Temperature, test conditions and test layout is missing.

[atorkmabrains]

We were trying to build a comparison against simulation results. And we are seeing some weird results against simulation. We wanted to make sure what are the test conditions that you have. Also, did you make multiple measurements on the same device?

Measurement temperature and test layout is missing from the repo. Could we get the layout to be able to inspect it to understand what's going on?

[mithro]

All the details we have about the layout can be found at the following links;

• https://github.com/google/skywater-pdk-sky130-raw-data/tree/main/docs/sky130-testtile-proprietary
• https://bit.ly/sky130-testtile-edit
• https://bit.ly/sky130-mpw1-testtile-notes

It would be best to publish the simulation results and let people help debug what is going on.

[akinakturk]

HI Amro,

The test conditions are written in header sections of the mdm files. See the section below.

One possible confusion is that the device name written in the etest die description is different from that of the open_pdk version. So we needed to guess the mapping of some names.

We made multiple measurements on some devices to verify that the results repeat. This was pursued when we suspected a problem. In general, most measurements were fine the first time around.

Tim provided information regarding the layout. There is a picture of the layout, and an accompanying excel file describing some of the test structures. However, there is no attendant gds file.

Please let me know if you have questions.

Thanks.

Akin

---

The files are named as follows: 'openpdk device name""width""length""multiplication factor""(module and pin numbers)".mdm".

For instance:

sky130_fd_presd_nfet_01v8_w5p40u_l0p180u_m1(4044_3_4_IDVD).mdm Device name = sky130_fd_presd_nfet_01v8 W = 5.4 micrometers L= 0.18 micrometers m = 1 (single device - not multiple in parallel) This device is in module 4044. Source and drain pads are 3 and 4, respectively. (The module number, and pad numbers are not necessary for understanding and examining the empirical data. They show the device location within the test die.)

The files are saved in MDM format using ICCAP (from Keysight). These are text / ascii files. They can be opened by any text editor.

The file structure of a mdm file reads as follows:

For example:

! VERSION = 6.00 BEGIN_HEADER ICCAP_INPUTS VD V E GROUND SMU2 0.1 LIN 1 0 1.8 73 0.025 VS V S GROUND SMU3 0.1 CON 0 VG V B GROUND SMU1 0.001 LIN 2 0 1.8 7 0.3 VB V C GROUND SMU4 0.1 LIN 3 0 -1.5 2 -1.5 ICCAP_OUTPUTS ID I E GROUND SMU2 B IB I C GROUND SMU4 B END_HEADER . . .

First couple of columns of the header section provide aliases given to each source measurement unit (SMU) voltage and current input/output ("VD", "VS", "ID", ....), and specify whether the SMU is operated in voltage or current mode (V or I).

Next few columns provide SMU number, compliance and sweep mode, i.e. linear, constant, etc.:

SMU2 0.1 LIN: SMU # 2, current compliance is 0.1A (since it is operated in voltage mode), and this sweep is linear.

Last columns show sweep details. For example: 1 0 1.8 73 0.025:

Sweep order # is 1. Voltage start value is 0. Voltage end value is 1.8V. There are 73 voltage steps of 25mV.

Following the top header section, there are mini headers or subheaders for each subsection.

BEGIN_DB ICCAP_VAR VS 0 ICCAP_VAR VG 0 ICCAP_VAR VB 0

VD ID IB

0 1e-014 -5.2e-013 0.025 -9.6e-013 1.15e-012 . . . 1.8 -1.21e-012 3.8e-013 END_DB

BEGIN_DB ICCAP_VAR VS 0 ICCAP_VAR VG 0.3 ICCAP_VAR VB 0

VD ID IB

0 2.24e-012 8.3e-013 . . .

These subheaders show the variables kept constant during the measurements listed following these subheaders.

On Wed, May 3, 2023 at 10:10 AM Amro Tork @.***> wrote:

We were trying to build a comparison against simulation results. And we are seeing some weird results against simulation. We wanted to make sure what are the test conditions that you have. Also, did you make multiple measurements on the same device?

Measurement temperature and test layout is missing from the repo. Could we get the layout to be able to inspect it to understand what's going on?

— Reply to this email directly, view it on GitHub https://github.com/google/skywater-pdk-sky130-raw-data/issues/25, or unsubscribe https://github.com/notifications/unsubscribe-auth/ACPNITT44QOTX66MFSAZ4YDXEJRPBANCNFSM6AAAAAAXUPBFDM . You are receiving this because you are subscribed to this thread.Message ID: @.***>

--

---

Akin Akturk CoolCAD Electronics LLC 5000 College Avenue, Ste. 2105 College Park, MD 20740

[miesli]

Hi, using DMT we managed to obtain simulation results fairly similar to the measurements, at least in DC.

Here is the link to the repo: https://gitlab.com/dmt-development/skywater130-pdk-dutlib

Until now, it only covers the first set of measurements. But in general we had the same issues (for especially the missing layout) as you, that's why we did not integrate all new batches to this DutLib.

[atorkmabrains]

@mithro I have reviewed all of these documents before and I did that again to double check. But some of the information that I'm looking for can't be found or missing. It would be better to get the actual GDS file to be able to measure any missing information that we would like to get from the layout directly. That would be really helpful.

