Anyone have suggestions/recommendations on my new coax light setup

[andrew...@gmail.com]

Have printed this test rig.

Camera mount for ELP720 camera.

Holder for 60x40mm 50% splitter glass

Holders for diffuser and LEDs

4x Targets (Nozzle, 144TQFP, Cut tapes,  HASL PCB)

For now it is only going to be my UP camera.

I know/realise that I only have perfect illumination in the centre of image (evident in the HASL PCB image)

I also know that the black felt is not great at absorbing black and this is ruining my contrast.  The final version will be painted in Musou black paint.

The images look much better than my previous lighting rig.

Was wondering if anyone has any suggestions for me before I go and 3d print a solid version I can mount on my machine?

[simpl...@tuta.io]

>> Was wondering if anyone has any suggestions for me before I go and 3d print a solid version I can mount on my machine?

In my opinion, it is better if the camera is on-axis, as otherwise you may see slight artifacts due to the refraction of light by the mirror.

In addition, a slight bevel on the side of the housing opposite the light source is helpful. With 3D-printed enclosures, sticking dark sandpaper there can lead to good low-reflection results. Musou Black or BLK3.0 are suitable, even if they are never as extremely black as, for example, in a Keyence coaxial light.

When designing the housing, you should also make sure that the camera can be easily adjusted. The settings for focus, center, and rotation are best made using a calibration rig before mounting the whole thing on the machine. The less correction required by the host software, the faster the image processing. 

Here is an old bottom vision example with an IMX335 camera (640x480 pixel):

[svetozar161]

Pay attention to the Pixie Placer project, there are links to it in the openpnp wiki. There's a coaxial light option, I made it for myself, but I didn't like it or I couldn't set it up. Maybe it will help you

среда, 3 сентября 2025 г. в 04:18:00 UTC+3, andrew...@gmail.com:

[Clemens Koller]

Hi, Andrew!

It looks good! Can you share the information what kind of diffusor you are using
as well as which mirror you use? The LEDs are usually a bite more distant
from the diffusor to be able to use a less dense diffusor / get more light out
(disadvantage: more volume needed). All interior of the light path will be
highly reflective white, later on... camera surrounding black?!

The image quality you achieve seems pretty decent. You could experiment with
additional AR-coated or polarizing glasses, if you want to get a little bit
further and have no limit in budget or so. The overall size of the cube might
get smaller depending on the fov of your camera.

I am lucky to have a CCS LFV3-CP as well as a ring light (and a
telecentric lens for the bottom camera). The image quality is not much better
than yours.

Regards,

Clemens

> CoaxTestRig.jpg
> CoaxTests.jpg
>
> --
> You received this message because you are subscribed to the Google Groups "OpenPnP" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to openpnp+u...@googlegroups.com <mailto:openpnp+u...@googlegroups.com>.
> To view this discussion visit https://groups.google.com/d/msgid/openpnp/9dc77fc9-3ffb-430d-9aff-f970de1dac23n%40googlegroups.com <https://groups.google.com/d/msgid/openpnp/9dc77fc9-3ffb-430d-9aff-f970de1dac23n%40googlegroups.com?utm_medium=email&utm_source=footer>.

[andrew...@gmail.com]

@simpl...

Thank you.  Those images are with the camera on the same axis as the target.  They still have a slight double image from the back surface reflection.  Most noticeable on the HASL image with the bright pads having a ghost about one 0603 pad above the bright pad.

I have already tried a sloped backstop painted with Mosuo Black.  The layer lines in the 3d print still show through.  My current test setup has the backstop moved 30mm out of focus and I am using Mosuo Black painted MDF.  It increases the contrast a lot doing it that way.

Also my latest setup I have moved to having the camera path bent 90 degrees rather then the light.  There was slightly more ghost - but it has allowed me to go from the 8mm lens to a 12mm one.  Much less barrel distortion and I can now get the full 3:4 720p image evenly lit.

I have not thought about how to make the camera adjustable in the housing yet.  I can see a need for it as the cheap cameras do not have the sensor on very flat by the look of it.

@Clemens

The diffuser is a product called "visiled"  It is an acrylic sheet with beads of slightly different refractive index acrylic mixed in.  My next light is going to have more smaller LEDs and I hop to be able to reduce the total height of the lamp/diffuser by half.

