Public Lab Research note

Honlian 60s LED nail polish hardener, various LED fluorescence updates

by warren | March 10, 2016 00:04 10 Mar 00:04 | #12823 | #12823

A nail polish hardening, USB-powered UV lamp at 398nm, for under $4. Read on!

What I want to do

I've been continuing to test cheap LED electronics for use in the Oil Testing Kit, even as we discuss the pros and cons of ultraviolet LED lighting in this great thread in a post by @stoft. One of the takeaways so far is that some published research shows that lower wavelength input (excitation) light can yield more distinctive data when comparing crude oil fluorescence spectra. That is, a 350nm light source causes more clear differences between different oils than a 400 or 450nm light source.

After @dhaffnersr posted a great comparison of LED vs. laser pointers, I ordered a 385nm and a 365nm high-brightness LED, and send one of each to @ethanbass in the hopes that he'll be able to test and compare them. I'm happy to order more, but you can get them here at about $8 each:

We're hoping lower wavelengths will also overlap less with the region we're trying to measure fluorescence in (also discussed in @stoft's post's comments, above).

Redesign of cuvette stack

Anyhow, on the construction side, I had been trying to build an enclosure for an Adafruit Trinket, an Arduino compatible tiny 3.3v microcontroller, which I was using to power a 405nm LED i'd bought. The stack (which would go inside the cuvette frame exterior I documented here](/notes/warren/01-25-2016/ongoing-design-of-compact-cuvette-frame)) looks like this, and might still be useful if we go with a smaller component (I did find some, which I'll talk about in a bit):




Cheaper, off-the shelf parts

But in the same comment thread above from @dhaffnersr's post, I'd found a few products which cost around $4, were USB powered, and had an on-off switch, which are designed to cure nail polish with UV light. These are great, and I ordered some to see what wavelength they are, and if they were bright enough. The answer: the white one pictured in the lead image is definitely bright enough, and is ~398nm. You can get that one for $3.84 here:

I also found one that is just a tiny USB stick with 2 LEDs on it, which was at about 403nm, but much dimmer. But a promising form-factor for very small work:


The data is here, showing the bigger white one at 398nm with and without a vial of fish oil (just to proof of concept the fluorescence) and the 403nm tiny board (called Perfect Prime):

Data here:

Anyhow, I then tried shoehorning the white LED device into the cuvette frame, such that it's bright enough, the button is accessible, and the cord fits. It worked all right, actually, and is MUCH less complex than the build above (note, there's a cuvette in there, with fish oil in it):


Also, kudos to Brian Degger (@briandegger) and Jeff Hecht (@jeffh) for suggesting the nail polish curing LEDs back in December:

Why I'm interested

Ideally the lower wavelength 365nm LEDs will work. But this one is already edging out of the important range (published research such as the Downare and Mullins 1995 piece discards data within 30nm of the excitation wavelength, so that'd get us 430nm+, whereas 365nm would get us 395nm+ -- although we don't know how narrow bandwidth their source light was). And this is a very very easy thing to use -- no soldering or electronics necessary, and it even has an auto-off after 30 seconds or so. So I just want to know, and who knows, maybe we could swap out the LED or something.

Questions and next steps

Can we reproduce oil grading using either one of these off-the-shelf LED setups, or with a lower wavelength LED? Can we make a very easy-to-construct, reliable and bright cuvette frame for fluorescence work?

We need to test out these various LEDs, and @dhaffnersr's recent post on the consistency of LED wavelengths also means we can't take the exact wavelength for granted between LEDs, even from a single manufacturer -- at least more than to within about 8nm. It remains to be seen what effect this has on fluorescence values.

I did this Help out by offering feedback!

People who did this (0)

None yet. Be the first to post one!


@briandegger @jeffh - thanks for the tips on LED sourcing! Thought you'd enjoy seeing this.

Reply to this comment...

@dhaffnersr and @stoft, if you'd like I could send you the lower-wavelength LEDs as well for comparison testing?

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

Reply to this comment...

Sure ... what's the LED's packaging style/type? SMD, leaded, ...?

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

Reply to this comment...

Hey Jeff! Sure I'd be happy to test any LEDs you want to send. I do want to throw an idea your way, I think using the higher wavelength LEDs are still a good idea and here's why, it's easy to design a circuit where you can control the output dc voltage like in a step motor, the concept begin that voltage is relative to wavelength and so we can control that, a lot easier than replacing LEDs all the time. What do you think?

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

Reply to this comment...

Unfortunately, LED wavelength does not change with applied voltage or current -- their center peak wavelength is not variable or controllable and PWM only controls intensity. Any correlation between observed Vf (forward voltage drop) and peak center wavelength is purely manufacturing variation.

That manufacturing variation results in different center peak wavelength as a result of bandgap variations in the semiconductor material; once constructed, it doesn't change. Also, the Vf forward voltage drop can be different between devices of the same product; same reason. The only two spectral changes are 1) a very slight aging at a rate of about 0.0001nm/hr of on-time and 2) with some devices, an increase in the FWHM (bandwidth) as current increases from the initial threshold voltage point to primary conduction with 20mA or more. With UV LEDs this later effect seems to be primarily a spread of the LED spectrum slope on the longer-wavelength-side, with little on the shorter wavelength side. However, once the current is set (fixed) even this effect does not change.

Reply to this comment...

Hey Dave, well I was thinking along the lines of when I built a wind speed indicator with a small dc motor and LCD screen attached and calibrated it by having a friend drive at different speeds while I logged the voltage as related to that speed (rough way of doing it but it worked.)

I understand what you are saying about the PWM increasing intensity, but that's the whole point with fluorescence, that's also why I hooked up with this person, Dr. Friedrich Menges in Austria, the developer of spekwin32. I also understand the manufacturing process, I use to be an assembler level 3 at KC automation (which is know bought out by Borgtwalt in Germany.) we did 90 percent in house, I built from the ground up, the 20 port cigarette smoke analysis machine and the Side Stream smoker analyzer(which analyzes second hand smoke.) that's what we did there building Lab and industrial quality high precision equipment.

This kind of work I have never done but I have always been very facsinated by it and a friend of mine had a sample of something that they wanted to see if I could find out what it was, and I just happened to come across this spectrometer kit, since I just didn't have 30,000.00 to get a decent one. now I'm hooked and I'm learning a lot, and I'm sure I will continue to learn and refine my prep and technique.

Perhaps your correct about voltage and wavelength begin separate, I'd like to know how some of these spectrometers can have their wavelengths adjusted, I'll find out.

Reply to this comment...

Hey Jeff, I saw your spectra of the nail polish machine and there it is the tell tail "hump," so I exported it and processed it and here is what it really looks like:


Thought you would be interested.

Reply to this comment...

Hi, Dave - what is the tell-tale hump? Do you mean from the LED input light? If so, it's at 398nm, not 385nm. What processing did you do?

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

Reply to this comment...

Hey Jeff, I use spekwin32 for all my spectral data processing, first here are 2 references that may help you!divAbstract

below is a new graph detailing what I am talking about and below that will be another reference graph detailing it even further



Reply to this comment...

It's me again Jeff, Dave H, sorry I posted the wrong graph i made, this is the right one


Reply to this comment...

Login to comment.