TJIIRRS: Number 5C [New] of an Ongoing Series;

“Theorie und Praxis IIA”:
Revamping the “DKDIY” Laser

Part 1

(15 August, 2006, ff)

This page details the construction of a nitrogen laser that is a follow-on to the “DKDIY”design I published here a few months ago, along with a “How-To” page. Because this material is being written substantially as a historical track of the project as it is taking place, it is not necessarily organized logically. When the design is fully stabilized I will try to provide a “How-To” page for those who want to build a laser of this type.

(Note, 2006 September 27: Between the “DKDIY” laser and this “DK-Plus” laser, I experimented with a larger design, which operated, but not at the performance level I had expected. This appears to have been caused by several factors, some of which I may explore [and, I hope, correct] by returning to that laser and rebuilding it, now that I have this one working well.)

(Note, 05 October, 2009: I am reworking this laser, and I hope to get somewhat better performance from it than I originally did. There are a number of issues involved in the rework, which I discuss at one or two places on this page, and in a second page specifically devoted to it.)

I would strongly suggest that you read through at least the early sections of this page and all of the following page before you attempt to build one of these, as that will help you avoid some “gotchas” that I encountered.


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

Origins and Design Rationale

The first version of this laser was intended to provide performance at least as good as that of the classic Scientific American “Amateur Scientist” nitrogen laser design, but using “doorknob” capacitors instead of circuit board, and with a triggered spark gap as a switch, rather than a free-running gap. It is a “Voltage Doubler” circuit, often mistakenly called a Blumlein. I chose the doubler circuit because it is relatively easy to construct, and because it parallels the circuit of the SciAm design.

That first “DKDIY” laser used a dozen laser-grade SrTiO3 doorknob capacitors. It reached threshold at about 12.4 kV, and developed a little over 100 kW at about 20 kV. The channel was 22 mm across and about 45 cm long, and the electrodes were pieces of extruded aluminum carpet edging.

This revised version will use 16 doorknobs, and will have electrodes a bit more than 80 cm long, with capacitors along the middle 45 cm or so. I originally intended to use a piece of brass shim stock as the cathode; shim stock has a fairly sharp edge and so should provide some preionization by generating corona early in the discharge cycle. It’s a bit thin in comparison with the anode, but may work. If not, I can always try a packed-blade cathode, or add some extra preionization.

(Note, added later: I soon abandoned the idea of a thin sharp-edged cathode, and shifted to two identical electrodes.)

My original intention for this enlarged version was to use carpet edging again, at least for the anode; but when I examined the pieces I had on hand I discovered that they were not very straight. Straightness being a real issue, I changed my mind and turned to something that does have a straight edge: an aluminum ruler. The ones I’m using were originally 4 feet long, but I have cut them down to 32 inches, a reasonable length for this head. They are 2" wide, which fits well, and 1/8" (~3.25 mm) thick, also very reasonable, though that is significantly thicker than the previous electrodes, which had a cylindrical edge only 1.5 or 2 mm in diameter. My hope is that the capacitors I’ve added will be enough to let me pump the additional channel volume. (It is important to deposit sufficient energy into the discharge, in order to be sure that the laser will operate well above threshold. Typical high-performance nitrogen lasers seem to dissipate at least 40 joules per liter of active discharge, and some exceed 100 j/l.)

Following an idea developed by Jarrod S. Kinsey, I am constructing the sidewalls and spacers of this new head from wood, a construction material that is widely available, extremely tractable, and relatively inexpensive. (I may provide windows through the sidewalls to permit observation of the discharge in the channel.) Wood tends to be slightly conductive at high voltages, but I will be using walls that are varnished, which should reduce the conductivity somewhat. If some conductivity remains, there is a small chance that it may provide a bit of preionization, in much the same way that a semiconductor plate does.

(Addendum, 27 September, 2009)

Most wood is a bit porous, which is a problem. (It turned out to be a problem even with the varnished yardsticks I used in the head of this laser.) Consider a nitrogen laser running at 50 Torr; if there is a leak that admits 2 Torr of air, the gas mix contains more than 3/4 of 1% oxygen, which is too much for optimal performance. (A small amount of oxygen is okay, and in fact at roughly 1/3 of a percent you may see slight improvement in performance. Beyond about half a percent, however, oxygen degrades the performance of the laser.)

If you use wood, make sure that you either coat the interior surfaces with something that is relatively impervious to air, or soak the wood in something that will solidify and make it impervious. One way to do this is to mix epoxy with isopropyl alcohol (at least 91% pure; 99% pure is better), and brush it on. You probably want the epoxy to be just liquid enough that it brushes reasonably easily; if it is too dilute it has a tendency either to fail to cure, or to cure to a soft and rubbery condition. After the first coat of epoxy has had time to cure, repeat the process until it no longer soaks in; this will almost certainly take several coats. When all of the epoxy has fully cured, the wood should be sealed well enough to be usable.

There are certainly other ways to accomplish this, but you will want to think it through carefully. Some paints, for example, continue to emit solvent vapors for months. These may (or may not) interfere with lasing. There is only one real way to find out; and if the answer is ‘yes’ you will probably be obliged to rebuild the head, which is (believe me) a real annoyance.

Circuit and Physical Layout

Here is the circuit diagram of the laser:

The capacitor labelled “Start Cap” is present mostly in order to produce a current of a few dozen Amperes in the spark gap very quickly when it is triggered, to help develop a substantial conduction channel in it. The manufacturer recommends pushing at least 10 Amperes through the gap to get it to switch properly. This probably takes only a few dozen pf, which can be furnished by a rather small doorknob; it is, though, important to keep the connections fairly short and the inductance down, as you have only a brief time in which to accomplish the job: a good spark gap should switch in a few tens of nsec. (Please note that although it is possible to make subnanosecond spark gaps, the designs I’ve seen were pressurized to about 1500 psi and were built into cylindrical transmission lines.)

I have shown a charging inductor across the channel, but a charging resistor works better in some designs, and I will probably try both to see which is appropriate for this laser. I have also shown a small capacitor that connects to a dot near the cathode. The dot represents a thin wire that is strung along the entire length of the head, and serves to preionize the laser. This is discussed in the text, and is the initial configuration I’ve chosen to try; I may move to a different preionization method later.

(Note, added much later: I did. See below.)

The sidewalls of the channel are wooden yardsticks, about 1.5" wide and just over 34.5" long, cut down from their original length of 36". I have widened the lower sidewall by gluing smaller pieces of wood to its edges, as it was just a bit too narrow to work well — I want to have the electrodes 25 mm apart, and I need to have room to attach spacers to separate the electrodes from the sidewalls so the discharge doesn’t track along the surface. (At the bottom of the left photo below you can see the 1" spacer sitting on one of the rulers, before I added extra wood at its edges. You may be able to tell that the spacers, which are visible in the photo on the right, are farther apart than they could have been with the unmodified ruler.)

The electrodes are 32" long, but the capacitors are restricted to the center 23" or so. This should help avoid sparking at the ends of the electrodes, where the longer current path provides higher inductance. We hope. [[NOTE: As of 19 August, 2006 I am revising my thinking about this, and moving toward a slightly different head design. See below.]]

Here is a photo showing the underside of the roof, with one of the gas ports visible near its left end; the bottom wall of the channel, with its spacers to position the anode and cathode; and the left end of the anode.

I have to decide, very soon, whether I expect to run this laser under vacuum. If so, I will need to coat the inner surfaces of the wood with something, to reduce outgassing. If I’m only going to use atmospheric pressure helium with a small amount of nitrogen, on the other hand (as I did originally), I don’t think I will worry about that. The rulers are already varnished or coated in some manner, and the other bits of wood I’m using are fairly small. I am tending to think that atmospheric pressure may be a good bet, as I am not sure how well the rulers, which are of low quality, would stand up to vacuum and its attendant stresses.

(Note, added much later: they stand up just fine to the stress, but they leak unless carefully sealed.)

Yet Another ReDesign (YARD)

(19 August, 2006)

As I continue to think about this laser, I am changing my mind about how to proceed with it. (This should come as no surprise to anyone.) The base design takes a replaceable head, and I have at least three ideas for heads. The one I now think I will build first has 1/8" thick aluminum rulers as both electrodes. These will be separated by 5/8", just under 16 mm (I did a rough calculation last night that suggests better chances of good performance with tighter spacing than I used in the original DKDIY laser), and will be about 35" long, to help prevent sparks at the ends. Because of the narrower channel, the rows of capacitors will end up being 4" apart instead of 4.5"; I will probably put the “lower” row 1.5" upscreen, rather than at the 1" position shown in the diagram.

I am becoming more and more convinced that preionization is the key to high performance in low-pressure nitrogen lasers (in TEA lasers, it appears to be essentially mandatory), so I am also going to add a preionizer, a thin nichrome wire in the first version. I intend to attach the upper sidewall in such a manner that it is easily removed. That will let me test other preionization methods without totally rebuilding the head.

My rationale is that this head will be extremely easy to construct, and should swiftly provide some much-needed information about preionization methods. It should also work well, assuming I can preionize it sufficiently.

Here is the tentative layout of the laser, based on an 8" x 34" brass kickplate I acquired at the hardware store:

The asymmetry, with wider open area “above” the “upper” row of capacitors (in case the print in the picture is too small to read) facilitates making connections to the spark gap, which is not shown in this diagram.

I also acquired some square extruded aluminum tubing to go under the baseplate to stiffen it — the head is long, and probably not stiff enough by itself to force everything to stay lined up.

The aluminum rulers I’m using for electrodes are just 2" wide, and the spacing between them is 5/8", which sets the spacing between the rows of capacitors — I have to be able to bolt the electrodes to them. Very fortunately, that spacing just accommodates the wooden rulers I am using for sidewalls with a wee bit of “wiggle room”, as they are 1 & 7/16" across. If it had been any closer, I would have had to widen the laser channel or adopt a different design.

Sidebar: Preionization Techniques

For preionization by wire, there are two obvious modes: passive, and active. I have discussed these on the root page of this series, but it’s easy enough to do a brief review here.

Passive: You typically use a single wire for this. It is strung along the channel, typically off to one side so it isn’t in the middle of the discharge (though people have actually used wires spang in the middle), and usually considerably closer to one electrode than the other. You connect the wire through a capacitor to the more distant electrode. When you fire the laser, the rapidly-rising voltage across the channel appears on the wire (the little cap is uncharged, and essentially looks, at least for a brief period, like a dead short), and a corona discharge develops. This provides ions and UV, which preionize the channel.

There are two modes of thought about positioning and connecting the wire. One says that you should put the wire close to the cathode and connect the small capacitor to the anode, because you want your preionization to be at the cathode. The other says that the cathode emits electrons, so you should put the wire close to the anode and connect the little capacitor to the cathode, which makes the preionizing wire effectively be the cathode at the beginning of the discharge cycle. I think I fall into the former category (wire near the cathode, capacitor to the anode), but I have not made any study of this, so you are advised not to trust my opinion. On the other hand, that is the configuration I am building for the initial tests of this head.

An alternative that doesn’t involve wires or capacitors is to use a sheet of semiconductor, off to one or both sides of the discharge. The conductivity of the semiconductor has to be chosen moderately carefully, but this technique (which was developed with CO2 lasers, works quite well. I have made one of these by putting a thin coating of epoxy onto a sidewall, and applying fairly fine silicon carbide abrasive grit to it while it was still wet. I also left an empty channel somewhere in the middle of the wall, so that sparks could jump from one semiconductor “electrode” to the other, but I doubt that this is necessary. On the other hand, it could provide an easy way of compensating for semiconductor material that’s a bit too conductive.

Important note: passive preionization steals energy from the main discharge. If you don’t steal enough, you don’t get adequate preionization. If you steal too much, the main discharge doesn’t get enough. Either way, performance suffers.

Active: You can do this with one or two wires. Active preionization requires a separate power supply, and draws about 1 mA of current, either between one wire (or two, if you can make sure they both source current) and one of the electrodes, or between two wires placed at either side of the channel so that the corona discharge crosses through the channel.

An oil-burner ignition transformer is a convenient source of high voltage (this technique tends to require only about 5-10 kV). If your oil-burner transformer has a centertapped secondary winding with the centertap connected to the case, the obvious circuit is a bipolar full-wave rectifier, similar to what you might build for a low-voltage bipolar supply. Most neon-sign transformers are constructed the same way, and a small one could be used for this. Here’s the circuit:

Because this has two outputs, it suits itself well to the use of two wires.

In use, we would like to forget about the ground connection, but it is important to be careful because one side of the HV supply for the laser is also grounded. I am guessing that if you put a large resistance to ground from the centertap of the transformer, and only moderately large resistors (for current limiting) to the corona wires, you should be okay. Remember that all resistors need to be able to withstand rather high voltages. Remember also that your HV rectifiers should be able to withstand 6 to 8 times the rated voltage. Here’s the logic:

Granted, this is expensive; but it will save you much grief.

As I mention above, I am going with single-wire passive preionization for the first version of this laser.

The Build Continues

(20 August, 2006)

I have marked and drilled the holes in the electrodes, and I’ve almost completed the sidewalls.

I had set the electrodes out on the bench with the spacer between them...

...and was about to RTV the floor of the channel on, when I realized that after I flipped the resulting assembly over, I’d be trying to connect to the doorknob caps through the painted numbers. This was clearly suboptimal, so I set the floor aside and configured the roof. (I will have to remove the anodized coating from the surface anyway, so it isn’t that big an issue, but I think it will be instructive to have the numbers visible when the laser is finished.)

Because I intend to use a wire preionizer for this iteration, I wanted to have a narrow space for the wire to sit in, down between walls of some sort, to help prevent the discharge from heading for the wire, which it might otherwise do. This is a wooden head (much like my own, argh), so I decided to use two strips of wood; I pushed the first one into place with a straightedge and held it while the CA (cyanoacrylate) set. Then I put pieces of cardboard between it and the second one while gluing that into place:


The result appears to be a moderately straight groove, which will probably serve. Now I have to figure out how to glue a 0.0031"-thick nichrome wire into place in the bottom of that groove, without covering the wire with glue. I’m thinking about diluting some epoxy with isopropyl alcohol, painting a thin layer of it into the groove, letting the iso evaporate, and then laying the wire onto the surface of the epoxy layer, which will be quite thin by that point. Because I don’t need much physical strength, I don’t have to worry if the epoxy starts to harden before I get the wire onto it. In fact that might even help, as the wire would be less likely to get covered if the epoxy is already a little stiff.

The real problem with this arrangement is that the wire is a wee bit too close to the cathode. I should have used a slightly wider spacer on the cathode side. Such, however, is life. We’ll see how well it works, or doesn’t.

Once I have the gas ports and the wire in place, I will RTV the roof on.

(Early AM, 22 August, 2006)

I have dropped the 3-mil nichrome wire into the groove I made for it, drilled the holes for the gas ports and installed the connectors, and RTVed the roof to the head. If I do this again I may design the thing to make installing the wire easier, and I may also opt for wire of larger diameter. Still, it’s an interesting test, and I have another wooden ruler if I decide that I need to fall back to semiconductor preionization.

Once the electrodes and the roof of the channel are firmly attached to each other, I can set them on the baseplate to check the hole positions. (I don’t want to drill the baseplate until I’m sure I’ll be able to assemble the laser.) Then I get to:

At that point, I think I’ll be ready to start testing.

(24 August, 2006)

Here’s a test of what the machine will look like when I assemble it:


Notice that there are no end windows on the head yet, the preionization wire is not connected to anything, and the spark gap is not present. Several other things also need to happen before this will be a laser, but at least we can now get a sense of what it will be like.

Because the rulers I’m using for electrodes are anodized, it was necessary to clear the areas where they are going to connect to the capacitors. I figured that a wire brush on a Dremel™ or other rotary tool would do the job, and a test showed that this is correct:

I have brushed the 16 contact areas for the main caps, on the undersides of the electrodes, and 2 extra areas on the top of the anode, one at each end, for pieces of brass shim that will go to the caps that drive the preionizing wire. As of now, I am planning to use 760 pf doorknobs for those, partly because I actually expect the wood to do some of the work for me.

Sidebar: Capacitor Improvement 101

It seems extremely likely that doorknob caps with broad terminations have lower inductance (and are therefore faster) than doorknobs with narrow terminations. Sometimes, there isn’t much you can do, but a few types can be improved fairly easily. Here is a little MuRata 760 pf cap:

You can easily see that although the termination is just a narrow ring around the #8-32 threaded hole, there is actually much more metal present. Unfortunately, it is covered with epoxy. Fortunately, you can do something about that. First, you get to clean the epoxy off:

Unfortunately, I only have two hands, so I can’t show you how I hold the knife with one hand and rotate the capacitor with the other; but I trust you get the idea.

Here is what the cap looks like when it is thoroughly clean:

You then use silver-loaded conductive epoxy to put a #14 brass washer (or a 6mm washer, as I discovered when I ran out of #14s) on the cap. The next photo is just to give you a sense of size and fit; I have not yet epoxied the washer on. If you don’t have (or cannot afford) silver epoxy, don’t worry about it. If you attach the cap to your circuitry with reasonable pressure, you will still get good conductivity. Do remember the washer, though.

The washer is just thick enough to match (more or less) the height of the original termination, and the hole is just a bit wider than the ring. Why the manufacturer didn’t simply make these caps with full-width terminations, so we wouldn’t have to go through this idiotic rigmarole, I have no idea.

We now return you to your previously scheduled programming...

(Midnight, 25/26 August, 2006)

I have almost completed the assembly of this head. The windows are on (though, as you will see in the photos, the RTV has not yet set on the second one, so it is still propped into place), and the preionizer is ready for use.


I have discovered a pleasant thing: an ordinary hole-punch for paper will (just barely) make a hole in 5-mil brass shim stock that is only a little bit large for a #10 machine screw. It was even tolerable for the shims on the preionizer, which currently take #8 screws. I don’t know how long the device will last in this service, but it is certainly convenient. Cost me a whole dollar at an office supply store.

Once all the RTV has set, I may be able to perform some preliminary testing on this design. I have a GP-70 that is already mounted on brass shims, and although it won’t be as fast as a properly mounted gap, it should give me some sense of whether this design is viable. Proper operation, however, will want a GP-14B mounted on much wider pieces of shim, for good tight coupling and better performance. That should be fairly easy to arrange, fortunately.

(27 August, 2006)

I have now tested this arrangement, using the GP-70 mentioned above. The device discharges, but does not lase at charging voltages up to 20 kV.

I have tried a variety of fills and pressures, and I have tried increasing the drive capacitance on the preionizer. There are some indications of sparking at the end of the channel that I can see, and I am now thinking about retreating to a head with glass walls, though I may very well use wooden flats as semiconductors for a preionizer similar to the one I used on the original DKDIY device. I may also sandpaper the edges of the electrodes to remove the anodizing from them.

Another possibility is to try 900-pf capacitors, which should be somewhat faster than the 2-nf doorknobs that are currently in place; but they also store less than half as much energy, and I doubt that the speed increase would be great enough to compensate — that should change with the square root of the capacitance, whereas the energy storage is linear at a given voltage.

Time to sit and think for a while...

(later, that same evening)

...but not for too awful long. I have just started construction of the new roof for this channel. If the roof isn’t enough, I can also build a new floor, but let’s do this one step at a time.

I have some glass bars, about an inch and a half across and 1/8" thick, but they are only 30" long, and I need something a lot closer to 36", so I am cutting and gluing. Clean, straight ends are easy to epoxy if you are not concerned too much about strength, and in this particular case I think most of the stress will be across the bar when I pull vacuum on the head, so I’m not going to worry about it too much. 1/8" of glass should be about as strong as 1/4" of wood, and the wooden roof stood up to the vacuum reasonably well.

I am going to try wooden preionizer bars on this roof, with silicon carbide grit on them. I will leave 1/4" separation between the flat wood pieces, and that should be enough to get good operation. It will also let me look inside the head and see a little of the discharge, which will give me some information that is hard to get by other means.

Here is a cross-section of what I am expecting the new roof to look like:

(Glass is shown in pale blue-green; wood in brown; electrodes in purplish gray; silver paint in blue; silicon carbide in gray. The pale pink vertical bar marks the centerline. The electrode spacing is unchanged, as is the floor of the channel, which is not shown in this illustration. Glue and sealant[s] are also not shown.)

(early evening, 27 August, 2006)

Alternatively, I could dispense with the silver paint and just put a silicon-carbide-coated strip down the middle of the open space, possibly 1/2" wide, or maybe even the full width (3/4"). This would obscure my view of the discharge, but would probably preionize the channel adequately.

(early AM, 28 August, 2006)

...And that’s exactly what I did. Here is a diagram:

Here are two photos, one taken with the top sidewall sitting in place, before I made the preionizer, and another taken after I put the preionizer in and put RTV onto the sidewall to seal it and hold it in place:


The wood I had on hand that was the correct width was in 2-foot sections; you can see the joint in the preionizer somewhat to the left of center in the second photo.

I still need to create the gas ports, but that should be relatively straightforward. Then I can resume testing.

I should note that while I was setting up the new “roof”, I also acquired some 1500-grit sandpaper and removed the anodizing from the upper edges of both electrodes. I wasn’t able to get the entire face, but at least there is now an exposed region of aluminum, and it is close to the edge of the preionizer, which should help that work correctly.

(early afternoon, 28 August, 2006)

I have now built the gas ports. With any luck I should be testing at room pressure within a few hours, when the RTV has begun to harden. My apologies for the slight motion blur in these photos; the light was less than optimal. In fact, I had to shine a white LED up into the end of the head to show the surface of the preionizer in the second photo.


Notice that the preionizer extends well past the ends of the electrodes. Also note the slanted bottom on the gas port section, which unfortunately is not easily visible in either photo. I built both of them this way, to minimize the exposed areas on the sides, and also for strength. As Bucky Fuller pointed out, a triangle resists deformation a lot better than a square does. (Yes, I know, the side-plate is a square, but that’s because I had it handy; it is covering a triangular hole.)

(early that evening)

Initial tests give me lots of bright sparks from the electrodes to the preionizer, and under some circumstances there are lots of very small sparks, a good sign. I suspect that the preionizer conducts a bit too well, as I do not appear to be getting a “regular” discharge, at least with helium at 1 atmosphere; but we’ll see how things go under vacuum ...after the RTV sets, probably tomorrow morning unless I just can’t stand it. (I certainly should stand it — the stuff is still quite soft now, and it won’t be even moderately hard for at least another 12 hours.)

In the meanwhile, here are four views of part of the channel, showing typical shots. The brighter sparks in the laser are pink because the head is filled with helium. If you want more detail, btw, change ".10c" in the filename to ".22c". That gets you the crop I made from the original image, unscaled. I think the upper ones are in better overall focus than the lower ones, but the right sides of the lower ones are okay.



One saving grace of this head design, btw, is the fact that I can scrape the current preionizer off and put a narrower one on, if I decide that it really is just too conductive. It wouldn’t be fun, but I could do it. Alternatively, I could just raise the roof a bit by adding more spacers, which would put the preionizer further from the electrodes. Again, annoying; but certainly possible to do, which is good because I expect to be doing it in a day or so.

(early AM, 29 August, 2006)

Based on what you can see in the photos above and on my general sense of the behavior I observed it seemed pointless to wait, so I have just removed the “roof” and added an additional 1/8" of spacing. It was certainly annoying, but I knew I might have to do it, so I’d set up the head to make it as easy as possible under the circumstances, and it really didn’t take very long. It’s done now, and when I wake up in the morning I should be able to resume room-pressure testing. Later on we’ll see about leaks...

(around 0900, 29 August, 2006)

Initial results at room pressure (arcs and sparks that don’t actually appear to be going to the preionizer) lead me to think that 1/8" may actually have been a bit much, and I may eventually remove those spacers and replace them with 1/16" ones; but I want to do some testing at reduced pressure first, as that could change the behavior significantly. Have to wait until the RTV gets much stiffer, though, before I will feel good about pulling vacuum on this head. Maybe tonight...

