Code
library(dplyr)
library(tidyr)
library(photobiology)
library(photobiologyWavebands)
library(photobiologyFilters)
library(photobiologyLEDs)
library(ggspectra)
photon_as_default()
library(ggplot2)
theme_set(theme_bw())
Note
This page is still a draft, not only of the text. Data are incomplete and may contain errors even if I have been careful. As any work in progress the contents of this page may drastically change, even in some of the conclusions. The methods are described in detail, and, I hope already fully reproducible.
Warning 1: Ultraviolet-A radiation dangers
Ultraviolet-A (UV-A) radiation at high irradiance can be damaging to eyes and skin, especially when there is little or no visible light together with it. Eye protection is recommended in all cases. It is safer to use only the so called “black light” corresponding to UV-A1 at wavelengths near 365 nm. However, even in this case when using flashlights or any other sources of UV-A that emits a concentrated beam, use of safety goggles is a must. When selecting goggles or safety spectacles pay attention to their UV-protection rating (e.g., UV400 CE, or increasing protection from ANSI U2 to U6 markings) as some goggles are certified to protect only from mechanical impacts (e.g., ANSI Z87+ with no U marking) (Figure 1). It is important that goggles and other safety eye wear provide protection also from the sides.
Folded code chunks
In this page code chunks can be “folded” so as to decrease the clutter. Above each plot, table or other R-code output you may find one or more “folded” code chunks signalled by a small triangle followed by “Code”. Clicking on the triangle “unfolds” folded chunks making the R code that produced the printed values, plots or tables visible. Clicking on the same icon on an “unfolded” chunk, folds it hiding the code.
The code when visible in the chunks can be copied by clicking on the top right corner, where an icon appears when the mouse cursor hovers over the code listing.
The </> Code drop down menu to the right of the page title makes it possible to unfold and fold all code chunks in the page and to view the Quarto source of the whole web page.
Introduction
I have been using a full-spectrum modified Olympus E-M1 camera since 2016. The modification to increase the range of wavelength sensitivity was based on replacement of the built-in “UV IR cut” filter with UV and NIR transparent glass. The RGB filters of the Bayer array remained on the sensor. This results in a colour cast in the VIS region and false colour in the UVA and NIR regions.
The colour cast in normal visible photography can controlled by means of an external filter úsed to substitute for the removed internal one. In the tests described here, no such filter was used, and the extended range of walengeths to which the camera responds after modification was available.
Interpreting the false colours in UV and “far-red” is rather difficult as they tend to vary vary to some extent between camera models. The interpretation would be easier with knowledge of the spectral response of the RGB channels of the modified camera. A low cost approach is to use LEDs emitting at different wavelengths instead of a very expensive spectrograph (see page Camera objectives for digital UV photography). Of course the approach has limitations as the emission peaks of LEDs are rather broad and the nominal wavelength of emission can differ from the actual one. The “loose” LED wavelength specifications are common in generic LEDs from small suppliers, but these suppliers do carry what seem to be “off-type” LEDs or from unusual wavelength bins. The actual wavelength needs to be measured, but otherwise these “generic” LEDs provide a way of obtaining LEDs emitting at wavelengths in between the typical ones.
A recently announced commercial spectral imaging system is based on the use of nine types of LEDs (eight colours and white) switched-on sequentially instead of an spectral camera (Rayn vision system camera from Rayn Growing Systems). It is based on a 1/4” 1‑Megapixel (1280 x 800) sensor (OV9281, Omnivision, Santa Clara, CA, USA), as used in some cameras for the Raspberry Pi available for around 25–35 u$s ready to plug-in with lens included. The camera sensor choice would be reasonable for a cheap system as it has a global shutter. However, other sensors from OmniVision could offer higher definition (5 Mpix or 9 Mpix), better dynamic range and better low-light and infrared sensitivity. The example images from the Rayn system are of individual small plants, the ability to image multiple plants or single large plants with high enough resolution can be extremely useful for phenotype screening and condition assessment, respectively.
The Rayn vision system uses eight wavelengths ranging from 475 nm to 940 nm at the LEDs’ peak of emission. This system seems overly expensive at a list price of over 8000 u$s. It is delivered with open-source software. On the other hand, the approach seems easy to implement at a relatively low cost. In fact, nothing prevents the use of a camera with a sensor with higher resolution and better light sensitivity. The range of the Rayn system LEDs does not cover the UVA-violet-region and the blue region only in part, but extends into the NIR.
