Coaxial Cavity BP Filter Design Using Synmatrix and HFSS Integrated Workflow

Hi, this video is about Coaxial Cavity BP Filter Design Using Synmatrix and HFSS Integrated Workflow. Symmetrix is a powerful tool for designing cavity filters as well as diplexers and multiplexers. The purpose of today's video is to design a coaxial cavity filter.

However, we will use that as a basis for explaining how to use the tool. The following chart shows a few of the options available in Symmetrix. We will go through them one by one.

When you open Symmetrix phase one, filter synthesis, users have the option of designing a filter or a diplexer, a band pass filter, band stop filter, low pass filter, or multi band. You can also choose the topology. These are the different kind of topology available in Symmetrix.

You can choose the topology of the filter, and you can choose the filter you want to use. This is the whole interface of Symmetrix. Start first by saving your project. We choose to design a single filter and its band pass, going from one, that's the center frequency, and the band is 50 MHz.

You can also switch to start and stop if you want. We will choose a filter of order five, and this is the return loss we want. This goes in the black box here and this is the return loss we want. We can sort out.

And as you add information, each input and output of the filter can have the same value as the input. Click on the filter, choose input function only. You can also select the level for your advantage, or effort.

You can also add obvious crispness at the lower profound window method, if you do, your 32 MHz is disadvantage. We're going to click on the line with four values. I'm going to click on left is right, you don't need many values here. I'm going to commission some value.

The direction is tight, until 2100, so I think that'll do. When I click and press, if I click here, I'm going to add and here it is.

This is how you add zeros, which means you want a sharp drop and you want to have almost negligible insertion at a specific frequency because there is something else running in this band for our case we'll go without transmission zeros. You can see the results here.

This is the S parameter you calculate, all this is the group delay, and this is the power analysis, the dispersion is shown here. It's possible for the user to specify dispersion correction to be applied. This is for applying a dispersion correction.

We will talk about dispersion at the end of the video. To the right, we have what we call the normalized coupling factors. These are calculated by the same metrics based on the specifications you entered here. Now instead of the normalized coupling, you can also look at the bandwidth coupling.

This is the equivalent of the normalized coupling. There is a relation between them. It's equal to the coupling divided by the bandwidth, which is in our case 50 MHz specifications. This is for entering more stuff related to thermal. We're going to have a special video for that in the near future.

For the normalized one, you can also edit the metrics. You can modify these numbers. You can also edit the sign if you want to change the sign. Changing any one of these numbers, of course, will change the shape of your filter, so be careful when you do these changes.

To an s2p file or a touchstone file, you can export the metrics. It's solved. You can also load data.

So if you have a response and you would like to know the coupling coefficient of that response, you can load the data and Symmetrix will automatically read the s2p file or the touchstone file and derive the equivalent coupling parameters for that file.

Now that we are happy with our filter response, we are happy with all these entries, we move to the next phase, which is filter design 3D modeling. The following options are available with Symmetrix. You can have a coaxial cavity, a waveguide cavity, or a cylindrical cavity.

You can have a substrate integrated waveguide cavity filter or a planar coupled resonator. In our case, we will choose coaxial cavity. Confirm and start a new design. Say yes again. The first thing you do is you save this a new design. Now we'd like to save. Say ok.

Based on the specifications of the filter, Symmetrix goes ahead and calculates the different dimensions for the cavity. So we're talking about the cavity width, the cavity height, resonator radius, resonator height, and this is a range for analysis. Same thing for the height.

This is a range for you to calculate to sweep along to see the effect on the resonant frequency and the Q. You have the tuning screw. So that's the radius of the tuning screw and that's the depth of the metal disk computer with a gaming in this net construction.

Your guillotine thickness is 0.1 pixel axis error. When thenej originator's transmittance powerful PDK filter rooms here, we'll save this.

