Hi everyone So last Sunday, Ferenc and I brainstormed on how to achieve smaller feature sizes with a UV wave length, which limits us to 400nm or so. We discussed splitting a lower feature size layout into multiple masks and then doing overlapping exposure steps. Ferenc pointed out, that we might run into an offset issue. I've been thinking about this for a while and the solution came to me on the toilet just now, and it's actually surprisingly simple. Since we're using a DMD and not physical masks, we can do the multiple exposure steps all at once, without any stepping in between. while(!eow()) { // end of wafer expose_mask(1); expose_mask(2); step_to_next(); } cheers -lev
On Fri, Apr 10, 2020 at 8:37 PM David Lanzendörfer <leviathan@libresilicon.com> wrote:
Hi everyone So last Sunday, Ferenc and I brainstormed on how to achieve smaller feature sizes with a UV wave length, which limits us to 400nm or so. We discussed splitting a lower feature size layout into multiple masks and then doing overlapping exposure steps. Ferenc pointed out, that we might run into an offset issue. I've been thinking about this for a while and the solution came to me on the toilet just now, and it's actually surprisingly simple. Since we're using a DMD and not physical masks, we can do the multiple exposure steps all at once, without any stepping in between.
while(!eow()) { // end of wafer expose_mask(1); expose_mask(2); step_to_next(); }
https://www.youtube.com/watch?v=n-_D-cVIsLk
cheers -lev_______________________________________________ Libresilicon-developers mailing list Libresilicon-developers@list.libresilicon.com https://list.libresilicon.com/mailman/listinfo/libresilicon-developers
David Lanzendörfer schreef op vr 10-04-2020 om 19:35 [+0100]:
Hi everyone So last Sunday, Ferenc and I brainstormed on how to achieve smaller feature sizes with a UV wave length, which limits us to 400nm or so. We discussed splitting a lower feature size layout into multiple masks and then doing overlapping exposure steps. Ferenc pointed out, that we might run into an offset issue. I've been thinking about this for a while and the solution came to me on the toilet just now, and it's actually surprisingly simple. Since we're using a DMD and not physical masks, we can do the multiple exposure steps all at once, without any stepping in between.
while(!eow()) { // end of wafer expose_mask(1); expose_mask(2); step_to_next(); }
This is not going to work. What you basically will be doing is adding two sinusoidal intensity profiles with half the period shifted on top of each other giving almost not modulation of the intensity anymore. There is not reason why this would only be possible with DMD and not with two physical masks. greets, Staf.
Hello, Did anybody have time to check on Jade Bird Display? It appears to me that the company is still functioning after all, despite their EE and testing people leaving. The real man behind the enterprise, Qiming Li, is quite a well known guy with LED and materials people. On Sat, Apr 11, 2020 at 1:51 AM Staf Verhaegen <staf@fibraservi.eu> wrote:
David Lanzendörfer schreef op vr 10-04-2020 om 19:35 [+0100]:
Hi everyone
So last Sunday, Ferenc and I brainstormed on how to achieve smaller feature
sizes with a UV wave length, which limits us to 400nm or so.
We discussed splitting a lower feature size layout into multiple masks
and then doing overlapping exposure steps.
Ferenc pointed out, that we might run into an offset issue.
I've been thinking about this for a while and the solution came to me on the
toilet just now, and it's actually surprisingly simple.
Since we're using a DMD and not physical masks, we can do the multiple
exposure steps all at once, without any stepping in between.
while(!eow()) { // end of wafer
expose_mask(1);
expose_mask(2);
step_to_next();
}
This is not going to work. What you basically will be doing is adding two sinusoidal intensity profiles with half the period shifted on top of each other giving almost not modulation of the intensity anymore. There is not reason why this would only be possible with DMD and not with two physical masks.
greets, Staf.
_______________________________________________ Libresilicon-developers mailing list Libresilicon-developers@list.libresilicon.com https://list.libresilicon.com/mailman/listinfo/libresilicon-developers
Hi Pavel I haven't yet had time to look into this company. Also, my new boss *sobbing again*, pointed out that we might have slightly different intensities with each micro LED, which might lead to uneven exposure. That's why I'm eyeing at DMD chips again. Cheers -lev On Saturday, April 11, 2020 1:29:54 AM WEST Pavel Nikulin wrote:
Did anybody have time to check on Jade Bird Display? It appears to me that the company is still functioning after all, despite their EE and testing people leaving. The real man behind the enterprise, Qiming Li, is quite a well known guy with LED and materials people.
In defence of that, wouldn't be that even easier to correct with an LED matrix than with anything else? Those matrices are not 1 bit, each pixel has a CMOS driver on the backside. And the man behind the company should be quite knowledgeable in materials, and the process side himself. Lastly, the resolution should favour the LED matrix. There is a hard limit on DMD density, not so much on LEDs. On Sat, Apr 11, 2020 at 4:13 PM David Lanzendörfer < leviathan@libresilicon.com> wrote:
Hi Pavel I haven't yet had time to look into this company. Also, my new boss *sobbing again*, pointed out that we might have slightly different intensities with each micro LED, which might lead to uneven exposure. That's why I'm eyeing at DMD chips again.
