M
Michele Ancis
- Jan 1, 1970
- 0
Hi all,
I have a matching problem, maybe somebody here can give me a hint. I have a
low noise Hetero Junction FET (NEC3210S01), I would like to match it for
optimum return loss, not for optimum noise. I've therefore calculated the
input and output loads to be presented to my transistor for simultaneous
conjugate match. The transistor must be matched from 9 to 10 GHz, so we're
talking about a 10% bandwidth, not narrow, not broad.
Moreover, the matching networks should be as simple as possible: I would
need the networks to introduce as little phase shift across the band as
possible. That is, I can't afford having multiple matching stages, as it
would for sure broaden my matching but also introduce an unwanted phase
shift. Of course we're talking about distributed networks, microstrip on a
Rogers 4003 0.5mm subsrate. What I have, from the optimization, is that I
can't get any better than -10 dB input return loss at the edges of the
band, while the output matching is OK. I would be fine with -15dB at the
input as well...;-)
While investigating the behaviour of my "optimum" input gamma, that is how
my generator should look like, I found an interesting thing about which I'd
like to hear from someone more experienced. "Normal" impedances follow a
clockwise path in the Smith chart, when going from low to higher freq. In
other words, the phase slope of "normal" impedances is negative. The only
way to produce a positive phase slope is being in the neighborhood of a
zero...However, since the conjugate input gamma is, roughly, conj(S11), one
has that the optimum impedance he should synthesize is something going the
other way...counter-clockwise with frequency. This is because the optimum
gamma is - roughly - the conjugate of the input gamma, which being "normal"
goes clockwise...
I would be interested to hear from anybody who has "observed" this thing
and maybe thought about it...
Finally, being constrained not to use multiple sections, a possible
approach would be to use some feedback and maybe loose a bit in gain, which
I could afford, for an "easier" input matching...Pity I haven't the
foggiest idea how to properly feed back my transistor, given that I can't
use lumped elements at 10 GHz (maybe resistors, but how? I don't want to
destroy my NF)...Again, any insight or idea would be greatly appreciated...
Thank you in advance,
Michele
I have a matching problem, maybe somebody here can give me a hint. I have a
low noise Hetero Junction FET (NEC3210S01), I would like to match it for
optimum return loss, not for optimum noise. I've therefore calculated the
input and output loads to be presented to my transistor for simultaneous
conjugate match. The transistor must be matched from 9 to 10 GHz, so we're
talking about a 10% bandwidth, not narrow, not broad.
Moreover, the matching networks should be as simple as possible: I would
need the networks to introduce as little phase shift across the band as
possible. That is, I can't afford having multiple matching stages, as it
would for sure broaden my matching but also introduce an unwanted phase
shift. Of course we're talking about distributed networks, microstrip on a
Rogers 4003 0.5mm subsrate. What I have, from the optimization, is that I
can't get any better than -10 dB input return loss at the edges of the
band, while the output matching is OK. I would be fine with -15dB at the
input as well...;-)
While investigating the behaviour of my "optimum" input gamma, that is how
my generator should look like, I found an interesting thing about which I'd
like to hear from someone more experienced. "Normal" impedances follow a
clockwise path in the Smith chart, when going from low to higher freq. In
other words, the phase slope of "normal" impedances is negative. The only
way to produce a positive phase slope is being in the neighborhood of a
zero...However, since the conjugate input gamma is, roughly, conj(S11), one
has that the optimum impedance he should synthesize is something going the
other way...counter-clockwise with frequency. This is because the optimum
gamma is - roughly - the conjugate of the input gamma, which being "normal"
goes clockwise...
I would be interested to hear from anybody who has "observed" this thing
and maybe thought about it...
Finally, being constrained not to use multiple sections, a possible
approach would be to use some feedback and maybe loose a bit in gain, which
I could afford, for an "easier" input matching...Pity I haven't the
foggiest idea how to properly feed back my transistor, given that I can't
use lumped elements at 10 GHz (maybe resistors, but how? I don't want to
destroy my NF)...Again, any insight or idea would be greatly appreciated...
Thank you in advance,
Michele