RFSim99 anomalies
Hi all,
Every once in a while I come across an anomaly in RFSim99. I'd be happy if someone could tell me that I'm doing something wrong in the following: Fire up the simulator. Tools--Design--Attenuator. Pi Section, 50R input impedance, 50R output impedance, 10dB attenuation. Simulate. !Current schematic will be overwritten; OK. Now note that the graphics window displays S11 as 2E+31dB. In the circuit window, edit the values of the two 96.25 ohm resistors; they have apparently been simulated to more decimal places than shown. Just manually enter 96.25 ohms for each. Now the magnitude of S11 and S22 show up as -102.69dB; that seems reasonable. The phases for S11, S12, S21 and S22 are all zero. Now, from the graphics window, File--Save results as S Parameter File, and save the file in some convenient place, as type S parameter files (*.S2P). Now delete the resistors from the schematic and instantiate a two-port in their place. In the 2 port properties window that comes up, select "Load File" and load the file you just created from the attenuator simulation. Notice that the phase is 89.994 degrees for each of the S parameters, at each frequency. Or at least that's what I'm getting. Continue on and connect up the schematic and notice that it simulates as 89.994 degrees. Edit the .s2p file with your favorite text editor and notice that it really does have 89.994 degrees in it. Boy, I wish I could make a frequency-independent 90 degree phase shifter that easily in real life! More similar fun: create a 10dB 50 ohm attenuator as before and simulate it. Add an 80nH inductor across the top, from port 1 to port 2, and put a 32pF capacitor between the bottom of the two 96.25 ohm resistors (connected together) and ground. Simulating that should give you an S21 that's 0dB at low frequency, smoothly transitioning to -10dB at high frequency, with -5dB at about 120MHz. But S11 and S22 have strange magnitude and phase behaviour. Change the 96.25 ohm resistors to 96.2 ohms, and the S11 and S22 magnitudes behave OK, though their phases are still strange at very low frequencies. Save as an s-parameter file again. For convenience, start another copy of RFSim99 so you can compare the results of a new simulation with the one you just did. In the new window, instantiate a 2-port and load the file you just saved into it. Notice in the file that S11 and S22 angles appear to match the simulation you just saved from; there's no 90 degree phase shift this time. But when you connect ports 1 and 2 to the 2-port you just placed and simulate it, you find the phase of S11 and of S22 are all messed up. Huh? Has my copy of RFsim99 gotten corrupted, or do others see this behaviour too? Cheers, Tom |
RFSim99 anomalies
K7ITM wrote: Hi all, Every once in a while I come across an anomaly in RFSim99. I'd be happy if someone could tell me that I'm doing something wrong in the following: Fire up the simulator. Tools--Design--Attenuator. Pi Section, 50R input impedance, 50R output impedance, 10dB attenuation. Simulate. !Current schematic will be overwritten; OK. Now note that the graphics window displays S11 as 2E+31dB. In the circuit window, edit the values of the two 96.25 ohm resistors; they have apparently been simulated to more decimal places than shown. Just manually enter 96.25 ohms for each. Now the magnitude of S11 and S22 show up as -102.69dB; that seems reasonable. The phases for S11, S12, S21 and S22 are all zero. Now, from the graphics window, File--Save results as S Parameter File, and save the file in some convenient place, as type S parameter files (*.S2P). Now delete the resistors from the schematic and instantiate a two-port in their place. In the 2 port properties window that comes up, select "Load File" and load the file you just created from the attenuator simulation. Notice that the phase is 89.994 degrees for each of the S parameters, at each frequency. Or at least that's what I'm getting. Continue on and connect up the schematic and notice that it simulates as 89.994 degrees. Edit the .s2p file with your favorite text editor and notice that it really does have 89.994 degrees in it. Boy, I wish I could make a frequency-independent 90 degree phase shifter that easily in real life! More similar fun: create a 10dB 50 ohm attenuator as before and simulate it. Add an 80nH inductor across the top, from port 1 to port 2, and put a 32pF capacitor between the bottom of the two 96.25 ohm resistors (connected together) and ground. Simulating that should give you an S21 that's 0dB at low frequency, smoothly transitioning to -10dB at high frequency, with -5dB at about 120MHz. But S11 and S22 have strange magnitude and phase behaviour. Change the 96.25 ohm resistors to 96.2 ohms, and the S11 and S22 