Using a full-spectrum converted or other suitable camera it would be possible to extend the wavelength range further than that provided by any single hyper-spectral camera that are commercially available. With suitable filters on the camera lens to block the light used for excitation and powerful enough LED-based light sources one could image fluorescence excited by different wavelengths, i.e., obtain spectral data using fluorescence as a reporter instead of reflection. This approach has been used in vitro to characterise plant photoreceptors. More relevant to whole plants, chlorophyll fluorescence informs about photosynthesis activity and state and has also been used as a reporter of UV transmittance of leaves’ epidermis.
Thus I aimed at preliminarily testing the suitability of camera, lens and LEDs. The specific aims of the test were: 1) to test the range of response of the camera with a lens I normally use, 2) measure the relative sensitivity of the red. green and blue sensor channels to radiation of different wavelengths, and 3) assess if spectral imaging would be feasible using LED illumination as sources of “monochromatic” light.
Methods
R code
The code below attaches R packages to be used and reads in the measured response data and the metadata describing the LEDs and filters used.
Code
# read files saved using RawDigger: RGB values, shutter speed, aperture and ISO
main_samples.df <- read.csv("sampled-regions.csv")
green_samples.df <- read.csv("green_samples.csv")
new_samples.df <- read.csv("new_samples.csv")
nir_samples.df <- read.csv("NIR_827_samples.csv")
new_samples.df$Id <- new_samples.df$Id - 1L
pixel_samples.df <- rbind(main_samples.df, green_samples.df, new_samples.df,nir_samples.df)
# colnames(pixel_samples.df)Code
# each raw image file has a matching entry with information about LED and filter used
if (file.exists("frames-metadata.csv")) {
frames_metadata.df <- read.csv("frames-metadata.csv") |>
mutate(LED.Id = paste(LED.Id, Wavelength, sep = "."))
} else {
# save template for manual editing
filenames.df <- unique(pixel_samples.df["Filename"])
write.csv(filenames.df, file = "frames-metadata.csv", row.names = FALSE)
}
# colnames(frames_metadata.df)Code
# The description of each LED is in a separate file
if (file.exists("leds-metadata.csv")) {
LEDs_metadata.df <- read.csv(file = "leds-metadata.csv")
} else {
# save template for manual editing
LEDs_metadata.df <-
unique(frames_metadata.df[c("LED.Id", "LED", "Wavelength", "wl.real")])
write.csv(LEDs_metadata.df, file = "leds-metadata.csv", row.names = FALSE)
}Code
# Join data with metadata
camera_response.df <- left_join(pixel_samples.df, frames_metadata.df)
# colnames(camera_response.df)Camera and lens
A mirrorles digital camera (E-M1 (Mk I), Olympus, Japan) camera from 2013, converted to full-spectrum (by DSLR Astro TEC) was used. The camera has a 4/3” sensor with an effective resolution of 16 Mpix. For the tests a Sigma 30 mm 1:2.8 DN A lens was used. This modern autofocus lens transmits UV-A1 radiation rather well, much better than any other modern lens for this system. For the test at 340 nm, a different lens with deeper UV transmission was used: a Soligor 35 mm 1:3.5 from the 1960’s on M42 mount (s.n. 9700437) was attached to the camera with a helicoid adpater and aperture and focus set manually.
The camera was mounted on a copy stand (MS-Repro, Novoflex, Memmingen, Germany) through a macro focusing rail (MFR-150, Sunwayfoto, Zhongshan city; China). On the camera an Arca-type base-plate specific to the camera model (generic from AliExpress) ensured alignment.
LEDs
On the copy stand a solid support for the LEDs was assembled using Arca type rails and clamps (Haoge HDR-400, FITTEST D290 angle clamps, XILETU QR-50B two way clamps, generic NDR-200 nodal rails, all from AliExpress), and small “magic arms” (UURig, AliExpress). The LEDs were kept in place by magnets making swapping them fast and easy. LEDS in SMD packages were used, soldered onto star-boards, and the star boards mounted onto small aluminium heat-sinks (SA-LED-113E black anodized, OHMITE, Rolling Meadows, Illinois, USA) using self-stick pre-cut thermal transfer pads (LP0001/01-TG-R373F-0.25-2A-ND, T-Global Technology, Lutterworth, U.K.). A small self stick magnetic attachment point (13 mm, Magsy, Fryšták, Czech Republic) was fixed to the back of the heat sinks. Small but strong magnets (13 mm NdFeB pot magnets, Magsy) were used on the “magic arms”.
