https://open-source-silicon.dev logo
Join SlackCommunities
Powered by
Hi all, I was wandering if anybody could give me s...
# sky130
k
Hi all, I was wandering if anybody could give me some feedback about the layout of the following photo diode I have made in magic. The outer ring is a p-tap contact, then we have pwell and nwell rings over dnwell with respective contacts.
l
It looks like a Photo Diode to me. Make sure your Nwell widths and Pwell widths meet all DRC rules.
k
@Larry Harris the photodiode is exactly what I try to make, the desing is DRC clean
l
Your drawing shows the Nwell going all the way down to Deep Nwell. If that is indeed true. The Deep Nwell is shorted out to the Nwell.
k
In Magic there is a a photo diode in device generator and the connection between dnwell and nwell is exactly the same
l
If you intended for the Deep Nwell to be independent and not shorted to the Nwell, you have a problem.
k
here goes other design
image.png
what is strange that if You analyze the photo diode generated by magic You will find that the dnwell is not a continuous surface but rather a ring
l
If the Deep Nwell is just a ring, then it doesn't short to the Nwell. You may want to draw you own Photo Diode and not use the generator in Magic.
k
The layout is not generated, I have made it on my own trying to reproduce the diode from the first image
l
What kind of problems are you seeing with your photo diode?
k
as You can see the dnwell is almost directly connected to the contact Vdnw, what in sky130 has to be made using nwell and respective difussion implant.
then Youu have concentric rings of p-well and n-well (i suppose) with respective contacts
In general there are a few methods to make a photodiode however sky130 recoemnds only one based on dnwell. Moreover in sky130 there is a layer (or marker) called photodiode which seems to be dnwell
g188.png
this one is generated by magic
l
Your concentric ring layout looks like a photo diode as well as the generated photo diode. A photo diode is simply a P-N junction. I am new to Sky130 and have not yet reviewed the way they make Deep Nwell mask. You may want to talk to someone who knows how the Deep Nwell mask are made and if they do an "area file" if the Deep Nwells are too close together.
Again, what kind of problems are you seeing? Do you have silicon back for your photo diode?
k
In genera I know that there is a variety of methods to make the photo diode however I was told that it necessarily will work in this process. So I am trying to have some more feedback
l
What you have will work as long as you treat the Deep Nwell and the Nwell as being shorted together.
k
Here for example I have laid out a photodiode with a readout circuit which is simple PN junction (ndiff over substrate)
l
The photo diode is very small area. I am use to using Photo Diodes that are on the order of 1000u x 1000u. What king of photo sensitivity do you need ?
k
I agree that it s small, it my first layout of a photo diode
There is also a pinned photodiode
image.png
and I would also like to give it a try. The basic idea is to make kind of comparative study between the different PD using the sky310
The other question is about the "phtodiode" layer in this technology, which seems to be a reference to dnwell however I was told that it is just a marker (whatever marker is). Do you know something more about it ?
l
I don't know how the Mask shop uses the "photodiode" layer in order to generate the Photo Mask. You may want to talk to Skywater about how they use the "photodiode" layer.
k
The do not have time to respond those questions
what I've found in the tech file is the following
Copy code
types
# Deep nwell
  dwell dnwell,dnw
  dwell isosubstrate,isosub
  dwell photodiode,photo
what makes me believe that photodiode makes reference to dnwell
l
My guess is that they are using the Photodiode layer and added it to the DNwell layer. Check and see if they are also using the Photodiode layer with the top level passivation layer. They may be removing some of the top layer oxide of the photodiode just like they open up the passivation layer over bondpads. The top level passivation layer will act as an optical interference filter and block some of the photons reaching your photo diode. This optical interference filter will be wavelength dependent.
k
ok, processing at FEOL level is other issue, one have to make a clear view for the light to reach the photo diode
look at that "In magic, you can use the "fillblock" layer to avoid any fill over the device (you will need to run the command 
tech unlock *
first (I should have thought to add that to the photodiode device generator)."
l
Yes, you want to use "fillblock". Make sure the metal fill routines don't put metal fill on top of you photo diode.
k
where can I check those routines ?
l
I Don't know where to check the routines or who to ask. I am new to Efabless. However, I am not new to the Industry. I retired last year after 43 years of working in the Semiconductor Industry. I have a lot of experience with designing Opticial Sensors, Magnetic Sensor, Thermal Sensors, ADCs, Voltage Reference, and much more. I got bored with being retired, now I am back designing my own circuits with this Efabless stuff. I can answer your circuits and device physics questions but I am not yet familiar with how Skywater make their mask sets.
🌍 1
k
Thank you Larry. I am pretty new to the IC design, tools and methodologies surely I will ask you more questions.
and what do You think about the new open source wave ?
l
Well, I wish there was more formal support for all the CAD Tools as well as with any question about the Fab process and how they make the Photo Mask.
On another note, if the CAD tools weren't free, I wouldn't be here.
b
Regarding the pinned diode -- you'd have to violate design rules to get anything close, see e.g. https://iopscience.iop.org/article/10.1088/0268-1242/30/4/045002/pdf
k
@Boris Murmann thank You for your feedback, I am working in magic in order to lay out some more structures, i will post it in the afternoon here looking fore some more comments about
here it goes, the names of the files explain the structure, any feedback more than welcome
3 Views
Open in Slack
PreviousNext
Powered by