tallface65
Full Member level 6
- Joined
- Dec 8, 2009
- Messages
- 343
- Helped
- 190
- Reputation
- 394
- Reaction score
- 173
- Trophy points
- 1,323
- Location
- land of the free
- Activity points
- 3,407
I read a post that was describing the frequency sweeps in HFSS, and I noticed that some of the information was out-of-date. But the poster had good intentions :grin:, so this is a follow up with updated and correct information. Latest Release: HFSS13.0.2.
There are 3 frequency sweeps in HFSS: Discrete, Fast, and Interpolating.
Discrete Sweep: This sweep returns fields and SYZ parameters for all frequencies of interest. The time to solve is proportional to the number of frequencies desired. For a Discrete Sweep, the highest frequency in the band is desired to be the Solution Frequency for the Adaptive Mesh Convergence. Each frequency solved in a Discrete Sweep uses the mesh from the Solution Frequency, therefore the time will be proportional to the number of frequencies and the time of the last adaptive pass. Peak RAM will be close to the RAM needed for last adaptive pass. There is no bandwidth limitations for the Discrete Sweep. DSO option scales linearly with Discrete sweeps.
Fast Sweep: This sweep is ideal for narrow band antennas or for resonators that have just a few resonances in the band. This sweep returns SYZ parameters and field quantities for all frequencies of interest. The fast sweep is based upon an extrapolation routine and thus is limited in its bandwidth of use to no more than a decade, ie 1-10GHz or 500MHz to 5GHz. For a Fast Sweep, the operating frequency of the antenna should be the Solution Frequency (Or Center frequency if resonance is unknown). This is due to the fact that the sweep is extrapolating so one wants the most accuracy in the operating frequency of the antenna or center frequency of operation as the extrapolation is equal in both directions. If one solves and after solving wants more frequencies in a bandwidth, there is no resolving as it merely re-samples. The time required for a Fast Sweep will be faster than a Discrete Sweep and retrieve all the same information. However, the RAM needed for a Fast Sweep can be larger than the RAM for the last adaptive pass, so if one is RAM constrained or near the limit, a Discrete Sweep may be better if fields and SYZ are needed. DSO cannot be utilized.
Interpolating Sweep: The Interpolating Sweep uses a rational polynomial fit to the S Parameters vs frequency. The sweep is an iterative process that solves discrete frequencies until the fitting function changes, from frequency to new frequency, by a percentage criteria. The Interpolating Sweep only returns SYZ parameters, however it is ideal for these as it has no bandwidth limitations and has been well improved for DC extrapolation! Since the Interpolating Sweep is fitted by a continuous function, if one solves and after solving wants more frequencies in a bandwidth, there is no resolving as it merely re-samples the continuous function. The solution frequency of the Adaptive Mesh should be the highest frequency in the band of interest. Interpolating Sweeps are the most commonly used sweeps, particularly for simulations requiring broadband S Parameter export or SPICE models as passivity is easily seen. ALSO, it is highly discouraged to break a frequency sweep into two separate sweeps. In HFSS 10 there was advantage to this, however in the latest version, a single sweep will solve faster AND be more appropriate and accurate for post analysis. DSO can be utilized very well to reduce necessary solve time!
In summary, if one compares these frequency sweeps within a decade of bandwidth, they will overlap. Each frequency has its strengths and computational requirements (time/RAM), but hopefully this will help a user determine which sweep is best for them! :razz:
There are 3 frequency sweeps in HFSS: Discrete, Fast, and Interpolating.
Discrete Sweep: This sweep returns fields and SYZ parameters for all frequencies of interest. The time to solve is proportional to the number of frequencies desired. For a Discrete Sweep, the highest frequency in the band is desired to be the Solution Frequency for the Adaptive Mesh Convergence. Each frequency solved in a Discrete Sweep uses the mesh from the Solution Frequency, therefore the time will be proportional to the number of frequencies and the time of the last adaptive pass. Peak RAM will be close to the RAM needed for last adaptive pass. There is no bandwidth limitations for the Discrete Sweep. DSO option scales linearly with Discrete sweeps.
Fast Sweep: This sweep is ideal for narrow band antennas or for resonators that have just a few resonances in the band. This sweep returns SYZ parameters and field quantities for all frequencies of interest. The fast sweep is based upon an extrapolation routine and thus is limited in its bandwidth of use to no more than a decade, ie 1-10GHz or 500MHz to 5GHz. For a Fast Sweep, the operating frequency of the antenna should be the Solution Frequency (Or Center frequency if resonance is unknown). This is due to the fact that the sweep is extrapolating so one wants the most accuracy in the operating frequency of the antenna or center frequency of operation as the extrapolation is equal in both directions. If one solves and after solving wants more frequencies in a bandwidth, there is no resolving as it merely re-samples. The time required for a Fast Sweep will be faster than a Discrete Sweep and retrieve all the same information. However, the RAM needed for a Fast Sweep can be larger than the RAM for the last adaptive pass, so if one is RAM constrained or near the limit, a Discrete Sweep may be better if fields and SYZ are needed. DSO cannot be utilized.
Interpolating Sweep: The Interpolating Sweep uses a rational polynomial fit to the S Parameters vs frequency. The sweep is an iterative process that solves discrete frequencies until the fitting function changes, from frequency to new frequency, by a percentage criteria. The Interpolating Sweep only returns SYZ parameters, however it is ideal for these as it has no bandwidth limitations and has been well improved for DC extrapolation! Since the Interpolating Sweep is fitted by a continuous function, if one solves and after solving wants more frequencies in a bandwidth, there is no resolving as it merely re-samples the continuous function. The solution frequency of the Adaptive Mesh should be the highest frequency in the band of interest. Interpolating Sweeps are the most commonly used sweeps, particularly for simulations requiring broadband S Parameter export or SPICE models as passivity is easily seen. ALSO, it is highly discouraged to break a frequency sweep into two separate sweeps. In HFSS 10 there was advantage to this, however in the latest version, a single sweep will solve faster AND be more appropriate and accurate for post analysis. DSO can be utilized very well to reduce necessary solve time!
In summary, if one compares these frequency sweeps within a decade of bandwidth, they will overlap. Each frequency has its strengths and computational requirements (time/RAM), but hopefully this will help a user determine which sweep is best for them! :razz:
Last edited:
