Custom IC Design - Recent Threads

create bus labels from CSV file

Darrell L — Tue, 29 Sep 2026 19:39:07 GMT

We have a digital control circuit that uses 1232 hardwired inputs to set the trim values for an RF circuit. The hardwired inputs are organized as 176 busses with 7bits each. After simulation and lab testing, we will have a csv file with the optimum values for each bus. I would like to have some automated way to label the bus connections to each input based on a csv file values. IE if the csv file has 1001000 for bus x, and 0110011 for bus y, then I want some way to label the wire connected to bus x '1,0,0,1,0,0,0", and bus y "0,1,1,0,0,1,1". Is there any way to do this?

Alternatively, is there some way to create a schematic block with labeled wires and pins based on the csv file? I could then place that block in the main schematic to set all the trim values.

Thanks in advance,

Darrell

RE: create bus labels from CSV file

IN2FAB — Tue, 06 Oct 2026 14:42:22 GMT

Actually, this bit:

allBusses = foreach(mapcar datum data
busData = parseString(datum ",")
busName = car(busData)
busBits = buildString(cdr(busData ","))
)

Should read:

allBusses = foreach(mapcar datum data
list(car(parseString(datum ",")) buildString(cdr(parseString(datum ","))))

RE: create bus labels from CSV file

IN2FAB — Tue, 06 Oct 2026 14:22:48 GMT

It should be fairly straightforward to write some Skill code to do this. You'll probably need to do a little processing on each bus label once you've read it in to separate out the bus name from the bits, assuming it's all csv. If each csv line reads "busx,1,0,0.1,0,0,0" after you pull in from the file, it would be something like:

data = <function to read the file>

allBusses = foreach(mapcar datum data
busData = parseString(datum ",")
busName = car(busData)
busBits = buildString(cdr(busData ","))
)

That will give you a list containing each bus in two parts:

'(
'("busx" "1,0,1,0,0,0")
'("busy" "0,1,1,0,0,1")
etc.
)

Then step through each one to find the appropriate wire on your schematic and add the label. Getting the wire that connects to a pin rather than a random wire on each net will give a neater result.

foreach(bus allBusses
<find the pin based on the name in car(bus)>
myWire = <find the wire that connects to the pin>
myCoord = <get your coordinate based on the wire>
myLabel = dbCreateLabel(cellId "device" myCoord cadr(bus) "lowerLeft" "R0" "stick" textSize)
schGlueLabel(myLabel myWire)
)

Label attachment is done through parent/child connections in a schematic so you can attach the net that way if you want to stick with db functions.

Once you've done all this, you'll need to add tieHi and tieLo cells to your schematic and label the wires attached to them as 1 and 0.

Your proposed schematic will work electrically but it's going to cause you problems in the layout. If you connect all the 1 and 0 bits together as single nets, you'll end up with two wires with over 600 connections that each hang off one resistor. You're much better off creating local 1 and 0 connections for each bus and tying them off separately.

Given this is for trim, you might also consider adding a tieHi or tieLo connection for each bit of every bus. That would allow you to adjust the trim based on post-layout extraction simply by flipping the tie cells from one to the other.

- Tim

Netlist error

DA202602028132 — Thu, 01 Oct 2026 08:58:10 GMT

Hello

I am currently running a test containing emx extraction (s param file)

How have I assembled the test ?

to do so, I have created schematic and schematic_emx cellviews

schematic_emx just has instance analogLib/nport with s param file attached
schematic is just a schematic (metal resistors etc)

To run the test, I have assigned schematic_emx cellview in corner setup (and not the schematic cellview)

How have I run the test ?

I have created testbench schematic (in ADE assembler, right click on the test Netlist>Create... or Netlist>Recreate...), using the schematic cellview, not the schematic_emx cellview, which was successful

Then I have proceeded to run the simulation (using the schematic_emx cellview according to corner setup) and got netlist error, because the schematic_emx was not checked & saved

My question is - How can I define the Netlist>Create... (or Netlist>Recreate...) to create the netlist according to the corner setup?

Thanks

Daniel

How to save PEX post-sim waveforms the same way as pre-sim？

dogman4 — Thu, 17 Sep 2026 09:39:00 GMT

Hi friends

I extracted the PEX parasitic netlist using Calibre v2016.2_39.26 and then performed post-simulation analysis in Assembler.

