Flats box

[Stargazer33]

As I have almost finished my Christmas project of building a mains power supply box, I got to thinking about what I could use the 12v supply sockets for.

Well, with Christmas being over and the shops/garden centres selling off their unsold Christmas lights, I thought, why not use a string of white LED Christmas lights to make a light box.

With a length of 100 - 240 LEDs you must get a reasonable amount of scatter to even out the light. I thought of wrapping the lights around the centre cyliner that the OTA would slip inside. The lights would then reflect off of the inner surface of the main box.

Something like this:

[Bizibilder]

If you make the box from white foamboard the reflected light system will work. I must admit that my own lightbox to a similar design only uses four LED's (one in each corner) and is quite bright enough!!

[Stargazer33]

Hi Roger,

So do you think a string of 40 LEDs would be too bright? I thought that I would set the camera to adjust automatically for the exposure setting or do you need a particular level?

I presume that foam board is card faced quickly looking at adverts for it on the web. Do you not find that the dampness makes it go soft? Or is it not outside long enough for the damp to effect it?

I think I remember seeing your post earlier in the year - probably stored in the back of my mind which is where I got the idea from!

[Bizibilder]

I find my DSLR is quite happy with the four LED's - I did make a dimmer switch but don't use it - just full on and set the camera to Av and it sets the exposure for me. Foamboard seems to be "plastic" rather than card - with mine I only take it outside at the end of the session just to take the flats. I doubt if it would survive sitting on wet grass for any great length of time, but I may be wrong.

[fwm891]

Roger - followed your design with 4 LED's - works fine with DSLR nad CCD's

[Stargazer33]

Looks like I need to put in an order for some Foamboard and some LEDs, holders and wire.

Does it matter what the wavelengths of the light output is for coloured pictures. I read a post that some LEDs don't put light in the red wavelength. Would this effect the flats or are they purely for counteracting dust bunnies and vignetting?

I was thinking in my design I may put a thin piece of Foamboard down the entire length of the four corners of the box and possibly along the four edges at the top. Ideally I would like a dome that is the same diameter as the box but I don't think I am going to find something like that - unless I can find a white mixing bowl or something of the right diameter.

[fwm891]

Use 'White light' LED's

[IanL]

I tried white LEDs and I wasn't satisfied that they were giving out sufficient red output (plenty of green and blue though). I then rebuilt using tri-colour LEDs (basically a red, green and blue LED in a single diffuser package) but in the end it didn't make a huge amount of difference, since:

a) The red led element had a much lower output than the blue and the green ones. I did think about re-doing the resistors to each element to try to balance them better but haven't got round to it yet.

My Canon DSLR is much less sensitive in Red (the detector is anyway I think, and the IR filter blocks a lot of red too).

Does it matter? I don't know yet, since building it the weather has been so bad I haven't been able to test it on real images!

It is pretty clear to me that if you are using a mono camera with individual filters it should not matter at all. You're going to shoot a separate flat for each filter, and provided your LEDs give out some amount of red light, you are just going to expose that flat for longer than the others to get a reasonable ADU count across the flat.

If you are shooting OSC, it is more of an issue since you cannot shoot a single exposure that will get comparable ADU counts across all three channels.

If your processing software is sane, it should be able to cope with a bayer-matrixed flat being applied to a bayer-matrixed light. Basically it should be separating the red, green and blue pixels of the flat in to separate images (without debayering it), normalising each set (mapping the ADU ranges to a number range between 0 to 1) and then applying the result to the relevant pixels in the light frame. If it does that then you should be okay, but I guess you could still have problems if the flat's red pixels are too dim.

If your processing software is dumb, it will debayer the flat and the light first, then normalise the flat and apply it to the light. In that case not having a full-spectrum light source will cause you some issues; specifically the debayered flat will have a colour cast (mine are very blue if I manually debayer them), and applying it to the light will give it a colour cast on the opposite side of the colour wheel (I would expect it to be far too red in my case). It is obvious why if you think about it; since the flat-field process assumes all pixels should be evenly illuminated, but my red pixels are much dimmer than the other two colours, then they get boosted way too much compared to the blue and green.

