If you have a couple of micrometer stages for precise positioning, then you can couple light out of the SMA connector to free space with a lens, and then couple it to FC/PC connector from free space with another lens. Common glass lenses will work well with different wavelengths in visible and near-infrared ranges.

It is necessary to provide a very accurate calibration using a micrometer using a SMA Connector Specifications and Interface Dimensions specifically chosen (see compatibility table) data sheet .

There are.... There are several SLED sources, especially from Denselight that can emit wavelengths having a central wavelength of 1550nm and 1440nm with a bandwidth of 35-40nm. The output is coupled via FC fiber connector. If indeed you are using a separate LED, be assured there will be severe amount of losses, during coupling, but you will require a minimum of two axis micrometer screw stage to couple the single mode fiber end.

Thanks Alexander and Gerson for your suggestion, but my application does not allow to use any micrometer stage. I need a direct connection from the source nothing in between. I was interested in some company made LED source which allows FC connector to be connected. 

I have been using the M590L3 from Thorlabs, directly attaching it a fibre holder, such that the distance between the LED and the fibre entrance was of few mm. This allows the LED to fill the whole numerical aperture of the fibre.

Below the list of components you could use for this (all from Thorlabs). You can of course find plenty of other suppliers for the same functionality, I just give you a list of what has worked for me...

M590L3 (you'd need also the LED and controller and power supply for this LED)

SM1T10

SM1FC

SM1L05

Thanks J. Paufique. I will try this one

The efficiency of coupling light from an LED will always be low for very fundamental reasons.

For example if you try to couple light from a hot emitter such as a light bulb filament, you can not launch more than 1/2 KT per mode,and a single mode fibre supports 2 orthogonally polarised modes, so only KT can be coupled in (where K is Botzmanns constant and T the Temp).

The same is true when using an LED which is also a multi-spatial-mode source, with some equivalent "temperature". Provided the LED simultaneously fully fills both the core area and the acceptance angle of the fibre, exactly the same power will be launched whatever the intervening optics. This is a fundamental property. The only relevant factor is then the equivalent temperature of the LED, the brighter the better, but note that a smaller LED with higher total output power, will launch exactly the same power providing the conditions above are fulfilled.

The smaller LED will provide a higher launch efficiency (waste less power), but is no disadvantage in power launched. The inefficiency is simply the ratio of source modes to fibre modes, (and this also applies to launching from LEDs into multimode fibre). This is the reason why we almost always use lasers to launch light into single mode fibres. Apart from measurement purposes, LEDs and SM fibre never go together.

If you perform spectral measurements of launched power V wavelength with any such spatially incoherent source, you will note the power steps up by a factor of 3 when the wavelength is short enough for the fibre to support additional spatial modes, it jumps from 2 x KT to 6 x KT.

I hope this will save you wasting time exploring various optical coupling methods, searching for some fundamentally impossible increase in efficiency.

If you are using a conventional surface-emitting LED, then coupling efficiency will be low for the reasons explained by Richard.

Super-luminescent diodes are a better option if you need a broad spectrum source with efficient coupling to a single mode waveguide.  Sometimes referred to as edge-emitting LEDs (ELED's), they can be thought of as laser diodes in which reflections within the cavity are suppressed.  Light is confined to a single transverse spatial mode, but multiple longitudinal modes are supported.  The small fraction of spontaneous emission guided by the cavity is subsequently amplified by stimulated emission.

Thorlabs offer single mode pigtailed SLDs delivering up to 40 mW at 1550 nm, and there are other suppliers for both 850 nm and 1550 nm sources.

Connectors are typically FC/APC to minimise reflections.  In the absence of an optical isolator, any reflections back into the device can lead to oscillation at one or more discrete wavelengths, effectively turning your broadband source into a laser.