Before running the pre-simulation, I add the signals I want to monitor to the "Outputs Setup" and set "Save options" to the selected mode. After the simulation is complete, I can directly plot the waveforms within Assembler.

Now, I want to adapt this workflow for post-simulation: adding the target waveforms to the "Outputs Setup" and then select selected for "Save options".

Does anyone have advice on how to do this?

My virtuoso version: "@(#)$CDS: virtuoso version 6.1.8-64b 09/03/2024 19:11 (sjfhw316) $"
My spectre version: Version19.1.0.063 64bit--10 Aug 2019

页面翻译

RE: How to save PEX post-sim waveforms the same way as pre-sim？

Andrew Beckett — Mon, 21 Sep 2026 20:08:39 GMT

The best solution would be to move to IC25.1 and then you can use IDSPF views (if using Quantus), or if you're using parasitic extraction from Calibre, you can then use the DSPF2ADE flow. This then would allow the output names to remain as the schematic names and it can automatically map them.

This is covered in this Rapid Adoption Kit: DSPF2ADE flow (IC25.1)

Andrew

RE: UMC 65LL: snfp_ll_lvt12 vs snfp_ll_lvt12_corner — which Section for "Single Run, Sweeps and Corners" vs Monte Carlo?

Andrew Beckett — Mon, 21 Sep 2026 16:34:34 GMT

You should probably ask UMC this (or consult their documentation). We have to be careful about giving details of a foundry PDK/model library in a public forum.

Andrew

UMC 65LL: snfp_ll_lvt12 vs snfp_ll_lvt12_corner — which Section for "Single Run, Sweeps and Corners" vs Monte Carlo?

MT202609188310 — Sat, 19 Sep 2026 05:07:53 GMT

In the Model Libraries setup (ADE Assembler), when I add the device model file, the Section dropdown shows two entries for the same LVT 1.2 V core device:

snfp_ll_lvt12
snfp_ll_lvt12_corner

I'd like to understand which section to select for each run mode:

Single Run, Sweeps and Corners — which section?
Monte Carlo (local mismatch + global process variation) — which section carries the statistical block?

Specifically:

What is the actual difference between snfp_ll_lvt12 and snfp_ll_lvt12_corner in the UMC 65LL model set?
Does snfp_ll_lvt12_corner carry the mismatch/global-variation statistics needed for Monte Carlo, or is it strictly for the "Single Run, Sweeps and Corners" mode?
Which section should be active for each of the two modes so I don't, for example, run MC against a fixed corner?

Can clip() support multiple time windows?

Johanny Saenz — Thu, 17 Sep 2026 11:31:23 GMT

Hi all.

I want to calculate a waveform metric while excluding specific time intervals containing switching glitches.

The glitches occur when a control signal changes state, so I can identify the relevant time points using cross(). For a single interval, I can use something like:

clip(waveform t_start t_stop)

However, I would like to evaluate the waveform over multiple non-contiguous time windows, effectively excluding the intervals around each detected transition.

Does the Cadence Calculator clip() function support multiple time ranges, possibly using the output of cross()? If not, what is the recommended Calculator/SKILL approach to calculate, for example, ymin() while ignoring these glitch intervals?

Thanks in advance

RE: Can clip() support multiple time windows?

Andrew Beckett — Mon, 21 Sep 2026 16:33:06 GMT

Nothing built-in as far as I know. There is my article 20374821 which provides a function abFilterByValue - this allows you to filter out any values outside a given range. It also provides a way to filter out based on X ranges too - but it only supports a single X range at the moment (and anything in that range is omitted; you might want the opposite) - it might be a good starting point for what you want though.

Regards,

Andrew

RE: How to save PEX post-sim waveforms the same way as pre-sim？

dogman4 — Thu, 17 Sep 2026 15:38:51 GMT

Furthermore

Currently, for post‑layout simulation, I first use Save All, then select the desired nets and send them to Outputs Setup for plotting. However, Save All generates excessive data, so I need to switch to the selected mode for subsequent runs. When using selected mode, the saved signals cannot be plotted after post‑simulation.

Therefore, it would be acceptable if I could place the post‑simulation net names into Outputs Setup, set Save to selected , and directly plot the waveforms after running post‑layout simulation.

