I'd like to drive a ~400pF load ("electrode" below) at 1-10MHz,
100Vpkpk. I had a previous posting about this, and I got some excellent advice about doing this with a RF amplifier (50 Ohm out, which I have) + transformer (for voltage step-up) + resonator (for Z matching to 50 Ohm):
. ____________ . G n1:n2 ,-----+------------+---+--)___________}---| . __|\\__ T | | | | coax electrode (400pF) . | |/ | || # # _|_ | 1pF __ . Vac # || # # L C --- '-||--)__ coax carries RF . |______| || # # | voltage-monitor . gapless | | Cdc | sig to cap. divider . pot core '--+--+---||--(#)--+-- gnd . | \\ \\ . | gapped current transformer . dc bias --\\/\\/--' pot core or small sense resistor
I'd like to know if it's possible to just drive this 400pF load with an op-amp or common-emmitter amp. instead (so that I don't have to impedence match). For example, I've seen the APEX amps, but those can't quite reach 10MHz (good solution for 1 MHz though). I'd like to be able to do the following:
__|\\____ electrode (400pF to ground) | |/ Vac | gnd
and be able to vary Vac from 1-10MHz with the amplifier putting out
100Vpp across the electrode. No coax necessary here. Am I dreaming when I'm basically asking for a 10MHz 100Vpp op-amp?
Yes, I recognize my ASCII-circuit drawing. You're welcome.
No coax? How are you going to transport the 100V 10MHz signal? Also, how do you arrive at 400pF, that's a lot for an electrode, which would make one think some coax was involved, at 30pF/foot.
OK, the slew rate of a 100V sine-wave signal would be S = dV/dt = 2pi f V, where V is the peak voltage of 50V, so S = 3200V/us. The current required by your capacitor would be i = C dV/dt = C S = 1.25A. For this sine wave you'd need a wideband amplifier that could deliver 1.25A peak and 50V peak, with a slew rate exceeding 3200V/us. I know such a beast, ah, amplifier, could be built, but I've never seen one.
If your 1 to 10MHz signal is not a sine wave, for example if it's a SWIFT waveform with components extending up to 10MHz, you need to know the maximum amplitude of the near-10MHz components. Then you can calculate a new lower Imax and slower required slew rate. Perhaps the amplifier you really need may be a bit more practical.
each pair of jumpered electrodes in the picture is ~25pF to ground, and there are 16, which gives a total of 400pF. The 4 electrodes in this cross section have the same DC voltage (Vdc(i)), but each set of 4 will have a different DC voltage. All pairs E1+E4 will have 1MHz, 100Vpp. For running at 1MHz, I was thinking
Where A1 is an APEX PA98 op-amp, C1>>25pF (DC blocking cap so that DC doesn't get on RF supply), R1~1MOhm or higher, and Vdc(i) is one of your DC supply circuits (they work great!). I was going to build this all on a PCB board, which will be mounted next to our vacuum system. The RF+DC lines will come out on some multi-conductor amphenol mil-spec connectors that screw right onto our vacuum system feedthroughs (very short distance, small parallel capacitance compared to the electrodes, so negligible). Therefore, reflections shouldn't be a problem here. Line to line capacitive pickup shouldn't be an issue, since everything is at the same frequency.
Now here's the trick -- we wanted to be able to run this system anywhere from 1-10MHz with not too much tuning. I didn't think that I could get this kind of tuning range from the pot core set up that you suggested earlier.
Yes, I agree that that sounds crazy. The APEX that you suggested should work fine at 1MHz for us though.
I assume a SWIFT waveform just has harmonics at different frequencies? I know groups who have trapped atoms (ions) in these traps with square waves even, though they say the trap stability goes down. Perhaps the best way to have tunability from 1-10MHz is still to go with your transformer+resonator? If so, could you suggest a good place for some pot cores? Of course, the dream was just to be able to dial in a frequency and have it run there.
Hi Joerg, it's continuous. So, we'll be supplying a sine wave at say 1 MHz, and then at some point, would like to switch over to say 5 MHz, without any sort of massing tuning or rewinding of transformers.
If you need something more broadband, perhaps you could use something like a "T-coil" to couple the amplifier output to the capacitive load. Done right, this should get you about a 2.7x reduction in drive current. This was done, for example, in oscilloscope vertical amplifier outputs (driving the capacitive deflection plates). The trade-off is that the Apex amps want to see a capacitive load, and the T-coil inputs look resistive.
Yes, even with T-coils this would be monstrous, and really hard to achieve. Would your arrangement be any more feasible if you split the amplifier into pieces? Using separate amplifiers for each electrode set? This might reduce current levels to where real physical transistors might exist that could support these frequencies, currents, and voltages. As it stands it looks pretty tough...
Even if you used sinusoids, switching resonanting inductors wouldn't be easy, you would have to use a bunch of relays and the whole distributed nature of the system would "add challenges".
