13. Lasers
INSTRUMENT REPAIRS & MODIFICATIONS (IN ALPHANUMERIC ORDER)
# Action Status Problem Manufacturer Model Function
1. Repair Fixed Low power PhotoSynergy ? 532nm laser 200mW
2. Restore Ongoing No o/p Laser Export LCS-DTL-374QT 355nm laser
3. Restore FIXED PSU blew ? ? HeNe ???nm
1. PhotoSynergy DPSS 532nm laser (2021)
Manufacturer: PhotoSynergy, UK
Model: S500
Type: YV04 810nm > 1064nm + 532nm
Mode: CW
Wavelength: 532nm (810nm, 1064nm, 532nm)
Output power: 532nm ≤500mW
Polarity: unknown
REPAIRED 11/05/21 LIBS6 P.17
I bought this because it was only £40. I did not expect it to work and if it did, I expected low power and hoped I could repair it which proved to be the case.
The top of the laser says 500mW max but lasers never reach the power stated on the label and there is no documentation to say what the power level should be.
New text box
New text box
New text box
New text box
New text box
New text box
New text box
New text box
New text box
New text box
New text box
New text box
New text box
New text box
New text box
New text box
New text box
New text box
2. Russian DPSS 355nm laser (2024)
Manufacturer: Laser Export, Russia
Model: LCS-DTL-374QT
Type: YV04 810nm > 1064nm + 532nm > 355nm
Mode: Q-switched
Wavelength: 355nm (810nm, 1064nm, 532nm)
Output power: 355nm 30µJ 10ns / 355nm avg 50mW / 810nm, 532nm, 1064nm avg 420mW
Polarity: unknown
INITIAL REPAIR 04/03/24 LIBS7 P.171
I did not expect this to work, assuming it to be worn out or even just plain broken. It wasn't very expensive and I bought it out of curiosity as it is rare to find a 355nm source.
When it arrived there was a gaping hole in the back panel that I'd seen in the auction photos but could not fully make out. Something big was loose inside the box. This turned out to be a Pentium PC PSU that had once been GLUED to the case bottom using what looked like black potting compound.
The entire PSU cable loom was loose out in view, with just a few connections made with Molex connectors whose wire crimps were crudely attached with bare strands of wire visible between the crimp and the wire insulation, possibly torn from the weight of the loose PSU. Unused PSU wires were cropped and hanging around with heatshrink covering the bare ends, some of which was loose. The PSU mains lead was chopped off just inside the case. Immediately it appeared the unit had been hacked.
I traced the PSU wiring and found +3.3V, +5V and +12V leads plugged into a separate Russian pcb using Molex connectors. This pcb contains power supply filters and drivers for two LDs (Laser Diodes) and two TECs (Thermo Electric Coolers) on the laser heatsink assembly.
The PC Power Enable line and multiple supplies from the PSU plug into a front panel which has an Atmel AVR AT90S6535 8-bit µC: https://ww1.microchip.com/downloads/en/DeviceDoc/doc1041.pdf
The laser section is on top of a big heatsink cooled by two 12V fans, but no lead was present to power them. I connected a 12V mains adaptor to the laser heatsink fans and found they worked.
A permanent umbilicus attaches the controller to the laser, including an SMA RF coax cable which is usually indicative of a Q-switch, as was a button on the front panel labelled 'PUMPING ON/OFF'. The other end of the RF cable mates with a SMA connector halfway down the laser body.
On top of the lid is an orange LED that seems to be on regardless of output wavelength. The controller displays an error if the lid is absent, i.e. LED connector not mated, see below for its exact location:
Having paid just £45 for it together with 3 more items of which 2 worked: an analogue signal to 2-channel scope buffer and a transistor driven mains relay box, I really had nothing to lose by seeing if it would do anything. I repaired the bad connections and with the lid off, propped up the PSU, attached an RCD and applied mains. To my surprise it appeared to function with responsive front panel controls.