RE: How to save PEX post-sim waveforms the same way as pre-sim？

dogman4 — Thu, 17 Sep 2026 09:57:06 GMT

by the way, my pex is spectre format

RE: Automatically switch between "sparameter" view and "rfm" view

amitG22 — Thu, 10 Sep 2026 11:42:51 GMT

Hello Janko,

No I have never used MDIF myself. But that's a good suggestion. We will need to evaluate it and we can then use it in our future projects. It can help with maintaining all these different views more efficiently. Thanks for this tip.

Regards,
Amit

Automatically switch between "sparameter" view and "rfm" view

amitG22 — Wed, 15 Jul 2026 07:52:06 GMT

Hello Community,

When I generate sparameter files through EMX, EMX generates the sparameter views that is basically a touchstone file. There is in essence no separate schematic with a nport component. This works well except when you want to run transient simulation. Then, spectre tries to generate bbspice models for all these sparameter views. Now, I don't want spectre to generate its own models (why not is not relevant for my question). I want to generate my own RFM models and want spectre to use them during transient simulation without creating a separate schematic with a nport component and pointing to a rfm file. This is tedious and doesn't scale well when you are working on a big project.

My question is : Is it possible to do this with spectre in an elegant manner instead of all the tedious manual work?

A snapshot of the view generated by EMX :

If it indeed turns out that its possible to "hack" the normal WoW and force spectre to use my own RFM models then the following information may be handy : In the linux directory structure, I have a folder that is exactly the same as the name of the sparameter view. Within it, next to the sparam.snp file I have placed my sparam.rfm file.

Thanks in advance for all your tips/tricks and suggestions.

Regards,
Amit

RE: Automatically switch between "sparameter" view and "rfm" view

JankoK — Thu, 10 Sep 2026 11:21:09 GMT

Hi Amit,

For the first problem... I think that's what MDIF format is for. The tool supports it. You should be able to run the fitting and get your RFMs over corners. To be honest, I haven't tested that functionality myself, but you can find the workflow described here:

Generate and Simulate MDIF Files with EMX and Virtuoso ADE Corners

The quality of your RFMs will depend on the quality of your S-parameters and fitting. There are also a few fitting options to play with. Anyway, you should be able to use them with Spectre, that's the whole idea, to make life easier for you and the simulator. Please ping your local support for more detailed investigation or file an official case on our support (ASK) page.

/Janko

RE: Automatically switch between "sparameter" view and "rfm" view

amitG22 — Thu, 10 Sep 2026 09:10:25 GMT

Hi JankoK,

Thanks for your response.

As I already mentioned in my post, creating separate views with nports and then pointing it to the RFM files may work on a simple test-bench but doesnt work so well in a real transceiver design with multiple passive cells, where each cell has to be extracted over 5 BE corners and at least 3 temperatures. And then imagine creating 15 such cell views with nports pointing to RFM files and 15 cell views with just the EMX spar views. It is a nightmare and just doesn't work.

Regarding the RFM file generation, as far as I understood, the quality checking tool doesn't output a rfm file that we can directly use.

Regards,
Amit

RE: Automatically switch between "sparameter" view and "rfm" view

JankoK — Wed, 09 Sep 2026 13:07:34 GMT

Hi Amit,

You probably figured it out by now, but just in case...

Answer: Yes, it is possible. You can literally just point to your RFMs from nports. In case you need to create your RFMs, we got you covered there too... that's the fitting part of our S-parameter Quality Checking and Fitting Tool.

SPECTRE® S-Parameter Quality Checking and Fitting Tool

From IC23.1 ISR11, you have in the GUI too. It super easy to use:

How to use 'S-Parameter Quality Checking and Fitting Tool' and 'Comparison Tool' of SPECTRE in Virtuoso CIW

Have fun!

/Janko

RE: Why in 2026 still unable to rotate text labels by 45 degrees in symbol view?

Andrew Beckett — Tue, 08 Sep 2026 00:41:37 GMT

Hi Sjoerd,

This is set by the available orientations in the underlying OpenAccess database (from Si2). That only supports the 8 90-degree orientations.

Regards,

Andrew

Why in 2026 still unable to rotate text labels by 45 degrees in symbol view?

SH20260513571 — Mon, 07 Sep 2026 12:54:31 GMT

Hi,

Using latest and greatest Virtuoso Advance Node, I want to rotate some text labels by 45 degrees in a symbol view. Either by the GUI, or, if not possible (and prefered) using the SKILL backdoor. Reason being: to make the symbol be actual useful, i add diagrams to it, and for the pump-pad + IO cell I want to add text labels to the bump map, but the labels are overlapping. If I had the chance to rotate the labels, they wouldn't overlap anymore.

