Ad
Ad
Ad
Pages: [1]   Bottom of Page
Print
Author Topic: Shooting a Lamp's Light Spectrum  (Read 7346 times)
xpatUSA
Sr. Member
****
Offline Offline

Posts: 300



WWW
« on: April 02, 2013, 01:17:27 PM »
ReplyReply

Of course, most seem to prefer flash or halogen but I use 5000K LED floods for my own table-top purposes. They have a slight blue cast but my camera WB seems to handle it quite well. Next time, I might try the 4000K model which has less of a peak.



Do lamps age? Does their spectral emission change over time? Should you wait for CFL's to warm up? (Rhetorical questions, no need to reply!)

Here's how I checked my lamps spectrum, just out of curiosity:

I had made myself a toy spectroscope some time ago:



The razor blades form a slit. You shine a lamp near the slit, and a lens inside straightens out the resulting beam which then hits a diffraction grating at the other end - thereby creating a spectrum. Do please Google for the many ways to DIY.

I then shot the spectral image in RAW. Here's how it looks converted to a JPEG:



But how it looks in a JPEG is not the actual measureable spectrum. For that, the RAW has to be converted to a linear (gamma=1) TIFF file which is then opened with an analysis application. For a while I had some trouble finding the right applications and settings. Finally Dave Coffins' DCraw and ImageJ were good applications although early attempts didn't seem quite right. One problem was that the blue spike wasn't tall enough, compared to the manufacturers graph. Soon found out that the mighty Foveon sensor can capture quite a wide gamut of color and the spectrum was mostly out-of-gamut for sRGB.



So, I saved the file in ProPhoto color space (very, very wide), 16-bit, all 19MB of it from a 3.4MP camera!

Then it took a while to find the best way to measure the spectrum profile in ImageJ. I found that, although a linear 16-bit TIFF is displayed on-screen as a color image, the measuring tools think it is not. So I used ImageJ's "type" menu to convert the image to RGB. It still didn't look quite right until I discovered that the conversion was using "weightings" instead of a straight 1/3 deal.

So, finally:



From which not a lot can be deduced other than it can be done and was a lot of geeky fun!!

. . . and, hopefully, of interest to y'all.

« Last Edit: April 02, 2013, 07:17:32 PM by xpatUSA » Logged

best regards,

Ted
xpatUSA
Sr. Member
****
Offline Offline

Posts: 300



WWW
« Reply #1 on: April 02, 2013, 07:20:33 PM »
ReplyReply

Here's a Sylvania Mini-Craft CFL and a GE Halogen:




Logged

best regards,

Ted
MrSmith
Sr. Member
****
Offline Offline

Posts: 902



WWW
« Reply #2 on: April 03, 2013, 03:25:12 AM »
ReplyReply

I did wonder how spikey LED lighting systems are, I guess your example will cause problems when shooting certain colours?
I know flurescant tubes are not ideal either unless you get the Philips graphilux ones.
Logged
thierrylegros396
Sr. Member
****
Offline Offline

Posts: 653


« Reply #3 on: April 03, 2013, 05:58:50 AM »
ReplyReply

Very interesting, indeed !

Wanna check my new Solux 4700K to see if they are really that good  Wink

Thierry
Logged
xpatUSA
Sr. Member
****
Offline Offline

Posts: 300



WWW
« Reply #4 on: April 03, 2013, 11:30:17 AM »
ReplyReply

I did wonder how spikey LED lighting systems are, I guess your example will cause problems when shooting certain colours?
I know flurescant tubes are not ideal either unless you get the Philips graphilux ones.

I used the Sylvania Mini-Craft CFLs for several years without any problems, but I don't shoot models with demanding skin-tones under fixed lighting, for example. Watches and eBay knick-knacks are very forgiving, color-wise and human vision very forgiving also.

Also, most subjects, even the Macbeth card patches have fairly broad spectral reflectivity which helps with spikes quite a bit. Often, my eBay stuff is shot with the LED floods and an overhead Philips TL950 fluorescent and fill flash. But, like I said, my subjects are very forgiving . .
Logged

best regards,

Ted
ErikKaffehr
Sr. Member
****
Offline Offline

Posts: 7495


WWW
« Reply #5 on: April 03, 2013, 01:29:35 PM »
ReplyReply

;-) I Interesting stuff ;-)

Erik
Logged

Jim Kasson
Sr. Member
****
Online Online

Posts: 855


WWW
« Reply #6 on: April 05, 2013, 12:17:04 PM »
ReplyReply

Good work, Ted!

Have you tried the Cree 94 CRI LEDs?

