LED underlight for biological microscopes.

[HackedFridgeMagnet]

I has said I will do a LED replacement light for some biological microscopes. Basically a quick side project that I have already said yes to.

This is to replace 230V -> 12v Switched psu which drives a 12v halogen bulb.  I will do about 10 of them, and probably more as they burn out.

So I was thinking of spinning a LED board and a psu/LED driver board, To mechanically replace the existing module mounts exactly.

I will get some fairly pure white LEDs probably CREE, one for each. Don't need many lumens just say up to 3Watts worth.

Was thinking of just using a RECOM driver , http://www.digikey.com/product-detail/en/recom-power/RCD-24-0.50-W/945-1616-ND/2303636
and RECOM mains psu.  RAC03-12SE/277/W http://www.digikey.com/product-detail/en/recom-power/RAC03-12SE-277-W/945-2251-ND/5204549
Probably would prefer the through hole versions.

The systems needs to be dimmable using a 10kOhm pot which is on the microscope.
Was thinking of using an LT driver chip for the LED driver instead of the Recom but cant seem to choose one.

I doubt there is an all in one solution with any sort of approval but would like to be shown wrong on this.
There is also quite a bit of space for the electronics, and I more worried about saving time than cost.


If anyone has a better way of doing it, or any tips, then let me know.

[bobaruni]

Isn't that a bit of overkill?
You have 12V and access to a 10K pot, why not just roll your own simple 1 transistor job to drive the LED, a 3W LED shouldn't take much.
How many LEDs per unit?
Does it have to dim to completely off?
I would probably avoid PWM just in case they put a video camera on the microscope as this would create stroboscopic effects.

[SeanB]

Be careful of the LED module you use, it has to have a single die in it, as the optics in the microscope will be very carefully set so they take light from a small portion of the incandescent lamp filament, and use this to make the collimated beam to illuminate. Your LED must have the emitting surface of the die in exactly the same spot and it must have a single LED dice in it, otherwise you will have very uneven and colour changing illumination across the visible viewing area.

[danadak]

Seriously consider RGB replacement. Color, wavelength, can be used
to enhance the image. For example short wavelength light will highlight
edges of an object, long will smooth the appearance, acting as a filter.


Regards, Dana.

[Andy Watson]

Have a browse of this site: http://www.microscopy-uk.org.uk/  Articles and discussions about LED lighting and conversions frequenctly occur - often motivated by practical requirements.

[Kleinstein]

Using RGB LEDs could be tricky, as the diffrent colors will come from different spots. Depending on the optics this can cause only one part to be used. This should be problem only if a well focussed intense illumination is used to get high brightness, e.g. for high magnification.

Different colors are still attracttive, though I don't know an easy solution. An extra diffusor / defocussing to mix the RGB segments reduces the intensity.

The needed power also depends on the optics - for a well focussed system only a certain aerea of the LED would be used. So it's more about luminance and a suitable area than about power. How much power is needed also depends on the range of magnifications. HIgh magnification needs a high intensitiy in the center, thus enough luminace. With the same optics (no adjustment) a low magnification needs a suppficiently large source. With an good adjustable optics, one could get away with a much smaller LED, though a fixed optics might be more conventient.
 
So the project has two relatively independet parts:the first is finding a suitable LED and mountig it in place. The second is biulding / buying an adjustable current source to drive the LED.

[HackedFridgeMagnet]

Just thanked you all as I appreciate the input, as I will be pleased to get this little project off my plate.

So the project has two relatively independet parts:the first is finding a suitable LED and mountig it in place. The second is biulding / buying an adjustable current source to drive the LED.

Yes I should work it this way. I will get the LED sorted first then maybe that can refine my driver options.
When I can, I will put in a Cree XPG and a Cree MLE and see how they look, using a lab power supply to drive them, I have these LEDs lying around somewhere, both Warm White IIRC. Might be able to unglue/remove any lenses?
If I can get one of them looking good I will try to find a driver.

As far as RGB, not today, all they want is a direct replacement, maybe I can give them an option down the track.
Though maybe if I can score a microscope I will put something like this in, sounds nice.

Be careful of the LED module you use, it has to have a single die in it, as the optics in the microscope will be very carefully set so they take light from a small portion of the incandescent lamp filament, and use this to make the collimated beam to illuminate. Your LED must have the emitting surface of the die in exactly the same spot and it must have a single LED dice in it, otherwise you will have very uneven and colour changing illumination across the visible viewing area.

