Using surplus switchmode PSUs to power an argon ion laser

Warning - Switchmode power supplies use dangerous voltages, typically up to 350 Volts DC, and may retain a lethal charge on primary-side capacitors for several minutes after powering down, so take extreme care! Do not override fuses or other protective devices, as this could present a fire hazard. 

This information is provided 'as is'. Readers should only attempt to experiment with this stuff if they are competent at electronics work and understand the dangers of mains voltages. No responsibility is accepted for any damage, injury or death (however serious) or disruption of the space-time continuum caused by anything remotely connected to this website.
It is assumed that users are familiar with the potential hazards of lasers - if not, please go to  http://www.laserfaq.com first and read the safety information there.

This info is presented as ideas for experimentation, not a complete 'how-to' guide, but feel free to email me with any queries. However I have now found a UK source of these PSUs - see later.

Whilst rummaging around a scrapyard recently, I came across an interesting-looking laser head unit. The guy only wanted UKP5 for it so I thought at the very least it would make an interesting addition to my electric glassware collection.... After cleaning up the rust and water damage caused by it having been lying in the open yard for some time, and a little research on the net, I identified it as an Argon Ion laser, Cyonics/Uniphase model 2201-15SLB, and had 5000 hours run time shown on the internal indicator. Its wavelength is 488nm - a nice blue, with an output power rating of 15mW. (This is an end-of-life rating so output may be more than this - I have no way of measuring it.)  There is a huge amount of very useful information about this (and pretty much every other) type of laser at Sam's Laser FAQ, the ultimate source of everything you could possibly want to know about lasers.

The tube requires supplies of 3VAC at 25amps for the filament, and about 110VDC at 10 amps, with current regulation. The Cyonics head has an internal high-voltage igniter (starter) circuit, so this was one thing I didn't need to worry about.
I lashed up a simple capacitor-discharge circuit to test the tube in pulsed mode, and was delighted to find that it still worked! 
The 3V/25A Laser Filament supply was provided by overwinding 2.5mm² conduit wire round a 175VA toroidal transformer (left).

There are numerous power supply schematics at Sam's Laser FAQ, but many of these are aimed at USA 110V mains supplies. As I live in 240V-land, I needed to build a proper PSU, and didn't like the idea of having to use a huge, hard-to-find, hernia-inducing transformer, big smoothing caps and hot linear regulator, and decided that it would be good to be able to use a switchmode power supply (SMPS), which would be lighter, smaller and more efficient . Large SMPS's are readily available very cheaply in the surplus and Radio Rally (Hamfest) market, as they are mostly of little use out of the equipment from which they came - when was the last time you needed 5V at 80 amps ?

Unfortunately there are no mainstream applications for 110-V odd DC supplies,  so the possibility of finding a suitable supply at surplus prices is vanishingly remote. The 'ideal' PSU that you might actually stand a chance of finding would be a couple of 48V Telecoms supplies (wired in series) as these can often be tweaked up to at least 60V. Naturally, when I went PSU hunting at a recent radio rally, no such beasts were to be found. (I have previously picked up some Eltek 1.2KW telecom supplies, but at 20 amps each, these would have filled a 6U Rack, been impossible to lift and therefore considerable overkill).

What I did find, however, were some very cheap Weir WA505 SMPSs with the following specs :

Main output : +5V @ 80 amps
Aux outputs : +15V @ 10 amps, -15V @ 10 amps, -5V @ 10 amps, +24V @ 2 amps

The one pictured here is actually the one I blew up (see later), and there are a few parts missing.....

I bought three of them (UKP 25 the lot), on the basis that I could definately use all three in series, but hoped I could tweak them sufficiently to manage with just two.

Sep 2002 - WCN Supplies in the UK have new WA505's available at £30 each. This UK surplus supplier  had some used WA505's a while ago (Sep 2001) - not sure if they still have any. Martin Hopkins has a large number of WP605's for £15 (I have a couple of these - I've not tried modding them yet, but they look pretty similar, most of the differences seem to be on the mains side - I think they have power factor correction - but the exact mods may differ slightly).   I've just found these WA Series specifications

Details for the exact mods for this PSU are at the bottom of the page. These PSUs were used in some TV broadcast equipment, which seems to be occasionally surfacing on the UK surplus market - I recently bought a couple inside a couple of huge racks at a radio rally.

I was hoping to be able to crank the +/-15 volt outputs up to at least 25V, and then wire them in series with the main 5V outputs, but before we dive in, first let's look at a little bit of SMPS theory.....

