I want to make an invisible light floodlight, that together with a cheap webcam, will work as complete darkness CCTV system

The IR LEDs i see on most commercial CCTV cameras appear to be lit with slight red-ish color to the naked eye. Therefore i am looking for longer wavelength light

What are the wavelengths of those slightly visible IR LEDs and of completely inisible ones, but that cheap CMOS cameras are still sensitive to their light ?

And how would you implement such LED floodlight ? It have to be the cheapest solution possible in Lm/initial cost (efficiency don't matter, long term reliability don't matter - I expect to catch the culprit on CCTV within a few days and take the system down). I think of a protoboard with 5mm LEDs from Ebay. Any better suggestions ?

At first, if it is not visible, there will be just zero lumens. So you only may talk about radiated power...

The IR LED's are of two kinds, one is in the 800nm range, other 900+nm range.
Seconds what you need is something, which could be picked up by the cameras. And usually you want the same camera to be able to pick up the normal daylight as well. And that is the main problem - you would need a lens with exactly the same focal length for visible, as well as for the IR. And that is something you will never get with any cheap camera (there you are lucky it is barely maintains the visible range somehow in reasonable focus).
This is the main problem and in fact the main reason, why the supplemental IR illumination uses the near, shorter wave IR, which then slightly glows red/purple to the eye in the darkness, the eye just exhibits already some response to these wavelengths.

And for the camera: Normally the IR tends to excite the "blue" sensor (double wavelength) and normally that would destroy the color performance of the camera. That is the reason, why most cameras contain a filter blocking all but the visible range. So when you want to use it to "see" in the IR (and so sacrifice the color accuracy), you would have to remove that filter. With some models that is feasible, as the filter is a separate film behind the lens, so could be quite easily removed (however the camera then needs complete optical readjusting). But some cameras have that filter as a layer directly on the sensor chip and this you obviously can not remove.
The purpose made security cameras very often offer a 4-"color" sensor with one "color" being the IR (so IR-R-G-B), or with an optical system allowing to control the presence of the IR filter. This allows to use decent RGB for daylight, but a dedicated IR mode for "night" operation, beside offering way better resolution and image quality (but obviously costing way more).

And by the way I do not see much point to really make the artificial illumination not visible at all - the fact it is visible shows to the potential intruder the site is being monitored even in the darkness, so it would be way better for him to leave the place without even attempting to do anything bad there...

Medved, as usual, is 100% right!
I add my experience: longer wavelength ir led are less efficient in radiated power, so if you can get (cheaply) 850nm 1w/sr (I'm using this http://it.rs-online.com/web/p/led-a-infrarossi/7961772/?searchTerm=7961772&relevancy-data=636F3D3126696E3D4931384E525353746F636B4E756D6265724D504E266C753D656E266D6D3D6D61746368616C6C26706D3D5E5C647B367D247C5E5C647B377D247C5E5C647B31307D2426706F3D313426736E3D592673743D52535F53544F434B5F4E554D4245522677633D4E4F4E45267573743D3739363137373226 for cctv project), you won't find any 900+nm led with the same output.

Thanks for the replies !

I am thinking not as much radiated power (Watts), but power that the camera can see (according to its sensitivity curve). So it is sorta "Lumens" in a way, but measured in what the camera can see, not what we can see

I do have a suspect (which i am expecting to catch in the video)

 - I expect him to arrive in darkness. So if i can't get both dayligh and IR vision, i will choose the IR. (at most, i can just install side by side a 2nd cheap webcam focussed for day. But i think i won't need it)

 - I dont want to scare him away, i do want the evidence. Therefore making the light visible is not an option

The place is lit very dimly by light spill from a 6500K LED light installed nearby, but the amount of light present there is far too little to aid the camera

I do consider however, using the slightly visible 800nm range LEDs, and relying on the light spill to mask their light - place them under a large prismatic diffuser that will diffuse whatever visible light to an intensity that is not noticable over the background illumination level

I am considering this if the 800nm LEDs have significantly higher "camera Lumens" than the 900nm ones, otherwise i'll prefer going with the 900

I found this :
http://www.ebay.com/itm/DIY-940nm-850nm-IR-infrared-light-20W-High-Power-LED-Chip-Bead-Lamp-4-Floodlight-/261970719113

They are all 20W electrical
850nm - 1.8W output
940nm - 1.5W output

In Watts the output is not that different, but what about "camera Lumens" ? What is the sensitivity of CMOS webcams to either wavelength ?

