[swtnate]

I also made an impedance jig, that calibration is tricky too in REW, but I got good results and can test zobels:

it also seems the phase data is WAY off. don't know how to really get that reading correctly.

 

[swtnate]

I have three pair of the round SR Knowles driver that is a balanced armature BUT is designed to replace/ "upgrade" from a dynamic. My question... How in the world are you supposed to get the sound from the armature to the ear canal? There is no spout and just like a tiny .5mm hole in the center.... any advice would be of some help.

 

[Wgibson]

I have three pair of the round SR Knowles driver that is a balanced armature BUT is designed to replace/ "upgrade" from a dynamic. My question... How in the world are you supposed to get the sound from the armature to the ear canal? There is no spout and just like a tiny .5mm hole in the center.... any advice would be of some help.

I used heat shrink tubing - shrink a portion of it but use the non shrunk part to hold the driver. Or you can get aluminum driver holders made for dynamic, with a tube spout. Cheap enough, but you still have to wait for them to arrive.

 

[swtnate]

I used heat shrink tubing - shrink a portion of it but use the non shrunk part to hold the driver. Or you can get aluminum driver holders made for dynamic, with a tube spout. Cheap enough, but you still have to wait for them to arrive.

Or, I could 3D print it…..

 

[kkugel]

Anyone got ear impression scans into Autodesk Fusion yet? It worked for my left ear, but my right ear has some very tight creases, which makes the T Spline Faces intersect when thickening it, I cannot convert it to a body :/
I converted my STLs to quad meshes with Blender, this way Autodesk can convert it to T-Spline.

 

[mattmatt]

I have three pair of the round SR Knowles driver that is a balanced armature BUT is designed to replace/ "upgrade" from a dynamic. My question... How in the world are you supposed to get the sound from the armature to the ear canal? There is no spout and just like a tiny .5mm hole in the center.... any advice would be of some help.

Plop the tube flat around the hole then apply adhesive. Make sure you use industrial grade adhesives tho.

that's what I do with them.

 

[swtnate]

Plop the tube flat around the hole then apply adhesive. Make sure you use industrial grade adhesives tho.

that's what I do with them.

For y'all to play around with if you so choose. lol. That SR driver is so weird just gluing a tube flush... I felt better doing it with something like this. I attached the STL and the IGES so y'all can foul with it however you like. After you download them, change the extension from txt to either STL or IGES, same as with other components posted in here previously.

 

[ForceMajeure]

Anyone got ear impression scans into Autodesk Fusion yet? It worked for my left ear, but my right ear has some very tight creases, which makes the T Spline Faces intersect when thickening it, I cannot convert it to a body :/
I converted my STLs to quad meshes with Blender, this way Autodesk can convert it to T-Spline.

Are you importing your impressions before or after the digital trimming process (via meshmixer f.e)?

I assume once trimmed and smoothed, the tight bends and creases should not be there so it would be less of a problem.
You can also use Instant Meshes (this is can be downloaded from github: https://github.com/wjakob/instant-meshes ) to transform the obj to quads before importing to fusion360.

Speaking of which if anyone is interested, there is a rather nice android/ios app for sculpting digitally I can recommend called Nomad Sculpt. Meshmixer is really the best and simplest tool for trimming but the sculpting brushes are lacking (no easy masking option as well as some brushes intensity, behavior and control is not that great). One could export midway processing from Meshmixer to Nomad sculpt to do some arrangements with brushes and import back in Meshmixer for final smoothing, thickening, hollowing etc...

Nomad sculpt learning curve is much lower than Zbrush or Blender and does the job. The brushes are good and using low values allow for small work to be done nicely. The 2 main brushes I recommend are flattening and smoothing. Flattening mainly used as a scrapping tool ( a bit like you would Dremel regular impressions). Flattening in meshmixer is rather bad and I never found a a good balance with it. There is a nice Scrape brush in Blender that is quite nice but Blender's learning curve is annoying...

 

[swtnate]

For y'all to play around with if you so choose. lol. That SR driver is so weird just gluing a tube flush... I felt better doing it with something like this. I attached the STL and the IGES so y'all can foul with it however you like. After you download them, change the extension from txt to either STL or IGES, same as with other components posted in here previously.

Also, I made the ID of the sound spout coming off the SR to be equal to 1.14mm. Wanna know why??? Cause I bought I thought I'd be smart and get those teeeeeny tiny dampers from Knowles.... Yeah they fit into nothing really and kinda fall apart when you look at them. This gives me hope now. LOL

 

[JEHL]

Has anyone tried to implement bone conduction in a home made IEM before?

 

[swtnate]

Bone is a completely different listening experience. You just kinda feel it as opposed to hearing it. So, the frequency spectrums are all way different due to it mainly being solely conductive hearing.

 

[kkugel]

Are you importing your impressions before or after the digital trimming process (via meshmixer f.e)?