The 38 LEDs used are described in Table 1. The emission spectrum of each LED used was measured with an array spectrometer (Maya 2000Pro, Ocean Optics, USA) and are shown in the Appendix. If you intend to use any UV-A or strong violet LEDs please read Warning 1 beforehand.
The measured wavelength, in some cases differed by more than 10 nm from the nominal value, these differences were much smaller for LEDs obtained from big distributors like Mouser, Digikey, Farnell, TME, etc., and large only for some AliExpress sellers and LED types. Using nominal wavelengths results in a ragged curve, while using actual wavelengths resulted in a very smooth curve.
Code
knitr::kable(LEDs_metadata.df[order(LEDs_metadata.df[["Wavelength"]]) , c("Wavelength", "LED", "Type", "I.max", "Code", "Supplier", "Notes")],
row.names = FALSE, align = "cll")| Wavelength | LED | Type | I.max | Code | Supplier | Notes |
|---|---|---|---|---|---|---|
| 340 | SMD 3535 | 350 | DUV340-SD356L | Roithner | ||
| 367 | Osram | SMD 4444 | 1000 | LZ1-10UV00-0000 | Mouser | |
| 385 | Osram | SMD 4444 | 1200 | LZ1-10UB00-00U4 | Mouser | |
| 406 | Osram | SMD 4444 | 1200 | LZ1-10UB00-00U8 | Mouser | |
| 413 | Epileds | SMD 3535 | 700 | YTLEDONE | ||
| 432 | Epileds | SMD 3535 | 700 | YTLEDONE | ||
| 449 | Epileds | SMD 3535 | 700 | Juntang | ||
| 452 | Osram | SMD 4444 | 1300 | LZ1-10B202-0000 | Mouser | |
| 457 | Epileds | SMD 3535 | 700 | Juntang | ||
| 467 | CREE | SMD 3535 | 700 | |||
| 493 | CREE | SMD 3535 | 700 | XP-E | OTdiode | |
| 500 | straw hat | 700 | Leds-global | |||
| 516 | Epileds | SMD 3535 | 700 | Juntang | ||
| 523 | Osram | SMD 4444 | 1200 | LZ1-10G102 | Mouser | |
| 530 | straw hat | 350 | Leds-global | |||
| 547 | Epileds | SMD 3535 | 700 | Juntang | ||
| 548 | straw hat | 700 | Leds-global | |||
| 556 | straw hat | 700 | Leds-global | |||
| 580 | SMD 3535 | 700 | ICYYGO | a | ||
| 596 | Epileds | SMD 3535 | 700 | YTLEDONE | ||
| 605 | Epileds | SMD 3535 | 700 | YTLEDONE | ||
| 629 | Epileds | SMD 3535 | 700 | YTLEDONE | ||
| 630 | Epileds | SMD 3535 | 700 | YTLEDONE | ||
| 663 | Osram | SMD 7070 | 1000 | LZ4-40R208-0000 | Mouser | b |
| 680 | SMD 3535 | 700 | JH-3535R12G42-T8A-680 | Ledguhon | a | |
| 700 | SMD 3535 | 700 | JH-3535IR6G42-T8A-700 | Ledguhon | a | |
| 730 | Osram | SMD 4444 | 1200 | LZ1-10R300 | Mouser | |
| 740 | CREE | SMD 3535 | 700 | JH-3535IR6G42-T8A-740 | Ledguhon | |
| 780 | SMD 3535 | 700 | JH-3535IR6G42-T8A-780 | Ledguhon | a | |
| 805 | CREE | SMD 3535 | 700 | JH-3535IR6G42-T8A-810 | Ledguhon | |
| 850 | CREE | SMD 3535 | 700 | JH-3535IR6G42-T8A-850 | Ledguhon | |
| 852 | Osram | SMD 4444 | 1200 | LZ1-10R602-0000 | Mouser | |
| 905 | CREE | SMD 3535 | 700 | JH-3535IR6G42-T8A-910 | Ledguhon | |
| 940 | Osram | SMD 3535 | 1200 | LZ1-10R702-0000 | Mouser | |
| 940 | CREE | SMD 3535 | 700 | JH-3535IR6G42-T8A-940 | Ledguhon | |
| 1050 | Sanan | SMD 3535 | 700 | ICYYGO | a | |
| NA | no | NA | darkness | |||
| NA | Osram | NA |
The LEDs were driven using a custom-built constant current controller based of solid-state drivers (RCD-24-0.350, Recom Power, Gmunden, Austria) and voltage and current monitored.
