Joss Institute Projects:

A Charge-Transfer TEA Nitrogen Laser
That Is Driven by a Small Marx Bank

[Started on May 15, 2011.]


This page is intended partly as a followon to the “Easy TEA Laser” page, and partly as a standalone project. It describes a relatively powerful room-pressure nitrogen laser.

Before we get any further along, we need some safety information and a disclaimer.

!!   WARNING   !!

If you build this project you do so on your own responsibility, and at your own risk.

These lasers use high voltages, and capacitors that can store lethal amounts of energy. They put out invisible ultraviolet light that can damage your eyes and skin. It is extremely important to take adequate safety precautions and use appropriate safety equipment with any laser; and it is crucially important with lasers that involve high voltages and/or produce invisible beams!

In addition, this particular laser uses two open spark gaps, which will damage your hearing if you do not use adequate ear protection. I strongly suggest that you acquire and use at least a pair of sound-protection earmuffs of the type used by shooters at rifle and pistol ranges; they look about like this:

Figure 1: Hearing protection

(These cost me $35, and they are definitely worth it.)

Earplugs can also help, but by themselves are probably not sufficient unless they decrease the volume by about 30 db; I suspect that only special ones that are made to fit your own ears are good enough.

If you are not using enough hearing protection, you will probably get a nasty headache if you run the laser for a while. Take that as a warning, and get better protection! You can make a new spark gap, and you can make a new laser; but you cannot make new eyes, ears, or fingers.

Requisite Preliminaries:

The Marx Bank and the Design of the Laser

This laser is based on the Charge-Transfer circuit topology, in which there is ordinarily a single main storage capacitor (the “dumper”). The dumper capacitor is charged by the power supply and then connected, by a fast switch, to a smaller and faster secondary capacitor (the “peaker”). The peaker capacitor is charged very rapidly, and it then drives the laser channel, along with whatever current is still flowing from the dumper. (Because the dumper is larger there is usually quite a bit of current from it, which contributes a significant amount to the total. Because it is slower, however, its influence is somewhat limited.) The CT topology is typically less efficient than the more familiar “Voltage Doubler” circuit, but is often capable of delivering higher output energy.

I have chosen to modify the CT circuit for this laser to use a minimal Marx generator as its dumper cap. My hope is that this will make the discharge more energetic by providing higher charging voltage for the peaker cap than the power supply can provide.

[[The Marx Bank (or Marx generator) was invented by Erwin Otto Marx, in 1924. It is a fine way to increase the voltage of a pulsed system. The fundamental principle is that capacitors are charged up in parallel, and then connected by fast switches so that they discharge in series. A well-designed bank will get fairly close to an integer multiple of the initial charge voltage, particularly if it has only a few stages.]]

The design I’m using here is a stack of 4 capacitor plates. The bottom one is the baseplane of the entire system. (That is, it serves as the basis of the Marx bank, the peaker cap, and the laser channel. It is connected to the positive output of the power supply though an inductor. (More about polarity, below.)

Part of the baseplane is covered by a sheet of styrene plastic, 10 mils thick, which serves as the dielectric of the first part of the dumper. The next plate is on top of this styrene sheet, and is connected to the negative output of the power supply through a second inductor. It is also connected, by a piece of brass shim stock, to the lower side of the main switch, which is a triggered spark gap.

Then there is a dielectric layer of double or treble thickness, with the third plate on top of it. The third plate is lined up to be directly above the second plate. The extra-thick set of dielectrics allows the second and third plates to function as a small start capacitor for the spark gap, though I have added a very small (25 pf) start capacitor to it, on a just-in-case basis. [Note: this set ended up being 35 mils thick, and could be even thicker.] Like the baseplane, the third plate is connected to the positive output of the power supply, but through a separate inductor. It is also connected to the upper side of the main switch, via a piece of brass shim stock.

