I have a floating stainless steel tube enclosed in an apparatus used for atomic physics experiments. A feedthrough + 10ft coax cable allow us to apply a voltage to this tube. Until now, we have connected it to ground.

Now I need to apply a 10us negative square voltage pulse to the tube. The amplitude must be adjustable between 10V-80V. The rise and fall time must be less than 300ns for 50V (so say 200V/us). Most importantly (!) the top of the pulse must be very very flat, with minimal overshoot and ripple.

My problem is this: applying an 80V square pulse to the tube whose over/undershoot and ripple are <0.5V over the entire top of the pulse.

I've attached a quick drawing of my current setup. I need to throw this together quickly, so I'm using whatever I've got. The idea is to use an Agilent 33220a 20MHz Arbitrary Waveform Generator (which I can control with a computer) to produce a 0-5V pulse, and then amplify it to 0-80V.

For the amplification step, I'm using an Apex PA78 power operational amplifier (datasheet). I've taken a circuit from their applications note.

I've attached some scope traces with the probe attached right after the output of the PA78. In one trace, I have disconnected the stainless steel tube, while in the other two it is connected. (Note that I've zoomed in on the top of the pulse in two of the traces!) The circuit itself is wired up on a breadboard. 0.1uF caps are used for bypassing power to the op amps.

Question: How can I get a very flat top to my pulse? Are there any tricks I can try? Could this be from reflections due to the way things are terminated (or unterminated..). This sort of thing is not my specialty, so please let me know if I'm doing something dumb.

Thanks in advance!

 

While SOT23-5 is a hard package to mess with, the performance of the LM7121 is closer to what you might need. That LM318 is a very ancient design. Have you checked the drive of the LM318 into the PA78?

 

Question: How can I get a very flat top to my pulse? Are there any tricks I can try? Could this be from reflections due to the way things are terminated (or unterminated..). This sort of thing is not my specialty, so please let me know if I'm doing something dumb.

No, not doing anything dumb, you have a beautiful wave to start out with but you have some <snip> strict requirements there!

You show the wave at the tube end, does it look any better right at the amp output? If THAT point isn't good then it ain't getting any better at the other end of the line.

I am not a transmission line guy, but there are some radio guys here that could do it in their sleep. I believe you need to treat this as a transmission line and get it all matched up to have any hope here. Do you have any access to the tube end of the 10 feet of coax to add a termination?

(I'm gonna sit back & watch now as that's all I got here.)

 

the problem is- any conductor will have capacitance AND inductance. Those will always set up conditions which create overshoot and ringing. They will also limit the rise time. Coax itself is not the best choice for this, since its capacitance per foot is much higher than a single conductor without shield.

The flattest response you could get (ideal situation) would be from a battery with the DC level you require. Using an amplifier will introduce even more capacitance and inductance. Those are the two things you must reduce to a minimum.

Try ditching the coax and using WIDELY seperated single conductors as your current delivery method and see if that has any benefit for your needs.

 

Wow- thank you for the (quick!) responses. I'll work on this tomorrow, and follow-up on what I find.

While SOT23-5 is a hard package to mess with, the performance of the LM7121 is closer to what you might need. That LM318 is a very ancient design. Have you checked the drive of the LM318 into the PA78?

I checked the output of the LM318 when I first wired it up, and I "thought" it looked good enough at the time, but now everything is suspect. I'll have a chance to probe things tomorrow afternoon.

We have a stash of op amps, but they're all pretty old and slow. (In my first search through what was lying around, the LM318 seemed to have the fastest slew rate, at around 70V/us.) I'll look again to see if there is anything of higher performance. If it came to it, I might be able to arrange it to have something ordered on Monday, and delivered Tuesday.

No, not doing anything dumb, you have a beautiful wave to start out with but you have some <snip> strict requirements there!

You show the wave at the tube end, does it look any better right at the amp output? If THAT point isn't good then it ain't getting any better at the other end of the line.

I am not a transmission line guy, but there are some radio guys here that could do it in their sleep. I believe you need to treat this as a transmission line and get it all matched up to have any hope here. Do you have any access to the tube end of the 10 feet of coax to add a termination?

(I'm gonna sit back & watch now as that's all I got here.)

You (and beenthere) are right, I should check the point betweent he LM318 and the PA78. I'll have a chance tomorrow to fiddle with it.

the problem is- any conductor will have capacitance AND inductance. Those will always set up conditions which create overshoot and ringing. They will also limit the rise time. Coax itself is not the best choice for this, since its capacitance per foot is much higher than a single conductor without shield.

The flattest response you could get (ideal situation) would be from a battery with the DC level you require. Using an amplifier will introduce even more capacitance and inductance. Those are the two things you must reduce to a minimum.

Try ditching the coax and using WIDELY seperated single conductors as your current delivery method and see if that has any benefit for your needs.

I should come clean on the coax - the cable from the tube to the circuit is not a common 50 ohm BNC coax cable (of which we have tons of in the lab), but actually 10 plastic coated wires clad in thin metal foil shielding inside thick plastic casing. In my mind, it's a similar design to our usual coax cables, but with more wires inside, so I've been treating them the same. BUT, I didn't think to ground the metal foil shielding.

The metal tube is about 1 inch diameter, 4 inches long, and inside a vacuum chamber. At the wall of the chamber, the above cable either ends in a connector (10 pin?), or it feeds straight through into the chamber to the tube, in which case getting access to the cable at the wall of the chamber would mean cutting the cable (which is not out of the question. I will closely investigate what happens at the wall of the chamber (the closest accessible point to the tube), and see if the situation can be improved.