Cheers -lev
On Saturday, April 11, 2020 1:29:54 AM WEST Pavel Nikulin wrote:
Did anybody have time to check on Jade Bird Display? It appears to me that the company is still functioning after all, despite their EE and testing people leaving. The real man behind the enterprise, Qiming Li, is quite a well known guy with LED and materials people.
Hi Staf Of course it would also work with physical mask. The goal here is to eliminate the additional step of physical mask making. Removing the need to physically manufacture masks and then physically switching masks within the stepper aligner would save time and money. This said. The discussion right now is about on how to make the most of this new way of exposing. Of course, you can do virtually everything with a DMD chip, you could also do, if you had physical masks in there. The only advantage is a simplification of the manufacturing process and a cost reducting, due to the elimination of the mask making step. Cheers -lev On Friday, April 10, 2020 8:51:27 PM WEST Staf Verhaegen wrote:
This is not going to work. What you basically will be doing is adding two sinusoidal intensity profiles with half the period shifted on top of each other giving almost not modulation of the intensity anymore. There is not reason why this would only be possible with DMD and not with two physical masks.
David Lanzendörfer schreef op za 11-04-2020 om 11:10 [+0100]:
Hi Staf Of course it would also work with physical mask.
What I do want to say is that it will neither work for DMD as for physical masks. You can't increase resolution by doing two prints in a resist over the resolution you can get with a single print in that same resist. There have been tries to make the resist non-linear by doing a bake step in between the two exposures but this causes asymmetry in sensitivity to the image between the two exposures and has not made it in production AFAIK. AFAIK, only techniques where the first print is developed in resist and then transferred through etch in a mask layer and then doing a second exposure in a fresh resist layer has made it in production. greets, Staf.
Hi Staf Now I'm getting what you're trying to tell me! Good point! However. Double patterning is possible in this way: https://www.techdesignforums.com/practice/guides/double-patterning/ This already gives us the possibility to squeeze a bit more resolution out of the stepper. Right? Cheers -lev On Saturday, April 11, 2020 12:29:08 PM WEST Staf Verhaegen wrote:
What I do want to say is that it will neither work for DMD as for physical masks. You can't increase resolution by doing two prints in a resist over the resolution you can get with a single print in that same resist. There have been tries to make the resist non-linear by doing a bake step in between the two exposures but this causes asymmetry in sensitivity to the image between the two exposures and has not made it in production AFAIK. AFAIK, only techniques where the first print is developed in resist and then transferred through etch in a mask layer and then doing a second exposure in a fresh resist layer has made it in production.
David Lanzendörfer schreef op za 11-04-2020 om 15:57 [+0100]:
Hi StafNow I'm getting what you're trying to tell me!Good point! However. Double patterning is possible in this way:https://www.techdesignforums.com/practice/guides/double-patterning/
This already gives us the possibility to squeeze a bit more resolution outof the stepper. Right?
The LELE is the generic technique for doing double patterning and what I was also hinting at but as name says it involves 2 litho and 2 etch steps. The SADP looks nice when looking at cross sections but has big restrictions on what one can print. It basically can only print lines of one fixed size and they are formed around the border of the first pattern. This means that you likely need a third litho + etch step to remove the lines in places where one does not want them and possibly a fourth litho + etch step if one wants to also have lines with a bigger width. This technique is mainly used for NAND Flash where you have these long parallel lines, AFAIK it is not used for random logic. greets, Staf.
But you need reproduceable movements. Try to move from right to left and from left back to right and find the offset. What holds true for a turning machine also holds true for a chip scale machine. The problem are similar but shrunk together with the requirements in dimension. So turning at 0.05mm is enough for most applications, and the chip needs to be 0.05um precise. :-) On 4/10/20, David Lanzendörfer <leviathan@libresilicon.com> wrote:
Hi everyone So last Sunday, Ferenc and I brainstormed on how to achieve smaller feature sizes with a UV wave length, which limits us to 400nm or so. We discussed splitting a lower feature size layout into multiple masks and then doing overlapping exposure steps. Ferenc pointed out, that we might run into an offset issue. I've been thinking about this for a while and the solution came to me on the
toilet just now, and it's actually surprisingly simple. Since we're using a DMD and not physical masks, we can do the multiple exposure steps all at once, without any stepping in between.
while(!eow()) { // end of wafer expose_mask(1); expose_mask(2); step_to_next(); }
cheers -lev
participants (5)
-
Christoph Maier -
David Lanzendörfer -
ludwig jaffe -
Pavel Nikulin -
Staf Verhaegen