' which allows you to apply autoeq presets but not import your own which is annoying; you have to add individual peaking filters and is hard to do by sight. It helped a little bit with my HD6XX but still not there.

 

[olieb]

the best way to make a binaural "space-simulation" for headphones would probably be creating binaural IRs in an anechoic chamber with a perfectly neutral system, but afaik noone ever did this

I am not so sure about that. Such an anechoic reproduction might bring out the problems of stereo (comb filter effects for frontal sources, change of spectral balance from center to sides) quite clearly. And then there is the lack of reverberant sound (all bones and no flesh ) and at last one will always be hearing with "wrong ears".
And with HRIR only there will actually be no "space" simulated at all, there will be only the room of the recording that is in the mix.

I do not know about EQApo or the other mentioned ways of doing it, as I am on a Mac. But if you can handle two stereo streams in parallel there is a nifty plugin that probably could do what you want.
http://anaglyph.dalembert.upmc.fr/
Anaglyph is quite amazing with all the stuff it incorporates (HRIR convolution and room simulation with free movement and distance enhancements and so on). You just have to use two instances of the plugin for left and right speaker. There is a bunch of HRIRs to choose from (IRCAM), but as far as I can tell they do not sound too good.
Luckily you can add your own choice in .sofa format. The Neumann HRIRs from Cologne are great (sounding).
http://audiogroup.web.th-koeln.de/ (HRIR_CIRC360.sofa is good for me)
And if you want you can still add room effects. Without room I found you have to EQ a lot to get a realistic sound, but maybe I did it all wrong.

In the end it was like a boosted version of the Meier-Crossfeed, for those who are into the intimacy of headphones that might be interesting. But how good it sounds will depend on how well one conforms with the ears of this "Neumann". Or maybe you can have your own HRIR and use that with proper EQ. Probably would be super analytical (bony) but true to the "original".
I liked it but I definitely prefer the Control Room 1 from WDR.

 

[olieb]

I've been comparing the BRIR wavs provided in this collection with what I assume are the source data (eg I'm assuming that rooms 3, 4 and 11 correspond to CR1, CR7 and SBS respectively from the data here) and it's evident that quite a lot of processing has happened just from eyeballing the wavs in an editor.

Thank you so much for pointing to these BRIRs.
I have been looking for a good binauralisation for quite some time, but something is always off.
Being on a Mac I could not use this

excellent set of equalization/convolution files

but the BRIRs I can feed into a convolution engine, no problem.

The FR of the result (L+R mono signal) is quite wavy and there are problems in bass (room modes that are not surprising in a small room, even a good one). One can easily hear this in bass-heavy music.
But with EQ and crossing the bass (<200Hz) to what smyth calls "direct bass“ (bass is just added to the convoluted signal) this is great. One gets the "realistic" room of stereo speakers with the superb clean bass of headphones.
To my ears it is even a bit better than the room virtualisation that is coming with the realiser (KU100 in BBC room).
Now, how do I get access to a world class room with my "own ears"?

But for now I am very happy with what I got from here, thanks again!