We have an old dedicated 50V HP pulse generator, but it is not programmable, and it is absolutely necessary that I be able to continuously vary the height of the pulse from about 10V-50V (or beyond). Since I know how to do this with the Agilent 5V generator, amplification seems like the most realizable solution.

 

I agree with Kermit2 -- you're going to run into inductance and capacitance problems. One nice thing about using the AWG is that if you have slow risetimes, you might be able to boost them by changing the waveshape correspondingly. But the ringing stuff is going to require more knowledge and tuning from the radio guys.

Another thought -- if you can find an old HP214 pulse generator (A or B model), you might be able to use that to drive things directly, as IIRC they could put out around 200 volts peak. This might let you get things physically close to the vacuum feedthrough, reducing transmission line problems.

I had to drive some thin film magnetic stuff with switching currents about 30 years ago and I used a pulse generator into a commercial 1 kW RF amplifier. It worked fine (we were studying electromigration), but we didn't need the voltage levels you do -- and I didn't care about ringing, as the RMS current was the important parameter.

Oh, one final thought. If you're doing this in vacuum and have a decent amount of room in your chamber (ha! -- who ever had a vacuum chamber of adequate size ), remember that you can build vacuum tubes easily in them (the canonical example is the bare B-A ion gauge). This might let you use some electronics you might not otherwise use. I never had to use this technique in my thin film career, but it was always in the back of my mind as a possibility (I was working at Varian at the time -- and the Eimac folks would have loved to help).

 

UPDATE: We're going to shorten the cable to the vacuum chamber, and use single wires.

As for the circuit, I'm trying something totally new.

I'd like to use just a simple ZVN210 MOSFET. I'll trigger the MOSFET with a 0-5V pulse generator. In the experiment, I'll set the drain to +60V, and I'll vary the source between +0-55V.

I've tested this, but on a smaller scale (0-20V). I've tried different circuit configurations. I've attached a bunch of scope traces (w/ circuit layouts for each), in case anyone's interested.

Getting rid of that "tilt" in pulse for all voltage levels seems to be the toughest part. If anyone has any advice on that, I'd appreciate it. And thanks again to all the help.

 

...(sorry for the double post) Here's the rest of them.

UPDATE: Some thoughts about these:

- c08 has ringing, while c06 does not. I think this is because in the case of c06, there is a path to ground from the "gate-side" of the capacitor, while in c08 it is only connected to the "+" output of a power supply set to 0V.

- the "new" cap to ground in c11 is 0.1uF

- from c12, it looks like the resistor between the gate and source needs to be small?

- c13 supports what I said about a path to ground for that "gate-side" cap

- c19 through c24 try larger/smaller caps to ground (1.0uF and 0.05uF)... and it looks like the larger one is better, but there is still tilt.

 

I wonder if you have thoroughly de-coupled the power supply feed to the drain load resistance. Any variation of the supply voltage here may give a tilt, so a fairly substantial capacitor to ground (say a 100nF, or maybe bigger) would be in order.

A quick look at the supply voltage with a scope should give you an idea if it is bouncing up and down.

 

I wonder if you have thoroughly de-coupled the power supply feed to the drain load resistance. Any variation of the supply voltage here may give a tilt, so a fairly substantial capacitor to ground (say a 100nF, or maybe bigger) would be in order.

A quick look at the supply voltage with a scope should give you an idea if it is bouncing up and down.

Good idea. I'll try that!

UPDATE: Unfortunately, the "tilt" problem persists with either a 0.1uF or 1.0uF between the drain power supply and ground.

 

Cable television companies use "tilt" for the station frequencies; as the higher the frequency, the greater the loss (attenuation) in the transmission line - therefore, the higher frequencies start off at a much higher signal level than the lower frequencies. The "tilt" is again applied each time the signals go through an amplifier station. By the time it gets to the subscribers' television, the signal strength is roughly equal all up and down the band.

You're seeing the waveform being rounded off quite a bit.
In order to get an "ideal" square wave, you need infinite bandwidth, as an ideal square wave is the sum of ALL of the odd harmonics of the fundamental frequency. The lower your bandwidth, the worse the square wave will appear.

Decoupling, impedance matching, etc. are all pretty critical towards getting that output looking square.

Coaxes that are commonly available are either 50 Ohms impedance (best for general purpose), or 75 Ohms (best for television reception). The impedance needs to be matched (terminated) at both ends, or you'll have insertion losses (attenuation) and return losses (reflected energy). RG-58u (50 Ohms) is commonly used for amateur radio communications (like CB radios). Longer runs use RG-8u, which is much larger in diameter and much higher in cost, but has far less loss than RG-58u.

A straight piece of wire has inductance, but little capacitance if nothing else is near. That could cause some problems, too.

Since you are driving this thing with an arbitrary waveform generator, I suggest that you try programming it with a "tilt" to compensate for what you're seeing; ie: since the rise of the waveform is slower than desired, start off by overshooting it, and then ramp down. Someonesdad already suggested similar to this in reply #6. If you can solve the problem with just a bit of re-programming, you will likely save time and effort.

 

A coax line that's not terminated at both ends of the cable (into its "Characteristic Impedance") is no longer a transmission line. It becomes an equivalent capacitance, inductance and resistance and sometimes an antenna! A perfectly matched line to source and line to load should ideally be seen by the source as purely resistive load. This isn't possible in the real world though.

Of all aspects of electronics, transmission line theory lives in a world of it's own. The ITT Bible devotes quite a bit of paper and ink to it. I would imagine a book as thick as the Digikey catalog could be written and probably has.

FYI, Time "Domain Reflectometry" is one method of analyzing a line and its load.

 

You should also make sure that the oscilloscope, and any probe used with it is good. For instance, an incorrectly adjusted passive probe can easily give an apparent tilt, even if there is none in the signal being measured.