I tried opening my webcam and have not seen any such film - It is probably on the chip here. The camera however does see "something" under the light of a few 5mm 850/880nm LEDs which i tried, so i guess with 20W (electrical) it will see fair enough ?

(What if i get an ordinary 20W 240V LED floodlight and replace the chip to this, with a matching driver if needed)

1+W of IR? Why so much?
Normally about 10mW (so about 10x 5mm LED's driven by about 20mA) is enough for the 2m distance.
If you want longer distances, a simple "webcam" will be anyway useless, mainly on account of very poor resolution for that purpose, the picture quality will be just unusable (on a normal webcam resolution you can not anymore identify anyone face from more than about a meter or two or so; plus with your case the colors will be completely missing and/or distorted, what will make the identification way harder, so the usable distance shorter, regardless of the illumination).

For the wavelength, you will have to try with your camera, each sensor model is different. But if it has the filter (some cameras were sold on purpose without that filter, as there were head movement following systems for many games using that camera sensing three small IR LED's mounted on special glasses) and that filter is not possible to take off, I don't think it will really work.

1
The camera is not bad at all in its picture quality. I can recognise faces in its footage from more than 2 m distance - under visible light atleast

2
And i want the IR light as bright as possible, as i expect that with good enough light, i will be able to see about as well in IR too (in black and white)

3
I want to light up the entire field of vision of the camera, not only a limited spot in front of it, as i dont know from which side exactly the culprit will approach..

For that i want the light coming in wide angle, so multiply the 10mW at least by (solid angle of camera field of vision / solid angle of 5mm LED). Actually more, considering that i dont use any precise optics to limit the light exactly to the field of vision of the camera - A lot of it will just be spilled where the camera cant see

4
I considering putting the camera and light behind a window

The window gets somewhat fogged in the cold. I can partition the light from the camera on the inside (a simple cardboard partition between the window panes through which each of them is "looking", on the inside of the window), so that the camera won't be blinded by the light refelcted from the fogged window. But the fogged window means losses and drop in contrast, which i want to try to overcome with better lighting if i choose this setup



The camera is an inexpensive webcam. i doubt it is designed in any way to track head movements

It does see fair enough under the light of 850/880 nm LEDs, so i guess it will be ok at least for the 800nm

For the 900nm im not sure - The camera does see the LED of TV remote lighting up, but it does not appear very bright. Now thing is, i dont know whether it is becuse the 900nm is on the low edge of the sensitivity of the camera, or just because the LEDs in the remotes are driven low (perhaps low duty cycle in the modulation ?) and/or are just inefficient..

The remotes drives the LED in two ways (depend on the system; the later is way less common):
- a rather short bursts (some 10's of ms) of 50% duty ratio at 38/45/51(/whatever the carrier frequency is), this keyed by the data packet, so again another 50%, so the overall duty ratio is about 25%. These often use no more than 20..30mA LED drive
- Very narrow (few us wide) rather high current (100mA or more) pulses spaced about 1ms (the spacing then carries the data bits), so very low duty ratios.

In both cases either the duty ratio is low and the burst length quite short, so the camera will have difficulties to capture that.

And other remark: The window glass uses to be treated to reflect IR (to prevent heat from going into the house during hot summer or escape the house during winters) and that coating may affect that IR as well (however the 1um is very close to the visible in that matter, but some such glasses even have a green tint as a side effect of that layer, so already influence the visible)

Otherwise let's try it, there is nothing better I may offer, I'm affraid...

1+W of IR? Why so much?
Normally about 10mW (so about 10x 5mm LED's driven by about 20mA) is enough for the 2m distance.
If you want longer distances, a simple "webcam" will be anyway useless, mainly on account of very poor resolution for that purpose, the picture quality will be just unusable (on a normal webcam resolution you can not anymore identify anyone face from more than about a meter or two or so; plus with your case the colors will be completely missing and/or distorted, what will make the identification way harder, so the usable distance shorter, regardless of the illumination).