The mirror is from this aliexpress store 

https://www.aliexpress.com/store/1103814098

I am just 3d printing some more "target holders" so I can take photos of how things look now with my latest improvements.

The image quality has gone up a lot I think.

Probably 2 or 3 hours and I will post latest photos.

[andrew...@gmail.com]

Musou black backstop 30mm behind focal plane.

Changed to 12mm lens (from 8mm)

Camera path now bends 90 degrees and light path is straight.

Changed to 3:4 from 16:9

It's looking good Vern.

HASL, ENIG, TQFP44, Nozzle, Coin, 0603R.  0603R in tape and 0805 LED in tape, Satin finish stainless ruler.

[Mike Menci]

Hi - looking very good - Just take care that you use proper heat resistant filament because the Led PCB will produce heat in closed compartment (what Voltage you will use ? 12- 24V DC ?)

there are some nice already made PCB with Leds to use ...
The   60x40mm 50% splitter glass - Ali link does not point to it ---- what thickness is it - you could have a double image if the thickness of glass !??  
My coax lights are all in Alu housing with inside wall "white " wall plastic covers for light reflection.... 
But the images you shared look impressive !
Well done...
but there is always room for ---- keep improving! 

[Mike Menci]

And here! 

[Mike Menci]

BTW..... here is Coax light which I use...

[andrew...@gmail.com]

Here is the link for the glass I use.

https://www.aliexpress.com/item/1005007541640404.html

Previously I linked the store rather than the item, because aliexpress item URLs always break.

The glass is the 60x40x1.1mm T5R5 item.

At the moment I am running 36 Lumiled 3030 high CRI leds hand soldered on a stripboard.  They are only running 20mA each for now and don't get very hot.  They are specced for 65mA and do get hot at that power level.  I run 24v so that is 6rows x 6 columns of LED.

Next board I am going to make with 144 smaller LEDs so I can bring the diffuser to LED distance down.  I might even try a JLCPCB aluminium substrate PCB for that one :)

This is going to replace my UP camera which is quite lacking.

In the future I am going to try optimize and shrink this coax light setup down as much as possible to replace my down camera.  However there is no rush for that as my down camera is already adequate.

[Michael Anton]

Here is a 50/50 splitter glass with AR coating on one surface from Aliexpress: https://www.aliexpress.com/item/1005005553336695.html.

Beam splitting cubes are also available: https://www.aliexpress.com/item/1005008868545722.html.  This one isn't AR coated, but they should be available.  I wonder what would happen if Mouso Black was applied to the sides of a cube?  Not sure if the internal reflection of the glass would then ignore the AR coating or not.

[andrew...@gmail.com]

Thank you.  For now I am happy with the image in terms of contrast and double reflections.  In future I will look into AR coatings and polarizers.

My current problem is being able to adjust the camera x/y/pan/tilt/rotation.

I remember seeing a green-3d printed USB camera mount that had those adjustments on it - Google search is now such a failure I can no longer find it.

Does anyone have links to good mounting systems to correct for the poor image sensor placement on this ELP camera?

[andrew...@gmail.com]

150x64x55mm (Wide, Tall, Deep)  So way too big for down/top camera.

Field of view is 20x27mm.

Focal distance is 26mm above the top surface.

Perfectly suitable for up camera, though I do need to source a thin sheet of glass to protect the 50:50 mirror from dust,

I could not find the person that had made a 3d printed camera adjustment rig so I just dialled in the X/Y rotation and tilt by 3d printing iterations of the holder section.  Was tedious but is done.

I may revisit the camera mounting part when I have some more important jobs out of the way.

I am happy with the performance.  It is easy enough to mount.

I have some cunning ideas to get the top/down camera much much smaller when I tackle it.

Only including final picture of HASL PCB.  It is the hardest thing to get right.  If you can get that looking good all the other things are fine :)

[simpl...@tuta.io]

>> HASL PCB.  It is the hardest thing to get right

Yes, i agree. 

An alternative to making your own down looking coaxial lights is to buy them second-hand, especially if you want something compact, lightweight, and reasonably shock-resistant.