(early that afternoon)

I decided to apply Jarrod Kinsey’s excellent methods: used a piece of polyethylene hose as a stethoscope to find the leaks, and stopped them up with a rather rubbery type of hot-glue. Took me perhaps 30 minutes to clean up just about everything (or so I thought; see below for more about this), after which I resumed testing. As far as I can tell, all I’m getting is bright (and dim) sparks from the electrodes to the preionizer, even with the spacing as wide as it now is. I am somewhat surprised, but that’s what I’m seeing. I want to try a variety of fill pressures and mixtures, but I have not been able to do much testing yet — the vacuum pump emits an oil fog that I really don’t want to breathe; I have to figure out a way to trap that stuff.

Interlude: Vacuum

As long as I have this problem, let’s talk about it.

I have a lovely commercial roughing pump. Only one problem: if it runs for more than a few seconds against even a light load (for example, a nitrogen laser), the exhaust becomes a nasty fog of oil droplets. In the best of all possible worlds, there would be a neat way to catch the fog and turn it back into liquid oil, which would then drip back down into the pump: after all, that’s where it belongs.

In the next-best world, there would be a way to catch the fog and remove it from the exhaust of the pump, even if there weren’t any good way to turn it back into usable liquid oil.

In the mediocre (real) world, one runs the exhaust hose to the great outdoors. Vacuum-pump oil of the ordinary sort is a relatively innocuous material, and there are bacteria in the soil that will eat it if they can get their grubby mitts on it. For various reasons, however, including the fact that vacuum pump oil is expensive, I would prefer to let those bacteria eat other things.

Possible Method #1: The Cottrell Precipitator

There is a very pleasant way to remove dust and aerosols from the air. It is called a Cottrell precipitator, and it is a fine use of HV DC. You can operate a Cottrell from an electrostatic source, and I believe that under some circumstances you can even do it with AC. Basically, you make an open-air capacitor; let the air flow past the plates, and even modest amounts of corona will charge the particles, which then stick to plates charged one way or the other. (I have several commercial electronic air filters that do this. They are quite good at removing dust; I think the usual rating is 97% per pass, and the size of the particles is not particularly an issue, the way it is with HEPA or other mechanical filtration methods. After room air passes through such a device a few times, there is very little particulate matter left in it. Unfortunately, a room-air cleaner is not easily adapted to this service, so I’m going to have to build something.)

I happen to have a nice old oil-burner transformer here, courtesy the very kind fuel-oil place a door or two down, which can serve as a source of high voltage. It is time to do a spot of construction...

(late that evening)

I suspect that I have the “plates” (actually pieces of brass screen) too far apart, as there is still fog coming out of the pump. I will have to figure out a way to reduce the spacing. I tried putting rectifiers in, to make DC instead of AC, but it did not appear to change the performance. Too bad — it would be really nice if this thing worked. I want to get back to messing with the laser.

(next morning, 30 August, 2006)

Well, hmmm. It may be simpler than I had thought. This PDF file shows a simple design, and comments that oil droplets (at least in cigarette smoke) tend to be positively charged, which tells me that I want DC and that I want the negative pole in the middle. When I get a chance, I will probably go get some PVC tube with a thin wall, and see how this configuration works. That, however, probably won’t be until tomorrow, as I am scheduled all day today with other stuff. (Sigh.)

(NOTE, added 06 October, 2009: That PDF file is suspect. Milan Karakas has found other information, and has verified it by experiment. It is clear that insulating the conductors is not a good idea, at least under ordinary circumstances. See below for a rebuild of this device during October and November of 2009.)

(31 August, 2006)

I took a piece of #8-32 threaded rod about 2 feet long that was lying around, a couple small pieces of fine-mesh brass screening, a piece of 3/4" PVC pipe from the hardware store, some plain copper wire, and a fitting to hold the pipe and sit on the outlet of the pump, and constructed a device.

First, I made a hole in the middle of each piece of screen and used hot-glue to attach them to the ends of the pipe. Then I used a nut and a lockwasher on the underside of the top screen, and an acorn nut and a lockwasher (and a drop of hot-glue to prevent the vibration of the pump from wiggling the acorn nut loose) on the upper side, to position the piece of threaded rod so that it hangs vertically in the center of the pipe. I left a tab on the upper screen, to connect to the negative terminal of my impromptu power supply. Then I wound the copper wire around the outside of the pipe, holding it in place with a tie-wrap at each end. That is now connected to the positive terminal of the supply. Here is an overview of the Cottrell on the pump, and then a closer view of the negative terminal:


Net result: Much less aerosol, but if I run the pump long enough, particularly if I am allowing some gas to flow through the laser head, I do see some. I will be looking into ways to improve this, as I need it to be better than “some”. It is possible that my HV rectifiers are not up to the task; I will be testing them when I have time. I am bidding on some others, which should have sufficient PIV rating to serve. (I have pairs of 15 kV rectifiers now, and they may not be stackable...)

(03 September, 2006)

I have rebuilt this precipitator, using about 4 feet of 1/8" brass rod inside a pyrex tube down the center, and with a much taller PVC tube on the outside. It seems to work a bit better. Also, I am running it with a commercial 20 kV supply, which probably helps.

(Note, added on 27 September, 2006: I did not get the Cottrell unit to work the way I wanted, and I will have to revisit it when there is time. I know that Cottrell precipitators can be made to work quite well, and I am very curious as to why mine doesn’t.)

(Note, added on 27 September, 2009: Milan Karakas has, as mentioned above,provided some additional insight, and I will be rebuilding the precipitator when I have time and materials. It appears that an uninsulated wire down the middle of a conductive tube works better than an insulated wire down the middle of an insulating tube. This is, in retrospect, not surprising, and we are wondering why the PDF file referred to above specifies otherwise. Be that as it may, I have started to build a new device, which will be about 60 cm tall and will use 3-mil nichrome wire as its corona source.)

...But enough of this. Back to the laser.

I put a strong UV absorber between myself and the laser, just in case (even though it has not yet shown any sign of lasing), and examined the discharge. I am getting bright sparks across the channel, which is both a good sign and a bad sign. I’m happy that they go across the channel and not just up to the preionizer; but I’m unhappy that there are bright sparks rather than a nice clean discharge. Have to think about what might govern this and what to do about it.

