wafer.space Community - project-template

@everyone

The project template update is now live!

Project Template 1.5.0

Switch to open_pdks for the PDK, managed by Ciel. This enables a number of new community IPs.

Add SCL, PAD and SRAM environment variables to the flow. This allows building a chip with 3.3V standard cells, 3.3V SRAMs, and 3.3V/5V compatible I/Os, for example.

Enable the new DRC runner. The KLayout DRC deck is now parallelized across rule decks while sharing some of the memory. For a best possible utilization choose: threads * workers = hardware threads. If RAM usage is too high, reduce the number of workers.

Rename the QR code ID, add readable shuttle and project IDs, add a marker. Only required if CoB option is selected.

Precheck 1.6.0

Switch to open_pdks for the PDK, managed by Ciel.

Add the --cob flag.

Enable the new DRC runner.

I've seen a significant reduction in runtime running the precheck on my machine with 24 cores: 0p5x0p5 slot: 0:06:36, and 1x1 slot: 0:40:14. Please note that the platform still needs to be updated to benefit of the new DRC runner and to make the CoB packaging option selectable.

11

3

2

1

1

Leo Moser (mole99) pinned a message to this channel. 2026-06-03 1:23 p.m.

If you are planning to tape out on wafer.space GF180MCU Run 2, please ensure your project is updated accordingly. There will be further updates before the tapeout. Though they won't be as significant as this one, so they can be easily applied to your existing projects.

1

Leo Moser (mole99)

If you are planning to tape out on wafer.space GF180MCU Run 2, please ensure your project is updated accordingly. There will be further updates before the tapeout. Though they won't be as significant as this one, so they can be easily applied to your existing projects.

just checking, is the 3.3v library there @Tholin's?

Rob Taylor

just checking, is the 3.3v library there @Tholin's?

Correct, it's the gf180mcu_as_sc_mcu7t3v3 SCL.

1

@Leo Moser (mole99) just checking, is it required to move to 1.5.0 for the run 2 tapeout?

Rob Taylor

@Leo Moser (mole99) just checking, is it required to move to 1.5.0 for the run 2 tapeout?

If you use the CoB packaging option, then yes. You need the new ID and marker macros in the correct positions in order to pass precheck.

Leo Moser (mole99)

If you use the CoB packaging option, then yes. You need the new ID and marker macros in the correct positions in order to pass precheck.

thanks!

Any idea why I might be getting this error when running the KLayout.XOR step locally (but not on GitHub - it's running fine there!):

ERROR: In /nix/store/x0ckywq495x29y1q4bldarafacwzzsxj-python3-3.13.9-env/lib/python3.13/site-packages/librelane/scripts/klayout/xor.drc: 'source': Stream has unknown format: /home/mdb36/wafer-space/ws02-tinyQV/librelane/runs/RUN_2026-06-06_09-45-39/56-magic-streamout/chip_top.magic.gds in Layout::read

Only thing I can think of is the nix install is old. But it's a bit weird it's otherwise working. Can I upgrade nix by rerunning the install script from https://librelane.readthedocs.io/en/stable/installation/nix_installation/installation_linux.html ?

Minor project template update!

Project Template 1.5.1

Update PDK to 9233c192.

Automatically fetch new PDK if hash has changed.

2

Leo Moser (mole99) pinned a message to this channel. 2026-06-06 3:27 p.m.

RebelMike started a thread. 2026-06-07 9:29 a.m.

Bugfix project template update!

Project Template 1.5.2

Update PDK to f6bfbd4d
- Fix port attributes on the gf180mcu_ocd_io__in_s cell, which previously caused OpenROAD to skip CTS (clock tree synthesis)! Thanks @RebelMike for reporting this issue!

Update LibreLane dev to latest
- Magic scripts use locking disable, which prevents the maximum open file descriptor limit from being reached.

2

Leo Moser (mole99) pinned a message to this channel. 2026-06-08 2:15 p.m.

Hmm, cautionary note. May be specific to my setup and something I goofed up. Discovered my project template was on an old verison of the PDK despite being on latest git commit and running make clone-pdk. Turns out the newer PDK version was downloaded but the symlink was still pointing at the old version. No idea how that happened, I might have interrupted it mid-updating or something.

BreakingTaps

Hmm, cautionary note. May be specific to my setup and something I goofed up. Discovered my project template was on an old verison of the PDK despite being on latest git commit and running make clone-pdk. Turns out the newer PDK version was downloaded but the symlink was still pointing at the old version. No idea how that happened, I might have interrupted it mid-updating or something.

I forgot to put the pdk root env into scope before I updated, then wondered 2 days later why something was buggy that I had thought to have been fixed, just to realize what I had done...

2

@Leo Moser (mole99) We need to use librelane dev branch for stuff to work correctly right ?

It would also be nice if the action on the repo could save the resuls ... I'd like an example GDS as "reference" since I'm using a different build process and trying to match it without having to build it myself.