This is C Q 30. 2. You can specify this plate where the conductivity is 6. 1. Silver is used because it has so many good features for filters, one of the most important one is the conductivity.

It has the highest possible conductivity, which allows us to have the highest possibly conductivity, that allows us to have the highest Q. Accept these dimensions, click calculate all, and now it's going to do the sweeping, the sweeping.

So for a fixed post diameter, okay, so this is fixed at 0. 9. We are varying the height from 4.288 to 5. 513. That's for the resonator height.

As you can see, Symmetrix selected this point where the frequency is 1.196 and the Q is 8. 260. What's important for us first is that the Q should be very high. And the second thing is that the resonant frequency is higher than the upper band of the filter. That's very important.

Now if you go to fixed post height and you change the radius of the resonator, you notice that the Q would like is higher within this area. So Symmetrix selected this point. We can change this one to 1. 74. Make sure to change this also to 5. 25. That's somehow extreme.

Just to see what is the minimum resonant frequency we can reach. So I do calculate all here, and the minimum is 0.956, which is lower than the lower band of our filter. So, this is the minimum. So, this is the minimum. So, this is the minimum. So, this is the minimum. So, this is a good cavity.

We can use this one. It covers the whole band. So now the next step is immediately after here, step number two is single cavity. So what we're trying to do here is to design our cavity by playing with the tuning screw depth.

So we discovered that the tuning depth can change the resonant frequency from 0.95 to 1. 16. We want to resonate. We want this cavity to resonate at 1 GHz. So, we're going to use the filter frequency of 1.25, which is the center frequency of the filter. Apply a next step. Very good.

So now we want to play with the height. We want to change the height. Apply next. And here we do the tuning. Because we are looking for that point, construct the model. Select this point. Select here. Now Symmetrix is going to create the model in the electronic desktop in HFSS.

This is the model created by Symmetrix automatically for you. And it will do the calculation. And the resonant frequency is right at 1 GHz. So we finished designing our cavity. The next step is to go to the coupling scheme.

So what we want to do, we want to be able to reproduce these numbers that we saw before. All of them are the same numeric curve. So, I want to bundle the number. Yourself. If you notice, it's somewhat filled. I'm going to select this group as well. I'm going to valley and see how it is going.

So this one is the numeric curve, and at orange, it is 900. Oh, I saw the number 3 except that color is red, so I'm going to simply... Worth 0. Where is that number. And see if that number is angelic. And notice it's the same.

Like I said, it's the same. 1. You want to have a step width that step height and also the shift apply and next. Construct the model the same way we did with the cavity. So we go to HFSS.

You have the coupling two cavities and notice here we are using the eigen solver to calculate the coupling between the two. You can also use a driven solution to calculate the coupling around simulation. We have a solution.

It's around 0. 265. The numbers we're looking for are 0.9 and 0. 68. In fact, you can see the band here in this green area. So what do we do? We do another parametric step. So we have the iris width. We have also the coupling screw depth and we have also the step height.

So 0.285 seems to be the right number. It doesn't have to be that accurate. We just need to be very close to the value. So 0.285, that's the number we're going to use for to reproduce 0. 68. So we're going to use 0.68 and we're going to use 0.68 for the step height.

So we're going to use 0.68 pine-type set 0.6825 coupling. Now let's look for 0. 9738. So we see here that 0.9 seems to be also the right number for the second coupling for the step height. Good enough. Now we go to the last step, which is designing our input and output section.

These are the numbers proposed by Semmetrics, Vol 1 of Setnet, Vol 3 of Setnet 2.10, Vol 3 of Setnet Mount, Roses, Setcount, Setcross, Onboard, Anodes, Setlifestyle, Set AOA of Sample Data, 0 Vol9%,d Updateund Resultado'柱表, Lsd1 of Setline, Automatic Aggregate Capilar, Modul, Sqlsd Virtu,i100lax, Normalize, Music Application Handler, Software Performance, Screen Image Macro, Dressssize Advanced, Auto Automatic Can be Upward or Higher, Performance Color Information Setting, Mikey Duong Computer, Resolution Guys, Same Metrics, and the most important number is the port height.

