Hi,
This is an interesting subject.
I have built...
And they all sound different.
Naturally they do, because they all implement different circuitry, different coefficients of cross coupling channels etc.
Also, non of these systems leave a flat frequency response for the "Mid" or "Mono" signal and non have the exact same change of frequency response, so in addition to "Crossfeeding" they act as unpredictable EQ.
Hence they all should sound quite different.
What I find worse, all these crossfeed circuits, while successfully counteracting the "in head" localisation lead to a stereo image that is very narrow, I call the result "wide mono".
In the 80's a team at the RFZ (of which I was a peripheral part being a very, very junior engineer) in Berlin worked on an improved stereo format that was intended to be backwards compatible with existing stereo equipment. The 80's were an exciting time for that sort of thing, we also saw Ambisonics developed back then.
The new format would be able to be passed through a suitable simple and easily mass produced decoding matrix to give a correct and improved presentation of space via speakers (with speakers the LF image is too narrow and the HF image width is too large).
Decoding to headphones would match a dummy head recording, while giving good stereo imaging results played without decoding.
There was an intention for a similar encoding matrix on the recording side, which would encode any common format, including dummy head stereo into the new stereo format. Grand idea really.
This target was never reached, but among other things crossfeed circuits were evaluated and modified.
In the end the best solution found on the headphone decoding side used a M/S system that processed only the "Side" channel. Coefficients were empirically determined by many Tonmeister's to offer the closest match to listening to speakers.
During the original tests there was a widely adjustable parametric EQ in the side channel with the mid signal untouched. Every Tonmeister was asked to adjust to their preference from the basic setting (which was some generic crossfeed). The comparison standard was a pair of large format studio monitors in the same room.
From there a consensus emerged which was radically different from previously published crossfeed circuitry.
Later a simplified fixed hardware circuit was tested based on an "average" of the different settings and found to to give satisfactory results to each participant's, with materially reduced audible colorations, compared to using the adjustable EQ.
The program was terminated even a while before the German Reunification, when it became clear that the new stereo format was elusive. The RFZ buildings were demolished in '06, most documentation and hardware went to landfills when the RFZ was deestablished in October '90.
As for making a stand alone unit, I generally think this is a terrible idea. It is rare that adding extra things in the signal path produces improvements in sound quality. Crossfeed should be designed into the actual headphone amplifier circuitry, not as an either external or internal "bolt on" section.
What needs to be done, in my view, is to re-evaluate the current crossfeed systems and to replace them with a new system, which should eliminate the undesirable frequency response changes and should produce an improved spacial rendering over the current "wide mono", instead of continuously rehashing circuits that are simply "wrong".
The professional DAW plugin "Waves NX" offers a fairly good result if correctly adjusted and it includes optional head tracking. I quite like it actually, mind you it's DSP and needs a fair bit of CPU power.
Sadly the default settings leave a lot to be desired, so much fiddelling is needed, we are almost back at the RFZ setup where everyone dialed up his version. A professional sound engineer may handle this with a wave of his hand, an average user? Less so.
Thus selling a widely adjustable system without a really good preset to "normal people" is unlikely to give the desired results.
And all that said, I note that too often people listen with their eyes and not their ears and they do not hear what is presented to their ears, but what their prejudices expect. This makes discussing these systems problematic.
A good example would be a company that included an "enhanced spacial presentation" crossfeed (which I designed). Sadly the feature got stuck with a name that most people think of as some gimmicky DSP or EQ.
And so instead of hearing an "out of head" presentation with an appx. 60 degree coverage angle on minimalist miked stereo recordings (which should have been the result), most pre-biased listeners (and I am sorry to say reviewers) heard whatever they expected, not what was presented...
And the funniest thing is, if you actually demonstrate the feature to unbiased listeners, that is not head-fi people who think they know about the tech and have definite opinions, but instead totally uncritical listeners who normally listen to earpods connected to their iPhone streaming apple music, the reaction is the same, always. Namely "Wow, this is like I'm in club/Concert hall/at a concert etc."
So the system actually works fine for people who just use their ears to listen and are not biased towards hearing something specific.
Ok, enough of my ramblings of past glories and ideas.
Good luck with your search for a standalone crossfeed.
I know Garage 1217 make the "Project Kameleon" standalone headphone EQ, which has at least in principle a crossfeed option, as Solderdude designed such an option. You might want to ask.
Given that the Garage 1217 Kameleon plus a module sells for 270 USD it is not even very expensive.
http://garage1217.com/new_website_017.htm
Of course, a dedicated headphone amplifier with enhanced spacial presentation crossfeed can be made to sell for much less (last time I looked one unit had a 150 USD retail price) but hey, given nobody likes that kind of crossfeed and given its not a stand alone unit, it would not be relevant to this thread.
Thor