I jerryrigged the same LCD + a random 50W UV lamp from amazon together. I was able to expose 0.2mm traces, double sided. I aligned the two side manually on the edge of the LCD itself. I used the same corner of the PCB. I simply aligned the masks on the opposite edges of the LCD to achieve that.  It was really easy to align. A proper case would add some corner bracket for a perfectly repeatable alignment. I found you also need enough weight on the PCB to keep it flat against the LCD. I did use the LCD upside down. To reduce the parallax. I was surprised by the quality of the result. I don't think my UV light is supposed to send the light rays parallel.

There is one thing odd though. If I put a piece of photoresist front of the UV light, it takes 10 seconds to fully expose it. But through the LCD it takes 15 minutes! That seems excessive. Its cooking the hell out of the LCD itself. Is it some sort of UV wavelength mismatch? How are MSLA printers doing it in few seconds?

edit: pictures of a the successful test. The hardest part is applying the photoresist film. I didn't do a good job.

https://ibb.co/xJt13fb
https://ibb.co/ggKt293

Love this Project.   If you could combine this with a simple Mill for vias this could be the next Othermill.   I was researching to do something like this and was scared of the complexity but one feature would be premade/presensitized double sided  boards with pre drilled fuducal holes so users can flip it over and do 2 sided without too much trouble.  Then a second (slow) CNC Drill can cut the holes, vias and board outline.  

Also It looks like there is a lot of light creep maybe a polarize lens should be what the PCB comes in contact with the glass on the lcd allows the light to scatter

Funny you should mention "CNC drill"/mill.  Parts are printing for that on my 3D printer as I type! :)

Simple double sided boards is my final goal.  I plan on experimenting with various frames/jigs to allow for this. Right now, the Qwicktrace software does have a "mirror" option, and it not only mirrors the image sent to it, but it does so on the opposite side of the screen.  Thus, one side is exposed with the upper left corner and left edge of board "squared" to the corner, and the other side is exposed with that same "corner" squared to the upper right corner of the screen. My hope is that any skew correction that may be required will be handled in software.  Same will go for the mill portion that drills the thru holes.  I'd like to avoid the REQUIREMENT to cut the board after the fact if possible. I have a small desktop table saw that is perfect for taking large pre-sensitized boards and cutting them down to usable sizes (like two 4" x 3" boards out of a 4" x 6" purchase, and further two 2" x 3" out of one of the cut 4" by 3"). I only want to have to mill the board edges for non square shapes.

Way back when I used to make pcbs, I found that using a UV light box with diffuse or multiple sources e.g. flouro tubes, multiple leds etc, results in poor edge definition.  Due to parrallax, any gap between the mask plane (which is up inside the lcd glass in your case, so quite big) and the photo resist plane, means that light can come from multiple directions past the edge of tracks, and end up in different places on the pcb i.e. the edges are blurred. (commercially vacuum printing frames, and putting the emulsion down onto the photoresist somewhat solved this)

I found that having the light source a good distance from the pcb gives much more reliable results, and a point source is best. i.e. have some high power leds with lenses focused on it from some distance away.  As a matter of interest, I just used the sun on a clear day quite often.

It is a simple optical lever so 

Distance / Wsource = Tlcd / Parallax

If you want Parallax<=0.05mm , and the Tlcd glass is 0.3mm thick, and Wsource is 100mm, then 

Distance = 100 * 0.3 / 0.05  = 600mm

WSource is the diameter/diagonal of  the diffuse/multi point source. If you put the whole source through a single lens, it will be the source size, not the lens size that counts as WSource.

If instead you used a single lensed led where the die itself was 2mm square

D = 2*sqrt(2) *0.3/0.05 = 17mm

Well thats too close, but using a single lensed led at 100mm is going to give you ~10um parallax. Using the most powerfull single led is probably going to be best

Great information!  Thanks.  I do have an "LCD only" with a backlight that I can remove, which would allow me to place the board directly on the mask with no glass in between.  My initial tests had problems with under exposure in certain places, but I was using an entertainment UV light that had the LCDs concentrated in a single area, but it was fairly close to the board.  You've given me some good stuff to contemplate.

