วีดีโอ

Hi there, Some initial questions: Is the absolute difference between measured and simulated voltage drop less than 10%? Do you measure more or less voltage drop compared to the simulation result? If you have different measurement values, it can be due to the following reasons: * The copper thickness of most manufacturers has a tolerance of +-10%, this can vary from batch to batch. * This 2D simulation does not include the vias. If you have a lot of vias in your design there might be some differences due to that as well.

@panire3 it's more than 10%. My measurement took around 50 mV and my simulation 17 mV like seems like a bit off. I watched another video PDN they are using a Elmer and have less than 10% error. I followed it but got stuck on mesh parameter always error. Sorry out of topic. If you have sometime try to consider making a video of 3D simulation for PDN.

Many DDR layouts have the GND pulled back, so in a way I agree that its kind of the industry standard for routing in multi-layer PCBs. If thats the "best" way I am not sure. In fact via stitching every signal could cause less emissions (see here: th-cam.com/video/jYVD-8Wgdq8/w-d-xo.html), however vias take up much space, so removing GND from the top layer is often done due to limited board space.

Far Field: I will upload the far field simulation process on udemy this week and will re-comment here once done. Determination of required timesteps: Yes, for many problem types there is a formula for the required timesteps, will upload it on udemy those next days and keep you updated. Min decrement: I sadly dont know a formula for that one, I typically use a very low value and limit the simulation time by the number of timesteps. Inductors / chokes: As of now the lumped inductor models seem not to work. So a choke still need to be modeled by via the 3D geometry. Thats kind of hard to do especially if you want to simulate a specific partnumber of a given manufacturer. I am working on a video for that, but this one is going to take a while.

​@@panire3thank you for your hard work to share your knowledge on this. Looking forward on your udemy course update soon👍

Hi again :) on udemy now are following updates: Lecture 20: I added an Excel sheet on how to calculate the required timesteps when injecting a step signal. Lecture 19: check this one out regarding crashes. PS: what exactly do you mean with crashes. you mean that you get error messages or do you refer to very long loading times? If its mainly about loading times, then the new lectures 16 and 17 might also be very useful. I will text again once the far field tutorial is uploaded.

@panire3 thanks! Just checked it out yesterday. Haven't gone through but thank you very much for the udemy updates. Regarding the crash, it was not related to the long loading issue. It was more on that, openems immediately terminates the session and an error message code that I have shared with you in previous lectures. It goes something like this: error: openEMS binary exited with error-code -1073740791 However, after playing around with the timesteps and min decrement and maintaining a fine mesh grid, i think i did around 0.05, the error message disappears. When I increase the grid to 0.1, same issue occurs but after playing around with the timestep and min decrement, it disappears. Probably it might have something to do that openems breaks itself due to the fine mesh with to big of a timestep but with small min decrement. So, i came to the conclusion that it has to correlate with the mesh size, timestep and min decrement. So far, that was the issue that i had last year and i think my current workflow was to use a smaller timestep like in your previous video, maintaining default min decrement, grid size of 0.5. The lowest i could try was at 0.3 before the same error code message occurred again. Looking forward on your far field video this week. Need to learn how to use for my current project actually😂.

Thank you for that input. :)

Hopefully, that was just a rhetorical comment. 😄 Welcome aboard this little open-source adventure!

Thank you for the kind words! 😊 I’m glad to have you here + enjoy the learning journey!

Your problem description makes me think you are not using freecad 0.19.4. Please use exactly this version, else you might have problems with the import button. Also please use exactly kicad 6.0.11 Even the slightest change to the versions will likely cause problems.

if this change does not help, dont be shy to ask for help instead of giving up due to frustration

@ I did get the import to work now, had to patch a few things, but the imported geometry isn't correct. It seems the importer doesn't process correctly vias where there is only an annular ring on connected layers and not all layers. It's a recent-ish kicad addition so I'm guessing it was never updated to support that.