magnitudes behave OK, though their phases are still strange at very low frequencies. Save as an s-parameter file again. For convenience, start another copy of RFSim99 so you can compare the results of a new simulation with the one you just did. In the new window, instantiate a 2-port and load the file you just saved into it. Notice in the file that S11 and S22 angles appear to match the simulation you just saved from; there's no 90 degree phase shift this time. But when you connect ports 1 and 2 to the 2-port you just placed and simulate it, you find the phase of S11 and of S22 are all messed up. Huh? Has my copy of RFsim99 gotten corrupted, or do others see this behaviour too? Cheers, Tom I have used that program... not to the exent that you have... and I have seen some "funnies" on occation. www.telstar-electronics.com |
RFSim99 anomalies
On 9 Sep 2006 11:23:33 -0700, "Professor"
wrote: K7ITM wrote: Hi all, Every once in a while I come across an anomaly in RFSim99. I'd be happy if someone could tell me that I'm doing something wrong in the following: Fire up the simulator. Tools--Design--Attenuator. Pi Section, 50R input impedance, 50R output impedance, 10dB attenuation. Simulate. !Current schematic will be overwritten; OK. Now note that the graphics window displays S11 as 2E+31dB. In the circuit window, edit the values of the two 96.25 ohm resistors; they have apparently been simulated to more decimal places than shown. Just manually enter 96.25 ohms for each. Now the magnitude of S11 and S22 show up as -102.69dB; that seems reasonable. The phases for S11, S12, S21 and S22 are all zero. Now, from the graphics window, ---snip--- Aarrgg! This is very disappointing! I usually check at least my values for LC circuits against the LADPAC programs In "Experimental Methods in RF Design", however. By the way, when I click on Tools/DesignAttenuator, the schematic that is in the pop-up window for default 10 dB shows the three resistor values as 71.151 and 96.248. When I go through the process you did, I get the same results. I saved the file, and closed and reopened the program. Then, I loaded the saved 10dBAtten.cct file, and of course got the same -102.69 dB for S11. Then, I Tuned the values back to 71.151 (which immediately displayed only 71.15), and tuned the other two to 96.248, (which displayed 96.25). When I simulated, S11 was now -107 dB! Regards, Ted KX4OM |
RFSim99 anomalies
On Wed, 13 Sep 2006 20:26:41 -0400, Ted
wrote: On 9 Sep 2006 11:23:33 -0700, "Professor" wrote: K7ITM wrote: Hi all, Every once in a while I come across an anomaly in RFSim99. I'd be happy if someone could tell me that I'm doing something wrong in the following: Fire up the simulator. Tools--Design--Attenuator. Pi Section, 50R input impedance, 50R output impedance, 10dB attenuation. Simulate. !Current schematic will be overwritten; OK. Now note that the graphics window displays S11 as 2E+31dB. In the circuit window, edit the values of the two 96.25 ohm resistors; they have apparently been simulated to more decimal places than shown. Just manually enter 96.25 ohms for each. Now the magnitude of S11 and S22 show up as -102.69dB; that seems reasonable. The phases for S11, S12, S21 and S22 are all zero. Now, from the graphics window, ---snip--- Aarrgg! This is very disappointing! I usually check at least my values for LC circuits against the LADPAC programs In "Experimental Methods in RF Design", however. By the way, when I click on Tools/DesignAttenuator, the schematic that is in the pop-up window for default 10 dB shows the three resistor values as 71.151 and 96.248. When I go through the process you did, I get the same results. I saved the file, and closed and reopened the program. Then, I loaded the saved 10dBAtten.cct file, and of course got the same -102.69 dB for S11. Then, I Tuned the values back to 71.151 (which immediately displayed only 71.15), and tuned the other two to 96.248, (which displayed 96.25). When I simulated, S11 was now -107 dB! Regards, Ted KX4OM I just ran the simulation in the LADPAC programs. I first built the attenuator in Ladbuild, with the resistors rounded off to two decimal places, and saved the result. In GPLA, I got (after resetting the scale to -120 dB) S11= -107.3. Then I edited the values to the 3 decimal place values from the original attenuator pop-up window of RFSim99, and got S11=-121.67. The circuit review feature of GPLA still showed 2 decimal places, rounded off, but saving that version of the attenuator and opening the saved program in Ladbuild showed three decimal place values when clicking on the component. It's surprising how a couple of thousandths of an ohm can change the S parameters that much. Ted KX4OM |
RFSim99 anomalies
Hey, thanks for verifying my results, Ted.