The targets, the camera and the LED positions remained unchanged durings tests, carried out in a darkened room, keeping fluorescent objects at a distance and for UV and NIR using band-pass and long-pass filters on the camera lens, respectively.
White, black and pattern targets
Two white-reference targets were used side by side (@fíg-BW-targets). Each one a slab of white PTFE 6 mm-thick with the surface sanded with 600 grit wet-dry sand paper to obtain a matte surface. A 6 mm-thick black Nylon-slab painted with “black deep mat” camera paint (Tetenal kameralack Spray 105202, seems no longer available) was used as the black reference target. A small printed circuit board, with a black-coloured mask and a clear plastic film ruler were also added to facilitate focusing.
Targets used photographed in NIR shown after white balancing. Upper scale on the ruler shows centimetres and the lower one inches. The spacing of holes in the printed circuit board is 2.5 mm.
Image sampling
As the target and light sources, or camera were not moved, the sampling was done at the same two locations in every image. Two well illuminated areas, 500 x 500 pixels in size were sampled with RawDigger (RawDigger Research Edition, version 1.4.9 for Windows, LibRaw LCC, ), and the results of the sampling saved to two files, as the photographs were acquired in two separate sessions.
Additional tests
All the test images were obtained using auto-focus and auto-exposure settings with no manual exposure compesation. In some cases bracketing was used, but the bracketed were not better exposed than the ones exposed according to camera automation. Focusing worked in all cases, except 340 nm, where the Sigma lens did not transmit enough to obtain a well exposed image even at ISO 26500. At 940 nm, autofocus hunted to some extent, but did work most of the time. The target included high contrast features.
Results
Normalise data
Code
camera_response.df |>
mutate(RGB.sum = Ravg + Gavg + Bavg,
Ravg.rel = Ravg / RGB.sum,
Gavg.rel = Gavg / RGB.sum,
Bavg.rel = Bavg / RGB.sum) |>
select(Filename:Id, ISO:Aperture_Value, LED.Id:Bavg.rel) -> camera_response_rel.dfThe normalised data describe the relative sensitivity of the red (R), green (G) and blue (B) channels of the camera sensor at different wavelengths (Figure 5). At each wavelength, the average pixel values are expressed relative to the sum of the pixel values for R, G, B channels. These values even if not describing the quantum efficiency of the sensor, can help interpret the colours obserbed under nearl monochromatic light. This is specially useful outside the visible range of wavelengths that create a false-colour rendition on the photographs.
Code
camera_response_rel.df |>
select(-RGB.sum) |>
filter(! Filename %in% c("_U250126.ORF", "_U250126.ORF", # sigma lens blocks 340 nm
"_U250091.ORF", "_U250088.ORF", # fluorescence in image
"_U250128.ORF") # underexposed
& Id == 1) |>
pivot_longer(cols = c("Ravg.rel", "Gavg.rel", "Bavg.rel")) |>
ggplot(aes(Wavelength, value, colour = name)) +
geom_line() +
geom_point() +
# facet_wrap(facets = vars(Id), ncol = 1) +
wl_guide(ymin = -0.02, ymax = 0) +
scale_color_discrete(direction = -1) +
scale_x_wl_continuous(n.breaks = 6) +
scale_y_continuous(name = "Raw pixel signal (relative to RGB sum)")
As response values are normalised, an alternative presentation approach is to use stacked area plots (Figure 6).
