Jump to content
UltravioletPhotography

Cheap 365nm LEDs from eBay - spectral measurement


JMC

Recommended Posts

Having been inspired by Bruce's amazing UV light source(http://www.ultraviol...dpost__p__23920), I'd like to build a similar, with 5 different wavelengths, all individually controllable, and ranging between 340nm and 395nm.

 

I've not worked with the newer style LEDs before - to me, LEDs are little plastic domes with 2 legs sticking out the back, so before I tried to do anything with expensive components, I thought I would buy some cheap eBay 365nm LEDs to experiment with. Specifically to see what type of spectra they give, and what they are like to solder and work with.

 

The 365nm ones I bought were about £2 each, and were a typical 'no brand' LED from China. I did also buy a cheap DC-DC regulator, but that failed after about 20mins, so I drove the LED straight from the bench top power supply I have. As with many things in life, you get what you pay for.

 

SImple setup, soldered leads to one of the LEDs and attached it to the power supply. Turned it up until it came on and didn't look to be changing in brightness (at least in the visible spectra, and I was wearing UV safety glasses). Spectra measured with an Ocean Optics FX spectrometer, with a 600um fiber, and cosine corrector. Set up for Absolute Irradiance measures, and with the cosine corrector held about 1cm from the LED. I did 2 scans, one with the LED on, one with it off.

 

This was the full range scan;

post-148-0-51964100-1540293980.jpg

 

And zoomed in more on the UV and near visible region;

post-148-0-38308000-1540293979.jpg

 

I also noticed it seemed to have a bit of a bump in the visible, so a final plot, with a much lower max irradiance value to show that;

post-148-0-82708900-1540293981.jpg

 

So, a few observations, it's definitely not a narrow peak, and look to be maximum at about 370nm, with a long tail at longer wavelengths, and into the visible. Not sure what to make of the small bump in the visible. I don't think it is an artifact of the measurement. The noise at either end of the scan is due to the limits of where the spectrometer is sensitive.

 

Some caveats with the results. The spectrometer does need calibration for peak position, although based on current work, I expect this to be at most 1-2nm adrift for wavelength. I did not have the spectrometer on for long before taking the measures (had it warmed up fully, probably not). However the LED was easily visible to the eye, so it was producing lots of visible light. I think the tail to the left of the main peak is due to stray light in the spectrometer, so probably isn't real.

 

Strangely enough, given the shape of the irradiance curve, it looks like it might be quite a nice broad band UV source for 370-400nm. As the camera sensitivity drops, from 400nm to 370nm, and this rises the two phenomena may well counteract each other. Although for my final light source I would like LEDs with much better defined peaks in roughly 10nm intervals. I could imagine 20 or so of these in a grid pattern might be quite a useful near UV spotlamp, and be a cheap light source.

Link to comment

Jonathan,

I know its not ideal, but to me that looks really good for the cost. $2 and you got a 370 to 400nm light. Place a U340 2mm filter on front and many wavelengths are covered. You may just need to add the more expensive 340nm Leds and you would be all set.

Hopefully other will chime in, but how much voltage did you apply? My limited knowledge of the Leds seems to indicate the higher the voltage the broader or longer the wavelength. You may need to drive these with very low voltage to get a tighter 370nm peak.

Link to comment

Jonathan,

I know its not ideal, but to me that looks really good for the cost. $2 and you got a 370 to 400nm light. Place a U340 2mm filter on front and many wavelengths are covered. You may just need to add the more expensive 340nm Leds and you would be all set.

Hopefully other will chime in, but how much voltage did you apply? My limited knowledge of the Leds seems to indicate the higher the voltage the broader or longer the wavelength. You may need to drive these with very low voltage to get a tighter 370nm peak.

 

It is correct that LEDs normally tend to get longer wavelengths for more current and also with increased temperature

 

LEDs should be driven with a controlled current. The voltage above them is not a good parameter to judge how hard they are driven.

They are diodes and a small change in voltage will hive a huge difference in current.