A typical SMPS takes the mains voltage, rectifies and smooths it to get a high voltage DC supply (about 340V, depending on line input voltage), which it then chops at high frequency into a ferrite-cored transformer. The transformer usually has several secondary windings, the outputs of which are rectified and smoothed to provide the main and auxilliary outputs. A regulation circuit on the secondary side monitors the output voltage, and sends a feedback signal back to the primary side (via a small transformer or opto-isolator) to tell it to increase or decrease the input power as required to maintain the output voltage as current demand changes.
Most large SMPSs have one 'main' output, usually 5V, plus several auxilliary outputs, most commonly some combination of +12, -12, -5 and +24V. These auxilliary outputs usually have much lower output current ratings then the main output, and in some cases are less well regulated. The main regulation loop from the secondary to primary side of the PSU is normally on the main output, with other methods being used to regulate the auxilliary outputs, often using clever tricks like saturable inductors, magnetic amplifiers and other magnetic wizardry that I don't claim to really understand... (here is the only decent explanation of magnetic amplifiers I could find) fortunately, we don't  need to know the details of this, but there are a couple of points arising that we do need to know about :

1) In order to get anything like the full output current from the auxilliary outputs, the main output usually has to have some loading, often a significant proportion of its design load. This is because if not loaded sufficiently, it will not be telling the primary controller to shove enough power into the transformer.

2) Depending on the regulation scheme used, pulling a lot of current from an auxilliary supply when the main output is lightly loaded, especially if you've cranked the voltage up, can cause the main output to rise to the point where an overvoltage trip kicks in and shuts everything down. Fortunately I found where the overvoltage trip was being sensed, and disconnected it (The 5V output now runs at about 7.5V). High-power SMPSs usually have sense inputs, used to compensate for voltage drops over the cable, and voltage regulation is typically done from these sense inputs, but overvoltage protection will usually be sensed on the output, so if you look for connections from the main + output (as opposed to the +sense terminal) to the regulation circuitry, chances are this will be the overvoltage sense, and a simple track-hack may be all that's needed to stop this pesky critter causing trouble. Regulation and control circuits on big SMPSs are often on seperate vertically mounted sub-boards, so it's usually easy to spot the connections between them and the high-power stuff. 

In the PSUs I used, the main 5V output was floating with respect to the auxilliary outputs. This meant I could wire it in series with the aux outputs, which (a) boosted the total voltage, and (b) provided the required loading to allow the auxilliary outputs to give enough power. The only problem was that although this worked at full power (8-10 amps), it became unstable at lower currents, due to the lack of loading on the 5v output. I solved this by connecting a 3R3 loading resistor across the main output to draw a minimum current and keep everything happy down to about 5A, which was the minimum current that my laser tube would run at.

The usual arrangement for SMPS regulation is for each output to have a potential divider, which reduces the output by a ratio to give a particular reference voltage (often 1.2 or 2.5V) when the output is at the correct level. There is often a trimmer which adjusts the ratio of this divider to fine-adjust the output voltage, the exact arrangement will differ between different SMPSs.

The first thing to do was to see how far I could crank up the +/-15V outputs. They had 35v electrolytic filter capacitors, so this was the upper limit before things would get messy. By tweaking the 'adjust' trimmers, the outputs would go up to +/-20V. The adjust trimmers had  series limiting resistors (R12 in schematic), and by reducing this, I got the voltage up to +/-30V, which was what I needed, and was safely below the 35V rating of the output electrolytics. See the PDF of controller schematic for the detail of the feedback divider in the PSUs I used.

One important word of warning when messing with output voltage settings:  if the feedback path disappears (like when the resistor you tacked on falls off..), the output voltage can rise dramatically as the PSU tries to pump more and more power in to get the voltage up, and is likely at the very least to make the output electrolytics fail in a spectacularly explosive manner. Don't stick your head over the PSU when testing!!!!! Something similar happened to one of my PSUs, resulting in two ballistic capacitor cans, lots of steam, and a nasty smell for a few hours. As well as the caps, something else seemed to have suffered a major melt-down, so this PSU was scrapped - I still had two left!

OK, so now I had two PSUs, each giving up to 30-0-30 and +7.5 volts, i.e. about 130VDC total, at 10 amps. There were just two mods needed - reducing the value of the +/-15v feedback trimmer limiting resistors to increase the +/-15V outputs to +/-30V, and a track cut to disable the 5V overvoltage trip.