Depressing 'PUMPING ON/OFF' illuminated its orange LED and '0.30' flashed on and off the LED display when set to mW. I fed the laser into a Laser Precision Rjp-735 energy probe coupled to a RJ-7610 energy radiometer but no laser output was evident. When I fed the output into a Rkp-575 power probe coupled to a Rk-5710 power meter it indicated 10mW.
Intrigued, I measured the wavelengths with my Stellarnet spectrometer, first through a B-cube attenuator and then with just a cosine collector lens on the FO offset to the beam. It detected a significant peak at 808nm, a tiny peak at 1064nm and no 532nm or 355nm. Having a 16-bit detector, the spectrometer has a maximum amplitude of 65k units. With the 808nm peak maximised at 64k, the 1064nm peak was around 30 times smaller at ~2k.
Unhappy with the loose PSU and its mess of wires, I decided to secure it to the case before investigating further. First I removed the black potting compound from the PSU and case. I then opened the PSU & rotated its mains plug so I could rotate the PSU 180° to hide its unused wires inside the case wall cavity. I didn't have a 90° plug so I repaired the chopped mains lead too.
The rear panel is secured to the extruded case sides using 4 countersunk screws which are obscured by the plastic fascia covering the panel. A previous owner had pulled this up at the corners to expose them, clearly to open the box. I drilled access holes for the screws and Superglued the fascia back down.
The case has a thick aluminium plate below the PSU with a square hole cut in it next to a case vent that I assume was for the original PSU. The PC PSU had been glued to this base but the PSU fan outlet didn't fully align with the hole. I now corrected this by fabricating a 90° bracket to secure the PSU over it, tapping a couple of holes in the false aluminium base to secure the bracket.
While I was at it, I replaced the needlessly long yellow safety interlock lead with a wire link.
Excellent teardown of similar Laser Export LCM-DTL-347QT:
https://krazerlasers.com/lasers/347QT/
The above site includes an explanation of operation and indicates the presence of an AOM (Acousto-Optic Modulator) based Q-switch inside the laser. This is driven by the RF output from a small aluminium enclosure to the rear left of the PSU, from which an SMA connector protrudes at the rear of the case.
I traced the wiring to it, opened its lid and looked inside but could see no indication of damage.
Power to the AOM driver comes from the PSU +3.3V line via the driver pcb independent of all other circuitry as it just provides 2 LF 100µF 16V electrolytics and 2 accompanying HF MLC caps underneath.
.jpg)
3.3V seemed a bit low so I looked on the web for an idea of the driving voltage for an AOM but could not find any, probably because companies that make them don't want you to know how to do it.
A wire from the front panel µC leads into the AOM driver with a 5V 10kHz pulse train on it consisting of a 2µs wide logic 0 negative pulse occuring every 100µs, tying up with the kHz measurement display on the front panel, which can be altered with the parameter value MIN/MAX buttons that seem to be associated with the selected display: mW, µJ, kHz. Power (and frequency) is modified by altering the PWM distance between the 2µs pulses. The power settings were by default set to max.
Following this signal through the AOM driver pcb, on the far RHS of the AOM pcb is a metal can that houses a 28.67MHz crystal, and the RF circuitry somehow increases this to produce the 78.8MHz signal which is turned off during the 2µs logic 0 of the 10kHz input. The 78.8MHz signal feeds into the input of the large Mitsubishi M76643H RF AMP IC at the top of the pcb. The datasheet at the bottom of this section is for the M76643L which covers 68-81MHz; the H variant here is identical but covers 77-88MHz.
The output of the RF AMP goes to the SMA connector that leads to the laser AOM cell. There is also a resistor/inductor feedback at this point, which is why when the RF cable is absent the front panel indicates this with an error number.
Below, tracing the signals (the output is driving an external 14W 50Ω RF load on the SMA connector):
Left, blue probe is 10kHz signal and yellow probe is RF amp pin 1 input (78.8MHz switched by 10kHz).
Right, yellow probe is RF amp pin 1 input and blue probe is RF amp pin 5 output (no signal).