For the wavelength, you will have to try with your camera, each sensor model is different. But if it has the filter (some cameras were sold on purpose without that filter, as there were head movement following systems for many games using that camera sensing three small IR LED's mounted on special glasses) and that filter is not possible to take off, I don't think it will really work.

uhm... 10mW is an led diode, so 100mW is quite sufficient for 2m.
But, if you have to control more space, you'll need a lot of more power.

Also camera sensitivity is an big factor.

If you want to control larger space, you need not only more power, but mainly way better camera resolution. And by the resolution I do not mean only the Mpixel figure, but mainly the optical accuracy, capable to project all the details onto the sensor chip. And there I would see the weakest point of the "webcam" products. And the main reason, why even the cheapest real security cameras cost many times more than the webcams...

If you want to control larger space, you need not only more power, but mainly way better camera resolution. And by the resolution I do not mean only the Mpixel figure, but mainly the optical accuracy, capable to project all the details onto the sensor chip. And there I would see the weakest point of the "webcam" products. And the main reason, why even the cheapest real security cameras cost many times more than the webcams...

Obvious 
At the moment, i experimented some low profile webcam (trust) with good results (ok, poor optics, but for 2m^2 they are good, 10 ir led with 120° aperture), now i replaced them with raspberry pi noir camera (better results, lower optical distortion and more sensitivity).
Now i can't afford 350€ for an (basic) axis camera, but i saw some night images and, extra price means extra quality in this case.

That I understand, but I've seen few comments from the police detectives, all concluding the pictures from the webcams or so are practically useless, even quite stupid lawyer can blow that out from the court as "too many people would look the same, so it does not prove it was my client". And even for the police it is not that helpful - again the confidence of the person matching is very low.

That I understand, but I've seen few comments from the police detectives, all concluding the pictures from the webcams or so are practically useless, even quite stupid lawyer can blow that out from the court as "too many people would look the same, so it does not prove it was my client". And even for the police it is not that helpful - again the confidence of the person matching is very low.

Absolutely true, that's why i'm using webcams (or similar) only for small spaces.
I'm trying raspicam noir for backdoor night recording (with good results), but i need some optics over bare led.

Before getting OT, do you know anything over uv illumination?

Look into putting an IR pass filter such as LEE gels or similar ontop of your IR LED. Most humans are unable to see far-red light beyond 700nm. With such a filter the LEDs aren't dimmed more than a few % in the IR, but humans can't see anything. Search for a cutoff of 720nm. Better yet, if your IR emitter or camera has filter threads similar to a camera lens, just find the IR camera filter with said threads!

The problem is not the radiation in "visible" wavelength range, the LED itself already radiates only narrow band in the IR.
The problem is, the 850nm is so close to the visible, the eye already starts to see it - it sees the quite strong light as very dim glow, can not focus it that much, but in the darkness that is already sufficient to reveal the lantern.

..., the LED itself already radiates only narrow band in the IR.

I assumed that everyone here knows about the bell-shaped-curve of the spectral output of general purpose LEDs. Most LEDs don't put out descrete wavelegths but "bands", usually broad, even the single color LEDs (well just until maybe a few years ago). "IR" LED spectral output has the left foot of a bell-shaped-curve tapering off to just a fraction of a % in the visible 700nm area. This is the only reason we see anything. This is why I suggested a cutoff filter over the IR emitter. Now I would suggest one with a cutoff @ 750nm since filters (even the dichroic interference type filters) also have a bell-shaped transmission or rejection profile.

The problem is, the 850nm is so close to the visible, the eye already starts to see it - it sees the quite strong light as very dim glow, can not focus it that much, but in the darkness that is already sufficient to reveal the lantern.

Ok same..Naturally I speak for myself: I put a dark 720nm cutoff filter (available for IR photography) with a grating between my eye & a DX coated MV lamp. With the lamp just inches away I can only see the 691nm & 700nm "spectral lines" (from the DX coating) and LITTLE ELSE beyond 710nm.

It's similar with an HPS arc-tube. I can't see the strong 818.3/819.4nm NA doublet, see my spectra pic, but this is easily recorded by CCD or cmos chips (sans any filter).

So, Ash... You can use an HPS flood and filter it with a 750nm cutoff gel (cutoff meaning: cutoff visible but transmit IR). If your camera receiver is indeed IR capable, then voila!