One of the most important parameters that we are looking for is temperature. At what temperature did they measure this? I was never able to find that. Also, Do they have temperature controlled setup? Please point me to the document that might have that information if it's available.

We are planning to release all our work here as a PR later. We have created an interactive plot measurement vs simulation for all the nfet_01v8 sizes in the test structure as an example in Jupyter Notebook. But code could be generalized for all devices easily.

I'm waiting until we have done our due diligence of review and finding any issues that we might have missed on our side. Because we are seeing differences for some devices and we want to understand why they are different from the simulation?

@akinakturk Thanks for the info. GDS file would be extremely helpful we could measure any information we need from the layout directly rather than speculation how did you layout the device?

@miesli We just found out about this git repo yesterday. We have seen you have done some assumptions there that we are not sure where did you get that information? We are trying to replicate that. Also, you only took a smaller subset of the devices.

[akinakturk]

The devices were measured at room temp, which was approximately 25C.

Akin

On Thu, May 4, 2023 at 5:33 AM Amro Tork @.***> wrote:

@mithro https://github.com/mithro I have reviewed all of these documents before and I did that again to double check. But some of the information that I'm looking for can't be found or missing. It would be better to get the actual GDS file to be able to measure any missing information that we would like to get from the layout directly. That would be really helpful.

One of the most important parameters that we are looking for is temperature. At what temperature did they measure this? I was never able to find that. Also, Do they have temperature controlled setup? Please point me to the document that might have that information if it's available.

We are planning to release all our work here as a PR later. We have created an interactive plot measurement vs simulation for all the nfet_01v8 sizes in the test structure as an example in Jupyter Notebook. But code could be generalized for all devices easily.

I'm waiting until we have done our due diligence of review and finding any issues that we might have missed on our side. Because we are seeing differences for some devices and we want to understand why they are different from the simulation?

@miesli https://github.com/miesli We just found out about this git repo yesterday. We have seen you have done some assumptions there that we are not sure where did you get that information? We are trying to replicate that. Also, you only took a smaller subset of the devices.

— Reply to this email directly, view it on GitHub https://github.com/google/skywater-pdk-sky130-raw-data/issues/25#issuecomment-1534404705, or unsubscribe https://github.com/notifications/unsubscribe-auth/ACPNITS4H235VMIDW6ZZVZTXENZYFANCNFSM6AAAAAAXUPBFDM . You are receiving this because you commented.Message ID: @.***>

--

---

Akin Akturk CoolCAD Electronics LLC 5000 College Avenue, Ste. 2105 College Park, MD 20740

[atorkmabrains]

Thanks @akinakturk.

Could you share the GDS as well?

[miesli]

Yes, we just made the assumptions and wanted to see how far we can get. But also we only had limited time to spend on that project, which was only a proof of concept. As we proved the concept, we are basically able to add all measurements and simulations of all devices relatively quickly. Adding the other devices is a matter of adding a few lines of script.

Doing this, would only require more assumptions and already from the simple example one can already see that the model approximates the DC drain current with a visible error. In #4 and #9 we discussed the naming of the measurements and were looking for more information.

Our conclusion was to wait for the new test chip with an available layout and new measurements. There, a model verification (which was our initial goal) makes much more sense in our opinion.

[akinakturk]

Hi Amro,

Skywater has not provided the gds file. They provided the chip and the accompanying excel file providing descriptions for some of the modules.

By the way, I am working with NIST as well. We are about to publish 4K data for some of the devices on this test die under certain bias conditions. I believe that the paperwork associated with publicizing the low temp data is close to getting done.

Thanks.

Akin

On Thu, May 4, 2023 at 11:12 AM Amro Tork @.***> wrote:

Thanks @akinakturk https://github.com/akinakturk.

Could you share the GDS as well?

— Reply to this email directly, view it on GitHub https://github.com/google/skywater-pdk-sky130-raw-data/issues/25#issuecomment-1534956666, or unsubscribe https://github.com/notifications/unsubscribe-auth/ACPNITU7BZJWM7AGC6NRMHDXEPBNVANCNFSM6AAAAAAXUPBFDM . You are receiving this because you were mentioned.Message ID: @.***>

--

---

Akin Akturk CoolCAD Electronics LLC 5000 College Avenue, Ste. 2105 College Park, MD 20740

[atorkmabrains]

That's great news @akinakturk Looking forward to getting that data.

@mithro Could you check with Skywaters to see if they can share the GDS file?

[QuantamHD]

The GDS cannot be shared. That's an unfortunate limitation of the Skywalker test tile

[atorkmabrains]

@QuantamHD I'll share our results in the next few days. But simulation results are way off the measurements. The first thing that came to my mind that we missed something. Anyhow, I'll share everything for more scrutiny from the community. I hope we are wrong.

[mithro]

@QuantamHD is correct that we can not share the GDS for the SkyWater original test tile.

This is why @msaligane has been working with NIST to create a fully open source test tile - see https://github.com/google/skywater-pdk-sky130-raw-data/tree/main/docs/sky130-testtile-open -- We are still waiting to get these back from manufacturing.

[atorkmabrains]

@QuantamHD and @mithro I'll close this ticket as it's addressed at #26