Code
camera_response_rel.df |>
select(-RGB.sum) |>
filter(! Filename %in% c("_U250126.ORF", "_U250126.ORF", # sigma lense blocks 340 nm
"_U250091.ORF", "_U250088.ORF", # fluorescence in image
"_U250128.ORF") # underexposed
& Id == 1) |>
pivot_longer(cols = c("Ravg.rel", "Gavg.rel", "Bavg.rel")) |>
ggplot(aes(Wavelength, value, fill = name)) +
geom_area(alpha = 0.67) +
# facet_wrap(facets = vars(Id), ncol = 1) +
scale_fill_discrete(direction = -1) +
scale_color_discrete(direction = -1) +
scale_x_wl_continuous(n.breaks = 6) +
scale_y_continuous(name = "Raw pixel signal (relative to RGB sum)",
sec.axis = dup_axis(name = character(),
breaks = c(0, 1/3, 2/3, 1),
labels = c("0/1", "1/3", "2/3", "1/1")))
Discusion
Approach
The approach of using LEDs works well as long as the wavelength value used for calculations and expressing the observed RGB values is the exact one, measured on the LEDs actually used, rather than the nominal value reported in data sheets or informally in sellers’ descriptions (or measured on other LEDs of identical type and brand). The peaks of emission of LEDs are rather broad, with half maximum full width (HMFW) in the range 10 to 20 nm, limiting wavelength resolution of the obtained camera spectral response characterization.
As the results from the two different sample areas were almost identical, expressing the results as normalised relative sensitivities makes the method rather insensitive to small changes in luminance and setup geometry. The data are much easier to collect than actual quantum efficiencies but are not useful to predict responses under light sources that include a broad range of wavelengths.
For wavelengths at which camera sensor sensitivity is low or LED emission weak room stray light can be a problem to be aware of. In particular, under UV, visible fluorescence induced by absorbed UV radiation needs to avoided.
One thing to be aware of is that the voltage drop across LEDs depends on wavelength: as the energy per photon increases towards shorter wavelengths the voltage drop increases. Some sellers correctly advertise typical 3535 SMD LEDs as having a power rating of 1–3W. These LEDs are rated at a maximum current of 700 mA consistently across wavelengths, and consequently, IR LEDs are close to 1W power while blue, UVA and white emitting LEDs are close to 3W power. In contrast, Ledengine/OSRAM LEDs in the LZ1 series have a more even power rating across wavelengths as the maximum allowed current increases with wavelength.
LED suppliers
Buying LEDs from AliExpress sellers has the advantage of availability of LEDs emitting at a broad range of wavelengths, including those intermediate between the usual ones available through major distributors in small quantities from stock. On the other hand, while LEDs bought from distributors and reputable sellers are well described in technical data sheets, the information provided by AliExpress sellers is usually incomplete or even totally missing.
This means that wavelength tolerances can be broad and power ratings suspect in the case of some AliExpress sellers. However, in my experience in all cases LEDs from AliExpress sellers have been fully functional.
Camera response
The response curves obtained are similar to what can be expected by normalising those published for sensor detector arrays. By using a converted camera and extending the test into the UV-A and NIR regions, it is possible to see why the false colour for UV-A at 340 nm is yellow and purple at 385 nm. We can also see why at wavelengths of 840 nm and longer, images taken with this full-spectrum converted camera are monochromatic.
Next steps
The wavelength resolution in the range 950 to 1050 nm is less than at other wavelengths but as response seems flat the wavelenth coverage is now good enough. For quite a few LEDs I was using wavelengths from specifications, but I have measured them with a spectroradiometer and I have updated the computations, Figure 5 and Figure 6 with their true peak wavelengths. I still need to update the LED figures and table in the Appendix with the newly measured emission spectra and information on all LEDs used.
The method is fast and simple enough to be applied to other cameras, full-spectrum converted and off-the-shelf, for comparison. Automation would require electronic switching of LEDs in place of the current manual swapping of LEDs. Attaching them with magnets as I am doing helps a lot in speeding the process. The sensor response to exposure seems to be linear when extraction the raw sensor “counts” with RawDigger, so as long as there is no clipping, the normalised relative sensitivity of the three channels does not change wihtin \(\pm 2\) EV or more (the data points from bracketed exposures are in the figures but invisible because they overlap).
Ultimately, the aim is to develop an approach to spectral imaging based on the use of light sources of different wavelengths and a normal camera instead of a spectral camera. Thus, tests with plants, flowers and other objects will be needed. These objects are more difficult than the PTFE slabs and black-painted objects used for the tests reported here as they can fluoresce.