Having decided to aim for efficiency (and having found a really nice case into which the two PSUs would fit, but had no room for a big heatsink), I decided to use the SMPS's own regulation facility to control the tube current, instead of a linear current regulator, which would have had to burn off a lot of heat.

I connected a 0.5 ohm ballast resistor in series with the laser tube - this helps limit the peak current, and reduces the slope of the voltage/current characteristic, which should make it easier to get stability (the laser tube presents a load of about 10 ohms). It also provides a current-sense voltage of 0.5 volts per amp across the resistor. At 10 amps, this resistor dissipates 50 watts, so a metal-clad wirewound type was used, mounted in front of a fan. 

"Plan A" was a sneaky idea that didn't quite work, but would have been really neat if it had... I tried connecting the 5V output's sense inputs,  across the ballast resistor - the idea was to make the PSU adjust its output to maintain 5V across the resistor, effectively maintaining constant current. Unfortunately this gave the PSU a serious stability problem and it just oscillated.

The second approach was more sucessful. See the PDF of controller schematic.
This uses a differential amplifier (U1B) to measure the voltage across the ballast/sense resistor, and then a second amplifier (U1A) to measure the difference between the measured current and a setpoint, and uses this to adjust one of the aux output feedback ratios via Q1 to achieve current regulation. The picture shows the prototype lash-up.

The value of R1/R2 and R3/4 was chosen to reduce the common-mode voltage to below the op-amp's 5V supply voltage. A 5V supply was used as the regulation process can make the supply rail vary from 8 to 30V, and a stable supply is needed to avoid variations changing the current drive into Q1 when U1A's output is saturated. U1 needs to have rail-to-rail input and output on a 5V supply, but is otherwise non-critical.  R5/C1 filter out high-frequency noise, and C3 limits the control loop bandwidth to ensure stability.

A similar approach should also work with light sensing feedback schemes, replacing the current amplifier with a light sensing circuit.

One word of warning on laser safety - even when the curent limit is set low, there will be a momentary pulse of full power when the supply power is initially applied and the laser strikes, until the control circuit reduces the current. "Do not stare into beam with remaining eye..."

<The current regulator board fixed to the side of the PSU, and wired onto the PSU's feedback regulator sub-board. 

> The finished PSU & laser. Case is a very compact 14 x 6 x 12 inches - it's pretty tight inside but there is plenty of. airflow over the SMPS's and ballast/load resistors from the 120mm fan on the front. Note that large SMPSs like this are often rated assuming they are run with forced air cooling, so a fan will usually be necessary.

 
Some fun with the laser, some oddments from my box of interesting optical bits and a can of artificial smoke from Maplin.

Details of modifications to Weir WA505 PSUs

There appear to be two different versions of these PSUs - I have just received some which have an additional PCB (with a 10 way box-header on it) near the auxilliary output terminals. The regulator sub-board also has a different version number, but I can't see any difference that is significant for our purposes.  This makes access to the regulator sub-board more difficult, but it can be simply removed without causing any problems, and in fact it may be a good idea to remove it anyway to avoid potential smoke problems when the output voltages are cranked up. Before removing the board, the PSU did seem a bit touchy as the voltage approached +/-30V. I have updated the first picture blow so it applies to both main PCB versions (versions 998907 and 998909 - printed on silkscreen, near centre of PCB). If you find that your laser needs a higher voltage to start, you may be able to solve this using a boost circuit - see the section on this below, in the Omnichrome 543 PSU description.

Main PCB, near +5v output studs :

Track cut to disable overvoltage protection on +5V rail (Both PSUs). Just below C204.

Set +5V adjust pot to maximum (fully clockwise).




< Sub-board near output terminal block, view from output terminal block side.

Add 1K5 resistor to boost -15V rail adjust range to about -30V (Both PSUs)

Set +30V and -30V adjust pots on PSU 2 to max (fully clockwise)

Use the -30V adjust pot on PSU 1 to get the required voltage range - this acts as an absolute current limit, and will limit the peak tube current on startup, as well as fault current if the control circuit misbehaves. Set to the minimum position at which the maximum desired tube current can be achieved.


Rear of sub-board near output terminal block. you may need to bend sub-board towards terminal block for access - take care not to bend too much or you may damage pads on connector.

Add 1K5 resistor across R12 (marked 222) to boost +15v adjust range (R12 is component reference on control cct schematic). (Both PSUs)

On PSU 1 only : remove +30V trimmer P10.

Yellow dots indicate connection points on PSU 1 for the current control circuit.

Bigger laser... "More Power Igor"...!