It shouldn't be surprising that tiny variations in resistance cause large dB changes. Remember, you're essentially unbalancing a bridge, and looking at the unbalance over 100dB down from full "un-balance." That is, -100dB represents a one part in 10^5 unbalance. 40dB return loss is really quite good, and something you expect only in a pretty carefully controlled setup. This same sort of thing reaches out and grabs you if you try to make a direct conversion receiver with I and Q paths to balance out the unwanted sideband. You need balance of not worse than ten parts per million to get 100dB sideband suppression, and the components you are using probably have, at BEST, temperature coefficients that can make things worse than that with only a one degree C change. On the up side, it's exactly this sensitivity that is a strong plus for bridge measurements. Also, I'm not about to throw RFSim99 out of my toolkit. I just need to be aware of its anomalies, and I'll be very cautious about saving results as an S-parameter file for sure. I see that even an attenuator that's detuned enough that it has only 60dB return loss still gives 90 degree phase shifts! But if I add even 1fF of capacitance across the 71.151 ohm resistor (in the 10dB attenuator, as designed by the attenuator tool), the saved S-parameter file is OK. S12 and S21 are zero degrees (till you get out to really high frequencies). S11 and S22 still show -90 degrees phase, but with a capacitive load (every so slightly capacitive), that seems like the right answer. By the way, a trick that seems to work to get very accurate component values is to put parts in parallel or series. For example, to simulate a 57.6325nH inductance (ridiculous though it is in practice), invoke a 57.00nH inductance in series with a 632.5pH. Unfortunately, that doesn't work for sub-ohm resistances; milliohms are converted to ohms, and only two decimal places. Double-checking to see that the 57.6325nH inductance simulates correctly, I resonated it with 1nF. I expect resonance at 20.964591MHz, and RFSim with 401 points displayed from 20.965 to 20.967, shows resonance at 20.96460. Changing the simulated inductance by 0.1pH changes the resonance as expected. Cheers, Tom Ted wrote: On Wed, 13 Sep 2006 20:26:41 -0400, Ted wrote: On 9 Sep 2006 11:23:33 -0700, "Professor" wrote: K7ITM wrote: Hi all, Every once in a while I come across an anomaly in RFSim99. I'd be happy if someone could tell me that I'm doing something wrong in the following: Fire up the simulator. Tools--Design--Attenuator. Pi Section, 50R input impedance, 50R output impedance, 10dB attenuation. Simulate. !Current schematic will be overwritten; OK. Now note that the graphics window displays S11 as 2E+31dB. In the circuit window, edit the values of the two 96.25 ohm resistors; they have apparently been simulated to more decimal places than shown. Just manually enter 96.25 ohms for each. Now the magnitude of S11 and S22 show up as -102.69dB; that seems reasonable. The phases for S11, S12, S21 and S22 are all zero. Now, from the graphics window, ---snip--- Aarrgg! This is very disappointing! I usually check at least my values for LC circuits against the LADPAC programs In "Experimental Methods in RF Design", however. By the way, when I click on Tools/DesignAttenuator, the schematic that is in the pop-up window for default 10 dB shows the three resistor values as 71.151 and 96.248. When I go through the process you did, I get the same results. I saved the file, and closed and reopened the program. Then, I loaded the saved 10dBAtten.cct file, and of course got the same -102.69 dB for S11. Then, I Tuned the values back to 71.151 (which immediately displayed only 71.15), and tuned the other two to 96.248, (which displayed 96.25). When I simulated, S11 was now -107 dB! Regards, Ted KX4OM I just ran the simulation in the LADPAC programs. I first built the attenuator in Ladbuild, with the resistors rounded off to two decimal places, and saved the result. In GPLA, I got (after resetting the scale to -120 dB) S11= -107.3. Then I edited the values to the 3 decimal place values from the original attenuator pop-up window of RFSim99, and got S11=-121.67. The circuit review feature of GPLA still showed 2 decimal places, rounded off, but saving that version of the attenuator and opening the saved program in Ladbuild showed three decimal place values when clicking on the component. It's surprising how a couple of thousandths of an ohm can change the S parameters that much. Ted KX4OM |
RFSim99 anomalies
Tom,
One other odditiy I found after replying was that with the 10dB basic pad loaded in in RFSim99, giving the bogus S parameter, by clicking on the up or down arrow for tolerance in the simulation mode, it would give what appeared to be reasonable results. However, clicking on the button to return it to normal gave the bogus value again. I guess I was optimistically thinking that after the tolerance iterations, it would see the error of its ways and give a realist value. Obviously, the software is not "self-healing", ala the topic of the recent IEEE Spectrum article. What you say about extreme sensitivity is on the mark. I later experimented with GPLA and saw the same effect. I hadn't thought of it in the balanced bridge context. The main reason I haven't homebrewed one of Rick Campbell's IQ designs is the need for those tight parameters, which translates to relatively high cost in the parts required... :( Some of the inductors for one of them, the IQ DC receiver, I think, are no longer available when I checked Mouser and Digi-Key. I suppose AADE still