After that is a final sheet of dielectric and the fourth capacitor plate, which is connected to the negative output of the power supply through yet another inductor. The fourth plate serves as the output of the Marx bank, so it is also connected, through a piece of brass shim stock, to the top electrode of a passive free-running spark gap, which switches into the peaker capacitor. (See the schematic diagram, below, for clarification.)

The inductors prevent the bank from short-circuiting when the main switch is fired. They also keep some of the EMP out of the power supply. I built them by stiffening the cardboard tubes from two paper towel rolls with cyanoacrylate adhesive, and winding wire with thick insulation around them. I made two inductors per roll, with 31 turns per inductor. Here they are, with the power supply:

Figure 2: Power supply and quad charging inductors

I got the wire at the hardware store; it has insulation that is 45 mils thick, and it is rated to handle 150 volts. I have no idea why it has such a low rating; I’ve been using a similar inductor on the “easy” TEA nitrogen laser for days now, at +/- 6 kV and more, with no problems.

Here is the schematic:

Figure 3: Circuit of the laser

Polarity note: at least with EG&G (now Perkin-Elmer) commercial trigatron gaps, it is suggested that for best operation the positive-going trigger pulse should go to the trigger electrode, and the surrounding (in EG&G terminology, “adjacent”) electrode should be connected to the positive side of the power supply. Because of the way the Marx stack is built, with the lower plate of the second capacitor sitting on top of the first capacitor, I was obliged to connect the positive side of the power supply to the baseplane, and the negative side to the top of the stack. I may, at some point, try reversing the polarity, especially if the laser doesn’t seem to want to work well with the circuit configured as it is in the diagram.

[Note, added later: I tried, and it didn’t seem to help, so I reverted to the original polarity.]

The Dumper[s]

The baseplane is a sheet of single-sided circuitboard, as with the previous lasers in this series, but brass shim stock would work just as well. I am using six 4" x 10" sheets of brass, in pairs, as the other plates of the main storage caps. Because I can use wide pieces of shim stock to connect the caps to the triggered and free-running spark gaps, I don’t have to worry about connecting the pairs of brass plates to each other for fast discharge. (This should be evident from the assembly photos, below.)

The dielectric will be 10-mil styrene, unless I can find 8-mil or 10-mil polycarbonate. Polycarbonate has good dielectric constant and excellent dielectric strength, but is hard to find with a plain glossy surface on both sides. It is available in 10-mil thickness with one side matte and the other side covered with adhesive, neither of which is desirable in this application, and that form tends to be expensive. (I may try it anyway, just to see whether it works.)

[Note, added later: I did, indeed use 10-mil styrene, but later I changed the peaker dielectric to 6.5-mil acetate, as detailed in the text below.]]

The Switch

It is, of course, possible to allow the main switch to free-run; but that makes it difficult to predict when the laser will fire. At this small scale, it is easy enough to build a triggered switch, and that is what I have done:


Figures 4 & 5: Preassembly of the main switch

The photos show a 5/16-18 carriage bolt, which I have drilled to accept a small piece of capillary tubing. The trigger electrode is a broken piece of jeweler’s saw blade. This parallels the design I used for the final versions of the Easy TEA Laser project, but I am using the head of the carriage bolt as the face of the electrode, rather than an acorn nut screwed onto the end of an ordinary bolt. It was more difficult to drill the hole in the steel bolt than it would have been in a brass bolt, but I only had to drill one hole. Also, I was able to start the hole more or less in the center of the electrode face, which is integral to the support column for the capillary. This also meant that I did not have to be as concerned about whether the hole runs down the axis of the bolt, as long as it doesn’t interfere with mounting. (In fact, it is only modestly off center at the back end.)

The small piece of steel that you can see in the photo on the right is a knockout from an electrical box; I found it in a parking lot, and sanded the paint off it. I am expecting to use it as the lower electrode of this gap. It is more or less just a strike plate, intended to protect the brass plate or shim stock beneath it from the sparks in the gap. If I can find another one, I will use it in the passive gap.

