search

jdber1 / opendrop

OpenDrop pendant drop tensiometry software
GNU General Public License v3.0
31 stars 18 forks source link

Fails to converge #2

Closed kc112358 closed 8 years ago

kc112358 commented 8 years ago

I've double checked all the physical parameters I provided for the density of each phase, and for the inner diameter of the needle. However, the software is failing to converge on any pendant drop image I've taken. Is there any way to understand why by observing the image?

cyclohexane softened profile cyclohexane softened residuals capture

ricotabor commented 8 years ago

Hi!

Your drop image looks a bit grainy at the edges (i.e. the drop profile isn't smooth) This could be the issue, but without the clean image it's hard to check. Also, you should use the outer diameter of the needle (as this calibrates the length-scale, it's not a fitting parameter/part of the Young-Laplace solver), although this isn't the problem here, it will give you inaccurate results to use the needle inner diameter.

If you send a typical image (without the attempted fit overlaid) to opendrop.dev@gmail.com then we can do some troubleshooting and get back to you with more info on how to fix this.

cheers, Rico

kc112358 commented 8 years ago

Hi,

Thank you for the information about the needle diameter. I'll correct that.

My equipment (and camera) are not the best quality. I attempted to clean up the image by removing the background, which is likely why the edges are grainy. I've attached an un-edited image as well as an edited image with no profile fitting.

Thank you, Killian

On Fri, Mar 18, 2016 at 5:02 PM, ricotabor notifications@github.com wrote:

Hi!

Your drop image looks a bit grainy at the edges (i.e. the drop profile isn't smooth) This could be the issue, but without the clean image it's hard to check. Also, you should use the outer diameter of the needle (as this calibrates the length-scale, it's not a fitting parameter/part of the Young-Laplace solver), although this isn't the problem here, it will give you inaccurate results to use the needle inner diameter.

If you send a typical image (without the attempted fit overlaid) to opendrop.dev@gmail.com then we can do some troubleshooting and get back to you with more info on how to fix this.

cheers, Rico

— You are receiving this because you authored the thread. Reply to this email directly or view it on GitHub https://github.com/ricotabor/opendrop/issues/2#issuecomment-198572610

Killian O'Connell Colorado State University Biochemistry & Molecular Biology, 2015 Tel: (434) 466-2441 killian.oconnell137@gmail.com

ricotabor commented 8 years ago

Hi Killian,

Can you please forward the images to opendrop.dev@gmail.com - for some reason attachments don't seem to work on the github message thingy.

cheers, Rico

kc112358 commented 8 years ago

Hi Rico,

Of course. I've sent them to the address you provided. Hope it works.

On Mon, Mar 21, 2016 at 4:08 PM, ricotabor notifications@github.com wrote:

Hi Killian,

Can you please forward the images to opendrop.dev@gmail.com - for some reason attachments don't seem to work on the github message thingy.

cheers, Rico

— You are receiving this because you authored the thread. Reply to this email directly or view it on GitHub https://github.com/ricotabor/opendrop/issues/2#issuecomment-199509848

Killian O'Connell Colorado State University Biochemistry & Molecular Biology, 2015 Tel: (434) 466-2441 killian.oconnell137@gmail.com

ricotabor commented 8 years ago

For future reference: we resolved this issue. It was the quality of the original image that caused problems with finding the droplet profile - specifically the diffuseness of the interface. This can be solved by binarising the image in ImageJ (or similar) and then proceeding, where the quality of fit becomes very good.