wafer.space Community

carlfk 2025-10-01 6:13 p.m.

Inside the Making of an Intel Chip – 360 Fab Tour for VR

@Leo Moser (mole99) / @Andrew Wingate - Just had a random idea which may or may not be very good. It would be nice if we could embed two pieces of information (probably optionally?) -- (a) The checksum of the source file and (b) a URL / Link to the GDS (maybe stored on archive.org or something?) Basically if someone is looking at the die being able to get a link to the source code for the project would be pretty rad....... Would probably need to be a QRCode type thing?

3

7:05 p.m.

There are also a number of lesser node QRCode versions, like the Micro QRCode (https://en.wikipedia.org/wiki/QR_code#Micro_QR_code) and the Rectangular Micro QR Code (https://en.wikipedia.org/wiki/Rectangular_Micro_QR_Code).

QR code

A QR code, short for quick-response code, is a type of two-dimensional matrix barcode invented in 1994 by Masahiro Hara of the Japanese company Denso Wave for labelling automobile parts. It features black squares on a white background with fiducial markers, readable by imaging devices like cameras, and processed using Reed–Solomon error correc...

7:06 p.m.

Apparently https://en.wikipedia.org/wiki/Data_Matrix is an alternative to QRCodes which store a lot of data in a small area too.

Data Matrix

A Data Matrix is a two-dimensional code consisting of black and white "cells" or dots arranged in either a square or rectangular pattern, also known as a matrix. The information to be encoded can be text or numeric data. The usual data size is from a few bytes up to 1556 bytes. The length of the encoded data depends on the number of cells in the...

I was just about to mention data matrix

7:07 p.m.

Another way to go about this is to create some kind of open source chip registry

7:08 p.m.

and give each chip a unique number

The most popular application for Data Matrix is marking small items, due to the code's ability to encode fifty characters in a symbol that is readable at 2 or 3 mm2 (0.003 or 0.005 sq in) and the fact that the code can be read with only a 20% contrast ratio.[1] A Data Matrix is scalable; commercial applications exist with images as small as 300 micrometres (0.012 in) (laser etched on a 600-micrometre (0.024 in) silicon device)

like you have ISBN for books

7:08 p.m.

ISOCI (International Standard Open Chip Id)

@urish - I was thinking about the same thing. Maybe something we could look at with the https://oshwa.org/ ?

Home

The Open Source Hardware Association (OSHWA) aims to foster technological knowledge and encourage research that is accessible, collaborative and respects user freedom.

7:09 p.m.

They have https://certification.oshwa.org/#

OSHWA Certification

Certification provides an easy and straightforward way for producers to indicate that their products meet a well-defined standard for open-source compliance.

uuid

anfroholic

[waferspace_die_sorter:main] 1 new commit

c319256 Create die_yield_helper.py - anfroholic

@Tim 'mithro' Ansell I got the webhook going. We can add others later if you'd like, or create a more general channel that just catches them all. \

4:53 a.m.

Example of what came out of that script. @Tim 'mithro' Ansell asked how I came up with yield numbers, so am showing work

1

anfroholic

[waferspace_die_sorter:main] 1 new commit

fc5a673 test duscussions - anfroholic

@Tim 'mithro' Ansell @Leo Moser (mole99) @urish These are all great! I have a few OSHWA certified things (I have been lazy about posting more) https://certification.oshwa.org/us002531.html If it's a machine readable code it would be nice if it were a URL This URL is over 40 chars which would put it in the 22-24x range. This density also just starts to look like a dot at macro scales and would fill the pin1 specification Data matrix are more meant for smaller things so I believe this should be the default choice. They are also read by my phone natively so that is a bonus. I think this puts us with wanting some kind of ISBN like @urish was saying, but this could just as easily be a serial number. Also having this point to a website also allows a place to put all the fabrication data, ie. process, run#, date and everything else. This has the added benefit I think of also having a way to point to source files--*or not depending if you want to**

US002531

Andrew Wingate

Example of what came out of that script. @Tim 'mithro' Ansell asked how I came up with yield numbers, so am showing work

Are dies good right up to the edge ? Or is there a min distance from the wafer edge for them to be viable ?

tnt

Are dies good right up to the edge ? Or is there a min distance from the wafer edge for them to be viable ?