Jim
Logged

EricV
Full Member
***
Online Online

Posts: 130


« Reply #7 on: April 05, 2013, 01:17:29 PM »
ReplyReply

Very interesting, thanks for posting.  Can you go one step further and convert your "inches" axis to approximate wavelength?  If you want to be precise, you could calibrate this with a light source (like low-pressure sodium) which emits only one narrow band, or you could find some narrow wavelength filters to insert in front of a broadband source.  For fun, you could look at the spectrum from a candle and see how it compares to a blackbody distribution.
Logged
xpatUSA
Sr. Member
****
Offline Offline

Posts: 300



WWW
« Reply #8 on: April 05, 2013, 03:19:00 PM »
ReplyReply

Very interesting, thanks for posting.  Can you go one step further and convert your "inches" axis to approximate wavelength?

For fun, you could look at the spectrum from a candle and see how it compares to a blackbody distribution.
Thanks, Eric,

Sorry, I won't be converting the x-axis although I've done it in the past and it can be a pain involving quite a bit of photoshopping. Usually the manufacturer's data gives the wavelength of the peaks and that doesn't change, being a property of the materials.

Taking a pic of a candle spectrum with my old Sigma SD10 would be quite a challenge, I reckon (they're not noted for their sensitivity . . ).

Jim,

No, haven't tried those, got a link?

Logged

best regards,

Ted
Jim Kasson
Sr. Member
****
Online Online

Posts: 855


WWW
« Reply #9 on: April 05, 2013, 04:01:42 PM »
ReplyReply

No, haven't tried those, got a link?

Sure. Look here to start. Not cheap, though. I use these in ceiling cans.

Jim
Logged

xpatUSA
Sr. Member
****
Offline Offline

Posts: 300



WWW
« Reply #10 on: April 05, 2013, 05:47:54 PM »
ReplyReply

Sure. Look here to start. Not cheap, though. I use these in ceiling cans.

Thanks, I went looking round the Cree site and it's hard to match a lamp model (at a glance) with their actual LED's but the green graph on page 3 here looks pretty good - almost like real daylight!

http://www.cree.com/~/media/Files/Cree/LED%20Components%20and%20Modules/XLamp/Data%20and%20Binning/XLampXPEHEW.pdf

Logged

best regards,

Ted
Jim Kasson
Sr. Member
****
Online Online

Posts: 855


WWW
« Reply #11 on: April 05, 2013, 06:07:06 PM »
ReplyReply

...but the green graph on page 3 here looks pretty good - almost like real daylight!

Yeah, it looks like they manage to tame that 440nm peak in the higher-CRI versions. If they use the 90 CRI LEDs in the cans, they may do some more filtering to boost the CRI still further. You can't see the LEDs directly in the PAR 38 lamps I have; it looks like the bulbs are mounted facing backwards into some kind of mirror, although I can't really tell for sure.

Jim
Logged

Jack Hogan
Full Member
***
Offline Offline

Posts: 242


« Reply #12 on: April 11, 2013, 01:38:34 AM »
ReplyReply

One problem was that the blue spike wasn't tall enough, compared to the manufacturers graph. Soon found out that the mighty Foveon sensor can capture quite a wide gamut of color and the spectrum was mostly out-of-gamut for sRGB.

Hi Ted,

Fun.  Two questions come to mind.  Most Bayer sensored DSLRs have gamuts much smaller than ProPhoto and Foveons have more approximate colors.  It'd be interesting to understand how much of the curves are limited by the sensor.  Also, it'd be fun to compare the differences between Bayer and Foveon results.  Do you have access to a Bayer sensored DSLR?

Jack
Logged
BartvanderWolf
Sr. Member
****
Online Online

Posts: 3681


« Reply #13 on: April 11, 2013, 03:11:12 AM »
ReplyReply

Most Bayer sensored DSLRs have gamuts much smaller than ProPhoto and Foveons have more approximate colors.  It'd be interesting to understand how much of the curves are limited by the sensor.

Hi Jack,

I don't think it's as much a smaller gamut, but rather a tri-chromatic sampler (with an IR-blocking filter) and demosaicing, that will cause the approximation. To get a better idea of what influence the sensor response has on the captured spectrum, one should calibrate the response based on a known spectral emission, e.g. an incandescent light source of known color temperature.

Amateur astronomers can use a simple tool like the StarAnalyser to analyze spectra. The SA-100 can be used on a DSLR with an adapter and step-up ring on a 85-150mm lens to show the spectrum of a point or line of light. Of course that will then combine the spectral absorptions of the lens + sensor-array.

Cheers,
Bart
Logged
xpatUSA
Sr. Member
****
Offline Offline

Posts: 300



WWW
« Reply #14 on: April 11, 2013, 11:51:19 AM »
ReplyReply

Thanks, Gents,

My SD10 has a one-screw removable dust/IR/UV blocking filter. Very popular with IR people.

So the next step is to increase the camera pass-band by removing the filter.

Then the sensor response will be:



I've been looking at DCraw's -D (document) option lately and it may well be a better solution for measuring spectral response in ImageJ with no attempts to convert the raw data into any color space.
« Last Edit: April 11, 2013, 01:14:02 PM by xpatUSA » Logged

best regards,

Ted
Pages: [1]   Top of Page
Print
Jump to:  

Ad
Ad
Ad