Tracing signals on the M76643 RF AMP IC indicated pin 3: output amp bias was 0V. On the pcb this comes from a SOT-89 to the left of the white label **'б.3.09/14 051', see photo.
**The '6' is actually Russian letter 'be':
The SOT-89 is labelled 'F5 32' but no SMD index I have [E42] lists a device with a pinout that matches the pcb layout wiring. The most obvious candidate is a voltage regulator but I could not find one with this SMD label.
Assuming the SOT-89 to be faulty I removed it using low melting point ChipQik solder and hooked up an Amrel PPS1202 programmable PSU in its place. With the +3.3V supply applied as before I slowly increased the voltage on the RF AMP bias pin and a signal appeared on its output that mimiced its input, indicating the RF AMP was functional.
Only now did it occur to me to try the removed SOT-89 on my Peak Atlas DCA Pro component analyser which revealed to be a fully operational bog standard +5V regulator!
The only logical conclusion is the amateurishly added Pentium PSU was of course not the original fitted to the laser controller. I assume the original PSU became faulty at some point and whoever previously owned it was unable to determine the voltage that fed into the AOM, so cautiously wired in the lowest available voltage. The other instruments originated at the University of St Andrews in Fife, Scotland, and Google AI found it mentioned in their laser inventory together with a basic spec:
'Laser2000 LCS-DTL-374QT 50486 Nd+SHG+THG pulsed 1064nm Avg 420mW 14kHz 10ns'
https://www.st-andrews.ac.uk/~pic/LaserSafety/RegisteredLasers.pdf
Why they didn't use the +5V supply when a 10kHz 5V signal was evident, is beyond me. Nobody powers logic with a voltage lower than its input signal as this is often met with that device's demise. Fortunately the RF circuitry is analogue and no damage resulted.
In fact, the M76643 is designed to run off +12.5V and is rated to 7W at 78.88MHz.
The DCA Pro indicated a dropout voltage of 1.54V at its test current of 3.56mA, meaning to obtain a +5V output it would require 5V + 1.54V = +6.6V minimum.
Clearly a +5V regulator would not run off a +5V supply, let alone a 3.3V one. I put the SOT-89 back in and substituted the +3.3V supply into the driver pcb with the Amrel PSU, starting at +7V and increasing to +12V, and at some point a green dot appeared on the output of the laser, indicating the presence of 532nm. With a +12V supply the AOM driver output signal was now ~76Vpp.
I replaced the +3.3V that fed into the driver pcb with +12V instead, salvaging the original connector using new crimps. I also twisted all used PSU power leads with their grounds to reduce noise.
Powering up, the laser again produced a green dot. I hooked up the Stellarnet spectrometer and found the 1064nm peak was now 15x bigger at 44k vs 58k for the 808nm (slightly more at 806nm but the latter the same intensity as before: I had to align the spectrometer FO afresh). The intensity of the 532nm was 682 units, 66 times smaller than the 1064nm. The actual values are lower than these because as with my original measurements, I forgot to align the noise floor to zero. There is still no 355nm but this is generated from an additional NLO (1064nm + 532nm > 355nm), which significantly reduces power.
Given the indicated 355nm power of 30µJ is for a 10ns pulse spaced 100µs apart and already knowing the overall 10mW power I measured was far below the maximum average power of 420mW on the label, I wondered if the 355nm was there but the spectrometer simply could not see it.
I fitted a VIS grating to a Bausch & Lomb high intensity monochromator on the laser output and selected 532nm, then fed the B&L output into the PhotonView 350nm-1550nm image intensifier I had bought for a bargain $75 [see Detectors & Visualisers]. This displayed an intense dot in the centre at 532nm confirming laser alignment. I swapped in a UV grating and set it to 355nm. This time I saw a line with a less intense dot on it, confirming 355nm presence.
