NuHertz is another tool available on the electronic desktop from Ansys. It's a powerful filter design tool. The interface is simple and easy to learn, and you can choose from tens of designs.
Figure 1 shows the NuHertz tool's interface. In a previous blog, we discussed the FilterQuick interface. In this blog, we will present the advanced interface. The user needs a lot of background on filters to use this interface.
Figure 1
Filter Type Section:
We have many options to choose from in the filter type section, more than the options we saw when we used the filterQuick interface. We will select the Chebyshev type I for today's lecture, which has ripples in the passband.
Figure 2
Filter class:
In the Filter class section, we have lowpass, highpass, bandpass, and stopband. We select the passband. It is a very famous option for wireless applications.
Then select the implementation: Lumped, distributed, active, Switched cap, and digital. We select the distributed option. This means RF lines on a PCB will be used to do the filter design, not lumped elements.
Figure 3
Filter Specifications:
In the filter specifications or attributes, the user needs to enter the requirements.
Figure 4
1- One can specify the order, the center frequency, the passband, and the ripple in the passband. This is enough information for the software to make the design.
2- One can also select to specify the stop band and the stopband attenuation, and in this case, the software will decide the order.
3- In these designs, the software assumes that the passband definition is that the attenuation at 200MHz, 100MHz on each side is -3dB. The user can change that, just deactivate this option.
4- The user can also specify the filter by edge instead of center frequency.
5- Add Tx transmission zeros, which are zeros, to the S21, in order to improve the group delay flatness. This topic will discussed in another blog.
6- One can request that the design should have arithmetic symmetry. It is another way to improve the group delay. If this option is not selected, the amplitude response of the filter will be symmetric, but the group delay will be bad. If the user chooses to use arithmetic symmetry, the group delay will be flat or almost flat, but the amplitude response of the filter will not be symmetric. And the filtering on the high-frequency side will be poor.
7- Add the asymmetric in the physical design to overcome this defect caused by the arithmetic symmetry. This will give the software the capability to maneuver and create a better and symmetrical filter.
8- The last thing to talk about in the specifications is the rippling modifications. Constrict ripple. What does that mean? Usually, Chebyshev and Elliptic filters produce a flat equiripple response. This increases the order of the filter. There are applications that require equiripple at the edge of the band. Restrict the equiripple condition to a section of the filter. For example, 50%, i.e. only to the top 50% of the filter for low pass, or 25% on each side for the passband.
9- Single point ripple is where we restrict the response to one point to reduce the load on the filter design. This can be accomplished by adding zeros at zero or at infinity. In a future blog, all these cases will be discussed.
10- And the last thing is the half band ripple. So, the traditional filter spreads its zeros across the band. For some applications, it is better to spread half of the zeros and leave the other half at zero or infinity. According to the manual, this eases the sharpness of the group delay and aids in group delay equalization, accommodates potential manufacturing restrictions, and accommodates some planar geometry requirements.
All these options require a strong knowledge of filter design.
Topology panel:
Figure 5
1- Because the distributed option was selected, many topologies are available to choose from. Choosing any one of them will change the entries in the panel. We selected to use the combline.
2- Tapped has different meanings for different filters. For combline, it means the user wants the input to tab to the first resonator. If not selected, then the input will be connected to the first resonator directly. See here an example.
3- Equal width approximation forces the use of the same width for all resonators.
4- set Center Zo if you want the filter line to have an impedance different from the input and output impedance.
5- specify the input and output impedance. We can also specify the ½ length frequency.
Substrate panel:
Figure 6
NuHertz needs this info to do the right calculations. Select the transmission line type in the dialog box. Depending on what was selected, the substrate parameters will change.
We'll go with microstrip. If the user wants, add a cover.
Select the conductor and dielectric thicknesses. Select the materials. Choose one from the list. If the user can't find the right material, pick the closest one. Choose the conductor material. Each dielectric material has its own tangent loss number. The user can select to override that.
Geometry panel:
Figure 7
1- This panel is only applicable if the distributed filters were selected. Use this section to add some constraints to the geometry. For example, in section 1, specify the geometry of the section that will be used to represent a capacitor. We do that by specifying the line width used as a function of the dielectric height. Then, the software calculates the necessary length to create any capacitive required. The same length is for the inductive sections.
2- Choose to split the stubs into two parallel stubs for wide stubs. Practically one upper and one lower. Usually, all the stubs are on one side, upper or lower. If chosen, the user needs to enter the width-to-dielectric height ratio that will trigger the split. If one selects 4, for example, then split if the width is greater than 4 times the dielectric height.
4- Check the alternate stub orientation if you want the software to alternate the subs, one up intentionally and the next one down with or without splitting.