Simple circuits with LEDs as for indication lamps normally consist of a LED with a current-limiting resistor in series.

That might be problematic for power-LEDs due to the high power-loss in the resistor.

Link to comment

It is really nice some of you have spectrometers that you test things out with and post results.

Ha ha, someone said to me the other day,

"Many folks try UV for about 3 months and then don't do it anymore"

I said,

"No, you are wrong, people try UV for 3 months, and then they buy spectrometers"

:D

Link to comment

Ok, so I was going off my nice old fashioned, analogue power supply, and I wasn't paying much attention to the voltage - between 3 and 4V. I didn't realise the voltage would shift the wavelength. I shall try and repeat with something a little more controlled.

 

David, you are right, with some 340nm in there, these could make a nice broad band UVA source. My premise with the full build though is to make something where I can tune the wavelengths individually, and create different 'colours' of UV. Ideally for that I'd be wanting something with narrower FWHM. These may well make a nice ring light or light panel by themselves though for 370nm to 400nm imaging, especially given the cost.

Link to comment

Jonathan,

Since you have a spectrometer. As per Cadmium uv photography senior grade status. You may actually be able to tune these LED in a controlled way to see what bands you get. You have 3 peaks and it may be possible if limited, that the Leds may give you 370nm, 380nm and 390nm.

In the 280bp10 filter thread we diverted discussion to 280nm leds. I found a cheapish one that when tuned did something similar at 255, 280nm and 300nm.

So you may be able to play with what you have and get out what you want.

 

Link to comment

Jonathan,

Since you have a spectrometer. As per Cadmium uv photography senior grade status. You may actually be able to tune these LED in a controlled way to see what bands you get. You have 3 peaks and it may be possible if limited, that the Leds may give you 370nm, 380nm and 390nm.

In the 280bp10 filter thread we diverted discussion to 280nm leds. I found a cheapish one that when tuned did something similar at 255, 280nm and 300nm.

So you may be able to play with what you have and get out what you want.

David, as a very eminent skin measurement scientist once said, "a fool with a tool is still a fool". I still feel as though I am in the foolish stage at times, especially compared to the depth of knowledge I see on here. When you say, "tuning the LED" in this context, sorry but I am a bit confused. This was one LED, so how is tuning done - by changing the voltage? I shall go back and read the 280bp10 thread.

Link to comment

Ok, so I was going off my nice old fashioned, analogue power supply, and I wasn't paying much attention to the voltage - between 3 and 4V. I didn't realise the voltage would shift the wavelength. I shall try and repeat with something a little more controlled.

 

The wavelength of LEDs is not widely tunable and then very dependent of temperature and current.

If you try to tune the wavelength of a LED you get an intensity variation that is very big. The wavelength shift is rather small, normally just a few nm.

 

A LED is NOT!!! a voltage driven device!!!

If you drive any LED with too much voltage it will change colour to infrared before the internally stored magical smoke is emitted.

The working operation time for this state is very short before the LED light fades to black.

Such an overheated LED will never recover.

 

A LED is normally current controlled.

Put a current limiting resistor in series with the LED please, or if your power supply has a well-behaving current limit control, use that to control the intensity.

 

For different currents and chip temperatures you get a slightly varying voltage over the LED due to the diode forward voltage and internal resistance.

When a LED is supplied with a variable voltage starting from zero and increasing, at first there is no current or light, until the diode forward voltage is reached. Then for a small further voltage increase the current is increasing very quickly only limited by the LED's internal resistance.

Link to comment

I love internally stored magic smoke. I have only seen it a couple times though. Once was when some CPU that were 300MHz were default set to 3Ghz. They were really fast. Then lost their magic smoke and never worked again.

 

Ulf, great advice. Its been 20 years since E and M for me, but I still remember V=IR. I don't know enough about LEDs though. I will watch more of these builds before I decide to jump in with my own.

Link to comment

Hehehe, I agree with Ulf (as I usually do) although I’m laughing at how emphatic the above post is.