I have bought several of them on eBay for as little as $50, replaced the LED circuit board, and adjusted the focus, rotation, and center using a simple calibration device so that optical errors are kept to a minimum and only minor software corrections are necessary.

It should also be mentioned that one must not forget to take into account the offset on the corresponding axis due to the refraction of light by the mirror—depending on the thickness of the mirror, this can be several tenths of a millimeter.

[Mike Menci]

You might use how I did it - add a M12 plastic nut behind lens and follow ; https://grabcad.com/library/pnp-up-looking-coaxial-light-housing-logitech-c270-1

[Mike Menci]

Andrew - normally your camera would be inline with object and light would be reflected 45deg.... 

[SM]

@Mike    often, there is not enough height, and in that case, an off-axis coax light also works - you can live with the ghost images if the contrast is high enough.

On our old machine, however, we had to mill the entire housing out of aluminum to get the thermal problems (the industrial camera gets quite warm) under control.
The main interior surfaces of the housing were painted with Musou Black using an airbrush (gives a nice matte and even finish).

[Slavko Kocjancic]

Hello...

I silently read that posts as I slowly build my pnp machine. Indeed I just realized that Maybe I fail at my coaxila lightning engineering. It's not finished yet but just got one tought about that.

The coaxial lightning have one bad thing in concept. The light come from LED and in half mirror is diverted to PCB and other half pass thru halfmirror and lighten the box wall usually black mate painted. 

The camera itself see PCB thru mirror but at the same time see reflection from that (highly illuminated) black painted area. And that reflection make a big role on image quality. 

 

Now what I just realized is: The distance from camera to the PCB should be as different as possible to the distance from camera thru reflection to the wall. In that case the wall will be out of focus and will not disturb much. If the distances are near same then the black wall will be in focus and any inperfection will be highly visible. 

...didnt test if that is true but by logic seems that it is. 

...sadly my design have near same distances to PCB and black wall.

--
You received this message because you are subscribed to the Google Groups "OpenPnP" group.

To unsubscribe from this group and stop receiving emails from it, send an email to openpnp+u...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/openpnp/a3299205-a031-4b13-9291-aa534a309495n%40googlegroups.com.

[andrew...@gmail.com]

@SM "It should also be mentioned that one must not forget to take into account the offset on the corresponding axis due to the refraction of light by the mirror—depending on the thickness of the mirror, this can be several tenths of a millimeter."

Is this what you mean?
If yes - then I don't think it is a concern to me as my only reference to the real world is the image captured.  I don't even know the exact x/y the camera sits at until I start calibration.

@Mike  "Andrew - normally your camera would be inline with object and light would be reflected 45deg.... "

That is the way I had it in the first photos in this thread.  When I changed to the 12mm lens to get less barrel distortion and better lighting I had to change it to this way as it would have been too tall in the first configuration.
After making the orientation change the ghost image from the 2nd surface reflection was worse, but after giving it the Musuo Black backstop the ghosting has almost  disappeared.

@Slavko

In the second setup the distance (Yellow Lines) from the mirror (Blue line) to the target/PCB and the original backstop (Green lines) where about equal.
Even though the target was very very black, it was perfectly in focus and very brightly lit.  I will try post a photo later of Musou Black painted 3d print in and out of focus.
I added the spacer and moved the backstop (Purple line) 30mm behind the focal distance.  This extra 30mm space (Pink line) made the backstop out of focus which was a vast improvement.
In the latest 3d printed version, you can see the backstop has been moved CLOSER to the mirror also dragging it out of the depth of field for the focus.
There is only a small amount of improvement I think I can make to the UP camera at this stage probably involving polarizing filters and maybe some fairy dust.
I DO think the the down camera is going to be a lot of work.

[SM]

>> Is this what you mean?

Yes, exactly.

After all, you're trying to align the camera (+lens/sensor) as perpendicularly as possible, and knowing the camera center is quite helpful when using a precise calibration frame, especially if you have limited leeway for mechanical correction. The view range of the head camera is particularly important when planning a machine – for example, whether you can still reach a feeder position or not.

The beam offset caused by the mirror can be calculated according to the law of refraction, but the glass coatings often interfere with this calculation :-)