Just an idea but if your still into the wideband idea something clever could probobly be done with a current feedback op-amp although slew rate is still not quite enough and as such one would have to get several of the highest slew rate (usualy very hard to use) and build an absolutly horibly complex feedback network, may not even be posible.
20 MHz HV Amp
I faced a similar problem when designing an amp to drive piezoelectric transducers with a 100 Vpp signal. The max frequency was only 2 MHz, so I opted for a wideband design. A 20 MHz wideband design would be considerably more difficult. Probably the best approach is class E driver and a narrowband output stage. The nice thing about a tuned design is that the load capacitance can be made part of an LC circuit that achieves a large voltage gain. In other words you can drive the tuned circuit with ~20Vpp and realize >200Vpp across the capacitive load. A 20Vpp 20 MHz squarewave is fairly easy to achieve using MOSFETs. You will definitely want to simulate the design before building it.
I faced a similar problem when designing an amp to drive piezoelectric transducers with a 100 Vpp signal. The max frequency was only 2 MHz, so I opted for a wideband design. A 20 MHz wideband design would be considerably more difficult. Probably the best approach is class E driver and a narrowband output stage. The nice thing about a tuned design is that the load capacitance can be made part of an LC circuit that achieves a large voltage gain. In other words you can drive the tuned circuit with ~20Vpp and realize >200Vpp across the capacitive load. A 20Vpp 20 MHz squarewave is fairly easy to achieve using MOSFETs. You will definitely want to simulate the design before building it.
Report On Probability A
Oh yes they do, so I would/wouldn't put up with too much nonsense from that quarter.
w
Goddam quantum physicists. Haven't they got any cats to torment?
Oh yes they do, so I would/wouldn't put up with too much nonsense from that quarter.
w
Goddam quantum physicists. Haven't they got any cats to torment?
Okay, okay, after sleeping several nights with the ARRL handbook under my pillow, I think I've gotten enough through osmosis to see what y'all are talking about.
Hows this look? It seems to work well in SPICE...
EDIT: The two 3-ohm resistors are just there so spice doesn't freak out with the ideal capacitor and inductor, the wire resistance and parasitic losses should take care of that in the real world I would assume
Any thoughts on construction techniques, aside from be very careful about lead positioning. I think I can do both inductors air core, no?
Hows this look? It seems to work well in SPICE...
EDIT: The two 3-ohm resistors are just there so spice doesn't freak out with the ideal capacitor and inductor, the wire resistance and parasitic losses should take care of that in the real world I would assume
Any thoughts on construction techniques, aside from be very careful about lead positioning. I think I can do both inductors air core, no?
Last edited:
OK, the series cap and paralled inductor match out to near infinity somewhere, and the tuned circuit resonates.
24 turns single-layer 20mm ID aircored closewound 1.5 mm copper wire will have an inductance of ~5.5uH and will be self supporting if you use hard-drawn antenna wire. You can stretch the coil to lower the inductance. It's quite a big coil, 2cm*4cm... about half the turns for the other.
I'd prefer to have some tuning caps in there for flexibility, coil tuning is a bit hit-and-miss, but you have to think about flashover. Don't touch the circuit when it's on, even if the power is low.
w
24 turns single-layer 20mm ID aircored closewound 1.5 mm copper wire will have an inductance of ~5.5uH and will be self supporting if you use hard-drawn antenna wire. You can stretch the coil to lower the inductance. It's quite a big coil, 2cm*4cm... about half the turns for the other.
I'd prefer to have some tuning caps in there for flexibility, coil tuning is a bit hit-and-miss, but you have to think about flashover. Don't touch the circuit when it's on, even if the power is low.
w
Actually the EOM cell will have losses so the resistors are pushing the model closer to reality.
Lots of RF voltage there, and RF burns HURT (and tend to take forever to heal, be careful).
Do get a VSWR meter, and learn to use it, and a dip meter will make initial tune up much easier. Is that matching network really right? Something about it looks a bit funky to me, it could be right, but it just looks a bit weird (L1 and L_match are functionally in parallel, maybe C_match should be the other side of L_match or something?).
One thing you need to check with the physics types: The lock in signal goes to both the laser and the RF amp? Does phase shift between whatever the laser is doing with the signal (mode locking?) and the EOM matter? You might need to cut some delay cables in here if the relative phasing is critical. Phase shift through the matching network will strongly depend on tuning.
Regards, Dan.