I recently chanced upon an Omnichrome 543 laser listed amongst a load of other electronic equipment in an auction catalogue, and a 7-hour round-trip, and UKP55 later, I was the proud owner of the beast pictured below (along with a car-full of other junk that was in the same lot, much of which went straight to the tip!)..

I found the schematic of the 532 head on Sam's Faq, and this appeared to be pretty much the same as this unit. This laser, like the more common 532 needs a supply of about 350V to power the igniter, and also charge a 'boost' capacitor, which dumps some extra power into the tube immediately after ignition, to aid plasma formation. I fired it up using the crude variac-based supply I used to test my Innova tube, and after fixing a dead SCR in the igniter, and building a crude supply for the 350v boost and igniter supply, I had it running.

I established that it needed about 200V across the tube (plus the 0.5 ohm ballast resistor I used on the Cyonics PSU described above), at up to 10A. It came with tuneable optics - the HR mirror being mounted on the back of a prism assembly, allowing the wavelength to be slected from one of the 8 argon lines by adjusting the rear alignment plate. I decided that while this was nice, multiline operation would be more fun, and removed the prism mount. I and was pleased to find that the HR optic would fit right into the plate it was mounted on, and all  that needed doing was to add was a small retaining plate for it. In this mode, it puts out about 120mW.

Fortunately, since building the Cyonics PSU, I had found a .source of more of the WA505 SMPSUs,. so ordered another four. I was hoping to be able to use three of them, and started hakcing about with the output circuitry, as the mods I did for the Cyonics supply would only bring the output to about 62V per PSU. While doing this, I found this good description of how the magnetic amplifier output regulators work (near bottom of page). After some creative rearrangement of the regulator, I did actually achieve this, however shortly afterwards I got hold of a copy of the WA505 datasheet, which showed that the PSU was only rated at 500W, and that was with forced-air cooling.I decided that trying to squeeze 700W out was probably pushing things a bit, so resigned myself to using four PSUs in series to get the required voltage.

One potential problem with using this many PSUs is that they often have a fairly high inrush current - this is the current surge taken from the mains when they are first turned on, due to the charging of the reservoir capacitors. Although the datasheet claimed that they had inrush limiting, it also quoted a figure of "<40A" for this current, and I didn't have any easy way to test it, so I decided that to avoid the risk of blowing fuses at power-up, it would be best to power them up one at a time with a short delay between each one. The way this was done was to use seperate relays to switch the mains to each PSU, the coil of each relay being driven from the +24V output of the previous PSU, so PSU <n> in the sequence would only be powered up when PSU <n-1> had started up. The final power supply controls a relay that enabled the 350V supply for the igniter.

After having the laser running well on the variac supply, I initially had problems getting it to start when running on the WA505 based supply. The only difference was the open-circuit voltage applied to the tube before it strikes - with the variac supply this was about 240V (I was using about 3 ohms ballast to make the current easier to control). After some experimentation, I found that in addition to the boost given by the high-voltage boost circuit on the igniter board, it also needed about an extra 40V above the normal anode voltage supplied for a fraction of a second after ignition. A 5,000 uf capacitor charged to about 40V seemed to do the trick, with the addition of a diode to carry the main tube current after ignition. I used a schottky diode from a scrap switchmode PSU to minimise losses due to the voltage drop across the diode. The capacitor is charged by outputs from PSU1, and when the igniter is activated ( about 4 secs after power-up), a relay switches the capacitor across the diode to provide the boost reserve supply. The second diode shown in the schematic avoids a negative bias on the cap.  As I'd already got quite a way through building the case by the time I found the need for this secondary boost supply, there wasn't room for a 5000uF cap, so I used five 1000uF ones in parallel.

Below is the final overall schematic. The current control stuff is as per the Cyonics PSU described above, and all PSUs have the voltage-increase mods and 3R3 loading resistors. The PSU +/-15V adjust pots are set to produce about 50V from each PSU (including the series-connected +5V output)

I used the original secondary winding of the filament transformer to provide a 24VDC supply for the on/off control, and also fed it via a relay to a small 24/240v transformer and voltage doubler to provide the 350v boost/igniter supply. On/Off control is by a simple thyristor latching circuit.

Inside view of completed PSU. A Vero Diplomat 19" case was used, with the PSUs mounted on plates near the top and bottom covers - the top plate can be hinged up for access. When in the normal position, there is a gap of about 50mm between the PSUs, to ensure good airflow from two 120mm fans in the front panel.