I'm asking for 45 degrees, but any multiple of 15 degrees would be even better, or multiples of 1 degree, if I may be greedy.

Why isn't something simple like this not possible in Virtuoso in 2026? Or if it is: why can't I find it?

With kind regards,

Sjoerd Herder

Cadence Extremely Slow Plot For DC Sweeps

AY202602197011 — Fri, 04 Sep 2026 08:30:46 GMT

Hello everyone,

I've recently noticed that plotting DC Sweeps in VIVA has become extremely time consuming. I can give you a concrete example.

Currently I'm using Cadence Studio IC25.1-64b.ISR5.36 and I've ran a -40 to 165 degrees temperature sweep of a schematic consisting of a few moderately sized blocks. In total I have 272 runs of this sweep across PVT. The plotting has taken more than an hour now and it's still struggling as of writing this post. It seems rather peculiar because plot 272 transient simulation of the same bench with a length of 200us takes less than a minute, sometimes it's almost instant. From the output log I see that I have 4543 signals saved. The whole psf output DB is about 433MB.

This problem first became visible in Cadence Studio IC23.1-ISR16, where the plotting was a bit faster and took maybe 15-30 minutes. I have to add that in both cases I'm simulating using Maestro and the problem is still visible when plotting from the Results Browser.

I really hope that someone has a solution to this problem because it's affecting multiple users and renders them almost useless on some days.

Best regards.
Antoan

RE: Cadence Extremely Slow Plot For DC Sweeps

Andrew Beckett — Sun, 06 Sep 2026 09:56:21 GMT

Antoan,

Please contact customer support (use the Case menu to submit a support case after logging into the ASK portal). We'll need to investigate where it's getting stuck and whether there's unusual about the data.

You might also want to check with IC25.1 ISR8 (since that's the latest) in case it was fixed in the meantime (I've only done a very quick check for reports, and I might have missed it).

Regards,

Andrew

How to setup calcVal (or different approach) when tests differ on number of Design variables

RH202503055935 — Sun, 06 Sep 2026 06:14:59 GMT

Hello,

I was unable to find any solution on this, which would be functional. I also tried some Skill scripting but nothing helped so far.
I have trimming procedure of oscillator.

Corner setup: process, biasing current

Test1 - Runs on 25deg and nominal VDD set in Design variable. Transient gives number of trimming code
Test2 - Runs on -40 and 125 deg + two VDD values. It should use the code from Test1.

This whole set should run across the Corners.

Reason for this is to use the transient from Test1 as IC for PSS and PNOISE or any other simulation within Test2.

I tried ?ignoreParams, ?matchParams in calcVal. I also tried to add dependency in run plan. It always fail mostly on different number of design variables.
I tried to create corner1 with just process + ibias, one temp and one VDD and create corner2 with process+ibias+vdd sweep+tempsweep. Applied matchparams and ignoreparams fails on error like> no points matched the?matchParams and ?ignoreParams specification.

Version of IC is: 23.1-64b.ISR15.37

Thank you very much for suggestions.

try_fast_op with SpectreX

SL20260319720 — Mon, 24 Aug 2026 02:18:32 GMT

Hi,

Is it possible to disable try_fast_op when simulating transient with SpectreX? I set have the option "no" selected under ADE Explorer -> simulation -> Options -> analog -> try_fast_op (see attached picture). After netlisting, I see the text "try_fast_op=no" as part of simulatorOptions. However, the transient simulation usually spend significant amount of time at the stage "Trying Fast DC 0 for initial condition." It appears that Spectre is still trying to do fast DC for initial condition.

Regards,

Su Lim

RE: try_fast_op with SpectreX

SL20260319720 — Wed, 26 Aug 2026 16:48:35 GMT

Hi Andrew,

changing the environment setting helps, thank you.

Regards,

Su Lim

RE: try_fast_op with SpectreX

Andrew Beckett — Mon, 24 Aug 2026 14:43:07 GMT

Su,

Firstly, that try_fast_op does not control that setting (there's more than one "fast" algorithm) - that's why setting it to "no" doesn't affect the "Trying Fast DC" message.

You can disable it by going to Setup->Environment and entering +fastdc=no in the User Command Line Arguments. Do be careful though (I see you are forcing the homotopy to gmin too) - assess in any example whether it really helps before setting this; never turn these on by default.

Regards,

Andrew