If you really need FC/PC and don't want to replace a pigtail connector, Thorlabs offer FC/PC to APC patch cords.

http://www.thorlabs.de/newgrouppage9.cfm?objectgroup_id=3902

http://www.qphotonics.com/Fiber-coupled-superluminescent-diode-7.5mW-850nm.html

http://www.superlumdiodes.com/news.htm

http://www.superlumdiodes.com/pdf/34hp.pdf

https://www.google.co.uk/search?q=850+nm+1550+nm+SLD

http://www.thorlabs.de/newgrouppage9.cfm?objectgroup_id=1634

Thanks Alan I had forgotten about ELEDs

J. Paufique, I plan to use your solution having in mind Richard and Alan's recommendations using a LED M1550L3. Did you use the controller for LED M590L3? or do you think is it easy to be controlled by my own source? And, it seems the LED warm up easily, but I am not sure if the main heat sink that it brings with, is through convection, radiation or conduction, if anyone knew it... Thank you.

Fabio:  Is this to couple into single mode fibre too?  You should have few problems aligning the 1 mm2 LEDs from Thorlabs, M1550L3, M590L3, but coupling efficiency will be very low.

The 31 mW M1550L3 has an approximately Lambertian radiation pattern, with radiance on-axis of order   L ~ 10 mW mm-2 sr-1.

Power coupled into a single mode fibre will depend on the mode field distribution.  We can make a very approximate estimate using the étendue of the core in the geometric optics limit  https://en.wikipedia.org/wiki/Etendue

Core area for standard single mode fibre ~ π a2 where core radius a ~ 0.004 mm.

Numerical aperture of standard single mode fibre NA ~ 0.12

Power coupled from extended Lambertian source ~ L π2a2NA2 ~ 23 nW.

The 590 nm LED has higher output power and a narrower radiation pattern.  On-axis radiance could be 10x higher, but this will be offset in part by the lower étendue (smaller core diameter) of fibre which is single mode at 590 nm.

Hope this helps.

Would it be feasable to couple 1mW (from 0dBm to -10dBm) of power from a high efficiency AlInGaP LED to a 62/125 multimode fiber?

I'm trying with 1mm ball lens right now with not excellent results...

Thanks

Enrique Alvarez : Have you any data on the radiance (W m-2 sr-1)from your LED. What is the fibre numerical aperture? Is it graded index, rather than step index? These parameters set an upper bound on the power you are able to couple into a multimode fibre from a Lambertian source.

Osram publish data on output powers, and luminous intensity (lm/sr), but typically do not include radiance or luminance (cd m-2)

https://www.osram.com/apps/product_selector/#!/?query=*&sortField=&sortOrder=&start=0&filters=producttype,LED&filters=color,Yellow%20%28580-595%20nm%29&sliders=beamAngle,63,214&sliders=electricalPower,select,low,0,0.15&deeplink=

We can make a crude estimate using values in Linder et. al. (2001)

Article High-brightness AlGaInP light-emitting diodes using surface texturing

Electro-optic conversion efficiency η = 10% (0.1 W/W)

Current density: J = 1 A mm-2.

Typical LED forward voltage vf = 2 V.

Fibre core diameter d = 0.0675 mm

Typical OM1 fibre numerical aperture, NA = 0.275

Assume that the LED acts as a Lambertian source. The total power radiated from a disk of diameter d is: η J vf π d2/4

The fraction of power captured by an aperture with acceptance half-angle θ is the square of the numerical aperture, NA2 = sin² θ

For a parabolic graded index fibre, the capture efficiency is reduced by a factor 2.

Power coupled to graded index fibre = η J vf (π d2/4) NA2/2 = 0.027 mW.

This is less than your -10 dBm target, but my assumptions may not be representative.

You may find LEDs with higher conversion efficiency. The Osram site shows devices with electro-optic conversion efficiency up to 17%, and current density higher than 1 A mm-2 may now be possible - especially if the active area is smaller than the 0.3 x 0.3 mm dimensions shown in figure 4 of the Linder (2001) paper.

Figure 4 suggests that some Osram LED chips have a central circular disk which is a bond pad, rather than an emitting area. If the device you have is similar, this might contribute to your difficulty in achieving good coupling using a ball lens.

Do you know the size of the active area in your LED? You may get better results with simple butt coupling if you have a uniform emitter larger than the fibre core diameter.