Yes this could be tricky, this is where I am out of my depth. I will have to experiment.

Isn't that a bit of overkill?
You have 12V and access to a 10K pot, why not just roll your own simple 1 transistor job to drive the LED, a 3W LED shouldn't take much.
How many LEDs per unit?
Does it have to dim to completely off?
I would probably avoid PWM just in case they put a video camera on the microscope as this would create stroboscopic effects.

The microscopes currently have a 3 legged 10k dimmer + mains on/off switch. This I have to use, but it should slot in easily if I use an analogue controlled LED driver. I guess it has to dim as well as the halogen did. Which I don't know. I will do a mock up and see if they like it.
Yes I could bring the mains down to say 5V and do it like that. Though I still have to use feedback, maybe a linear reg? I would prefer to do it in PWM though, as I could then use the same device in my stalled mppt solar charger project.
Any chip suggestions? I was playing with an LT3740 in LTSpice. I really only need one switching FET but one that also has a bootstrapped hi side driver was what I was looking at for the mppt.

Still looking through the microscopy website, seems to be a good resource, thanks.

[Marco]

It hardly seems feasible here, but if you collimate the light from RGB LEDs first couldn't you combine them through a prism?

[Kleinstein]

With an LED in the 1 W range, it is possible to use linear current control, especially if you start from a 5 V supply (e.g. wallwart) instead of 12 V. There is no need to use a high side switch, a low side switch is working as well. So the analog version could use a BJT or MOSFET and an OP. The PWM version could get away with even an NE555. Also simple switched mode regulators might work.

So it's a good idea to see how much power  current is really needed.

[LaserSteve]

Be aware that LEDs  often have huge gaps in their phosphor's spectrum. THIS can cause issues for Biology, Polarization based techiques,  and certain other uses.  You may also need UV spectrum to be present for Bio work.

Which is why we still use Xenon lamps on the Confocal microscope at work. Switching to LED word pay for itself in eight months or so for us. However, the spectral gaps were an issue.

Steve

[Someone]

Be aware that LEDs  often have huge gaps in their phosphor's spectrum. THIS can cause issues for Biology, Polarization based techiques,  and certain other uses.  You may also need UV spectrum to be present for Bio work.

Which is why we still use Xenon lamps on the Confocal microscope at work. Switching to LED word pay for itself in eight months or so for us. However, the spectral gaps were an issue.

Steve

High quality LEDs intended for colour rendering are similarly flat to xenon arc lamps across the visible spectrum, both have a factor 2 bump in them around 450nm (in opposite directions!). But as you note LEDs will have no output in the UV or IR and the intensity at the edges of the visible spectrum will be a long way down so they are inappropriate for some illumination tasks, of course for some of those fluorescence jobs a deep blue LED might be a better choice than a broadband source and a filter Its easiest to ask the user what they need, if its anything specific they'll mention it.

[TerraHertz]

As has been pointed out, typical 'white' LEDs are rather poor for microscopy, due to the spectral hole in the green region.
The next problem is trying to couple the LED light output to the microscope's optical path, to get target illumination that doesn't have any objectionable remnant of the LED physical shape or spectral emission variations. 'White' LEDs with multiple chips may look white on average to the eye, but they are really a bunch of bright blue dots on a yellow field. Putting this through any kind of lens system results in seeing exactly that.

Filament bulbs for microscopes generally have a filament form attempting to achieve as near as possible an even output over an aperture seen by the optics. But LEDs tend to have an electrode pattern, and are far from an even bright field.
The best way to fix this is with a diffuser sheet early in the optical path. You can find some quite good diffuser films in the backlight filtering stack of old LCD monitors.

Some of my microscope LED-lighting attempts here:
http://everist.org/NobLog/20150510_Olympus_bhm.htm
http://everist.org/NobLog/20150628_LED_lighting_mk3.htm
http://everist.org/NobLog/20151112_planning_vacuum.htm#refl


Even with the diffuser in the Olympus one, although there is no trace of the LED structure in the image, the green spectral hole results in a slightly annoying tendency for images to have 'bluish fringes' on contrast edges. At least, I didn't notice this effect with a filament illuminator.


My current microscope illuminator project is stuck, because I can't find concentrator lenses for the small 10W flat LEDs. Plenty of concentrator lenses for the big 50W LEDs, but no lens manufacturer seems to think it's worth making them for the 10W LEDs. And I need to fit four 10W LEDs and their lenses on each of three smallish heatsinks, so I can have a mix of different color LEDs combining to make a controlled-spectra light source.