Note, added later: brass conducts significantly better than steel, and if I build another laser of this sort I will use brass parts for the spark gaps. Brass will wear faster, but I would expect enough of an improvement in performance to offset the minor annoyance of having to replace parts more often.

Here is the assembled gap:


Figures 6 & 7: The Installed Switch

(I have added a 25-pf start cap, on a “just-in-case” basis. It is currently a bit far from the gap, and I may move it closer at some point. Note, also, the insulating structures behind the gap; I had to add these to prevent flashovers from the top capacitor plate.)


Initially, I am triggering the switch with a TM-11 or TM-11A trigger unit; but I am working on a design for a DIY trigger generator, and will post it when I get it up and flying. (Triggering of this type of gap may be slightly tweaky, and an automotive spark coil, on its own, may not be very good for this purpose. For this reason, the trigger generator may have to involve a bit of finesse.)

The Passive Gap

The power supply for this laser is the same type that I used in the previous lasers, a small unit that I took from an old electronic air cleaner. It delivers about 12 kV into an open circuit. When I trigger the main switch, the Marx generator should erect fairly quickly (probably within about 20 nanoseconds), and at that point it will be providing at least 20 kV at its output. The passive gap must hold off the initial charge voltage, which it is exposed to whenever the main store is being charged; but it must not withstand the full output voltage of the Marx, or the laser will fail to fire. As I write this I have not yet constructed it, but I expect to make it fairly easy to adjust. At least initially I expect to tweak it until it holds off the charging voltage easily, but will only take a modest overvoltage to fire.

[Note, added later: It turns out that the main gap is the one that needs to be adjusted so that it barely holds off the charging voltage; the passive gap wants to be nearly as wide as possible, so that it doesn’t begin to conduct until there is a high voltage coming from the Dumper.]

I eventually put a 200-pf start cap across the passive gap; it seems to help the performance a little, but may be on the large side.

The Peaker Capacitor

Unless I change my mind, I will use another 4" x 10" brass plate for this. It should, ordinarily, have about 1/3 as much capacitance as the dumper cap; for this reason and because it may have relatively high voltages across it, I am likely to use a thicker sheet of dielectric for it. (I have 15-mil styrene, which will work.)

Note, added later: The hobby shop changed my mind for me by running out of brass plates of the thickness I wanted, and I ended up using a copper plate instead. Also, as I have already mentioned, I tried a thick dielectric and was not satisfied with the resulting performance, so after some thought I shifted to 6.5-mil acetate. The acetate is too thin to handle the amount of voltage involved for any length of time, but because it is being charged for only a few dozen nanoseconds at a time it has withstood hundreds of firing cycles. It performs quite well — when the laser is fully adjusted, its unfocused output will pump a cuvette of “DTC” in 70% isopropyl alcohol that is several inches away from the end of the channel. (See Figures 29 and 31, below.)

For what it’s worth, the peaker cap measures about 3.4 nf on my DMM.

The Build

It took me about two and a half days to build this laser, and then about two weeks to get it to stabilize. It kept puncturing a dielectric or sparking where I didn’t want it to, but I finally got it more or less tamed out. It then took me about two more weeks to get fairly good performance from it.

Here is a sequence of photos, taken during the initial build. (You will notice that the peaker dielectric is a piece of plastic sign. Later on, I changed that out and substituted a piece of 6.5-mil acetate sheet, which has higher dielectric constant. The peaker now measures about 3.4 nf on my little DMM.)





Figures 8 - 18: Initial assembly sequence

Here are overviews of the completed laser, taken at an intermediate stage:


Figures 19 & 20: Overview of the Completed Laser

(The photo on the left has the spark gap and dumper capacitor at its left, and the peaker and the laser channel at the right. The photo on the right was taken from the usual output end, on the peaker side of the chassis, almost kitty-corner across from the other photo. You can see the cylindrical lens that I use when I want to focus the output, and a cuvette of dye solution.)