Yes, the project template currently uses LibreLane dev due to some required updates.

8:39 p.m.

The action in the project template actually builds all of the different slot sizes and uploads their GDS. But I think you can't access them because you're not the owner of the repository. You can fork the repository and should then be able to download the final GDS.

Mmm, is that a settings ? I mean on TT's action anyone can download the artefacts.

Maybe. At first, I thought that everyone could download them, but Tim E. let me know that he can't access them. I'll take a look at the settings tomorrow.

1

@Leo Moser (mole99) Oh my bad, I can download the artefact ...

tnt

@Leo Moser (mole99) Oh my bad, I can download the artefact ...

Good to know! I think you at least need to be logged in to GitHub.

1

hi, I hit the following w/make librelane on the latest template after a successful make clone-pdk: Writing src/generateddefines.svh SRAMDEFINE=SRAM_gf180mcu_fd_ip_sram librelane librelane/slots/slot_1x1.yaml librelane/macros/macros_5v.yaml librelane/config.yaml --pdk gf180mcuD --pdk-root /opt/chip/gf180mcu-project-template/gf180mcu --manual-pdk --scl gf180mcu_fd_sc_mcu7t5v0 --pad gf180mcu_fd_io --save-views-to /opt/chip/gf180mcu-project-template/final [14:29:35] ERROR Errors have occurred while loading the __main.py:160 design configuration file. [14:29:35] ERROR Unknown key main.py:162 'PDNCORERING_CONNECT_TO_PAD_LAYERS' provided. [14:29:35] ERROR LibreLane will now quit. Please check your __main.py:164 configuration. make: *** [Makefile:94: librelane] Error 1

Ian Hanschen

hi, I hit the following w/make librelane on the latest template after a successful make clone-pdk: Writing src/generateddefines.svh SRAMDEFINE=SRAM_gf180mcu_fd_ip_sram librelane librelane/slots/slot_1x1.yaml librelane/macros/macros_5v.yaml librelane/config.yaml --pdk gf180mcuD --pdk-root /opt/chip/gf180mcu-project-template/gf180mcu --manual-pdk --scl gf180mcu_fd_sc_mcu7t5v0 --pad gf180mcu_fd_io --save-views-to /opt/chip/gf180mcu-project-template/final [14:29:35] ERROR Errors have occurred while loading the __main.py:160 design configuration file. [14:29:35] ERROR Unknown key main.py:162 'PDNCORERING_CONNECT_TO_PAD_LAYERS' provided. [14:29:35] ERROR LibreLane will now quit. Please check your __main.py:164 configuration. make: *** [Makefile:94: librelane] Error 1

That is surprising, it suggests you have the wrong version of LibreLane. Have you installed locally rather than running through nix? (edited)

running through nix - I'm still digging, might have an old install

librelane --version should say LibreLane v3.1.0.dev1

[nix-shell:~]$ librelane --version LibreLane v3.0.4 [nix-shell:~]$ which librelane /nix/store/qjkqjsc147pnkgdr6lm288f5p757bliv-devshell-dir/bin/librelane

9:42 p.m.

it's my nix install, 1 sec

9:45 p.m.

good to go now - I think I had a newer nix installed, it was erroring updating the ciel package on startup

the error I had before was some sort of issue with the way nixpkgs is used vs the nix I had installed, so I removed/reinstalled using the curl command from the website. here's the error, if it's interesting: … while calling the 'import' builtin at «string»:1:2: 1| (import <nixpkgs> {}).bashInteractive | ^ … while realising the context of a path … while calling the 'findFile' builtin at «string»:1:9: 1| (import <nixpkgs> {}).bashInteractive | ^ error: file 'nixpkgs' was not found in the Nix search path (add it using $NIX_PATH or -I) uses bash from your environment

9:52 p.m.

template build is at step 63 so I think I'm set up, thanks

1

@Ian Hanschen Nice to see you here! I'm glad to hear that you managed to set up the project template.

1

@Leo Moser (mole99) f741b96babd3517a01e6cc9efc3ab1eef8cbacee from the "old" wafer space PDK doesn't seem to have been applied upstream.

11:22 a.m.

Change ANTENNAGATEPLUSDIFF multiplier from 2 to 15 in gf180mcu_fd_sc_mcu7t5v0 techlefs

11:22 a.m.

Kind of a useful one

@tnt Thank you! It looks like that one slipped through. I will update the PDK shortly. However, open_pdks won't be synced until tomorrow.

1

Andrew Wingate

Those boards are due to be delivered to the people who would bring it to the wirebonders on Jun 1

@Tim Edwards Is there any know time frame for your 3.3V OCD SRAM tests? I suspect a lot of people are eagerly awaiting your results

.

Egor Lukyanchenko

@Tim Edwards Is there any know time frame for your 3.3V OCD SRAM tests? I suspect a lot of people are eagerly awaiting your results

.

