|
|
|
|
Paul Burridge wrote in message . ..
On 22 Mar 2004 22:01:47 -0800, (Tom Bruhns) wrote: Of course, maybe you don't need such a high Q, Paul. Qu of 30 is quite reasonable for small SMT RF inductors, at least the type I use. In the following list, "series" means in series from the gate output to the next gate input, in order, and "shunt" means shunt to ground at that point. Anyway, try this (build it or SPICE it or RFSim99 it...add resistors to any simulation to account for the Qu. I'd suggest 3 ohms series and 12k ohms shunt for each 1.8uH.) Thanks, Tom. I've simulated the filter and posted the plot result to abse. Looks promising. See if it meets your expectations.... Hmmm...I simulated it, and it looked fine to me. I don't do a.b.* groups. I haven't actually built one; more pressing things to do. But I expect it will work OK. It assumes no inductive coupling among the coils. That's usually not too hard to get low enough in practice, if you orient the coils properly. BTW, check out Coil-Q inductors for commercial RF coils in small shield cans with decent Q. 7mm square can, Qu around 100 for inductances and frequencies in the neighborhood we're talking about here. |
|
|
Paul,
It's a series resonant circuit. The inductor is obvious. The capacitor is actually split between the one to ground at the gate input and the one on the other side of the inductor. Since it's a series circuit, the inductor and first cap can be swapped, of course; that might make it more obvious. The gate's RF input impedance (including a resistive part) and the bias resistors are in parallel (AC-wise) with the cap to ground, and provide a certain amount of damping -- lowering of the Q -- or the reason loaded Q is lower than Qu of the inductor. Since the tuning cap is in series with that cap, only part of the resonance voltage appears across it. Remember, in a series-resonant circuit, the voltage across the inductor or capacitor is much higher than the exciting voltage--that's where the voltage step-up comes from that's needed in this case. Splitting the net capacitance up this way lets you get a reasonably high loaded Q (to reject the other harmonics better) and control the output voltage, and provide the proper load at to the driving gate...as I recall, my design goal was about 100 ohms at the fifth harmonic, and a much higher impedance at the fundamental and the other harmonics. That way, the driving gate doesn't dissipate much power trying to drive those other harmonics into a heavy load, and only has to deliver significant power at the fifth harmonic. -- You could just use a cap from the gate input to ground, and an inductor to the driving gate's output (and then do away with the bias resistors too...), but then you can't so easily control the loaded Q. Yes, you should generally consider more than just the inductance of a coil. Q and self-resonant frequency are both generally important. At high frequencies, coils are the least ideal of our linear passives: Rs, Ls and Cs. Usually you can get by with ignoring the non-idealities of film and composition resistors and the types of caps usually used at RF, but inductors are a different story. It's good preparation for working in the GHz range, where pretty much all components have non-ideal performance. Cheers, Tom Paul Burridge wrote in message . .. On 23 Mar 2004 12:45:15 -0800, (Tom Bruhns) wrote: Hmmm...I simulated it, and it looked fine to me. I don't do a.b.* groups. I haven't actually built one; more pressing things to do. But I expect it will work OK. It assumes no inductive coupling among the coils. That's usually not too hard to get low enough in practice, if you orient the coils properly. BTW, check out Coil-Q inductors for commercial RF coils in small shield cans with decent Q. 7mm square can, Qu around 100 for inductances and frequencies in the neighborhood we're talking about here. Hey, give me a chance to keep up with your suggestions, will you? ;-) I've only just today taken delivery of some powedered iron toroids you tipped me off on (T37-10 for the time being) and am experimenting with those at present. Moreover, I replaced the factory, miniature, resistor-like inductors in my original 5X multiplier with hand-wound, air-cored ones for better Q, and guess what? *Huge* difference! Couldn't get a fifth *at all* before, as you may recall, but changing the inductors for the higher Q construction really brought it on big time! So much so I was convinced I'd made some fundamental mismeasure with the test equipment settings. Eventually it dawned that there was no error. The whole problem had been down to choice of coils - same values alright, but very different Qs. I've you to thank most sincerely for that revelation! Though the transistor-based multiplier and buffer/amp now works great, I'll still stick with your series filter solution as it saves on transistors and other components too. Just one last query, Tom: in your design between the ouput of the first inverter and the input to the next, you have, in series, a 0-10pF variable cap and a 20uH coil. Then you have your DC bias to the 2nd gate input and a 15pF cap shunted to ground at the same point. What was the purpose of that 15pF cap? Was it to provide an AC ground, bypassing the lower resistor (the one from input to gnd) or was there some loading function involved with it as well? Or was it intended to allow some independent control over the signal voltage level to the 2nd inverter input? Thanks again, Paul |
