Question: Question: How do I save the calibration data?

mokeymark is asking a question about spectral-workbench: Subscribe to answer questions on this topic

mokeymark asked on February 17, 2016 15:05
1,160 | 0 answers | #12701


What I want to do or know

I just put together the mini foldable spectroscope and captured an image of the spectra from a fluorescent light. I uploaded it, selected the cross-section, and clicked the calibrate button. When I hit the "Apply" button, I get the notification "The tag you've applied couldn't be saved, but it's been run locally. Clear it now." and the "Save" button just displays the spinning circle icon forever. I can't save a calibration.

This is on my desktop computer, running the latest Chrome browser. I also tried Internet Explorer with the same results. On Chrome on my Galaxy Note 4 phone, the browser just crashes whenever I try to perform a capture.

Background story



21 Comments

SpectralWorkbench_screen_capture.jpg


Hmm, your spectrum looks calibrated! It shows nanometer values, even in your screenshot. I'm not sure what the error was, but I checked the page myself and it looks ok now. Is there something missing?

Is this a question? Click here to post it to the Questions page.


The calibration looks like it's way off. Look at where the peaks are. This is a fluorescent light, so it looks like everything is skewed towards the UV end. I'm obviously doing something wrong...


Oh, sorry, i didn't look at whether it was a good calibration, just if the calibration itself seemed to save. Checking now...

So, two things stand out to me:

  1. I don't see any calibration operation in the Operations table. I'd think that means you copied an existing calibration in the Capture interface, but we recently deprecated that feature, so that doesn't seem likely. I'm not sure how your data got calibrated, actually -- any idea? Maybe somehow it got calibrated but the tags didn't save, but ... how'd the data get saved? Mysterious. See for example, this spectrum: https://spectralworkbench.org/spectrums/71660 which has a linearCalibration operation (actually it has two, so you should probably delete one of them).
  2. I see two crossSection operations. This is just not typical, so i'd remove them and just do a single crossSection.

You know, I wonder if somehow the range: operation is causing it to think you have calibrated data. Could you maybe remove that for now? I'd work on a fork of your data but that seems to be erroring for some reason. :-(

Is this a question? Click here to post it to the Questions page.


Thanks Warren, I'll start from scratch and try a fresh calibration. I'll be stuck in meeting purgatory for the next couple hours, but I'll jump back in this afternoon. We're having an open house on Saturday the 27th in my science building, and I'm giving a spectrometry demonstration. I cut 50 of the foldable mini spectrometers out of black card stock, and I want to have Spectral Workbench up on the giant monitor for the demonstration. I have a week and a half to become an expert. I'll also have an Ocean Optics spectroscope set up, with cuvettes of different solutions.


Oh, that sounds so great!

I'm curious, if you have access to an Ocean Optics, perhaps @ygstcu and/or @stoft would be interested in a calibrated scan of a Solux lamp they've been using for gain calibration tests... perhaps after your open house has passed.


I would be more than happy to help out!


While there will be some variation between individual lamps, even the spectra of one sample, taken with a commercial spectrometer, would be useful; and hopefully confirming. The Solux halogen lamp of interest is the 4700K; the only Solux model with a smooth, broad-spectrum response. If you find one to measure, allow 5+ minutes warm-up for stability. Thanks for your interest and support. -Dave


I will try to find a SoLux 4700k to measure! I could be mistaken, but those look like a direct replacement for the bulbs in my odd kitchen fixture, and two of them just burned out.


I bought mine from a Solux distributor, but I assume ones like this: http://www.amazon.com/18003-Daylight-Halogen-Degree-Kelvin/dp/B0002GS4OG are probably the same (although this one is 50W and 35W is more than enough). The 2-pin MR16 is a fairly common plug. You'll need a 12V supply rated for halogens and the bulb power.


Good deal. Gotta love Amazon. I'll use a laboratory power supply for the testing.


Also would be nice ..... while you have the Ocean Optics spectrometer available, capture the best resolution spectra you can of a 2700K CFL as that would make an interesting comparison for baseline noise, SNR, FWHM resolution and differentiation of the double-green Hg spike typical of CFLs.


Very good idea. I also have a large selection of high-CRI white LED's. Range from I think 2500k to 5500k. I want to compare results from Ocean Optics with cellphone/foldable spectrometer. I'd be happy to share those results as well.


Hi, @mokeymark - I'm tracking the bug here: https://github.com/publiclab/spectral-workbench/issues/338

Thanks for reporting, we hope to have a fix soon!


I tried the calibration again, on a fluorescent bulb spectra captured with my cell phone and the foldable spectrometer. I cleared the web browser cache, etc., cropped the image, and uploaded. I can't get the B1 and G2 points to line up, and the wavelengths are off quite a bit.


Fluorescent_calibration_issues.jpg


Hi, @mokeymark - I have a good theory as to the issue with your spectrum -- that the uploaded image is too big for our data storage system. I think we can get around this through not storing unnecessary precision (in https://github.com/publiclab/spectral-workbench/issues/335) but we're still finalizing that change, so please hang in there a few more days.

In the meantime, you could help us to test this theory by saving a lower resolution (say, 1000px wide, max) version of your un-cropped spectrum image, and posting a link here. Thanks for your patience and for identifying this bug.


This is the lowest resolution I can go on my cell phone (Galaxy Note 4). I believe it's 1152 x 2048

20160223_093745.jpg


Hi -- I believe this is resolved in v2.1.10 due to better precision management. Please try re-uploading the image and saving; I did so with a large image here: https://spectralworkbench.org/spectrums/72364

Thanks!


Yes! It worked with a full resolution image. Thank you!!!


Ah, I'm so pleased.


Log in to comment

0 Answers

Sign up or Login to post an answer to this question.