I bought the laser in 2023 and tested it in 2024. The only information I could find on it was the teardown site (the owner does not respond to enquiries and his links to information on it don't work). Later I found manuals for the LCS version (mine) and LCM version described in the teardown and Sam's Laser FAQ: https://www.repairfaq.org/sam/laserscl.htm#sclle1
A major difference is the LCM (M = Module?) PSU runs off 24V whereas the LCS (S = Standalone?) PSU runs off mains. The LCM also appears to have more intelligence such as an RS-232 link, and its laser LED provides BITE information, e.g. supply voltage status. Internally my LCS runs off 12V. I suspect internally the LCM also runs off a lower supply. However this should not affect operation since the source is a couple of laser diodes that likely run off no more than 2 Volts each.
A week after I got it going I powered it up again and found it no longer produced 1064nm or 532nm and Indicated power is now always zero. Initially the controller indicated it was READY but would not confirm PUMPING. The next day it decided it would PUMP after all. I confirmed the presence of the AOM drive signal but I could not determine any obvious reason for cessation of the harmonics. Maybe the 76Vpp signal fried the AOM? (I doubt it as pockels are similar and they take kV)
CW power is now zero; 1064nm and 532nm now absent. Below right, the 10kHz pulsed 78.8MHz AOM drive
signal is still present (MSO8k top blue Ch2 is SYNC signal from controller front panel, and purple Ch3 is 1MΩ input picking the signal off a T adaptor screwed into the RF SMA connector at the laser).
FOUR TRIMPOTS
Three of the trimpots at the end of the laser are annotated by green marker pen symbols I, II and T.
Powering it up again with PUMPING indicated, I confirmed 1/8th CW rotation of trimpot I increased the rear LD voltage by ~10mV and likewise trimpot II changed the voltage across the side LD. I suspect the voltage changes are a secondary effect and the trimpots in fact set the PD constant current limits.
I measured rear LD I = 1.727Vf and side LD II = 1.682Vf (HP3478A DMM).
I strongly suspect the trimpot marked T sets TEC Temperature.
There is a PD at the end of the optical assembly into which a small proportion of the output beam is reflected (see photo bottom right). This may be nothing more than a simple check for an optical output.
Alternatively it may be the sensor for an automatic power control (APC) feedback loop and the unmarked trimpot at the far end may calibrate its amplification. If so, the PD feedback signal likely feeds into a comparator to ensure LD output power is automatically regulated. However if this is the case it would render trimpots I, II non-functional, as well as disrupt any attempt at LD replacement.
I believe the LDs are worn out and need replacing (Sam hints they may be 5W emitters). Since I had nothing to lose given the useless power output levels even when it was working, I detached the side LD as I was aching to see what it looked like, see below. The LDs appear to be custom bars with 6 gold contact wires (only 4 are visible in the photos) and there is no chance of getting exact replacements.
WHAT IS PRINTED ON THE LABEL AROUND THE LD BODY?
Below: LDs are on PCBs LD pots & measured Vf Side LD removed LD bar has 6 gold wires
Below: LD focused by lens RAP cube combines LDs NPN transistors below the pots may drive the LDs?
Below, at the emission end, part of the beam is reflected onto a large square photodetector on a pcb. This may be a simple fault detector to check there is a laser output, but it may also be a feedback sensor to regulate LD power instead of a conventional constant current driver.
2026: Google AI revealed my blunder: the RF module is specified for a 12.5V supply, but the AOM should not have had more than ~20Vpp:
How a 12V Supply Produced 76Vpp:
In a Class C or similar resonant RF power amplifier the output transistors act like high-speed electronic switches, slicing the dc supply current into pulses timed to match the operating frequency.
The collector or drain of the output transistor is tied to the dc supply through an RF choke. When the transistor rapidly switches off, the magnetic field in the inductor collapses, forcing a large voltage spike often more than twice the dc supply voltage 24 to 28Vpk at the collector.
The module features an internal LC matching network that transforms the low output impedance of the transistor up to a standard 50Ω load, acting like an RF voltage step-up transformer.
A 76Vpp sine wave equates to 38Vpk (Vpp/2): Power = Vpk²/2x R = 14.4W.