Paul,
It's a series resonant circuit. The inductor is obvious. The capacitor is actually split between the one to ground at the gate input and the one on the other side of the inductor. Since it's a series circuit, the inductor and first cap can be swapped, of course; that might make it more obvious. The gate's RF input impedance (including a resistive part) and the bias resistors are in parallel (AC-wise) with the cap to ground, and provide a certain amount of damping -- lowering of the Q -- or the reason loaded Q is lower than Qu of the inductor. Since the tuning cap is in series with that cap, only part of the resonance voltage appears across it. Remember, in a series-resonant circuit, the voltage across the inductor or capacitor is much higher than the exciting voltage--that's where the voltage step-up comes from that's needed in this case. Splitting the net capacitance up this way lets you get a reasonably high loaded Q (to reject the other harmonics better) and control the output voltage, and provide the proper load at to the driving gate...as I recall, my design goal was about 100 ohms at the fifth harmonic, and a much higher impedance at the fundamental and the other harmonics. That way, the driving gate doesn't dissipate much power trying to drive those other harmonics into a heavy load, and only has to deliver significant power at the fifth harmonic. -- You could just use a cap from the gate input to ground, and an inductor to the driving gate's output (and then do away with the bias resistors too...), but then you can't so easily control the loaded Q. Yes, you should generally consider more than just the inductance of a coil. Q and self-resonant frequency are both generally important. At high frequencies, coils are the least ideal of our linear passives: Rs, Ls and Cs. Usually you can get by with ignoring the non-idealities of film and composition resistors and the types of caps usually used at RF, but inductors are a different story. It's good preparation for working in the GHz range, where pretty much all components have non-ideal performance. Cheers, Tom Paul Burridge wrote in message . .. On 23 Mar 2004 12:45:15 -0800, (Tom Bruhns) wrote: Hmmm...I simulated it, and it looked fine to me. I don't do a.b.* groups. I haven't actually built one; more pressing things to do. But I expect it will work OK. It assumes no inductive coupling among the coils. That's usually not too hard to get low enough in practice, if you orient the coils properly. BTW, check out Coil-Q inductors for commercial RF coils in small shield cans with decent Q. 7mm square can, Qu around 100 for inductances and frequencies in the neighborhood we're talking about here. Hey, give me a chance to keep up with your suggestions, will you? ;-) I've only just today taken delivery of some powedered iron toroids you tipped me off on (T37-10 for the time being) and am experimenting with those at present. Moreover, I replaced the factory, miniature, resistor-like inductors in my original 5X multiplier with hand-wound, air-cored ones for better Q, and guess what? *Huge* difference! Couldn't get a fifth *at all* before, as you may recall, but changing the inductors for the higher Q construction really brought it on big time! So much so I was convinced I'd made some fundamental mismeasure with the test equipment settings. Eventually it dawned that there was no error. The whole problem had been down to choice of coils - same values alright, but very different Qs. I've you to thank most sincerely for that revelation! Though the transistor-based multiplier and buffer/amp now works great, I'll still stick with your series filter solution as it saves on transistors and other components too. Just one last query, Tom: in your design between the ouput of the first inverter and the input to the next, you have, in series, a 0-10pF variable cap and a 20uH coil. Then you have your DC bias to the 2nd gate input and a 15pF cap shunted to ground at the same point. What was the purpose of that 15pF cap? Was it to provide an AC ground, bypassing the lower resistor (the one from input to gnd) or was there some loading function involved with it as well? Or was it intended to allow some independent control over the signal voltage level to the 2nd inverter input? Thanks again, Paul |
|
| All times are GMT +1. The time now is 03:22 PM. |
Powered by vBulletin® Copyright ©2000 - 2025, Jelsoft Enterprises Ltd.
RadioBanter.com