You can fix the incorrect colour balance during processing, but debayering then flattening is the wrong thing to do for other reasons too. Partly we use flats to deal with vignetting and dust bunnies, and a debayered flat will still do that. But also flattening deals with uneven pixel response. Not every pixel will respond equally, some will be dead, some will be 'hot', either full on or nearly so at all times, and all will have slightly different responses to incoming photons due to imperfect manufacturing of the chip, the CFA and the Microlenses (if a DSLR). Flats also compensate for these issues, and if you debayer, you are losing some of that information from the flat and the light (as the image pixels are derived from some combination of R, G, and B pixels in the raw image).

[Stargazer33]

Thanks for that Ian!

I did see some LEDs on the web that states their output is in the region of 5000K to 6000K. I know that an overcast day gives a light output of about 6500K. The manufacturer does call the LEDs 'Cool White' for this light level and say that they do give off a 'blue' light. I believe they are household bulbs too for fitting under kitchen cabinets etc.

I may look into fitting some 'Cool White' LEDs and some tungsten bulbs to balance out the red frequencies.

[IanL]

Trouble is that the colour temperature of LEDs can be misleading; even if they appear to be 'white' that doesn't mean they give out a continuous spectrum.

E.g. A single colour LED will only emit light at a specific wavelength based on the band gap, i.e. the energy level that the electrons jump between (and in the process of doing so emit a photon of a known wavelength in order to lose the energy difference between the levels).

So in my tri-colour LEDs, I can tune the output of the red, green and blue LEDs in the package by using different resistors until the light looks 'white' and to some extent achieve a cooler or warmer colour temperature by the same tuning process. Nonetheless, if I looked at the spectrum of the output, there would just be three emission lines, one in Red, Green and Blue.

Single white LEDs are a bit different. Typically the band gap is such that the LED emits photons somewhere in the UV wavelengths. These strike a phosphor coating inside the LED which absorbs the UV photons and then re-emits them as photons in the visible wavelengths. (It's the same process as you get in a fluorescent light tube or a low-energy CFL bulb, except in that case the UV photons are generated by passing a high voltage through low-pressure gas inside the tube. Those funky purple disco lights from the 80;s that make your socks and teeth glow bright white, and also the tubes in sunbeds have no phosphor coating or a much thinner/different one that lets most of the UV out without converting it).

Depending on the LED and phosphor coating, the range of wavelengths emitted by a white LED may cover a wide enough spectrum fully, or it may not. Unfortunately it mostly seems to be a process of trial and error since cheap white LEDs don't tend to come with a chart showing the spectrum (I expect you could go to a scientific supplier and pay a lot more for one that did). You don't actually need a continuous spectrum, but you do need to make sure you have sufficient output in wavelengths that match your CFA filters in a DSLR/OSC, or your filters for your mono camera. Where it gets really critical is if you are using narrowband filters, because in that case it would be pretty easy to end up with a light box that emitted no light in the wavelength covered by a particular filter.

[Stargazer33]

Thanks again Ian!

There is a lot more to this than you realise. No wonder commercial flats boxes are so expensive!

It looks like I will need to do a lot more investigation into which type of LED to buy. Or I may go down your route and get the tri-colour LEDs.

I would be using the light box with my Canon 1100D (unmodified) by the way.

Can I take a flat from the light box - when I have finished it - and measure the RGB wavelengths in a programme like Photoshop, Registax or DSS? If so then I could perhaps use those values to change the resistor values to get a better balance.

[IanL]

Can I take a flat from the light box - when I have finished it - and measure the RGB wavelengths in a programme like Photoshop, Registax or DSS? If so then I could perhaps use those values to change the resistor values to get a better balance.

You just use a histogram with the R, G and B curves show separately, Pretty easy to see what you've got. I'd recommend taking an overcast day t-shirt flat and comparing that histogram with the light box though. You won't get three even peaks at the same level even with a perfectly flat white source, since the camera elements have different sensitivities for the three colours, but you do then have a target to aim for in terms of colour balance.

[Stargazer33]

Great tip about the 'reference frame' flat Ian! I would have never thought of that.

As we are having plenty of overcast days (and nights) at the moment it shouldn't be a problem getting a reference frame!