has the stock, but building from scratch is obviously cheaper if you can get a direct source. One femtoFarad? Wow! I'll have to play around with this some more, and your idea of paralleling (or series-ing, I guess, as well), sounds interesting. It could be an indicator that the results are valid, I should hope. I really like RFSim99 as a first-cut design tool. Very simple and straightforward. It's very nice for L-networks. As with the AADE filter designer, I typically then run any design through at least one other tool for confirmation. I've had some problems on occasion with crystal filter design for certain configurations using XLAD, so I've done the design using another tool, such as the AADE design tool, and then entered the values in LadBuild, and continued to tweak in GPLA. We're very fortunate to have such programs at our disposal. 73, Ted KX4OM 14 Sep 2006 10:16:09 -0700, "K7ITM" wrote: Hey, thanks for verifying my results, Ted. It shouldn't be surprising that tiny variations in resistance cause large dB changes. Remember, you're essentially unbalancing a bridge, and looking at the unbalance over 100dB down from full "un-balance." That is, -100dB represents a one part in 10^5 unbalance. 40dB return loss is really quite good, and something you expect only in a pretty carefully controlled setup. This same sort of thing reaches out and grabs you if you try to make a direct conversion receiver with I and Q paths to balance out the unwanted sideband. You need balance of not worse than ten parts per million to get 100dB sideband suppression, and the components you are using probably have, at BEST, temperature coefficients that can make things worse than that with only a one degree C change. On the up side, it's exactly this sensitivity that is a strong plus for bridge measurements. Also, I'm not about to throw RFSim99 out of my toolkit. I just need to be aware of its anomalies, and I'll be very cautious about saving results as an S-parameter file for sure. I see that even an attenuator that's detuned enough that it has only 60dB return loss still gives 90 degree phase shifts! But if I add even 1fF of capacitance across the 71.151 ohm resistor (in the 10dB attenuator, as designed by the attenuator tool), the saved S-parameter file is OK. S12 and S21 are zero degrees (till you get out to really high frequencies). S11 and S22 still show -90 degrees phase, but with a capacitive load (every so slightly capacitive), that seems like the right answer. By the way, a trick that seems to work to get very accurate component values is to put parts in parallel or series. For example, to simulate a 57.6325nH inductance (ridiculous though it is in practice), invoke a 57.00nH inductance in series with a 632.5pH. Unfortunately, that doesn't work for sub-ohm resistances; milliohms are converted to ohms, and only two decimal places. Double-checking to see that the 57.6325nH inductance simulates correctly, I resonated it with 1nF. I expect resonance at 20.964591MHz, and RFSim with 401 points displayed from 20.965 to 20.967, shows resonance at 20.96460. Changing the simulated inductance by 0.1pH changes the resonance as expected. Cheers, Tom Ted wrote: On Wed, 13 Sep 2006 20:26:41 -0400, Ted wrote: On 9 Sep 2006 11:23:33 -0700, "Professor" wrote: K7ITM wrote: Hi all, Every once in a while I come across an anomaly in RFSim99. I'd be happy if someone could tell me that I'm doing something wrong in the following: Fire up the simulator. Tools--Design--Attenuator. Pi Section, 50R input impedance, 50R output impedance, 10dB attenuation. Simulate. !Current schematic will be overwritten; OK. Now note that the graphics window displays S11 as 2E+31dB. In the circuit window, edit the values of the two 96.25 ohm resistors; they have apparently been simulated to more decimal places than shown. Just manually enter 96.25 ohms for each. Now the magnitude of S11 and S22 show up as -102.69dB; that seems reasonable. The phases for S11, S12, S21 and S22 are all zero. Now, from the graphics window, ---snip--- Aarrgg! This is very disappointing! I usually check at least my values for LC circuits against the LADPAC programs In "Experimental Methods in RF Design", however. By the way, when I click on Tools/DesignAttenuator, the schematic that is in the pop-up window for default 10 dB shows the three resistor values as 71.151 and 96.248. When I go through the process you did, I get the same results. I saved the file, and closed and reopened the program. Then, I loaded the saved 10dBAtten.cct file, and of course got the same -102.69 dB for S11. Then, I Tuned the values back to 71.151 (which immediately displayed only 71.15), and tuned the other two to 96.248, (which displayed 96.25). When I simulated, S11 was now -107 dB! Regards, Ted KX4OM I just ran the simulation in the LADPAC programs. I first built the attenuator in Ladbuild, with the resistors rounded off to two decimal places, and saved the result. In GPLA, I got (after resetting the scale to -120 dB) S11= -107.3. Then I edited the values to the 3 decimal place values from the original attenuator pop-up window of RFSim99, and got S11=-121.67. The circuit review feature of GPLA still showed 2 decimal places, rounded off, but saving that version of the attenuator and opening the saved program in Ladbuild showed three decimal place values when clicking on the component. It's surprising how a couple of thousandths of an ohm can change the S parameters that much. Ted KX4OM |
All times are GMT +1. The time now is 10:31 PM. |
Powered by vBulletin® Copyright ©2000 - 2024, Jelsoft Enterprises Ltd.
RadioBanter.com