5- In the geometry limitations section, specify the manufacturing design rules. The minimum width and gap. Maximum width and gap. Select this option to enforce these limits. And select this option to allow the software to change the filter length to accommodate these limits. Adjusting the length on the limit means allowing the software to adjust the length of any structure used to mimic a capacitor or inductor and using the numbers on top or the minimum values used here for the width.
Radial panel:
Figure 8
1- The user can select the shape and angle of the stub. The standard is a 90degrees stubs. But the user can ask for radial or delta stubs.
2- For the radial ones, specify again if the software can split any large stub into upper and lower stubs. The user can also ask the software to alternate the stubs, one on top, the next on the bottom, and so on. The user can also implement offset stubs. It can be attached to the transmission line, or one can shift the stub away from the RF lines by offset. Why? To make it easy to manufacture.
3- Add limits to the minimum and maximum angles. For radial ones, keep it between 15 and 140, but for delta ones, use between 15 and 120.
Graphs:
Figure 9
The graphs are produced using the buttons shown below. These 5 buttons generate the 5 graphs.
In all of them, the user can:
1- rint the graph to a PDF, OneNote, or Orcad file on the top bar. You can also copy to clipboard, then paste into another application, like PowerPoint, Paint, or Word.
2- specify the y-axis limits. By the way, the x-axis limits are specified on the other side
3- see the curves in a text file.
4- Freeze will simply produce a new graph. Any change made to the design will affect the original graph, but the frozen copy will keep the old results.
5- zoom in, move left and right, and restore the original graph.
6- display the mag, phase and group delay in the frequency response.
7- show the graph in Smith chart pr polar.
8- if one clicks the right button, will add a marker. If click on the marker, one can adjust its location.
9- see the step, pulse, and ramp response for the time response. This is important for the lowpass filters. No value for bandpass or high pass filters.
10- see the poles and zeros of the filter. By default, the poles and zeros are of the S21 as the transmission response.
11- One can also see the poles and zeros of the reflection.
Figure 10
12- ask the limits of the x and y axis to be the same to see things on a scale.
Figure 11
13- change the grid to polar.
14- change the location of the poles, but before doing that, activate this option. As one moves the poles and zeros, the response will change in the other graphs.
15- Select the prototype option, which is a low pass filter with a cutoff frequency of 1radian per second. From this low pass, the bandpass was extracted. If one activates this prototype, the user can modify the location of the poles and watch the change in the bandpass filter. These things need an expert person in filter design. The last graph is the function used to design the filter. Display the prototype, which is the ideal lowpass filter with a cutoff of 1 radian per second.
Figure 12
15- display the function in its standard form,
Figure 13
16- in a cascaded Biquads form,
Figure 14
17- or parametric Biquads summation form. Biquads are because of the S square.
Figure 15
18- specify the maximum number of digits for each number.
19- And the fit, to try to fit the formula on the screen.
Synthesize:
When you click synthesize, the software will build the filter based on the entries.
Figure 16
Here, we see the filter schematic. You can see it in Layout form or 3D. You can see all the numbers and dimensions here. If you select the other information, you see the impedance of each stub.
Like all other graphs, one can print, copy, annotate, generate a netlist, and edit the filter dimensions.
Figure 17
The Monte Carlo is where the user can ask the software for sensitivity analysis. Fill out the dialog box to specify which parameters to change, the tolerance, and the number of trials.
Figure 18
The user can ask the tool to save all the runs of one run, not just the optimum. One can also ask the tool to record all runs if running Monte Carlo many times. Monte Carlo will be discussed more in a future blog.
Exporting:
We're finally getting to the fun part: exporting:
1- export the S parameters, Z parameters, or Y parameters
2- export it as a DXF
3- export to Ansys electronic desktop. The first thing one needs to do is set up the export. Here's the dialog box, with so many options
Figure 19
3.1- send the design to circuit, HFSS, or 3D layout.
3.2- want AEDT to simulate the filter after exporting.
3.3- want AEDT to calculate: S-parameters, group delay
3.4- specify graphs
3.5- to optimize the filter more? What are the goals? The information will be transferred to AEDT.
3.6- There are more options here for the side boundaries of the air, open or closed.
3.7- specify the substrate to extend to the edge of the metal.
3.8- specify the format, is it direct to AEDT, or a Python file that will generate everything in AEDT. Choose Python to learn how to do scripting in AEDT.
3.9- save and close this setup or go back to the default configurations, cancel all changes, and leave this dialog box,
4- The last form is 3D Data. This is a generic 3D file that can be used in lots of applications. It creates a txt file. Users can read this file directly into other applications or write a script to do it.