[[As far as I know, btw, “kitty-corner” comes from “catercorner” or “catercornered”, where “cater” appears to derive from French “quatre”; it originally seems to have meant simply “four-cornered”.]]

A note about adjustment: This design has a large number of adjustable parameters, and it is seriously not recommended as a first laser. You can drive yourself bats trying to figure out what you could possibly have done that caused some outré behavior. Alternatively, you can do something you’ve done several times with uniform results, and have something entirely different happen for no obvious reason. I can suggest a few rules of thumb, but you must take them with several grains of salt.

  1. The electrodes of the main switch should, typically, be set as close together as possible without too much self-flashing, so that it has low inductance and relatively low effective resistance when it fires. (I tolerate an occasional self flash, perhaps once every 5 to 10 seconds. This helps prevent the voltage on the dumper cap from getting too high between shots.)

  2. The passive gap wants to be set relatively wide, so that charging of the peaker starts only after the output from the Marx bank is approaching its peak voltage.

  3. It seems to me that the preionizers (I have them on both sides of the channel) work best when they are perhaps 2 mm farther apart than the electrodes, and are essentially parallel to them. (See the “Easy TEA Laser” page for preionizer shaping.)

  4. So far, it seems to me that the electrodes work best when they are very smooth, and very cleanly rounded. This agrees with the experience of Jarrod Kinsey, but not with that of Alfonso Torres Rodríguez. (Alfon’s laser designs are quite advanced, and they operate at somewhat higher voltages, so it is not entirely surprising that he obtains different results in some areas.)

  5. I have found that it is instructive to note the behavior as I adjust the spacing between the electrodes, and then to repeat with different preionizer spacing.

  6. If you find that you get sparking at almost any electrode spacing, it is likely that you are not getting very high voltage on your peaker capacitor.

  7. On the other hand, if you get a nice clean discharge but no lasing, your electrodes are probably too far apart. (There is a parameter called E/p, which is discussed in various technical articles about nitrogen lasers that have appeared in the scientific literature. It relates the voltage to the channel width and the partial pressure of nitrogen in the channel. The reported optimum for E/p is something close to 100 volts per Torr centimeter, so a 2-mm-wide channel that is operating at 760 Torr takes a little over 15 kV for best performance. As the channel width increases, the optimal voltage increases with it. If you are attempting to operate a 2-mm channel and your power source cannot provide this much voltage, the operation of your laser will suffer.)


An End-Mirror

A low-pressure nitrogen laser is typically operated with a mirror at one end of its channel. This can provide a large enhancement in output power, and a beam that is more tightly defined. TEA nitrogen lasers, in contrast, are only occasionally operated this way. With a channel that is >30 cm long, it would seem pointless to add a mirror, but I had to know whether it would help, so I got one and tried it. It is 15 mm square, and I think the coating is UV-enhanced aluminum, so it probably reflects 85 or 86% at 337.1 nm.

Figure 21: The Mirror

I quickly discovered that little plastic strips are not very good for tweaking the alignment, but at least they kept my fingers away from the high voltage during initial testing. I ended up using only one of them, and taping it in place; eventually I will figure out a better solution.

Here are some photos, showing misalignment and roughly correct alignment. They are not all precisely to the same scale, but it is definitely the case that when I get the mirror aligned nicely the bright central region of the beam gets wider, while the overall beam appears to become slightly narrower. (Compare the central part of the main beam as it appears in the first and fourth photos.)



Figures 22-25: Aligning the Mirror

(In order:

  1. The return from the mirror is far too low; if it were any lower it would bounce off the dielectric and appear above the main pattern.

  2. The return is somewhat low [notice the shadows of the electrodes!].

  3. The return is slightly high.

  4. The return is positioned at least roughly correctly.)

In sum, although this laser probably does not quite reach the MegaWatt power level, it provides substantial performance and can be used for various purposes.