Thanks Alan,

Fiber is as you assumed a graded index with 0.275NA. Led is a Vishay VLDR1235, no data given on size but typical 11cd in luminous intensity, it is very bright and can be modulated up to 1MHz (which is nice to me...)

Given the minima at 0º in the luminous intensity this led may also have that circular disc at the center that you mentioned. I'll open one and microscope it.

Would there be any difference in using a half ball lens vs a ball lens?

Hi Ernesto,

According to the Vishay data sheet, VLDR1235 is a 624 nm 14.5 cd device. VLDS1235 is 630 nm 11 cd output, both with ±11° angle at half intensity, and "high luminous flux and large chip size allowing a high DC forward current up to 70 mA".

https://www.vishay.com/leds/list/product-84280/tab/documents/

The polar radiation pattern is broadly consistent with a 0.3 mm wide source roughly 1.2 mm behind the apex of the lens, but that is not the only possibility. It does seem likely that you will only be able to couple a small fraction of the available output. https://en.wikipedia.org/wiki/Etendue

How much power can you couple with your 1 mm ball lens?

How does this compare if you simply align the fibre with the partially collimated output from the LED, without any additional lenses? Depending on the precise location of the chip inside the package, there may be a shallow alignment optimum between the lens and a point a separated by a few mm.

Do you specifically need a red LED source? Many of the devices on offer are optimised for use in general lighting/indicator applications, or for use with polymer optical fibres with much larger cores than 62/125 OM1 fibre.

Ermesto,

Thanks for the message explaining your project in more detail. What I say below may be of wider interest, so I am posting it as an additional answer.

Surface-emitting LEDs usually behave as approximately Lambertian sources. A fundamental aspect of imaging is that an ideal optical system will preserve the radiance of the source, but cannot increase it.

https://en.wikipedia.org/wiki/Radiance

https://en.wikipedia.org/wiki/Etendue

If you have a uniform Lambertian emitter which is larger than the core of your fibre, you cannot do better than simply butt the fibre close to the surface, with appropriate measures to minimise Fresnel reflections and other losses. In such cases, there is no benefit from lens coupling, other than an increase in working distance between fibre and LED.

It follows from my analysis of two days ago that the coupled power from a large area LED is limited by the etendue of the fibre (product of core area and square of numerical aperture), by the quantum efficiency of the LED and by the current density driving the LED.

One factor which constrains current density is the LED operating temperature and the rate at which heat can be removed. For a given junction temperature rise, reducing the active area of the LED allows higher power dissipation per unit area. When the active area is smaller than the fibre core, lens coupling can increase the coupling efficiency.

You are more likely to find near-IR LEDs optimised for OM1/OM2 fibre coupling since such silica-based optical fibres have lower attenuation at longer than visible wavelengths. One example is Hamamatsu L11368

https://www.hamamatsu.com/eu/en/product/light-and-radiation-sources/led/index.html

https://www.hamamatsu.com/resources/pdf/ssd/l11368_kled1060e.pdf

This is specified to couple >45 μW (typically 65 μW) into a 50 μm core graded index fibre. You would probably get > -10 dBm into a 62.5 μm 0.275 NA core, though this is not guaranteed.

Modulation bandwidth exceeds 35 MHz (50 MHz typical).

Edge-emitting LEDs, also known as superluminescent diodes, offer even higher performance. These are similar to single transverse mode lasers, with high spatial coherence, but low temporal coherence and a relatively broad spectral bandwidth, achieved by suppressing longitudinal cavity reflections.

Typically aimed at measurement applications, such as optical coherence tomography, they can deliver tens of mW into a single mode fibre, but are more expensive than surface-emitting LEDs. https://www.thorlabs.com/newgrouppage9.cfm?objectgroup_id=3902

An alternative to increased source luminance is to use a fibre with a larger core diameter. 1 mm core polymer optical fibre may be inconveniently large for use with small high speed photodiodes, but a wide variety of intermediate core diameters is available. https://www.thorlabs.com/navigation.cfm?guide_id=2284