(some hours later)

What I think is that I need to know just how much vacuum I am achieving. Without that information, I am floundering around in the dark.

I cleaned up one of my mechanical vacuum gauges so it gives sensible readings, and put it on the system. Couldn’t get any better than about 25" Hg. Went over the head with Jarrod Kinsey’s stethoscope method, and found two very small leaks. That bought me perhaps another inch. It seemed possible that I was getting some leakage through the wood, so I fingerpainted RTV on almost all exposed wooden surfaces. Here is a view of the underside of the channel:

When I get back from rehearsal this evening, I will check to see whether I get any better vacuum. If I can’t get to at least 29" Hg (that’s a little less than 23 Torr) I am not going to be able to do a reasonable test, because there will be too much air in the channel. (For reference, 1 Torr is the amount of pressure it takes to raise a column of Hg by 1 mm, so 760 Torr, which is one standard atmosphere, amounts to about 29.92" Hg. Needless to say, if you are in the middle of a low-pressure zone, and the barometric pressure is only 28.5", you are not going to pull 29.9" of vacuum. Likewise if you live at a high altitude. This kind of thing is why we prefer absolute pressure measurements.)

By dint of careful listening and wholesale application of hot-glue, I finally got to just slightly better than 29" on the gauge, which means (assuming that the gauge is not totally "outta whack") that I can now get down to about 20 Torr. That isn’t great, but it is low enough that I seem to get something that resembles a discharge a bit more than just a pile of sparks. I have succeeded in adding a small amount of helium without losing the character of the discharge, and I guess the next step is to add a small amount of nitrogen and see whether the device reaches threshold...

In fact, it does:

This is just the beginning, but at least it is a beginning. I was worried that there might be something fundamentally wrong with either the design or the build.

Next I get to do some optimization, assuming that I can get the precipitator to be fully functional.

(mid-morning, 01 September, 2006)

Here are two more photos. The first is the discharge as it appears with perhaps 75 Torr fill pressure, a mixture of helium and nitrogen (and some air). The second is the output. These are both slightly out of focus, and eventually I will try to take better ones.


The fact that the output looks like a donut is somewhat unsettling, and I am going to have to take a good hard look at it. I will also probably have to put a mirror on one end of the laser, to see what effect that has. I have acquired a piece of 3/8" thick glass that will serve as a shelf for the mirror mount to stand on, and I am attaching it to the baseplate with RTV. I also begin to think that it’s time to put a GP-15B spark gap on this laser, and see what it does at 30 kV.

(03 September, 2006)

Here is a photo of the output at 20 kV driving the fluorescence of a piece of bond paper. There is a mirror at one end of the laser, and the gas pressure is vaguely optimized. The mirror is quite difficult to adjust, and may not actually be precisely “on” in this photo, but it is fairly close. My apologies for the polyethylene tube that cuts off the top right corner of the output spot. It is the vacuum hose, and is difficult to move because it goes to the gauge, which is quite close to the target.

(Unfortunately, this is cropped straight out of the original; there is no larger image.)

Small amusement for those who have dealt with high voltages: because the baseplate is “hot”, I had to hot-glue plastic bottlecaps onto the adjustment knobs of the mirror mount, because otherwise I couldn’t really touch them with the power on. (I checked and found that even though the mirror mount is sitting on a glass plate that keeps it insulated from the base, I drew tiny sparks if I got my fingers within about 3 mm of any of the knobs, so I decided to be safe rather than sorry. It pays to be careful with these devices!)

Here is the focused output of this laser, pumping my homebrew cuvette to superfluorescence. The dye in the photo on the left is Rhodamine 6G of rather dubious purity (certainly not laser grade), and in the photo on the right is 7-Diethylamino-4-Methyl-Coumarin of somewhat better quality; both are dissolved in 95% ethanol:


The output is the greenish double stripe at the upper left, on the piece of paper. The first focusing lens is not readily visible in these photos, but you may be able to see the second lens, which is cylindrical. In the photo on the left, the laser is not well optimized. It had no trouble lasing the dye, however, even with no external mirrors, and with the walls of the cuvette deliberately misaligned from the front window so that they cannot function as mirrors. In the photo on the right, I have improved the focusing and adjusted the gas pressure slightly, and the laser is using a larger spark gap that is connected with considerably broader pieces of brass shim stock, so it should be switching slightly faster and possibly producing slightly higher output power.

(05 September, 2006)

In preparation for running this laser at higher voltages, I have painted the underside of the head and parts of the brass shims with HV insulating varnish. When that dries I will reassemble the laser and try it on the bench, where I have a larger HV power supply.

(later, that same day)

Here is the laser on the bench. Directly above and behind it is the electrostatic voltmeter. You can see part of the trigger unit in the foreground, and a commercial TEA nitrogen laser in the right rear. I have now run the new laser as high as about 28 kV, and it continues to perform quite well.

Here is something you just don’t get to see every day — I accidentally got the concentration of dye in this solution (it’s Rhodamine 6G) a little bit low, and as you can see in the photo, I am lasing it by pumping it longitudinally:

If you want to do this deliberately, you will probably find that it is easier with Fluorescein, which does not absorb particularly well at 337 nm. (I tested, and it worked on my first try.) As a start, you may want to adjust the concentration until essentially all of the pump light is absorbed in the first 1/4 to 1/2 of the cuvette; see how things go from there.

It is also possible to lase some kinds of fluorescent plastic sheet this way, and I once saw some video footage; but I rarely observe longitudinal pumping in my own setups, and have not yet succeeded in thresholding any of the samples of fluorescent plastic sheet I have. Perhaps as I get this laser better optimized I will try again.

Power and Energy Measurements

(05 September, 2006)

Pulsing the laser once per second, focusing the beam into a Scientech head, and running the output through a x100 instrumentation amplifier, I get a reading that corresponds to about 0.67 milliwatts. That’s roughly 670 microjoules per pulse. I am hoping that I will eventually get considerably more out of this laser, but it will take additional optimization, possibly including finding and fixing more leaks.