You may find the mechanical and optical mating elements of your project are the hardest. I'm not even bothering with the LED-drive electronics part of mine, until I'm happy with the optics.

[HackedFridgeMagnet]

mmm.. ok food for thought.
Maybe I should just get some sort of drop in replacements for the power supply.
There is plenty of space but they dont make the original PSUs anymore. Nikon. I haven't got the model number here.
They do run hot and they fail due to Cap leakage, no prizes for guessing which one. Which in turn does a good job of wrecking the pcb.
Got some LEDs on the way I guess I will give them a try.

[HackedFridgeMagnet]

Be aware that LEDs  often have huge gaps in their phosphor's spectrum. THIS can cause issues for Biology, Polarization based techiques,  and certain other uses.  You may also need UV spectrum to be present for Bio work.

Which is why we still use Xenon lamps on the Confocal microscope at work. Switching to LED word pay for itself in eight months or so for us. However, the spectral gaps were an issue.

Steve

High quality LEDs intended for colour rendering are similarly flat to xenon arc lamps across the visible spectrum, both have a factor 2 bump in them around 450nm (in opposite directions!). But as you note LEDs will have no output in the UV or IR and the intensity at the edges of the visible spectrum will be a long way down so they are inappropriate for some illumination tasks, of course for some of those fluorescence jobs a deep blue LED might be a better choice than a broadband source and a filter Its easiest to ask the user what they need, if its anything specific they'll mention it.

They are just for Uni Lab classes. Yes I will do a demo unit and give them a good chance to check their suitability. The longer the better...

[Zero999]

What about a deep red LED & violet LED + remote phosphor? That should cover most of the visible spectrum but designing the optics to mix the light could be challenging.

[HackedFridgeMagnet]

What about a deep red LED & violet LED + remote phosphor? That should cover most of the visible spectrum but designing the optics to mix the light could be challenging.

Thanks for the idea but optics is definitely not my thing and I am a bit busy to change that now. I am also wanting to use only visible light, to keep the WHS police happy so  I will try to avoid operating near the edges of UV.

My motivation for this is not the challenge, but just a quick and painless fix using modern LED technology to replace the globes and PSUs, which would then finally let the electronics run at a comfortable temperature.
I was assuming the existing light spectrum was fairly flat, and was then hoping the spectrum of the white LEDs would somehow be flat enough to make this work. Maybe it will.

I guess I was hoping that some of you guys had run in to this problem and already had a solution to it, well it sounds like some of you have run into the problem anyway.
Appreciate your help though.

 

[Someone]

I was assuming the existing light spectrum was fairly flat, and was then hoping the spectrum of the white LEDs would somehow be flat enough to make this work. Maybe it will.

Dont worry, the currently available high CRI white LEDs are just as smooth as (and without any narrow peaks of) an arc lamp. For general lab use white LEDs were fine 10 years ago if you could get the focus accurate, the iris controlled illuminators were quite a bit more forgiving than the directly dimmed lamps. Checking against a glass graticule is usually sufficient.

[HackedFridgeMagnet]

Just an update.

The clients inspected the first replacement and compared it against the original and were very happy which was nice. They were amazed at how cool it runs in comparison.

Gave me 10 scopes as the first batch and said they had 160 all up. So it seems I have a fair bit of ongoing work.

I used a Richtek RT8477 for the LED driver, a single Warm white Cree XPG LED and a RECOM 3W 240->12vdc power supply.

The RT8477 let me use the existing 10k dimmer pot.

As far as mounting the LED i just screwed a little LED sub board onto the existing stainless reflector. As far as centering goes, it seemed that it wasn't that critical, I deliberately moved it off centre and the LED still gives enough light that you can't actually notice the difference.

They had no problems with the evenness of the light and the spec sheet for the CREE show there is negligible UV, which is better than the Halogen.

For light level we decided that 500mA @ around 3V provided at least as much light as the 6W halogen.

I did a PSU - LED driver pcb in Altium that had the same mechanical format as the original (barring 1 mistake cured by a Hacksaw). So the boards are a direct swap which makes the changeover job fairly easy.

Will do a respin for the next batch. It was my first real board in Altium, normally use Kicad, had a few issues with the footprints. (Size of the annular rings was a bit too small.)