 

I hate when good and sometimes expensive components are wasted due to misconceptions or wrong handling and wanted to send a very clear message, instead of just presenting the technical facts.

 

The process of putting the magic smoke inside the semiconductors that makes them do their magic is a quite complex task.

It can involve some short wave UV-light too:

https://en.wikipedia...let_lithography

 

Compared to that an UV-C wave is very Loooooong. ;-)

Link to comment

Not to interrupt the electronics discussion, but Nichia LED's are available directly from Nichia if anyone ever needs those.

Last I checked, they are about $10 for the average 365nm.

Link to comment
The ones I bought were described as having a forward voltage of 3.9-4.5V, and a forward current of 700mA. My power supply does have a current limit control, and I used that to control the brightness, while I kept the voltage at 4V. What I would say, is that the specs on the advert for them were wildly optimistic. Above about 100mA there was no change in brightness, so I decided against taking it any further.
Link to comment

The NVSU333A look interesting at 3640mW for 365nm and 4640mW for the 385nm Leds.

 

http://www.nichia.co.jp/en/product/uvled.html

 

Interestingly, I don't see any products under 365nm.

 

I also couldn't see a price though and its bounces for me when I click were to buy. So I think if I have to ask, I can't afford it.

 

Update

Found them on AliExpress for $180 each. Yep can't afford it.

Link to comment

The ones I bought were described as having a forward voltage of 3.9-4.5V, and a forward current of 700mA. My power supply does have a current limit control, and I used that to control the brightness, while I kept the voltage at 4V. What I would say, is that the specs on the advert for them were wildly optimistic. Above about 100mA there was no change in brightness, so I decided against taking it any further.

Jonathan its an exponential curve. So what may be 100mA at 4V which is close to the low end of the spec, may ramp up as you get closer to max voltage. That actually sound about right. But I don't think I would push them hard as the magic smoke would definitely be released.

Link to comment
£2 UV LEDs should be the better quality ones from China, there are LEDs a lot more cheaper but they are trash, emitting tons of visible light and heat but a fraction of UV. I guess your choice is a good compromise between price and quality. In regard to how much we should push the current through these LEDs, I strongly suggest keeping it under the designed value. There is some gain in brightness once you push it above 700mA say the its forward current was 700mA, but the power efficiency is lower, simply that from per wattage of electricity there is more heat. Too much heat build up the temperature and make the LEDs even less efficient. I would rather install more LED chips to make the whole thing brighter than pushing the limit of supplied current.
Link to comment

The NVSU333A look interesting at 3640mW for 365nm and 4640mW for the 385nm Leds.

 

http://www.nichia.co...duct/uvled.html

 

Interestingly, I don't see any products under 365nm.

 

I also couldn't see a price though and its bounces for me when I click were to buy. So I think if I have to ask, I can't afford it.

 

Update

Found them on AliExpress for $180 each. Yep can't afford it.

 

http://www.nichia.co.jp/en/product/uvled.html

Inquire below. A nice person replied to me from their Torrance, CA, USA location.

https://www.nichia.c.../inquiries.html

Here are a few of the prices I got from them in 2016:

NCSU033B - Ua - $10.68 <100

NVSU233A(T) - U365 - $10.68 <100

NVSU333A(T) - U365 - $59.52 1-49

You would need to check with them for an up to date quote.

 

NICHIA AMERICA CORPORATION

LOS ANGELES SALES OFFICE - Torrance, CA, USA

TEL +1-310-540-5667

Link to comment
  • 2 weeks later...
I built also 365nm LED 3Wx14 lamp. But they emit a lot of white light as well. in your spectrum I see it's only ultraviolet, did you use UV pass filter?
Link to comment

I built also 365nm LED 3Wx14 lamp. But they emit a lot of white light as well. in your spectrum I see it's only ultraviolet, did you use UV pass filter?

No filter was used Avalon, but they are visible when switched on. The spectra in my first post shows a good chunk of light above 400nm.

Link to comment
  • 1 month later...