@@smunaut I was able to also work with only an annular ring when changing the grid size in x and y direction to 0.05mm instead of 0.1mm as shown in the video. Here are some screenshots on how to review the generated PEC dump, hope they help: drive.google.com/drive/folders/1BWAzePieKENgAA3u-ROWe8_CsoqpEZNL?usp=drive_link

Absolutely, there will be some differences. Later this year, I plan to make comparison measurements with real PCBs, which will provide an opportunity for us to refine the model.

The fuse copper does combine all nets in a single object, but this doesnt mean there is a short circuit somewhere if thats what you mean. So with "fuse" you still end up having multiple nets and multiple ground planes as defined in your kicad layout. If you encounter issues importing a large file, it’s helpful to break the process down-import one trace at a time until you identify the problematic part. Attention: The process here does not work for "microvias and burried vias (or the so called HDI PCB production process)". If you dont want to bother to figure out your own process on how to import microvias, I show how to do it in my udemy course.

Sadly not a full SMPS, because: openEMS as of now does not support diodes and mosfets. What you can do is to simulate the input and output filter of a SMPS.

​@@panire3 thanks for a detailed answer. I have another question, is this maybe possible to do a thermal simulation. For example a cree xp e2 led 4x4 mm with a ground plane, what temperature the ground plane would be and crystal of the led?

​@@salovamrani2084 The only free tool I sucessfully used so far for thermal simulations is Elmerfem. But I only managed to simulate single materials (e.g. only the heatsink only or asuming the leds are mounted on a pure aluminium PCB) with airflow - see my udemy course. If you want an exact simulation that includes the PCB layout, I havent figured out this yet.

Nice to hear that :) and I hope you will enjoy using the tools for your own projects.

I think the docker script idea of you is great, I will have a look into it, thanks for the suggestion. +Sure, i can add the full simulations to the end of the videos.

Hi there :) Sadly I also didnt yet manage to make the import tool work yet with FreeCAD 1. Change to FreeCad 0.19.4 or 0.21 and the import button should work.

Youre correct :) check out this newer video, it shows a much more automated approach using the 3D model of the PCB: th-cam.com/video/trNCL96CELo/w-d-xo.html

glad to hear that it helped you to speed up the process. :)

This might be the case, have fun testing it out with your own simulation! 😊 My concern with using thicker traces is that more of the signal might bypass the capacitor. If the traces are only as wide as the capacitor pads, the signal is more likely to flow through the capacitor.

Hi again :), hmm i know this error but i cannot recall in which exact scenario it happens. feel free to send me an example openems code via email and i can very likely tell you where the error comes from.

@panire3 hey panire! Sorry for the late reply. Finally got it working. The rootcause was due to the geometry. Had to fix the geometry and give it a try again, and the warnings were finally resolved🫠. It was frustrating and rewarding at the same time. I guess you faced more frustrating times on the tool too, i guess😄😄

​@@nurhaida1983 Glad to hear you got it working! 😊 yeah, I can feel your frustration 😅 and also I can feel your joyful moment after finally figuring it out. 🎉 ⚡

​​​@@panire3 i'm looking forward onto your next video. 😁😁👍. I know you mention that it'll be in a future video, but if I want to simulate far field, is the sine excitation given the same way as in your previous video?

@@nurhaida1983 the excitation method is the same, the only thing to worry is the grid size once the simulation gets bigger to keep simulation times at a reasonable level. I will show how to work with a variable grid size for the far field simulation

@@panire3 Oh wow when I saw that you mentioned 64GB of RAM It didn't catch my attention so much so but now that you've highlighted it again, that's a scary lot. I have a 32GB RAM workstation laptop which means I'd have serious difficulties. But what is responsible for demanding such RAM demand? Okay is there a way we can raise this issue with the KiCAD community?

@@saviouremmanuel3608 Yes, if we try to push the issue to KiCad, FreeCad or openEMS community, we might find out whats going on with such a high ram usage. After all, these three software people dont ever try to communicate with each other. we might be able to bring them together to talk about this and make an impact! @panire3 let me know if you want to make an issue on KiCad gitlab and share the link so we can support this, or I could do it instead :) (for me, I could not run the far field simulation of the board, my ram hangs up, I only have 16GB)

responsible is that the outter ring of the vias ia just 25 micrometers and FDTD is not good with moving from very fine meshes to rougher meshes (adaptive meshing).

