LED choice for video using microscope
Moderators: rjlittlefield, ChrisR, Chris S., Pau
-
- Posts: 142
- Joined: Tue Dec 09, 2014 3:25 pm
LED choice for video using microscope
CREE now have this LED http://www.cree.com/LED-Components-and- ... /XLamp-XPL but, as others have pointed out, once you get to a colour temperature of light with no colour cast (around 5500K-6000K) the CRI drops. But, when I have looked at the spectral power distribution, the higher CRI LEDs have more in the same wavelengths as the lower ones with a better output from 680nm but they are all missing 480nm. http://www.cree.com/~/media/Files/Cree/ ... ds-XPL.pdf
Which would make the best source for video? Brightness is key and colours do not have to be shifted if the colour of the LED doesn't have a yellow or blue tint. I think this is really important when shooting video as it is only 8 bit (256 colours) so doesn't stand up well to colour grading.
The single die of this LED just fills the field of view.We've tried the 20W CREE but this has 4 emitters and three are wasted.
Which would make the best source for video? Brightness is key and colours do not have to be shifted if the colour of the LED doesn't have a yellow or blue tint. I think this is really important when shooting video as it is only 8 bit (256 colours) so doesn't stand up well to colour grading.
The single die of this LED just fills the field of view.We've tried the 20W CREE but this has 4 emitters and three are wasted.
Andy, there must be a way to alter your white balance in the camera--if so, that would correct for the color temperature of the light. This is assuming you don't have mixed light sources, such as some natural light plus some LED light.
The color temperature of the 90-CRI is in the neighborhood of an incandescent light bulb, and appears to have the best spectral curve (matches most closely with an incandescent bulb). If you can adjust your white balance to this temperature range, it may be the best choice.
ETA: There was a recent thread on high-CRI daylight balanced LEDs. I'm not sure if their output is high enough for your needs, but it is worth a look. http://www.photomacrography.net/forum/v ... hp?t=29779
The color temperature of the 90-CRI is in the neighborhood of an incandescent light bulb, and appears to have the best spectral curve (matches most closely with an incandescent bulb). If you can adjust your white balance to this temperature range, it may be the best choice.
ETA: There was a recent thread on high-CRI daylight balanced LEDs. I'm not sure if their output is high enough for your needs, but it is worth a look. http://www.photomacrography.net/forum/v ... hp?t=29779
User zzffnn uses 40W Cree light, described in this topic to take videos of aquatic organisms and results seem to be quite successful.
-
- Posts: 142
- Joined: Tue Dec 09, 2014 3:25 pm
Correcting a strong yellow or blue tint by white balancing in camera means that quite severe post-processing is done at this time as two colour channels have to be shifted by large values.
So, there is a trade-off being have to do a large WB correction versus having a higher CRI LED. I'm wondering which way to lean as the higher CRI LEDs mostly get their better rating by largely having more in the wavelengths of light that are present in the brighter LEDs which are more neutral in their colour tint at around 5500-6000K.
Incandescent bulbs have spectral power distribution with all wavelengths present and do not have the 480nm hole of most LEDs.
So, there is a trade-off being have to do a large WB correction versus having a higher CRI LED. I'm wondering which way to lean as the higher CRI LEDs mostly get their better rating by largely having more in the wavelengths of light that are present in the brighter LEDs which are more neutral in their colour tint at around 5500-6000K.
Incandescent bulbs have spectral power distribution with all wavelengths present and do not have the 480nm hole of most LEDs.
- rjlittlefield
- Site Admin
- Posts: 23603
- Joined: Tue Aug 01, 2006 8:34 am
- Location: Richland, Washington State, USA
- Contact:
I worry that you're being misled by that description.Andy Davies wrote:Correcting a strong yellow or blue tint by white balancing in camera means that quite severe post-processing is done at this time as two colour channels have to be shifted by large values.
When the camera does white balancing, it is working with the raw sensor data. That data has a linear relationship with light intensity, which allows color balance to be corrected without artifact except for somewhat increased noise level in any channel that has to be greatly expanded.
The problem is completely different when you try to adjust color by post-processing a JPEG or TIFF image. The difficulty there is that pixel values have a strongly nonlinear and largely unknown relationship with the original light intensities, so it's very difficult to get accurate corrections at all brightness levels.
Summary: color correcting in camera is OK, color correcting during raw conversion is OK, color correcting from JPEG or TIFF can have big problems.
--Rik
Here is some small detail which is usually missed.
Camera sensor channels have different quantum efficiency. Therefore, excess of light in certain spectral ranges can be compensated by lack of sensitivity and otherwise.
So, higher CRI by itself works better for human eye, which has much wider dynamic range than camera sensor, which gives it more room for compensation. Therefore, technically, for digital photography it is more important to have better spectral evenness, normalized according to per channel quantum efficiency of sensor.
However, it requires pretty much effort to plot all these data and calculate it.
Camera sensor channels have different quantum efficiency. Therefore, excess of light in certain spectral ranges can be compensated by lack of sensitivity and otherwise.
So, higher CRI by itself works better for human eye, which has much wider dynamic range than camera sensor, which gives it more room for compensation. Therefore, technically, for digital photography it is more important to have better spectral evenness, normalized according to per channel quantum efficiency of sensor.
However, it requires pretty much effort to plot all these data and calculate it.
Asha, that is exactly why I'm talking about the hidden issues of compensation.
For example, here we can see an example of uneven spectral response of each channel of Sony CMOS sensor (see Figure 2).
Same amount of photons, reached sensor surface, produces least signal in Blue channel, a bit larger signal in Red and the largest in Green.