Have you looked at crt cathode drivers ? they have the voltage and speed, not sure what their current is offhand,
400pf at 10mhz is 40 ohms; 100vpk pk = >1amp pk you can get 3 in one package, maybe have one per electrode ? otherwise ive no idea how well they would work in parallel tho,
If you want that kind of luxury you will have no choice but to build a linear amplifier with a whole lot of power. Look at video transistors that drive the cathodes or gates of CRTs. Those are cheap. You'll need lots of them. Driving 400pF up to 100Vpp at 10MHz is like a tractor pulling contest. Old trucker wisdom: There is nothing that can replace lots of cubic inches, except more cubic inches.
How does the trap and the vacuum system go together?
I'm sorry Jesse, the APEX amps don't come close to handling your task.
I'd like to know more detail about those connectors and feedthroughs.
Yes, but you still want to shield those lines, transmitter/antenna city, you know, otherwise. But I do have a suggestion.
That's correct.
At 1MHz, yes, but sorry, not to 10MHz. Have you examined the specs with an eye to the required compensation, and how that slows down the slew rate? And the limited GBW product, which kills operation up to 10MHz. It appears the APEX amp is at least 4x short of your goal, even if it was driving only one 25pF electrode.
But I do have a practical solution for you, which I'll describe in my next post. Right now I'm going to walk the dog with my wife.
OK, here's my practical solution. The first issue to consider is that you'll find more high-performance amplifier ICs available at low voltages (and high currents), because that's where you'll find the manufacturability / customer-demand sweet spot. So the first part of my solution is to use a step-up transformer. It's easy to make 0.5 to 20MHz wideband transformers. Furthermore, we can use its leakage-inductance property to help us at 10MHz.
Second, we go looking for candidate amplifiers, such as the Analog Devices' AD815. This is a dual amp, good for 50Vpp into 50 ohms in bridge mode (500mA) up to 20MHz (G=2). See figure 52 in the datasheet,
formatting link
If we assume 40Vpp, a transformer with a modest 1:2.5 turns ratio would work well. We'll use this to drive two electrode pairs, say 70pF total including a short bit of coax. This capacitance would look like 438pF on the primary, or j36 ohms at 10MHz, and require 550mA to drive at 10MHz. Aha, just at the limit for the AD815.
Each amplifier sees a 20Vpp swing, corresponding to a slew rate of 630V/us at 10MHz, well under the AD815's 900V/us capability.
All transformers have leakage inductance, which limits their high- frequency performance. But, if we design the leakage inductance to resonate with the capacitive load near 10MHz, with L = 3.6uH, we can reduce the loading effect of the capacitance, and thereby reduce the increasing current requirement on the amplifier near 10MHz. If the leakage inductance is less than 3.6uH, we'll add a small discrete inductor. We'll also have to add a resistor in parallel with the transformer to damp the resonance, 330 ohms will be about right. This looks like 53 ohms at the primary, and that will be the amplifier's load at lower frequencies. The current load may actually drop a bit near 10MHz, which is nice. This will also help stabilize the amplifier, which would be unhappy driving a purely-capacitive load.
You could build eight of these, each in its own little Pomona box, mounted near its electrodes, to drive the entire linear trap. Or, given the operation near the limit of the AD815, you could relax the scene and make 16 of them... You could parallel amplifiers, as shown in figure 50. Or you could make two sets in each box...
Fortunately the AD815 is cheap, under $10. I suggest the 15-Lead through-hole SIP package, still available from Rochester, because it's easier to heat sink (Pd = about 4 to 5 watts). The wideband transformers will be small, inexpensive and easy to wind yourself.
Divide and conquer, eight or sixteen, that's my suggestion.
None. That means for example that Vos changes by 4mV as the output swings to 6V, which is less than 0.1%, and of little concern here. If you were making a precision A/D amplifier, it might be an issue.
Dammit, my lip's bleeding :). I've now looked at this datasheet twice and still can't get my head around the spec's being quoted. How does figure 15, page 5, of the datasheet fit into the grand scheme of things?. regards john
I thought of something that interesting that might work if you wanted to avoid the output transformer, you could have a chain of amplifiers, each one bootstrapping the power supply of the following amp. Bit like a voltage multiplier or rather a charge pump. Each amp output swing would add onto the previous stage, so for your nice AD815 with 20v swing 5 stages would give you 100v swing. each stage would nead a gain of g, 2, 3/2, 4/3.
However I got confused as to how much current each stage would take, first i thought it would be the same, then I thought it would be twice as much, now I think each stage might be the same after all (aprox), any thoughts ?
Instability might be an issue with the power supplies not being referenced to ground if theres much stray capacitance to ground. phase adjustment might be needed to keep each stage op in phase with its power supply.
+15+---+ | | | c* (rfc chokes or could use comon mode transformer) | c | c | +-----+ | _|_ | | -,- | __|\\___|____|\\___ ... 3 more stages |/ _|_ |/ | -,- | | | | | +-----+ | c | c | c* | |
Hmmm... The maximum slew-rate goes down with increasing load? Hmm. Or, the maximum output-current capability goes down with increasing frequency? Hmm...
ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here.
All logos and trade names are the property of their respective owners.