Hi :) 1) KiCad 6.0.11 2) FreeCAD 0.19.4 3) Sadly just by trial and error and a lot of patiente Apologies, I realize now that it's not clear which version was used. I should have included this information from the start, as it's crucial for anyone looking to replicate the simulation.

If it is not done, FreeCAD gives me an error when trying to cut out the copper from the airbox, so its just a workaround. I dont know the exact reason for the error, it might be an intersection problem.

Hi there, interresting request, after 35 years in this world i never heard about common source, gate/main loop inductance). 😅 I will have a read to see what it is about somewhen those next days.

Hi! Yes, I’m working on it, but haven’t gathered enough material yet to make a full video. 😊

@@panire3 Aw I see, I am planning on ordering my first proper PCB soon, so I would have loved to see that video. Would you mind stating what the technique is? Is it do with stitching capacitors?

​@@flamebull5 yes, layout-wise stitching is for sure one of the biggest improvements that i dont see on many boards.

Hi there, the lumped inductor model does not work yet and based on the github discussions group of openEMS it wont be implemented anytime soon. Simulating inductors as of now just works by defining them via their shape. I will make a video on that once for sure.

Hi there! I touched on this topic a bit in my last video. At some point, I'll likely need to make another video about it since I don't fully understand it yet. theoretical basis: The antiparallel layout style forces currents in opposing direction for each capacitor. The opposing currents reduce common mode current around the capacitors, but I’m still unsure how exactly this improves the S21 performance.

@@panire3 thanks. i am subbing you

Same problem here even on version 0.21 is not running

I face the same issue, which is why I always stick to version 0.19.4. The same applies to KiCad - even the slightest change leads to major problems. 😊 For anyone using openEMS for the first time, I strongly recommend using my exact setup without any modifications to avoid unnecessary frustration.

It looks like "import WebGui" has issue

if you comment above line out the button will appear enabled

You're welcome! Thanks so much for watching and even taking the time to leave a comment :D

Ah thats quite a bummer to hear - I think downgrading is not possible in KiCad. Alright promise, my next youtube video will be with KiCad 8. Reason why i used KiCad 6 so far: I started with it and once i tried to move overe to KiCad 8 nothing worked anymore, so I thought why not wait for KiCad 9 to make a transition as KiCad 8 is soon to be phased out anyway. Thermals: Sure I can add a heatsink example with and without airflow. What I am still strugggling even after months of failurees is to make a thermal simulation with multiple materials, so combining heatsink+PCB in one simulation still doesnt work with any free tool that i tried.

regarding 4:21: I didnt check this one yet. But will do it now and give you feedback. As the material of the ground layer and traces is set to PEC (perfect electric conductor) it will for sure behave differently than copper which has limited conductivity. 4:55 yes its not strictly necessary to make a union. And you are right that with a complex board this is quite a hazzle.

4:21 here is the current distribution inside the via. so even though its a full copper via, the current is mainly concentrated at the surface: drive.google.com/file/d/1sNpWKgx8_lCW1eQKQSZFhLtPzLg_GMwm/view?usp=drive_link

@@panire3 Excellent - thanks for checking it out. So this means the filled via approximation does not impact the RF performance. Solid PEC would also translate to an inner domain boundary (as long as we're not including finite conductivity in the material parameters). Next, it would probably be interesting whether a 3x3, 5x5 cell or other very small cube is sufficiently representative of a via barrel that may not be well-resolved spatially - or whether there are magic number / cylinder-to-cube conversions that produce close matches for the inductance to de-feature the round shape of a via.