The masking part of the LCD is on the inside, i.e. in between two sheets of glass. (above) I assumed 0.3mm for one sheet of glass. In the context of a negative for making pcbs thats _very_ thick. When we made film negatives, you made sure to print them mirrored, so that the emulsion side was against the photoresist, not on the outside of the film. The film was a lot thinner than LCD glass. 

BTW what parallax defocus really affects is your ability to make fine gaps between tracks (with negative photoresist) and to get the track and gap widths correct. 

It might also manifest in exposure and developer time becoming critical. i.e. your tracks and gaps are only OK, if you get exposure just right and the time in the developer right. When you have a high density/contrast mask, and properly collimated light source, neither is critical. i.e. you should be able to overexpose by 2x without tracks/gaps changing size.

While the LCD part I am using for the mask is a resin printer part, it in turn is actually a hi res screen from a smartphone. It is flexible and incredibly thin. That in turn is mounted to tempered glass to protect the screen.  The LCD by itself is about 0.7 mm thick. When you say “glass”, are you referring to the transparent but flexible material on this part?

In any event - my prototype tests were consistent with what you said re: exposure and/or developer times. I was dealing with highly concentrated UV in the center and not as much toward the edges, leading to uneven exposure for the same amount of time. 

My shipment of photo positive boards just arrived today (I consumed my previous supply with the prototype) so now I can repeat my tests with this new unit

> The LCD by itself is about 0.7 mm thick. When you say “glass”, are you referring to the transparent but flexible material on this part? 

Yes. There is glass, then polariser plastic stuck to the outside. 

https://www.sekisui-fc.com/en/application/images/pic_u21_02.png

The image is formed in the middle of the 0.7mm thickness i.e. 0.3mm above the photoresist.

> actually a hi res screen from a smartphone

So is it just a normal colour lcd then, and not a special monochrome one?

> So is it just a normal colour lcd then, and not a special monochrome one?

Yes, this one is color.  The monochrome LCDs are more expensive as monochrome MSLA printers are relatively new.  No reason a mono LCD couldn't be used, other than price.  This is intended to be a homebrew machine, so I am trying to see if I can keep costs down using less expensive parts.

I've done some additional digging, and have uncovered some interesting facts.  Many of the mSLA printers available are using sub-systems from a company named Chitu Systems.  They have a "parallel UV light" fixture which appears to use lenses to correct for the parallax and makes the UV beans parallel. It is a tad bit pricey at $75 USD, but not out of the ballpark. I've also located some other manufacturers that have UV lights with corrective lenses that are in the $35 USD range, which makes them a little more practical.  I've ordered a pair of both for experimentation.

Once UV is corrected for parallax, the primary difference between the use of color LCDs and monochrome LCDs appears to be exposure time. The monochrome LCDs allow more light through, and thus needs less exposure time. It seems, however, that the monochrome LCDs seem to be special purpose built for the mSLA printer guys, so they also make some adjustments to the polarizers.

I found this blog post from Chitu Systems helpful: helpful: https://www.chitusystems.com/2020/11/03/monochrome-lcd-screen-vs-normal-rgb-lcd-screen-whats-new/

What an awesome idea! How do you make sure that the board has full contact to the screen? I assume there must not be any gap between the two, right?

Have you already had time to test what the smallest pitch would be with sharp contours?

The copper board sits directly on the glass of the LCD screen (the screen is a replacement part for a Mars Pro SLA printer, so the screen is attached to the glass). The "Lid" part then sits on top of the board to hold it in place. I've thought about adding a weight to it, but right now, I don't think it's necessary. The lid really serves more to block external light than holding the board down.

My first prototype used a pre-assembled blacklight used for entertainment purposes, and the exposure was uneven across the board. This build uses UV LED strips, and I can control the intensity of the light, so it should be more even (though I really need to try and find a good diffuser layer).  I have a test pattern to print, and the results were just "OK" with the blacklight - but again, it was uneven intensity at 100% only.

I am going to re-run my experiments with this new build shortly. I am hopeful that a slightly longer exposure time at a less than 100% intensity will be the ticket to a perfect mask. A lower intensity should not leak thru the black masked areas as much I think.

> A lower intensity should not leak thru the black masked areas as much I think

See my comments above about exposure sensitivity above. If you are using a normal LCD backlight, it is almost certain to be causing major parallax defocus.