A quick update. After looking unsuccessfully for 340nm LEDs at a reasonable price, I decided to go ahead with a proof of concept and build a ring light based just on the 365nm LEDs. I used a Savarin mould, as outlined in Bruce's post here; http://www.ultraviol...dpost__p__23920

 

I went ahead and bought another 10 LEDs from the same supplier (which turned out to be a problem as we shall see later). Anyway, I made up 2 banks of 8 LEDs wired in series, 16 LEDs in total. The aim being to run these in parallel straight from my benchtop power supply (I ran these at 26V and 0.5A, as they didn't get any brighter after that and I had not desire to burn them out). Here's how it looks.

post-148-0-09298600-1546723257.jpg

 

post-148-0-63358400-1546723259.jpg

 

Not meant to be fancy, just a quick mock up for proof of concept. When I switched it on, there was obviously a problem - the 2 banks of LEDs (same supplier, but not bought at the same time), looked different. Even in the visible, the light they produced there was a difference in appearance. I haven't run the spectra of them yet, as my spectroscopy kit is currently all packed away. So I have 2 banks of LED's with different colours, and most likely different UV output.

 

Does it actually work though? Again, quick proof of concept in the workshop this evening. Nikon d810, modified for UV by ACS in the UK with one of their proprietary filters. I think it's similar to a Baader U, but as has been mentioned before they do not share their transmission spectra for it. UV Nikkor lens. Aperture f22, ISO1250. Lens focused as closest possible distance. Whitebalanced using PTFE disk in Darktable.

 

First, a Satsuma. 20s exposure. Full image (resolution reduced for sharing);

post-148-0-99200200-1546723559.jpg

 

I was focusing on the top where the stem would attach. A closeup of where I was focused,

post-148-0-97872300-1546723562.jpg

 

And with the UV light off, but the room lights on, again 20s exposure;

post-148-0-61529200-1546723563.jpg

 

On the plus point I'm getting a UV image. There's a weird colour split in the whitebalanced satsuma image, almost diagonally from top left to bottom right, which I guess is down to the two batches of LED's having different output. I think my original batch is the good batch, and the second batch is much closer to 380nm, but I'll update this with actual spectra when I next get all the kit out to measure it.

 

The moral of this story. Don't scrimp on your LEDs, and buy enough for your needs all at once. I'd like to redo this now with 16 better quality 365nm ones, or atleast, all from the same batch. I also want to get some 340nm ones too

Link to comment

Just as you said it looks like you got mixed wavelength-types.

I think one of them is "365nm" type.

If you had mixed the two types when placing them the colours would have been better.

 

It looks like the Savarin mould is just a reflector in your build. It is a good idea to couple the small LED-PCBs thermally to the mould.

The rounded notches in the metal-PCB is there for mounting screws.

A thermal paste between mould and PCB is also a very good idea.

 

Ideally the the small LED-boards should be mounted against a flat surface like a heat sink or thicker aluminum sheet to spread the heat and make the layer of thermal paste thin.

 

A cool LED chip shine brighter and lives longer.

It is likely that the heat prevented you from getting more light with a higher current.

Link to comment

Ulf, had I known the 2 batches of 365nm LEDs weren't the same I'd have mixed them up (or just not used them). As it was I made one bank of 8 from one batch, and the other bank of 8 from the other batch, which I think is leading to the unevenness in illumination.

 

I'd planned on using thermal paste behind the LEDs to attach them to the savarin mould, however the one I got was not as sticky as I thought it was going to be. Once they were all soldered together I couldn't get them to sit flat with the paste alone, so I stripped it all off and used a small amount of double sided tape beneath each one and stuck it to the aluminium mould. There isn't much flat space on the mould so getting good contact for them all like this wasn't possible, especially as I wanted to angled them all slightly inwards. If I go ahead and make a 'mk2' version with different LEDs, I was planning on getting some thicker alumiinum rings machined with flat portions on them so I can get good contact points for the LEDs.

Link to comment

Please sign in to comment

You will be able to leave a comment after signing in



Sign In Now
×
×
  • Create New...