Wow, can't believe it's been 2 weeks or more since I wrote that. Unfortunately his boards are currently still in China. I am hoping they will ship in the coming day(s) Then there's a few days for it to get to him. Yeah, we're all eager

but want to keep a bit of pressure off Tim

1

But those are just the COB ? Are you shipping some breakout to 100mil headers too ? Because else, he will need to design some kind of host board and getting any kind of results before the submission deadline sounds pretty tight. (edited)

tnt

But those are just the COB ? Are you shipping some breakout to 100mil headers too ? Because else, he will need to design some kind of host board and getting any kind of results before the submission deadline sounds pretty tight. (edited)

Good point.

Leo Moser (mole99) started a thread. 2026-06-15 2:32 p.m.

@tnt : I was thinking the same thing the other day. Have any host board designs been made yet that are compatible with the COBs? I could probably make do with a simple board that just routes the connector pins to a standard header.

I have some of these https://codeberg.org/rebelmike/ws-breakout, but they aren't tested yet as my COBs should arrive at roughly the same time as yours. Since I made that the breakouts in the WS repo have been updated: https://github.com/wafer-space/chip-on-board-wire-bonded-pcbs/tree/main/run-1/motherboards but I don't know if anyone has made any of those yet. I think various people have tested with the previous version of that (which as far as I'm aware was fine except the pin 1 marker was the wrong side)

The 1x0p5 COB is pinout compatible with the 1x1 COB in terms of power and ground. The signal pads are mostly in the same order too, but for (probably bad) reasons I reversed the order of pads 67-70. You can view the schematic here (pin/pad names are those on the default project template): https://kicanvas.org/?repo=https%3A%2F%2Fgithub.com%2Fwafer-space%2Fchip-on-board-wire-bonded-pcbs%2Ftree%2Fmain%2Frun-1%2F1x0p5-cob

RebelMike

I have some of these https://codeberg.org/rebelmike/ws-breakout, but they aren't tested yet as my COBs should arrive at roughly the same time as yours. Since I made that the breakouts in the WS repo have been updated: https://github.com/wafer-space/chip-on-board-wire-bonded-pcbs/tree/main/run-1/motherboards but I don't know if anyone has made any of those yet. I think various people have tested with the previous version of that (which as far as I'm aware was fine except the pin 1 marker was the wrong side)

I've used the long breakoutboards from that repo, seem to be working for the minimal tests I've tried. I have some of the PCBs unpopulated and some of the mezzanine connectors, happy to fedex you one if you want @Tim Edwards . Will have to do the soldering yourself since my soldering skills are terrible and it took me many tries to get one working

1

5:22 p.m.

(oh, mine are the pre-updated version. I think it was just the silkscreen that changed)

@BreakingTaps : Sure, I'll take you up on that offer. I assume that unless I'm really bad at soldering, I should only need one board. I can get the connector myself as long as somebody can tell me what the part number is from Digi-Key or Mouser.

1

It's 0.4mm pitch

Yeah, I got JLC to solder mine. Though having seen those photos of the CoB boards I might go double check they actually managed that..

i definitely regret thinking I could do it by hand

I believe mouser 798-DF40C2070DS04V51 is compatible.

Thanks for the heads-up on the pin pitch. That indeed will be difficult to hand solder.

5:47 p.m.

@RebelMike : Will JLC assemble from parts that you send to them?

Tim Edwards

@RebelMike : Will JLC assemble from parts that you send to them?

yes, it's expensive

5:55 p.m.

like 70$ for each 50qty, each batch seems to be billed in full iirc

5:56 p.m.

(I've done so)

But JLC can also get the part from LCSC directly.

yes

5:57 p.m.

if mouser or digikey have them it's different fee structure if you use global sourcing

At least when I ordered LCSC had the part in stock so it was easy to get them to use it. 5 PCBs and assembly was ~$25 + shipping and duties

Yeah, I just sent my PCB for fabrication via JLCPCB, including assembly with parts from LCSC.

GitHub - mole99/waferspace-main-pcb: Main PCB for wafer.space FABul...

Main PCB for wafer.space FABulous FPGA. Contribute to mole99/waferspace-main-pcb development by creating an account on GitHub.

Tim Edwards

@BreakingTaps : Sure, I'll take you up on that offer. I assume that unless I'm really bad at soldering, I should only need one board. I can get the connector myself as long as somebody can tell me what the part number is from Digi-Key or Mouser.

I should have some around as well I can send if you like. I consider myself a touch above average with soldering. I still mucked a few of them before I got a good one. 0.4mm is quite tiny.

Yeah, I'm decently good at soldering but I struggle with 0.5mm pitch and so I think I'll leave 0.4mm pitch to the experts.

4:17 p.m.

@Tim 'mithro' Ansell : You need to be selling these boards on the Wafer.Space site!