But also, if you think about normal LCD's , you would notice that they only really have best contrast at a single viewing angle. Again, getting a collimated / single point light source a reasonable distance from the LCD will probably also increase the actual contrast of the lcd masking significantly as well. (You might also be able to adjust the lcd contrast voltage to improve perhaps)

Further, polarisers are also tuned for visible light not UV/NIR. Making a photodiode sensor so you can actually measure the actual UV light on/off levels vs angle/contrast voltage is probably well worth the trouble.

@crun - For clarity: what the consumer resin printer guys do is take a hi res LCD meant for a smartphone, they remove the backlight, then they attach it to tempered glass to protect the LCD from the resin tank that sits on top of it. They then place a UV projection system under that. I am basically doing the same thing with the LCD part, but am using LED strip lights as the UV source in hopes of cost savings. 

Depending on the next round of testing, I could end up repurposing their UV apparatus (the ones I’ve seen use some specialized lenses which I presume helps with evening out the UV light). Of course, if I do that, I will have ended up building 2/3 of a resin printer. 

> They then place a UV projection system under that. 

Is it LED's or some kind of arc lamp? Do you know what wavelength, and what wavelengths your photoresist is sensitive to?

>they attach it to tempered glass

The thick sheet of tempered glass makes the parallax defocus much worse, so they need a very well collimated light projector. 

>the ones I’ve seen use some specialized lenses 

I would guess that fancy lenses would be a short-throw projector. They want a short box. Short is what makes it difficult. If you bring the distance out to 1x-2x the screen diagonal it is easy.

The key question is can you get a workable exposure time (<5mins) from a single LED at 6" from the screen. Start with a high power lensed LED. You can get leds with 60deg lens moulded on. You can also get 30degree external lenses. 

>which I presume helps with evening out the UV light). 

It is very easy to even it out: Multiply your pcb image x illumination map.

Since it is a visible colour LCD, the different colours will transmit different amounts of UV (and this will depend on the LED wavelength), so the illumination map is different for each colour.

 (This also means you need to do some testing with you photo diode to measure the transmission of 100% R,G and B images. It is also key that you actually do get your UV through the green and red pixels, and not just blue. Otherwise you are relying on defocus and what light goes through the blue pixel, and will never be able to get very sharp pcbs, no matter what you do.)

Thanks for the response! That answers it.

Be aware that the longer you expose, the more light "creeps" through at the edges which will make lines thinner than designed!

In the offset printing industry, before laser imaging of printing plates became a thing in the 90ies there was the need for a calibration process to ensure the correct dot size when "copying" the printing film to a plate. There were fairly big machines that produced a vacuum inside to ensure perfect contact between film and plate and a meticoulously fine tuned exposure time. The wrong exposure time would have produced wrong colors in CMYK rastered images…

Here some more insight on what I'm talking about:

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

As these machines are all out of need there must be a used market for plate making machines…

Btw: Why do you go through the pain of constructing the whole UV-light-construction? Is there any other advantage besides saving money?

> Btw: Why do you go through the pain of constructing the whole UV-light-construction? Is there any other advantage besides saving money?

My first prototype used a pre-assembled Black light flood light. It "sort of" worked, but the exposure was not even. The use of LED strips is an attempt to even out the exposure.  User @crun pointed out the problems with non-parallel light beams, so I've ordered two different UV assemblies created specifically for 3D resin printers. They range in price from $40 USD to $75 USD, so they do add to the cost of the build. We'll see if the extra money is worth it.

Trying to wrap my head around this: are you basically making a one-layer SLA printer, and printing the mask onto the PCB?

I am using pre-sensitized copper boards.  No resin is used. The SLA printer's LCD is re-purposed to mask the UV light directly on to the copper board. Take it off the exposure table, and drop directly into the developer to remove exposed etch resist, then drop that into etchant to remove the copper

Gotcha, thanks. But you know what -- my idea wouldn't actually be that bad. At least if you didn't have pre-sensitized boards for some reason ;-)

So, using actual resin for etch resist with an actual resin printer was the inspiration for this project. I saw a guy do that with success on YouTube (after some crazy hacks to his printing process). My thought  was that presensitized boards would be significantly less messy than using resin. That being said, today I was using my resin printer, and I had a few thoughts that may make the process less messy. I may in fact give it a try with this unit with one simple modification. My goal is the best traces for the least amount of effort and time from CAD to soldering.