As a caveat, assembling the “image cube” possibly requiring alignment will be a further challenging step, possibly doable with plantCV or other software. So, image analysis will need to be also developed based on existing tools.
Conclusions
The range of usable wavelengths into the UV region is mostly limited by the lens I used, and in the NIR by the camera sensor.
In the UV the sensitivity of the blue channel rapidly decreases with decreasing wavelength reaching nearly zero while that of red and green channels become nearly the same (\(\approx 0.5\)), yielding the typical yellow false colour at 340 nm. In the NIR, at the opposite end of the spectrun at wavelengths of 840 nm and longer all three channels had equal sensitivity (\(\approx 0.33\)), as expected from the lack of false colour.
The camera and lens automation worked well across the whole range of wavelengths tested, except that autofocus was sluggish at 1005 nm.
Thus, using LEDs for low-cost hyperspectral imaging should be possible, but will require the development of a system for switching LEDs and triggering the shutter: custom assembled LED arrays of 12 channels are relatively easy to obtain and not too expensive but do not have the best geometry as LED chips of each channel are in a row in the array.
On paper at least, the Ryan “hyperspectral” camera system seems to use an unnecessarily weak camera sensor compared to the price and the sophistication of other components. I may be wrong as the sensor has fast readout and a global shutter, that taken together should allow acquiring the images very fast. It uses eight colours and white light. For the test presented here I used LEDs with peak of emission at 39 different wavelengths. In actual use for spectral imaging, in most cases a smaller set of wavelengths would need to be selected to reduce the amount of data as contrary to most true hyperspectral cameras, we also have high spatial resolution.
It may appear counter-intuitive, but assessing colour is easier than assessing overall reflectance as it does not depend on irradiance. As long as we have a set of greyscale references usable across all wavelength or if we can assume a linear sensor response to ligh, we can estimate reflectance. With whole plants, with leaves displayed at multiple angles and facing multiple directions, the irradiance at their surface is very difficult to keep uniform, even in a fully diffuse light field because of mutual shading. In contrast, mutual shading changes the light spectrum only minimally, if at all, when using a single-colour light source. This is in contrast with the change in spectrum caused by mutual shading of plants in sunlight and in white light. An important question is if a diagnosis/image-analysis approach based only/mainly on colour would be effective in detecting all the features of interest.
Appendix
Code
knitr::kable(frames_metadata.df[order(frames_metadata.df[["Wavelength"]]) , c("Filename", "Wavelength", "LED.Id", "Filter", "Lens")],
row.names = FALSE, align = "cll")| Filename | Wavelength | LED.Id | Filter | Lens |
|---|---|---|---|---|
| _U250126.ORF | 340 | DUV.340 | none | SGM30 |
| _U250128.ORF | 340 | DUV.340 | Tangsinuo Venus | SGM30 |