That's the one we can play with in order to produce the coupling we are looking for, which is 1. 1208. So apply next. Let's see the nominal one. Let's construct the model again. Same metrics will construct the model in a chronic desktop. That's the one we can go back here.

We run the simulation and notice here that what we are looking for is the group delay. So that value of the coupling gets translated to the group delay and the number we're looking for is 10. 136. And the number we have is around 9.617 group for the group delay, of course, that's in nanosecond.

We can improve that by doing a parametric study. So we can see that our number falls between 2.475 and 2. 55. If we make it 2.5, now that we have all the dimensions, we can go now to the last step, which is full 3D modeling. And here we have all the information we want.

This is the topology we adopt. So if you click on any one of them, it's going to give you all the dimensions. Remember here, we said we're going to make this 2.5 port length. We're not going to change this. Same thing here. Now we have to check each cavity. Now we don't have any cross-coupled.

So we move to modeling and we ask Symmetrix to construct the matrix. And then we move to our next screen. So we have for us in HFSS. So now the model will be constructed. This is our filter. This is our initial design. And we run the simulation.

Now that we have a solution, we go to HFSS and we plot the results. As you can see, there is a response that is different than what we are expecting. That's okay. Now we can export the results. Make sure to select do not override solution renormalization.

So we go back to the same metrics and we activate optimization. Now we load the data from HFSS and you do extract matrix. Notice the difference between what we want and what we got from HFSS. And this is the difference between the two models we got from HFSS.

And this is the difference between the two. You notice the two matrices are not the same. If you click on arrow level, it will give you the Delta of the each coupling. So this is the arrow level. Now we can see that m11 and m55 are completely wrong. The Delta is huge.

So now we need to tune cavity 1 and 5. So that's what we're going to do. So we go back to the design and because this is positive, it means we need to push this more. So we need to push this to six, maybe six five. Let's try that. Modeling. Update the model. Then you run the simulation.

Then you upload the s parameters again and you look at the difference. We're going to see now the results of modifying the design. Simulation is done. So now we go to HFSS. We export data the same way we did with the original one. Let's give it a name, one one one one one.

And it says there are no 406 models left. We just need to make sure that XM long and long one we have improvement. We have improvement. So we're moving in the right direction. So we're going to repeat the same process again and again until we are able to tune all of them.

As you can see after a couple of trials, we are somewhere around here. Almost the change is minus one to plus one. You can even proceed further and see a nearly perfect match between what we predicted or what Simmetrics is predicting versus what HFSS is predicted.

Now you can go farther and reduce them almost to zero or close to zero in order to improve the return loss below -20 dB or even below -25 dB. Notice here that there's a difference between the solid line, which is the same matrix calculation, and the dotted lines, which are coming from HFSS.

And that's what we call the dispersion effect. So if you have a filter that's physically symmetric, then the curves will not be symmetric as you can see. You will see a fast drop at high frequency.

Now if you want them to be symmetric, then physically, like the red line, then physically, it will not be symmetric. It will be asymmetric. That's not what we want. So we would like to keep our filter physically symmetric. So we'd like to know what the filter is going to be.

And we're going to use the filter as the advantage of putting a solid line in this here. So look at the fly in the field. So there we have the table variable that is going to be antiquated here.

And we can do this and these are going to be the get edge network that's the get edge network is going to be local once again. And they're going to be the diversify first and hoping we don't worry that because we don't want them to get a half speak or four lines.

And we're going to give each line a reverse price and we call the get edgeii. This is how it works.

Now, if you don't want to use fresh volta, we're not going to let the data gets a half i am going to say uppercut in theão r place option between HFSS and Simmetrics by applying the additional dispersion. And by this, we conclude this video.