Here are four representative output pulses. These were taken at 25-27 kV (the last one is around 27, but I am not entirely sure about the others). The last one also shows what happens when I tweak the gas pressure. It is a little better than most, but still fairly representative. (My apologies about the blur. The camera seems to have some doubts about how to focus.) These are at 2 nsec per division on the screen. The sensor is a Motorola MRD500 photodiode, with bias provided by 6x 9V batteries in series, for ~54 VDC. The signal from the diode goes down a piece of 50-ohm coaxial cable that is about 8" long, directly into the input of a 7A19 vertical amplifier (600 MHz bandwidth) in a Tektronix 7104 mainframe (1 GHz bandwidth). The risetime of this setup should be 1 nsec or a bit less, and is most likely limited by the photodiode.


Please notice that the FWHM pulsewidth here is 9-10 nsec in the first two photos; 6 or 7 in the third photo (it kinda hangs at the 50% mark for a while before finally dropping); and 4 nsec in the last photo, largely because of the tall first peak. The peak power of the pulse in the last photo is much higher than the peak power in any of the others, but I suspect that the pulse energy is largest in the second photo. The fill gas for all of these, btw, was just nitrogen (except for any air that got in through leaks). I have not yet tried examining the pulse with nitrogen and helium together.

Here, just because it happened, is the result of putting too much energy into the photodiode. This was at 10 nsec per box...

I knew it couldn’t really be an accurate record of the laser’s output, but when I blocked the light I didn’t get a trace, so it was clearly a real signal, and it was coming from the photodiode. Took me a little while to figure out what was going on. I then put a piece of 1/4" window glass and some fluorescent plastic in front of the laser, which absorbed enough of the beam that the diode could handle what remained. I found, btw, that I had to choose carefully. Even a 1-mm thickness of the plastic that appears to be doped with Rhodamine B absorbs essentially all of the nitrogen laser’s output. (The plastic did not lase, partly because the output of the nitrogen laser was not focused. I will probably be trying again at some point.)

(evening of 06 September, 2006)

After dinner, I optimized the pressure by watching the scope as I fired the laser. With the current mix of nitrogen and helium, the traces were tallest and probably widest at 25.2". (I should note that I do not have gas flow meters, so I don’t really know what the precise mix is.) I then put the power meter head in front of the laser, and measured the energy. This turns out to be, very roughly, 1.2 millijoule per pulse. Given a pulsewidth of about 8 nsec FWHM, which may actually be a bit on the long side — see photos, below — we are looking at peak power on the order of 150 kW. (The 10-10 pulsewidth is considerably longer, and the average power is only perhaps 100 kW.)

Unfortunately, I did not get a picture of the optimized trace; but I have other traces here, taken at various pressures that I have labelled on the small images. Notice that with both gas mixtures shown here, I am getting best operation around 26.5 to 27.5" of vacuum; I don’t know how to reconcile that with the 25.2" that I was getting earlier, though that mixture may have had a significantly different ratio of helium to nitrogen.

[Note: if you click any of these, you’ll get a 1080x870 px enlargement. That larger image is a direct crop from my original, and is the largest size I have. The traces, btw, should be easier to see on the enlargements; they did not show up well, so I tweaked the green levels to make them more apparent.))

First set, nitrogen and helium:


Second set, just nitrogen:


There is considerable variation in the pulse shape (and the height) from shot to shot, as you can tell by the fact that the 24" trace with nitrogen and helium is slightly higher than the 24.5" trace.

Notice that there is clear evidence of a double peak in some of the traces, particularly around 27 to 26". You can compare those with these two pulses, showing the two gas mixes at the same pressure:


Another thing that I am beginning to observe (but have not yet had a chance to photograph) is that if I pulse the laser and then pulse it again a little less than a second later, the peak at the beginning of the second pulse is generally a lot stronger. In fact, if I start with a good first pulse at 27 or 27.5", the peak of the second pulse is usually off the top of the screen. It is very likely that this indicates inadequate preionization, and I may decide to grit my teeth, rip the roof off the head, and bring it back down to its original height. I dread the process of finding and fixing the leaks, but I really want to know...

(08 September, 2006)

I took off the lid and tried to remove the added spacers. This failed, so I built another lid. The main spacers are 3/16" thick, and the preionizer is on a piece of spruce that is 1/8" thick, so it is back to its original location, 1/16" above the height of the electrodes.

The preionizer is warped, but it seems to be of uniform width, so the total gap is about the same all along even if one side is wider than the other. I am seeing some bright sparks, and I was prepared to build yet another roof for this head if necessary, but the device is definitely a laser — I have already used it to pump and tune some R6G. When I get a chance, I will make some measurements on it, but first I want to find and fix any leaks I can. Even moderate quantities of air are bad for performance, and I can tell that I have to let a little more gas into the head now, before I see lasing. I’d like to get it at least as vacuum-tight as it was earlier.

(some time later)

I decided, after measuring less than 900 μjoules per pulse, that this roof was less than satisfactory. After some thought I decide to cover the entire underside of the next roof with SiC, which meant that I’d be unable to see the discharge in any case, so I made a new roof out of wood. Here’s a view of the “carborundum carpet” on the “ceiling”:

Reasonably nice smooth coating of carborundum, 7/8" wide. Sorry the photo is slightly blurry.

(afternoon of 09 September, 2006)

I have found three smallish (but clearly significant) leaks, and applied RTV to them. Tomorrow morning when I get up (or around 2 am, if I’m still awake), when the RTV on the third one is reasonably firm, I will check again to see how well the head holds vacuum. So far, I am not seeing very much change as I find and fill these things; but they have all been right near the vacuum port, so that isn’t too surprising.

(afternoon of 10 September, 2006)

It was essentially impossible to fix the leaks, partly because I had not clamped the roof onto the head when I attached it, and the weights I used were not heavy enough; the new roof is slightly warped, and it sits a wee bit up in the air ...until I pump on it. The motion was reopening the leaks, so I finally gave up, removed and reglued the vacuum port, and started the de-leaking process again this morning.

That went well, so I fingerpainted RTV over almost all of the exposed wood on the top. I just measured about 230 kW output, and that’s with a small amount of the beam obscured by some RTV on the inside of one of the windows. (See photos, below.)

