[Moved]: EM simulation of a combiner in Sonnet.
I am simulating a simple combiner in Sonnet but I am getting strange results (specially at DC). Let me describe the combiner shown in the Figure 1. The combiner has a width (metal 5) of 18.5 um that will be connected to bases of transistor through vias because the contacts for the bases of transistor are on metal 1. I need to use metal 1 also as a ground plane for the design of my power amplifier that is why I shorted it to the sonnet box. The ground plane is separated from port 2 and port 3 of combiner by small gap allowed by the technology I am using. I have defined all ports as standard with some reference plane.
Can someone please help me and point out what is going wrong.
I recommend you to use "Frequency Sweep Combinations" in analysis setup in Sonnet.
Select this option and sweep the frequency in 2 steps. DC and High frequency.Try..
This is what I did. Thanks for your reply. I figure it out. the problem comes if you change the default frequency to 1 Hz or any other small frequency for DC.
Since you work at 170-200 GHz, why you consider DC point ?? HF EM simulators have less accuracy toward to DC.
You can check your structure in ADS too because ADS recognizes DC better than Sonnet to my past experiences..
Yes, I compared it with ADS momentum and both gave same results. The sonnet took a lot of time in calculating DC and RF frequencies. I am designing the matching network in a way to include the biasing circuit inside matching network that is why I am simulating DC.
I am not convinced about ADS EM simulator. Is there any way to get highly accurate results in ADS momentum at mm waves. Since, Sonnet takes a lot of time and memory.
No, ADS gives pretty accurate results at any frequency even at DC.All you have to do is to configure Momentum well ( meshing,simulation mode).I have done maybe more than 50 serious designs with Mometum and the measurements were quite consistent.
EmPro is also accurate simulator if you can use.
Thank you so much. I will use ADS then. Are there any publication at high frequency about IC design done with ADS. If you have know could you please share it with me.
Yes, Sonnet's method has a minimum frequency limit, and analysis fails below that, when the cell size is too small compared to wavelength. That is why Sonnet developers implemented a special "DC" option that technically is a safe very low frequency value. Low enough frequency to be close to DC value, high enough to give reliable results. See Sonnet help, topic "DC Frequency".
Momentum uses a slightly different method and can analyze down to DC. Both solutions work well from my experience.
Thanks I will read that Topic. Are there any online materials about EM simulation of ICs at very high frequency in ADS momentum ? I am designing analog integrated circuits at mm waves so far I am using Sonnet and it is taking too much time and memory.
I'm sure the Keysight guys have reference materials to show.
I'm working as an independent EM specialist in RFIC at RF/mm-wave, and was a Sonnet reseller in the past. For fairness reasons, I want to stay want away from that claimed accuracy marketing battle.
If you are looking for appnotes:
http://muehlhaus.com/support/ads-application-notes
Another thing which I almost forgot, is this definition of local ground correct ? Is it compulsory to make the sonnet box far from the circuit ? If I want to make it far from the circuit then, I have to increase the length of the I/O lines and increase the reference planes which will require more memory and increase simulation time. ?
You don't have increase Sonnet Box, all you have to care about is its Resonance Mode.There are some materials about how to choose the right dimensions for Sonnet Box.
I don't remember but I saw one it once..
It means that if there are no box resonances then we can accept the results depending on the mesh size and port setting. I simulated the same structure in ads momentum with 45 cells/wavelength but the results were different for S11 as you can see in the attached file from 160 GHz to 220 GHz.
On the sides with the port feed, port calibration will remove the side wall effect. Feedline should be long enough so that higher order modes from the DUT decay, a few line width is usually fine.
But you need to care about the other side walls, don't put them too close! Your example above is perhaps a bit too close. Double the distance and check what it does to results. Don't go too far - make sure you have no box modes.
Regarding Momentum mesh, the cells/wavelength is not always the best setting to ensure a good mesh. Also check edge mesh (especially when simulating thin conductors). In RFIC technologies where everything is small to wavelength, I often use an absolute value for maximum mesh size. In Momentum, you can set the mesh globally, or per layer, or even for each polygon individually. Visually check that the mesh looks fine enough.
I tried it by making the distance three times the substrate thickness as suggested in the manual but it gave almost same results. However, in this case it took more time. How about the local ground plane ? if we want to define a local ground plane then we just have to connect that metal layer to the sonnet box side wall as I did in this example ?
That's fine, so we know the distance is large enough.
If we are exact, you then have two ground paths: (1) the box walls and (2) your drawn ground path are in parallel. However, with the much larger path length in the box path (and resulting series inductance), the current will flow almost exclusively through your drawn ground. So that is fine.
You could also use an isolated ground, with properly defined (-) terminals for the ports, but then we need to be careful with deembedding the port parasitics. I would use your method as shown above.