Easy AutoEq and Crossfeed with CamillaDSP

[micha]

Thank you for the positive feedback Kent!

I looked into Jan Meiers crossfeed implementations and actually feel like I should probably email him There is so many presets floating around with his name on it that I feel like I should ask his opinion which DSP configuration he actually is in favour of.
That beeing said I now have a new release candidate which includes a "Natural Crossfeed" which is my take on reproducing what Jan Meier explains in his articles. I am sure that there a knowledgable people here who might want to contribute their insight on the "proper" DSP setup.

For all who feel like spending a bit of time in the "Crossfeed Rabbit Hole" here is some reading material:

Starting point for many is Siegfried Linkwitz's publication:
https://www.linkwitzlab.com/headphone-xfeed.htm

This was modified and evolved by Pow Chu Moy like described here:
https://headwizememorial.wordpress.com/2...mplifiers/

Then there is Jan Meier's take on the topic in discussion with Chu Moy:
https://headwizememorial.wordpress.com/2...crossfeed/
... and a modified version:
https://headwizememorial.wordpress.com/2...ed-filter/

And finally the work done by Boris Mikhaylov in his widely used "bs2b" which seems to be the go to source for DSP implementations of the above (analogue based) publications:
http://bs2b.sourceforge.net/

The new release candidate is available here: https://github.com/Ebert-Hanke/autoeq2ca...v0.4.0-rc3
Maybe someone wants to listen into the variants and give some opinions
The signal chains can be found here: https://github.com/Ebert-Hanke/autoeq2ca...ee/develop

Thank you and best
micha

[micha]

@micha thanks for the very interessting article from Mikhail!
Did you ever play with the IR's from ASH-IR-Dataset in a pipeline like below? :


The full acoustic or a room is maybe a little bit over the top for crossfeedback, the impact on the sound is way to much.
But at least those will 'pull' the music out out of my head and place it in front of me. Combining it with the information from Mikhail will be interesting.

Thank you, this looks really interesting! I will test it.
In the end of course it is the question what one wants to achieve in terms of headphone listening. Me personally still want headphone listening to be a very different experience of music listening compared to speakers but this of course is absolutely a matter of individual taste ... and thats so fun about all audio topics I find

There is also an Mid/Side effect worth trying out for a deeper soundstage which you might already have tested.
Build a M/S Mixer:
M = (L + R) * 1/2
S = (L - R) * 1/2
L = M + S
R = M - S
And in between delay either Mid or Side channel by a very low amount like 0.03 ms to introduce a bit of phase shift. I am currently experimenting with this.

[DRONE7]

Maybe someone wants to listen into the variants and give some opinions

Natural crossfeed seems to give a more out-of-head experience though a little biased to the right and a few instruments at more muted levels.

Compared that and the original ChuMoy and preferred...some of both!  

However...then compared both with no compensation at all...ie flat and simple in/out through camilladsp and was surprised at the detail missing in the processed versions....

So after some thought I suspect, in the following order,...My old AKG k240-sextett...My aged ears.. the compensation files from HK for the AKGs.

Then found a compensated Frequency Response for the AKGs and built a new impulse file and used that...
https://medium.com/@jaakkopasanen/make-y...bd567832a9

Different again...

So a little lost... next to try is the flat (no compensation) with the various crossfeeds. Perhaps your next version could include that option ? (No headphone compensation..choice of crossfeed)

Will report.

https://news.ycombinator.com/item?id=28799910

[micha]

@DRONE7
Thank you for taking the time to listen into the different presets and share your impressions! Very cool.
I think it is an interesting question if the use of (Camilla)DSP in general degrades SQ (which I guess in itself is a very personal thing).
Personally I dont like Crossfeed with every record and also AutoEq is not always better depending on the headphone ... Still trying to figure all that out a bit more.

I now released the 0.4.0 version but it is very possible that the presets could be tuned and improved. If there are people here who are deeper into DSP and/or Crossfeed in general any input is appreciated.

If one wants to test without the EQ it is easy to delete the EQ part from the Pipeline Steps either in moOde or in the .yml file. Just for reference the Crossfeed presets can be found here in the repository: https://github.com/Ebert-Hanke/autoeq2ca...n/src/data