For example, I wanted to know how accurate the timebase on our old Tektronix 7104 oscilloscope is, and as I do not have ready access to NIST (they are some miles from here, and I’m sure they have better things to do than calibrate my antique 7B15 for me), I decided to fall back upon fundamental constants: in this case, the speed of light. The protocol here is to produce two light pulses with a separation in time that is known, and to compare that known timing with the timing displayed by the scope. Fortunately, the pulse from a room-pressure nitrogen laser is typically less than 1 nanosecond long FWHM [Full Width at Half of the Maximum amplitude], which makes it easy to use for this purpose.

To create two pulses with a delay between them, I ran one of the output beams of the laser (this was before I added the mirror) into a structure rather like a Michelson interferometer, but with grossly unequal pathlengths, which I built on top of the oscilloscope because the cable on the photodetector is short:

Figure 26: Scope calibration setup overview

(Sorry about the view angle!)

Please note the carefully chosen “close-in” reflector, a threaded connector nut:

Figure 27: Scope calibration setup detail

A real mirror would have swamped the detector, so I picked something that was a vaguely specular reflector but not a very good one, and lucked out — it works quite nicely. Note, also, the beamsplitter, which is a 1-inch sapphire window intended for a sewer camera. I chose sapphire because it has high refractive index, and thus reflects about 14% at normal incidence.

When I took the photos above, the pathlength from the beamsplitter to the far mirror was about 16 inches. I redid the measurement for the scope trace photo below with a pathlength of 38 cm. We have to double that, because the light goes off to the mirror at the far end and then has to return. The speed of light is about 299,792,458 meters per second (a wee bit less in air, not that it makes any difference for this measurement), so it took slightly more than 2.5 nanoseconds for the light to travel the 76cm two-way path.

Here’s what the scope gave me, after I got everything aligned and applied a bit of optical attenuation at the far-end mirror to get the peak heights to be fairly similar:

Figure 28: Scope calibration trace

I’ve marked the positions of the peaks; the scope indicates that they are just over 2.8 nsec apart, which means that it is about 10% off calibration when it is displaying 2 nsec per major division on the screen. Such is life; at least now I know what the error is. (There is an adjustment for this, but it is already at the end of its range.)

[Note, added much later: It is somewhat astonishing to me that this worked as well as it did. The EMP from the laser was actually interfering rather badly with the measurement, and I eventually had to put the pulse splitter and the oscilloscope in the next room in order to get a stable reading on the screen. I am hoping to show that version as one of the demos in a YouTube video about this design family, and I will try to put a link here if and when I get that video finished and posted.]

There are, of course, other uses for a nitrogen laser. One of the most common is as a pump source for small organic dye lasers. Here is a cuvette of dye solution, sitting about 8 inches off the end of the channel, being pumped by the unfocused beam of the laser:

Figure 29: Unfocused output, lasing DTC in 70% Isopropanol

(If you look really closely at the beam pattern on the front of the cuvette, you can tell that the return beam from the mirror is very slightly high.)

I also decided to see what would happen if I focused the beam more tightly. We happen to have a few polished metal mirrors, and I put one into the beam after it left the cylindrical lens. If I let this combination focus the beam down to a dot, I get only very diffuse lasing; but as long as I let it make a bar that is at least a millimeter or two long, it directs the output fairly nicely:


Figures 30 & 31: Tight-focus setup; tightly-focused output, lasing DTC in 70% Isopropanol

[For more information about this dye, see my report on inexpensive dyes for the DIYer.]

Back to the previous nitrogen laser project, a straightforward and simple “first laser”

Back to the laser projects index on this server

To the Research Reports index on this server

To the top of this [updated] mirror

To my current research homepage


This work was supported by
the Joss Research Institute

Contact Information:

Email:, where you can replace a with my first name (jon, only 3 letters, no “h”) and b with joss.

Phone: +1 240 604 4495.

Last modified: Tue May 9 12:49:48 EDT 2017