Which means, that histogram of Blue channel of original RAW image will always be shifted closer to zero (left), while Green will always be shifted right, compared to Red somewhere in the middle. Then, camera (or RAW development software) compensates for it when producing an actual "human-viewable" image. Doing that, camera takes in account white balance setting. In a bit oversimplified words, it multiplies each pixel value in each channel by certain number (taken from white balance setting). And less this multiplier is, better (smoother) gradation that channel will have. Also, it will have less noise, because lower bits, having worse signal/noise ratio will not be shifted into midtones.
So, practically, it means, that certain excess of blue is probably still okay, if there is still enough greens and reds in light to properly expose each channel without using large multipliers to get proper white balance.
For example, here we can see an example of uneven spectral response of each channel of Sony CMOS sensor (see Figure 2).
Same amount of photons, reached sensor surface, produces least signal in Blue channel, a bit larger signal in Red and the largest in Green.
Which means, that histogram of Blue channel of original RAW image will always be shifted closer to zero (left), while Green will always be shifted right, compared to Red somewhere in the middle. Then, camera (or RAW development software) compensates for it when producing an actual "human-viewable" image. Doing that, camera takes in account white balance setting. In a bit oversimplified words, it multiplies each pixel value in each channel by certain number (taken from white balance setting). And less this multiplier is, better (smoother) gradation that channel will have. Also, it will have less noise, because lower bits, having worse signal/noise ratio will not be shifted into midtones.
So, practically, it means, that certain excess of blue is probably still okay, if there is still enough greens and reds in light to properly expose each channel without using large multipliers to get proper white balance.
And here is an article on the same topic, but written from a bit different point of view - regular photography with daylight (high CRI and high color temperature light source). It gives certain illustrations of histograms without suppressed green and with it. In case of LEDs, we don't need to suppress green light since they naturally have "green pit", while we do have extra blue to get better exposure of Blue channel.
So, my point is that we'd better look for LEDs with better deep reds and not so bad greens rather than chasing high CRI by itself.
So, my point is that we'd better look for LEDs with better deep reds and not so bad greens rather than chasing high CRI by itself.
Thanks for your response, Bushman.K! Between yours and Rik's posts, plus my own experience, it seems to me that the shift in light temperature would be easily compensated for, without too much loss in the perceived color "truth". In addition, just thinking about the Planck curve for varying Kelvin, the curves basically have the same shape, so it is simple corrective term... http://cimss.ssec.wisc.edu/goes/blog/wp ... nction.jpg
ETA: this was in response to your post at 8:13
ETA: this was in response to your post at 8:13
-
- Posts: 142
- Joined: Tue Dec 09, 2014 3:25 pm
Many thanks for all the interesting discussion. Thanks Rik for pointing out that white balance correction is done linearly in 14 bit at the time of shooting video.
So, I'm left pondering whether an LED such as this one Cree XLamp XP-L XPL Cool White LED 1079LM 200lm/W V6 1A at 6000K with a CRI of 65 is better or not than the less bright U3 or U2 at 3000K and 736LM with a CRI of 95?
CREE data sheets are here http://www.cree.com/~/media/Files/Cree/ ... ds-XPL.pdf
So, I'm left pondering whether an LED such as this one Cree XLamp XP-L XPL Cool White LED 1079LM 200lm/W V6 1A at 6000K with a CRI of 65 is better or not than the less bright U3 or U2 at 3000K and 736LM with a CRI of 95?
CREE data sheets are here http://www.cree.com/~/media/Files/Cree/ ... ds-XPL.pdf
I am wondering if the article should be read with caution--the camera he used has a Bayer filter (2 green, 1 red, 1 blue in a block pattern). Cameras with different kinds of sensor designs (eg a stacked sensor such as Foveon) may not behave the same way.Bushman.K wrote:And here is an article on the same topic, but written from a bit different point of view - regular photography with daylight (high CRI and high color temperature light source). It gives certain illustrations of histograms without suppressed green and with it. In case of LEDs, we don't need to suppress green light since they naturally have "green pit", while we do have extra blue to get better exposure of Blue channel.
So, my point is that we'd better look for LEDs with better deep reds and not so bad greens rather than chasing high CRI by itself.
Andy,
I wouldn't be so sure about "14-bit processing" in general case - for example, it is well-known fact, that Panasonic GH3 and GH4 cameras have different sensor readout bit depth in the same mode to achieve higher speed on GH4. It's just up to manufacturer. So, that should be a subject of test to make this statement in case of shooting video.
But even if it is not the full bit depth of sensor ADC, it is way better than working with compressed output (JPEGs or video stream).
Speaking of LED choice, I agree with Asha - judging by spectral distribution chart, warmer LEDs have better coverage of Red channel and just a bit smaller "green pit".
I wouldn't be so sure about "14-bit processing" in general case - for example, it is well-known fact, that Panasonic GH3 and GH4 cameras have different sensor readout bit depth in the same mode to achieve higher speed on GH4. It's just up to manufacturer. So, that should be a subject of test to make this statement in case of shooting video.
But even if it is not the full bit depth of sensor ADC, it is way better than working with compressed output (JPEGs or video stream).
Speaking of LED choice, I agree with Asha - judging by spectral distribution chart, warmer LEDs have better coverage of Red channel and just a bit smaller "green pit".
This article should not be taken as universal, for sure (I wouldn't call it "with caution" - it is just normal to understand limits of applicability of any scientific theory). But how much non-Bayer sensors and sensors with significantly different per-channel quantum effectiveness are on market?Asha wrote: I am wondering if the article should be read with caution--the camera he used has a Bayer filter (2 green, 1 red, 1 blue in a block pattern). Cameras with different kinds of sensor designs (eg a stacked sensor such as Foveon) may not behave the same way.