The M67743H RF amplifier output power is rated 7W nominal, 10W absolute maximum. Nominal supply is 12.5V, 15V abs.max; 14.4W likely occurred because its incoming RF drive signal attenuation was too low.
This may explain the white unknown pot like object below left of the RF amp which I suspect attenuates RFVin and thus RFVout, and should be set with a 50Ω dummy load before connecting the RF to the AOM.
Acousto-Optic Modulators are current-driven and power-driven devices rather than voltage-driven ones, relying on RF power to create acoustic strain waves inside a typically Lithium Niobate optical crystal.
Small AO Q-switches typically require 1W to 5W of RF power to reach maximum diffraction efficiency (saturation).
76Vpp (14.4W) would have rapidly overheated the AOM crystal, likely cracking it and blocking the optical path inside the cavity, halting diffraction of the internal laser beam to generate the necessary
Q-switched pulses, without which the nonlinear harmonic crystals cannot convert the fundamental infrared light into 532nm green light because harmonic crystals need 5 to 10W minimum to convert (1064nm to 808nm Nd:YVO₄ typically only needs 45mV to 300mW).
If 12V produced 76Vpp (14.44W), which Vpp would produce 1W, 3W and 5W?
_____
Vpp = 20 x √ power
1W = 20 x √1 = 20.00Vpp
3W = 20 x √3 = 34.64Vpp
5W = 20 x √5 = 44.72Vpp
8.12W = 20 x √8.12 = 57.00Vpp (from below)
14.44W = 20 x √14.4 = 76.00Vpp (the overload)
What is the likely original supply voltage Vcc to RF amp?
In a saturated RF amplifier stage, the output power scales with the square of the supply voltage
P ∝ Vcc². The current scaling constant k = P/Vcc² = 14.44W/12V² ~ 0.1003
Vcc = √(Power/0.1003)
1W = √(1/0.1003) = 3.16Vdc
3W = √(3/0.1003) = 5.47Vdc
5W = √(5/0.1003) = 7.06Vdc
8.12W = √(8.12/0.1003) = 9.00Vdc
14.44W = √(14.44/0.1003) = 12.00Vdc
78L05 regulator Vdropout is 1.7 to 2V, what would it be at 5V on typical M6774H pin 4 bias current?
Mitsubishi M67743H pin 4 Vgg bias current would typically be 5mA to 20mA.
At this current 100mA 78L05 Vdropout drops to ~1.2V to 1.4V; 5V + 1.4V = 6.4V.
A minimum supply of 7.0V would limit RF power to 5W
A maximum supply of 9.0V would limit RF power to ~8W
To maintain a high Supply Voltage Rejection (SVR) and eliminate HF input ripple, the 78L05 datasheet states line regulation is optimal when the input rests between 8.5V and 20V. This suggests the original supply was 9V, keeping the bias supply completely clean and quiet.
A 9V supply rail provides a perfect 2.6V of headroom for clean 5V regulation and allows the M67743H to output a solid 8.12W without exceeding its 10W absolute maximum.
Originally when it still worked, I thought of replacing the worn out 5W LDs with cheap TO-5 1W LDs from eBay China. Google AI now explained why this laser has a combined LD power of 10W.
1W 808nm LDs would be converted to 1064nm by the first stage Nd:YVO₄ but this would not be converted to 532nm (KTP) or 355nm (LBO) by the secondary stage KTP/NLO NLOs because Nd:YVO₄ has a low conversion threshold of 45mW to 300mW, but KTP and LBO have a much higher threshold of 5W to 10W:
At low powers, KTP and LBO crystals exhibit an incredibly low conversion efficiency (often much less than 0.1%).