This is something I actually have no knowledge in. It should be pretty easy to change when we do know. The entire thing was an artifact of creating pick maps and the sort

6:13 a.m.

also it looks pretty :)

1

tnt

Are dies good right up to the edge ? Or is there a min distance from the wafer edge for them to be viable ?

It depends on your definition of good, but a mature process should have defects evenly distributed across the wafer. If they aren’t then there is a systematic problem in the process that the foundry would work to eliminate.

1

@Valgamaa What I meant is I recently saw some wafers ( might have been at IHP can't remember if it was those one ) where there was a visually apparent borders of a few mm around the physical wafer edge were obviously some of the processing wasn't done right up to the edge and so any die in that area would obviously be scrap. But I don't know if this is standard and how it is for GF180mcu specifically. (edited)

anfroholic

[waferspace_die_sorter:main] 1 new commit

a4d92bc die marking discussion - anfroholic

I have compiled a bit of this discussion and will try to keep it up to date and be a little more persistent. https://github.com/evezor/waferspace_die_sorter/tree/main/discussions/id_fiducial_other (edited)

waferspace_die_sorter/discussions/id_fiducial_other at main · evez...

Contribute to evezor/waferspace_die_sorter development by creating an account on GitHub.

Leo Moser (mole99) started a thread. 2025-10-02 7:48 a.m.

tnt

@Valgamaa What I meant is I recently saw some wafers ( might have been at IHP can't remember if it was those one ) where there was a visually apparent borders of a few mm around the physical wafer edge were obviously some of the processing wasn't done right up to the edge and so any die in that area would obviously be scrap. But I don't know if this is standard and how it is for GF180mcu specifically. (edited)

It seems to be quite normal these days that the steppers stay within the radius of the wafer with a small margin. I guess there is no point in exposing to the edge where part of the chip lies off-chip, that is a waste of time. This isn't (shouldn't be) because the processing 'quality' deteriorates towards the edge, just a practical limit for handling the wafer. Having said that, I did one see a bipolar process that used epitaxy for the base and growth rate was higher towards the centre of the wafer. This resulted in devices where each wafer had the full process spread depending on where on the wafer the die were taken from; higher ft at the edges, higher Vceo towards the centre. It was soon discontinued!

You can see the edge of a wafer in pictures on the website.

1:12 p.m.

1:13 p.m.

That is a GF180MCU wafer

Ooh I see my die in there, on the edge

1

Makes me wonder - if you chop the top corner off a TT die, would it still be functional to some extent (as long as the controller, muxes, power pins, the mux control pins and a few IO pins are still there...)

My dies are in almost all the pictures and very recognizable...

2

algofoogle (Anton Maurovic)

Ooh I see my die in there, on the edge

@algofoogle (Anton Maurovic) - Really!?

Tholin

My dies are in almost all the pictures and very recognizable...

@Tholin - Which one is yours?

2:09 p.m.

@Tholin - You did send me the largest number of die

I put a logo on both my designs

2:11 p.m.

That’s what makes them recognizable

1

Yes

I have a high-res here: https://tholin.dev/content/static?id=E2zxGmncnZDV

Tim 'mithro' Ansell

@algofoogle (Anton Maurovic) - Really!?

Yeah, there it is…

1

@Andrew Wingate Here is what ChatGPT came up with regarding dicing saws

message.txt

7.95 KB

Tim 'mithro' Ansell

@Andrew Wingate Here is what ChatGPT came up with regarding dicing saws

Cool!! I clicked on all the links. Seems there is a lot of convergent design, I can't really tell the difference between any. Wonder how true chatgpt's numbers are.