Afterthoughts

When messing with lasers, you often want to do it in the dark, so an illuminated current meter is a good idea.
It may be better to connect the current meter across the sense resistor (via a suitable resistor and calibration pot) instead of the output of the current amplifier, as when the control circuit is powering down, the output can bounce about a lot, giving strange readings. This seems to be more of a problem when using sequenced power-up.

On my Omnichrome PSU, the mains input current can be up to 16A (depending on line voltage, and tube temperature) - as the standard rating of a UK mains socket is only 13A, I added a facility to switch the meter to measure AC line current, using a small current transformer.

Safety Precautions

...of course every laser setup must have suitable safety signage....   (This high-res GIF will print OK up to about A5 size)

Postscript

After getting the above PSU running, a few days later the laser began refusing to start. Warning - the following message contains material that some laser-lovers may find uncomfortable, however please be assured there is a happy ending!

After my first cry for help on the alt.lasers newsgroup, Phil Fostini wrote me this very helpful message :
"With the Omni's they tend to like to strike down the return path of the tube. This may cause some outgassing and later make the tube harder to start. Boosting the prestart voltage up may help. Also if it does restart keep it running for a while ( 5 hrs) at a 6 amp current. Short on times on tubes that have been sitting may also cause outgassing to happen. When allowed to run for some time it will clean up and you should see a change in the tube voltage. If all else fails there is the car battery trick but I strongly advise against that."

I tried increasing the boost voltage, but no joy. It appeared that the boost cap was being discharged, and removal of the optics (so it couldn't lase) and looking down the bore seemed to confirm this - a bright-ish flash but no proper discharge. The igniter strike did seem to veer off to one side, instead of being a nice bright spot in the centre of the bore.
...so I now knew that if I could get it to strike just once, it could probably be recovered.... so how to get it to strike....!
I had heard of highly risky last-ditch tactics involving car batteries from Phil and others, but decided that this was somewhat extreme, and I was not (quite) that desperate yet.
To summarise, below is the list of things I did to attempt to revive it. I don't know how many of these things helped (indeed some may well have made things worse..!) but the bottom line is that it is has now struck properly, and will be running for the next day or so until I get the courage to turn it off....

I ran it with just the filament on for a day.

I put it in the freezer overnight - I figured if it was very marginally overpressure the slight reduction by reducing the temerature might help. Anyway is was easy to do, so why not... I was slightly worried by the possible thermal stresses and condensation-induced tracking, but went ahead anyway. No joy!

I tried connecting a cold-cathode fluorescent lamp inverter to the heatsinks, to capacitively couple a high frequency HV field into the tube to help ionisation - this produced a nice glow in the tube, but no help in proper striking.

I connected the LOPT-based EHT supply from my Marx generator, and this produced a continuous glow, with about 60v across the tube when struck, decreasing as I increased the input power. I also observed that the tube voltage decreased as the filament voltage was increased.

I decided to try cranking up the filament a bit, but not yet to car-battery extremes. The only suitable supply I had to hand was the 5V output of one of the 500W switchmode PSUs I'd used for the power supply. I hooked it up and tweked up the voltage so the filament was more yellow than orange. One thing I then noticed was an intense white glow at the filament end - it appeared to be ionising across the filament! I couldn't decide if this was a good or bad thing as regards restoring tube pressure, but figured out that this is probably why ion lasers don't use DC for filaments. I had a 110V-primary Coherent filament transformer, so I hooked this up to a variac to allow the output to be increased (My mains is 240V. I didn't worry about minor details like transformer saturation..)

I disconnected the tube from the original igniter, and hooked up a microwave oven transformer to a variac, with a bridge rectifier on the output, and a 10R series ballast resistor. I also used an old Coherent igniter unit to strike the tube.

I was surprised to see that the tube struck at about 800V, without the help of the igniter. It maintained a weak glow (about 50mA) dropping out at about 500V. After cranking the filament up til it was a fairly bright yellow, the current started to increase slightly, and the striking voltage dropped. I left the filament bright for a minute or so and the striking and dropout voltages reduced significantly.

Something had clearly changed in the tube, so I decided to reconnect the original supply & see what happened. Although it didn't strike, it did now maintain a continuous, slightly unstable low-current discharge. This looked promising, so I then decided to crank the filament up again, and was delighted to see the needle of the PSU's current meter shoot up to 7 amps - BINGO!

I don't even care that it will probably take me a couple of hours to realign the optics when I put them back on!
I'm truly a happy bunny!

Many thanks to Phil Fostini and Chris Leubner for helpful advice.