Lots of RF voltage there, and RF burns HURT (and tend to take forever to heal, be careful).
Do get a VSWR meter, and learn to use it, and a dip meter will make initial tune up much easier. Is that matching network really right? Something about it looks a bit funky to me, it could be right, but it just looks a bit weird (L1 and L_match are functionally in parallel, maybe C_match should be the other side of L_match or something?).
One thing you need to check with the physics types: The lock in signal goes to both the laser and the RF amp? Does phase shift between whatever the laser is doing with the signal (mode locking?) and the EOM matter? You might need to cut some delay cables in here if the relative phasing is critical. Phase shift through the matching network will strongly depend on tuning.
Regards, Dan.
The network does look weird - but the two coils are not in parallel. I had to write a MATLAB script to prove it to myself. The two 3 ohm resistors are critical, and they must be the same otherwise we end up with a reactance before the matching network (though this can in theory be taken care of with the trimmer). Assuming the resistors are *exactly* the same, and the tank RC is tuned to *exactly* 20MHz (I know, neither of those hold in the real world) the impedance of the whole tank is about 11k ohm, real.
Then the matching network just has to match 11k ohm with 50 ohm, the coil is selected to get 50 ohm real (in parallel with 11k ohm real) and the capacitor is selected to cancel out the reactance of the coil. I think I see the light! 😎
Here's the MATLAB
Code:
f = 20e6; % 20 mhz
C = 30e-12; % 30pF - EOM in parallal with a trimmer
L = 2.11e-6 %1/(30e-12 * (2*pi*f)^2) % 2.11 uH - air core
% Tank is EOM (with trimmer to 30pF) and 2.2uH coil in parallal
z_l_tank = ((1j)*2*pi*f*L) + 3; % 3 ohms in series
z_c_tank = ((-1j)/(2*pi*f*C)) + 3; % 3 ohms in series
z_tank = (z_l_tank * z_c_tank)/(z_l_tank + z_c_tank)
L_match = 6.1e-6 + 3e-8 % 6.1uH - air core
C_match = 10.4e-12 % 10.4pF - this is a trimmer
% 'Match' is 'Tank' in parallal with an inductor, all in series with the cap
z_l_match = ((1j)*2*pi*f*L_match) +3e-8
z_c_match = ((-1j)/(2*pi*f*C_match))
z_sub_match = (z_l_match * z_tank)/(z_l_match + z_tank)
z_match = z_sub_match + z_c_matchsome selected output:
z_tank = 11.7199e+003 +210.5505e+000i
z_sub_match = 50.2787e+000 +766.1106e+000i
z_match = 50.2787e+000 +942.5820e-003i
So the final overall shebang ends up with 50 ohms and some very small reactance (+0.942j ohm) that can be tuned out with the trimmer capacitor C_match.
The 3 ohm resistors are dropping a constant 800mA pk-pk (280mA RMS?) so they should be rated for at least 1/2W right?
I need to find my SWR meter, although I'm pretty sure it's 75 ohm and not 50. I guess they make 50 ohm ones... Off to eBay...
Yes! I built the circuit illustrated above:
😀 And the resonant tank was resonating, but was having nightmares with matching the impedance of the amp, even with a SWR meter/antenna tuner in the loop. We ended up getting 4-500v standing waves on the transmission line, but only 30v across the device! Very much backwards... But ended up being able to test the device by hooking up a "T" in the input line and a dummy "load" on the device, and using the SW to drive the EOM. Very messy signal though...
So I ditched the matching network (the larger inductor and adjustable capacitor) and re-built the resonant tank smaller and using off-the-shelf inductors and capacitors, enclosed in an RF housing. This is a very very simple circuit, just the EOM and an adjustable capacitor (~35pF) parallel with a small inductor (~1.5nH). That's it. No resistors, no matching anything, hooked directly to the amp via 50ohm cable.
With this version it worked perfectly! We got to 400v pk-pk (twice the design spec) at 50% input. Very little distortion in the sine wave, and what little there was can be tuned out with the SWR meter/antenna tuner. Now, the SWR is unmeasurably low, and we have full power forward.
The part that I still don't understand is that (theoretically) the impedance of this load should be VERY high (ideally infinite, but at least several hundred k ohm) and the amp is 50ohm out, yet there are almost matching issues. Perhaps this is a special amp, or perhaps I was over-thinking things from the beginning. Anyway, let this be a lesson (once again) in KISS.
Thank you all for the help!
Jacob
- Status
- Not open for further replies.