| _U250137.ORF | 340 | DUV.340 | Tangsinuo Venus | SLG35 |
| _U250139.ORF | 340 | DUV.340 | Tangsinuo Venus | SLG35 |
| _U250142.ORF | 340 | DUV.340 | Baader U | SLG35 |
| _U250091.ORF | 367 | LZ1.367 | none | SGM30 |
| _U250095.ORF | 367 | LZ1.367 | Tangsinuo Venus | SGM30 |
| _U250088.ORF | 385 | LZ1.385 | none | SGM30 |
| _U250102.ORF | 385 | LZ1.385 | Tangsinuo Venus | SGM30 |
| _U250081.ORF | 406 | LZ1.406 | none | SGM30 |
| _U250084.ORF | 406 | LZ1.406 | none | SGM30 |
| _U250075.ORF | 413 | Epileds.413 | none | SGM30 |
| _U250066.ORF | 432 | Epileds.432 | none | SGM30 |
| _U250061.ORF | 449 | Epileds.449 | none | SGM30 |
| _U250056.ORF | 452 | LZ1.452 | none | SGM30 |
| _U250051.ORF | 457 | Epileds.457 | none | SGM30 |
| _U250047.ORF | 467 | CREE.467 | none | SGM30 |
| _U250038.ORF | 493 | CREE.493 | none | SGM30 |
| _U250117.ORF | 500 | Weili.500 | none | SGM30 |
| _U250034.ORF | 516 | Epileds.516 | none | SGM30 |
| _U250023.ORF | 523 | LZ1.523 | none | SGM30 |
| _U260143.ORF | 530 | Weili.530 | none | SGM30 |
| _U250031.ORF | 547 | Epileds.547 | none | SGM30 |
| _U260147.ORF | 548 | Weili.548 | none | SGM30 |
| _U260152.ORF | 556 | Weili.556 | none | SGM30 |
| _U180189.ORF | 580 | CREE.580 | none | SGM30 |
| _U180190.ORF | 580 | CREE.580 | none | SGM30 |
| _U180191.ORF | 580 | CREE.580 | none | SGM30 |
| _U180192.ORF | 580 | CREE.580 | none | SGM30 |
| _U180193.ORF | 580 | CREE.580 | none | SGM30 |
| _U250015.ORF | 596 | Epileds.596 | none | SGM30 |
| _U260158.ORF | 602 | Epileds.602 | none | SGM30 |
| _U250010.ORF | 629 | Epileds.629 | Heliopan 25 | SGM30 |
| _U250019.ORF | 630 | Epileds.630 | none | SGM30 |
| _U250003.ORF | 663 | LZ4.663 | Heliopan 25 | SGM30 |
| _U250005.ORF | 663 | LZ4.663 | Heliopan 25 | SGM30 |
| _U180186.ORF | 684 | CREE.684 | none | SGM30 |
| _U180187.ORF | 684 | CREE.684 | none | SGM30 |
| _U180188.ORF | 684 | CREE.684 | none | SGM30 |
| _U180180.ORF | 702 | CREE.702 | none | SGM30 |
| _U180181.ORF | 702 | CREE.702 | none | SGM30 |
| _U180182.ORF | 702 | CREE.702 | none | SGM30 |
| _U180183.ORF | 702 | CREE.702 | none | SGM30 |
| _U180184.ORF | 702 | CREE.702 | none | SGM30 |
| _U180185.ORF | 702 | CREE.702 | none | SGM30 |
| _U252050.ORF | 730 | LZ1.730 | Heliopan 25 | SGM30 |
| _U252046.ORF | 740 | CREE.740 | none | SGM30 |
| _U252047.ORF | 740 | CREE.740 | Heliopan 25 | SGM30 |
| _U180218.ORF | 740 | CREE.740 | none | SGM30 |
| _U180219.ORF | 740 | CREE.740 | none | SGM30 |
| _U180220.ORF | 740 | CREE.740 | none | SGM30 |
| _U180221.ORF | 740 | CREE.740 | none | SGM30 |
| _U180198.ORF | 760 | CREE.760 | none | SGM30 |
| _U180199.ORF | 760 | CREE.760 | none | SGM30 |
| _U180200.ORF | 785 | CREE.785 | none | SGM30 |
| _U180201.ORF | 785 | CREE.785 | none | SGM30 |
| _U180202.ORF | 785 | CREE.785 | none | SGM30 |
| _U180203.ORF | 785 | CREE.785 | none | SGM30 |
| _U252053.ORF | 800 | CREE.800 | Zomei IR760 | SGM30 |
| _U180210.ORF | 800 | CREE.800 | none | SGM30 |
| _U180211.ORF | 800 | CREE.800 | none | SGM30 |
| _U050005.ORF | 827 | CREE.827 | none | SGM30 |
| _U050006.ORF | 827 | CREE.827 | none | SGM30 |
| _U050003.ORF | 827 | CREE.827 | none | SGM30 |