I assume once trimmed and smoothed, the tight bends and creases should not be there so it would be less of a problem.
You can also use Instant Meshes (this is can be downloaded from github: https://github.com/wjakob/instant-meshes ) to transform the obj to quads before importing to fusion360.

Speaking of which if anyone is interested, there is a rather nice android/ios app for sculpting digitally I can recommend called Nomad Sculpt. Meshmixer is really the best and simplest tool for trimming but the sculpting brushes are lacking (no easy masking option as well as some brushes intensity, behavior and control is not that great). One could export midway processing from Meshmixer to Nomad sculpt to do some arrangements with brushes and import back in Meshmixer for final smoothing, thickening, hollowing etc...

Nomad sculpt learning curve is much lower than Zbrush or Blender and does the job. The brushes are good and using low values allow for small work to be done nicely. The 2 main brushes I recommend are flattening and smoothing. Flattening mainly used as a scrapping tool ( a bit like you would Dremel regular impressions). Flattening in meshmixer is rather bad and I never found a a good balance with it. There is a nice Scrape brush in Blender that is quite nice but Blender's learning curve is annoying...

Thanks so much for the thorough advice! I used Blender to convert to quads now. This kinda destroys the sharpness of cuts, but I just plane cut them again in Fusion now

 

[ForceMajeure]

Thanks so much for the thorough advice! I used Blender to convert to quads now. This kinda destroys the sharpness of cuts, but I just plane cut them again in Fusion now

I really don't recommend doing the trimming part in Fusion...I have no idea about the smoothing that way and what features of the ear it can crush along the way.

It's also usually computationally intensive, especially when importing raw scanned files (even more if not smoothed already) and converting quad to tspline>correcting errors>tspline to brep...It can crash easily depending on the quad counts.

The trimming, thickening and hollowing should be done with another software such as meshmixer. It's much easier then to import and play with.


You can easily close holes and repair the file with meshmixer using the inspector tool in the analysis section. Much easier than repairing while doing the quad to brep to tspline conversion in fusion. It will repair the bulk of errors and will leave you with potentially a small error count once imported in fusion, that will be easier to repair before the final tspline to brep conversion.

Once you are done with the the trimming and created the shells (that you could normally print at this stage), you could easily then convert to quads using Instant Meshes (you can control the number of quads and shapes with Instant Meshes to keep the features of your impressions as well).

Here is an example of an ear impressions that has been worked on via MM and another software for brushes as well as thickened and imported in Instant Meshes

I can then use the target vertex count to allow for more resolution for the quad count with higher subdivisions.
Here I have bumped up the vertex count and solved. Once extracting the mesh I use pure quad meshes and you can see the result

Here you can see the import in fusion

after a small repair job inside fusion that is the tspline result

The bulk of the work is actually trimming the raw impression in MM and other software to get a clean result.
It requires experience trimming in the physical realm especially if you want clean looking fp shapes and allow for enough space inside

I really don't recommend doing any kind of trimming inside fusion.

 

[JEHL]

I'd like to think that bone conduction works on the same principle as traditional hearing in that vibrations get turned into electrical signals that are fed to the brain to get processed into what we recognise as sound. The big differences being that:

1. The eardrum is not used much if at all. Instead the cranium does the job of transfering vibrations. Presumably this means not just each head may have a unique response to the frequency range, but different parts of the head may have different responses as well. Although since what I'd like to try is a DD+BC hybrid I assume this means the concha bowl is the part that will transfer the vibrations.

2. There's no air in between since this actuator want to rest as directly into the cranium as possible to achieve bone conduction... This presumably means measurement microphones won't do us any good here.

Am I just hopeless in the end?

 

[ForceMajeure]

Bone conduction is a subject that is actually still undergoing some development in the hearing aid world. Usually it used when there is an issue with the middle ear (where small bones that transfer and amplify the mechanical energy from the tympanic membrane to the inner ear) not being able to transfer air conduction but the inner ear still works.

I have seen a few reviewers trying to measure the bone conduction effect on a frequency response with a regular mic. Unfortunately this is not the way to do it ...(although those bc application indeed had a very minimal impact on the actual sound but that's not the point).

You need an artificial mastoid to measure the vibration and convert it into an actual forced transferred that mimics the human skin and bone structure (there are solutions from B&K)... you can then convert the output to an actual dbspl graph. ofc it's like the iec711. it is a approximation of the avg human skin and bone but everyone is different.

The first issue you face is of course mechanical impedance mismatch when using a bc device inside a shell and not directly against the bone structure. This will vary depending the contact area, where it lands with each individual etc...so you'll have a hard time evaluate it's effectivity.

It will vary between folks depending on how the shell rests against their skin and bone. Basically it's like not having a good seal with a tip...So the effect one would experience will be vastly different from the effect another one would depending on anthropomorphic features and matching with the shell. And that's assuming you've taken into account the optimal scenario.