(NOTE, added much later: the paintable RTV that I used for this seems to be the wrong type. If you must seal anything with silicone caulking, the kind that smells like vinegar as it cures is probably a better bet.)

There is a distinct problem about this: the RTV should not actually be in the beam path. The beam is shaped like one or two curved lines (depending on pressure and voltage), and I am convinced that this is because the faces of the electrodes still have the anodizing on them, so the discharge actually goes from upper edge to upper edge and from lower edge to lower edge. I have acquired a small diamond grindstone at the hardware store, and the next time I take the roof off this laser, I will attempt to get the anodizing off the electrode faces with it. That may not help, btw. If it increases the volume of the discharge too much, the laser may not be able to pump the gas as well, which would actually reduce the output. Mind you, I doubt that this will turn out to be the case, but it is definitely a possibility.

Here is a look at the window, so you can see the RTV, and then a look at the output. Unfortunately, I had the camera a bit far away from the target, so the 512x384 enlargement is as big as I have right now.


You can see the mountain shape all too clearly on the target. As I say, if I take off the roof again I will remove the excess RTV, but I hope it won’t actually be necessary.

If I have measured the power correctly, it should be possible to create a spark on a metal surface by focusing the beam. I will be trying that this evening, and will attempt to photograph it if I can do it.

...And, in fact, it is possible. Here’s the target, the target with a spark, and then tight crops, one from the second photo and another (even more out of focus, alas) that I took with the room lights out.


I have removed the RTV from the inside of the window, and attempted to remove the anodizing from the edges of the electrodes, with indifferent success. The head is now reassembled and the RTV is setting, so I won’t be able to do any further testing for nearly 24 hours. Such is life; when there is any news worth reporting (for example, output of significantly more than 250 kW, or if I find that I can pump a dye laser with the unfocused beam), I’ll report it, probably with photos.

(evening of 11 September, 2006)

The de-leaking process continues. I have eliminated several fairly gross holes, the latest one only a short time ago, so the RTV on it is still wet. I should be able to resume leak-testing some time tomorrow.

In the meanwhile, I should note that I will be constructing a new head for this laser. It will have a channel about 25 mm (roughly an inch) across, so I can see whether that works better than the relatively tight spacing (about 16 mm, roughly 5/8") of the current head. The initial preionization method will be essentially the same — a “carborundum carpet” on the ceiling, this time just over an inch across, and again 1/4" up.

I bought two painted rulers today, in the hope that they are not anodized, as paint is going to be a lot easier to remove from the working edges. They are now cut down to 35" length, just a bit longer than the ones in the first head. This means I have to position the mounting holes slightly differently, by 1/4", but that’s easy enough.

(27 September, 2006)

I built the new head with 1" channel spacing, which was probably a bit large for this laser. Nonetheless, it put out almost 240 kW. At that point I decided to return to the charge-transfer laser on the third page of this series (005b1.html), to find out whether I can coax better performance from it. I am also working up the “How to Build This Device” page. (See links, below.)

Rework, September-November, 2009:

(27 September, 2009)

I need a nitrogen laser for one of my projects. Because I have disassembled the one I’m currently working on so I can do a major rebuild of its head, which is complicated and will take considerable time and effort, I decided to reconstruct this one. In the process of doing so I have been having a very difficult time “de-leaking” the head, which prompts both the addendum about wood porosity above, and this second addendum.

Note: in addition to the porosity issue, it appears that the plastic fittings that I used in the initial version are not really suitable for use in vacuum systems. They perform nicely with pressure in them, but under vacuum they allow too much air into the head unless, perhaps, you operate with a large amount of helium in the gas mixture. For this rebuild I have substituted ordinary brass compression fittings, but with one difference: instead of the brass compression rings that came with the fittings or delrin rings (available at the hardware store), I am using pairs of small o-rings. This allows me to tighten the nuts by hand if I am careful. I’m not sure whether Delrin rings would permit you to tighten the nuts by hand, but they may be viable [with a wrench] if you can’t find appropriate o-rings.

The rings I’m using are 1/16" thickness, and have inside diameter just a bit smaller than the polypropylene or polyethylene tubing that I’m using for gas and vacuum, so I have to stretch them a little to get them on. I use them in pairs because a single ring does not provide an adequate seal; the nut goes all the way onto the fitting without compressing the ring. That is, the second ring is just a spacer, and you could substitute something else for it if you wanted; but it seems simpler just to buy them in pairs.

(02 October, 2009)

Meanwhile, I have found about as many leaks as I can by just dipping the ends of the head into water, and I have built a trough that should allow me to dip the entire head. When all of the caulk sets, I will give it a try.

(05 October, 2009, evening)

Having completed the trough I dipped the head into it, and discovered that in addition to a very small leak at one end and a somewhat larger leak at the other end, there were various leaks along the upper sidewall. I originally attempted to seal this sidewall with RTV, but as I mention above I used a type that can be painted. This turns out to be a mistake, and I cannot recommend it. In fact, I strongly suspect that RTV is not a truly optimal sealant for large areas of wood, and may not be a particularly good sealant for wood in general.

Last night I sealed two more wooden yardsticks with thinned epoxy, with the thought that I would use them to replace the sidewalls on the existing head. As of this afternoon, however, I have decided to put that effort on hold and make a new head first, using plastic sidewalls instead of wooden ones. (See the follow-on page for details.)


Citations for some interesting papers about nitrogen and other lasers...

To the first page in this set, a general discussion of the issues involved in designing and building a high-performance nitrogen laser

To a page about my initial effort to produce a high-performance nitrogen laser

To a page about my continuation of that effort, which resulted in a laser that puts out about 100 kW and can operate without a vacuum pump

To a “How-To” page about that laser

To an interim page about my effort to scale up a published design in order to enhance its performance

To the next page about this laser

To a page about my current (late 2006) effort to build a less-expensive laser with even better performance

Back to the Index

To the Joss Research Institute Website

To my [updated] mirror

To my current research homepage


This work was supported by
the Joss Research Institute

Contact Information:

My email address is, where a is my first name (jon, only 3 letters, no “h”), and b is joss.

My phone number is +1 240 604 4495.

Last modified: Wed May 10 14:54:38 EDT 2017