​@@hardwareful The cube approach is used in some 3D export tools (unfortunately, I can't recall which tool was exporting cubes), so that's definitely an option, and I have seen it being used. For rough estimations with fast simulation times, I am certain this would work.👍

And thanks for watching :)

The inductance is quite accurately represented through the geometry. To my understanding (based on what I have read on the Github discussions group), the option to select inductors in the FreeCAD to openEMS macro is just a placeholder for future functionality. The lumped inductor model does not work currently, and based on discussions on GitHub, it may not be implemented anytime soon.

Access has been updated, sorry for the mishap

Thank you and glad you think so :)

it performed slightly better than the others if used as combination of 0402 packages. But I am not 100% sure yet, the new video will be published today, hope it will be interesting.

Yes i strongly recommend using the exact same version FreeCAD 0.19.4 and also exactly KiCAD 6.0.11. I had big problems with even the slightest changes.

@@panire3 Hi, I bravely took the challenge and tried it with KiCAD 8 and FreeCAD 1.0. Actually it is very easy, although the import button of the macro doesn't get activated in FreeCAD 1.0 at all. But KiCad 8 allows for a STEP export of the board. You can easily import this step file into an empty FreeCAD document. It contains the board as one item and the copper including vias etc. as a bunch of other items. These can easily be combined in a Fusion and a simple copy created from them. Then I could proceed as you show in your video, i.e. delete all intermediate items, create air box, field probe layers and input port and assign the properties in the tabs of the macro.

@@gpetrowitsch Wonderful to hear it worked. This comment will help and encourage a bunch of people including me. 🤗

I had the same problem with another laptop using "Intel® HD Graphics 3000" but did not manage to fix it. It most likely has something to do with opengl support of your graphics driver i believe. -> Maybe a graphics driver update helps. With other laptops I had it working on specific windows versions only, for example it worked on windows 11 and 7 but not on windows 10 and 8.

Hi there! Whether it's worth it depends on your current monthly income. This video provides an additional option for you to consider. The main reason I started this channel was that a commercial license cost over $10,000 per month-several times my net salary. That left me with two choices: either go with free tools or forego the ability to simulate my circuits before production.

@@panire3 I understand, $10K is waaaaaay over my budget too, however, there has to be some kind of "middle ground" between those exorbitant packages and doing it "bare bones" like you are. Of course it's nice to be able to do simulations, plenty of good free software exists for that, what I think you are referring to is EM simulation and the status quo is either slaving away at the keyboard for ages or being a slave to those corporations. If you have the time, great, many like me simply can't afford the time spent on learning a whole bunch of new stuff just so I can see how to improve impedance bumps. Not that I wouldn't like to, but until it's more user friendly (OpenEMS or others) I will forego. I am an RF engineer, not a software engineer. An analogy, imagine having to build your car each time you want to drive it, I just want to get to work and do my job.

​@@paulov9626 Yes, I was referring to 3D EM simulation. And I agree-it’s quite challenging to get started with openEMS and also all the other tools on the channel. There could be a middle ground, although I personally had difficulty finding one. If anyone following this discussion knows of a good alternative and can share pricing details with everyone, that would be amazing. 🙂

@@panire3 Or maybe even scripts, pre-made libs and functions and also even macros to automate things and minimise all the typing, or at least reduce it. Since OpenEMS is open, is there not some scripting ability to enable the above? Then some clever souls could help out a whole bunch of people by writing them.

@@paulov9626 Yes scripting the most common usecases will be the way to go. Hopefully a clever soul, or even better, a group of clever souls will be sucessfull with it one day 😊

Yes, would be nice to have more things. For sure one day we can have an Arduino running in the simulation that drives some kind of motor. Multiphysics simulation would be also a nice thing to have, lets seee how far we will be in a few years.

Hi there, with boundary do you mean the physical simulation boundary? If thats what you meant, then yes, all fields in general extend to infinity (at least until they hit GND potential somewhere). I handle this by using a PML boundary all around the simulated components. PML means "perfectly matched layer" and it kind of works like an extension to infinity. This would also work well with a bigger port, also its important to keep some distance between port and the PML boundary

Hopefully next weeek 🙂🙃

@@panire3 im going to stay up all week next week refreshing

@@carsonau116 haha, to not take away your weeks sleep: it will be friday