I asked why it is possible to get a 1mW 532nm green laser pointer. The reason is the stacked NLOs have mirrors between them that massively amplify the signal before release to the next stage, whereas the Russian laser LD system is a single pass extracavity with no mirrors:
Reducing two parallel 5W LDs to two parallel 1W Chinese TO-5 LDs drops the raw optical power by a factor of 5. Nonlinear frequency doubling efficiency scales with the square of the fundamental power (Pgreen ∝ Pir²), dropping the input power by 5 times causing green conversion efficiency to drop by 25 times: photons passing through the KTP/LBO are simply too sparse to convert, and the efficiency collapses entirely to zero. Even with a working AOM Q-switch, there is insufficient power.
Two 5W pump diodes deliver 10W into the Nd:YVO₄. During the fractions of a millisecond that the AOM is closed, a vast mountain of energy is stored. When the AOM opens, it unleashes a peak power of ~4kW, meeting the KTP and LBO harmonic NLO threshold:
Peak power = Average power / (repetition rate x pulse width) = **420mW / (10kHz x **10ns) = 4.2kW
**From page 16 of University of St Andrews laser register, see pdf at bottom of this section.
With only 2W of total pump power, the resulting pulse width widens and the compressed peak power collapses:
When total pump power drops from 10W to 2W overall power reduces far greater than a factor of 5, directly into the inefficient region just above the Nd:YVO₄ crystal lasing threshold, consuming ~300mW of 808nm pump power just to overcome internal losses, clear thermal hurdles, and begin lasing at 1064nm.
With a 10W pump this leaves 9.7W of useable power generating high-efficiency 1064nm IR. With a 2W pump, only 1.7W is left and the Nd:YVO₄ crystal cannot accumulate a massive population inversion while the AOM Q-switch is closed. In a Q-switched cavity, if the pump power is weak, the giant pulse takes a very long time to build up and exit. Instead of a sharp, intense 10ns spike, the pulse stretches out into a wide 100ns pulse and the compressed peak pulse power drops down to roughly 10 to 15W.
Furthermore any drop in fundamental power undergoes a multi-stage compounding penalty:
Stage 1 (IR fundamental):
Dropping from 10W to 2W total pump power reduces the 1064nm fundamental beam by roughly 5 times.
Stage 2 (Green KTP Doubling):
Because green conversion efficiency scales with the square of the infrared power (Pgreen ∝ Pir²), dropping the infrared by 5x drops your green pulse generation efficiency by 25 times (5² = 25).
Stage 3 (UV LBO Tripling):
To get the final 355nm ultraviolet beam, the LBO crystal must physically mix the leftover 1064nm IR photons with the 532nm green photons. SFM efficiency scales with the product of both beams:
Puv ∝ Pir x Pgreen. A 5x drop in IR combined with a 25x drop in green causes the final UV conversion efficiency to plummet by a factor of 125.
Even with a brand new, pristine AOM running at a perfect 9V RF drive, swapping in 1W pump diodes would cause the final NLO stages to yield absolute mathematical zero. The photons would be too physically sparse inside the crystal lattice to combine.
SUMMARY
Obviously I'm embarrassed I wrecked it but I've learned a lot. It is highly likely the AOM is fried. If not, replacement LDs need to be no less than 5W.
In addition, the AOM RF module supply needs to be reduced to 9V and the ML7743H RF input signal needs to be reduced, more than likely by adjusting the large round component to its lower left. Lastly, if replacement 5W LDs are fitted, their current must be adjusted with any APC PD feedback system disabled. Further work is needed to confirm its presence and operation, see comments above.
Below: photos from Sam's website. The similar looking driver pcb suggests the manual's warning that only the original PSU be used is actually saying 'make sure this filter pcb is present', as the original PSU has to be a switcher to be small enough to fit under it (because that's what it looks like to me), so there is likely nothing wrong with using a P4 PC PSU in its place (overkill, in fact).
Right, photo of filter pcb from krazerlasers teardown webpage:
text
3. Hene lasers & PSUs x3 AS IS (1 dead hene)
New text box
µ Ω ± ° ⌠ ⌡ ∫ │ ─ √ φ θ Θ ∂ δ ζ ξ ς λ ψ ω τ µ Ω ∆ Δ ∑ ∏ π Ξ ○ ∝ ≠ ³ ² ±