| _U252037.ORF | 849 | CREE.849 | Zomei IR760 | SGM30 |
| _U180212.ORF | 849 | CREE.849 | none | SGM30 |
| _U180213.ORF | 849 | CREE.849 | none | SGM30 |
| _U252033.ORF | 852 | LZ1.852 | Zomei IR760 | SGM30 |
| _U252039.ORF | 907 | CREE.907 | Zomei IR760 | SGM30 |
| _U180216.ORF | 907 | CREE.907 | none | SGM30 |
| _U180217.ORF | 907 | CREE.907 | none | SGM30 |
| _U252025.ORF | 943 | CREE.943 | none | SGM30 |
| _U252030.ORF | 943 | CREE.943 | Zomei IR760 | SGM30 |
| _U252020.ORF | 952 | LZ1.952 | none | SGM30 |
| _U252021.ORF | 952 | LZ1.952 | Zomei IR760 | SGM30 |
| _U180208.ORF | 1003 | CREE.1003 | none | SGM30 |
| _U180209.ORF | 1003 | CREE.1003 | none | SGM30 |
| _U250110.ORF | NA | none.NA | Tangsinuo Venus | SGM30 |
| _U250111.ORF | NA | none.NA | Tangsinuo Venus | SGM30 |
| _U250112.ORF | NA | none.NA | Tangsinuo Venus | SGM30 |
| _U250113.ORF | NA | none.NA | Tangsinuo Venus | SGM30 |
| _U250122.ORF | NA | OSLON.NA | none | SGM30 |
Code
filters.mspct[c("Heliopan_Red_25_SH_PMC_2.2mm_30.5mm",
"Zomei_IR760_2.1mm_30.5mm",
"Baader_U_filter_1.0mm_48mm",
"Tangsinuo_Venus_2mm_30.5mm")] |>
autoplot(facets = 2,
annotations = list(c("-", "summaries", "boxes", "labels"),
c("+", "wls.labels", "peak.labels")))
Code
leds.mspct[c("Roithner_DUV340_SD353EL",
"LedEngin_LZ1_10UV00_365nm",
"LedEngin_LZ1_10UB00_00U4_385nm",
"LedEngin_LZ1_10UB00_00U8_405nm",
"Epileds_3W_460nm",
"Epileds_3W_470nm",
"Epileds_3W_495nm",
"LedEngin_LZ1_10DB00_460nm",
"Epileds_3W_420nm",
"Epileds_3W_430nm",
"CREE_XPE_480nm",
"Epileds_3W_520nm",
"Weili_3W.nominal.525nm",
"Weili_3W.nominal.550nm",
"Weili_3W.nominal.555nm",
"Epileds_3W_560nm",
"Epileds_3W_590nm",
"Epileds_3W_600nm",
"Epileds_3W_620nm",
"LedEngin_LZ4_40R208_660nm",
"LedEngin_LZ1_10R302_740nm",
"LedEngin_LZ1_10R602_850nm")] |>
autoplot(idfactor = "LED",
range = c(300, NA),
annotations = "colour.guide") +
theme(legend.position = "none")
Code
leds.mspct[c("Roithner_DUV340_SD353EL",
"LedEngin_LZ1_10UV00_365nm",
"LedEngin_LZ1_10UB00_00U4_385nm",
"LedEngin_LZ1_10UB00_00U8_405nm")] |>
autoplot(range = c(300, 450),
idfactor = "LED",
annotations = list(c("-", "summaries", "boxes", "labels"),
c("+", "peak.labels")))
Code
leds.mspct[c("Epileds_3W_460nm",
"Epileds_3W_470nm",
"Epileds_3W_495nm",
"LedEngin_LZ1_10DB00_460nm",
"Epileds_3W_420nm",
"Epileds_3W_430nm")] |>
autoplot(range = c(375, 575),
idfactor = "LED",
annotations = list(c("-", "summaries", "boxes", "labels"),
c("+", "peak.labels")))
Code
leds.mspct[c("CREE_XPE_480nm",
"Epileds_3W_520nm",
"Weili_3W.nominal.525nm",
"Weili_3W.nominal.550nm",
"Weili_3W.nominal.555nm",
"Epileds_3W_560nm")] |>
autoplot(range = c(450,650),
idfactor = "LED",
annotations = list(c("-", "summaries", "boxes", "labels"),
c("+", "peak.labels")))
Code
leds.mspct[c("Epileds_3W_590nm",
"Epileds_3W_600nm",
"Epileds_3W_620nm",
"LedEngin_LZ4_40R208_660nm")] |>
autoplot(range = c(550,725),
idfactor = "LED",
annotations = list(c("-", "summaries", "boxes", "labels"),
c("+", "peak.labels")))
Code
leds.mspct[c("LedEngin_LZ1_10R302_740nm",
"LedEngin_LZ1_10R602_850nm")] |>
autoplot(range = c(650,NA),
idfactor = "LED",
annotations = list(c("-", "summaries", "boxes", "labels"),
c("+", "peak.labels")))