Paper describing the 2D spectrometer design was published in The Journal of Physical Chemistry A (DOI: 10.1021/acs.jpca.0c00285). The accepted version of the manuscript can be downloaded here, along with the supporting information.
Our Millennia Prime stopped working, tripping the circuit breakers. I’m hoping it’s just power supply issues, and it’s out of warranty, so let’s take a look. Just don’t think about how much this thing costs…
Getting to the power supply unit doesn’t look like fun… It is mounted on the bottom left, underneath a couple of layers of circuitry. At this point it’s impossible to even see what sort of unit it is.
Let’s fast forward through the disassembly and see what it is. This is somewhat awkward and time consuming, but not hard once you know how. (Follow in reverse to see how to do the disassembly.)
So it’s a commercial unit, a Vicor PFC MegaPAC.
These units are customisable, using a switch mode front-end to produce 300 V DC, which is then converted down to the required voltages using up to 8 DC-DC converter ModuPAC modules. These modules are easily removable.
The unit is set up to use two high-current 3.3 V rails to run the two laser diodes, and a single 24 V rail to run the thermoelectric cooling system and other electronics. The “B” in the model name indicates they are a “current booster” module, and are connected in parallel to the equivalent non-B module. Note that the J40 system only has a single diode, so I would expect it to only have one 3.3 V rail, and perhaps one fewer of the 24 V boosters.
So the Vicor PFC MegaPAC can be bought new, but they are pretty expensive. We’re looking at over $4000 to replace this unit as configured, with at least a 6-week wait time. The DC-DC converter modules are probably fine though, and can be swapped into another front-end… and there’s a few options on eBay. We ended up getting one shipped from the US for just over $300 (yes, about a third of that was shipping costs). The DC converters swapped over fine and a quick check with a multimeter indicates it’s working OK and no longer tripping the circuit breakers.
So here’s the reassembly (the disassembly process is just the reverse of this). Bolt the power supply back into the chassis from the side and bottom.
Attach the DC terminals and the voltage trim pot connectors (the small Molex plugs with red and black wires).
Screw in the mounting bracket for the diode driver circuitry onto the top of the power supply.
Mount the diode driver circuit boards and plug in the connectors at the back. The black foam is to seal this chamber to direct airflow through the power supply and heatsinks on the diode driver boards. Try not to destroy it during disassembly!
Connect the high-current 3.3 V lines and diodes to the driver boards. Screw in the ground connection to the power supply, and plug in the connector on the left (this is the connection to the logic board to independently switch the 3.3 V rails on or off).
Fit the rear fan assembly using the nuts on the bottom, then re-attach the back panel. The fibre optic cables are armoured, but be careful here and don’t twist or kink them.
The logic board is then fitted on top of the diode driver boards. Connect the various Molex plugs.
Note that the rear panel DB-15 connector and the black wires (bottom-right and right side of the above photo, top-right and top of below photo) are actually attached to the front panel. The DB-15 needs to be unscrewed from the rear panel during disassembly.
As an aside, you can see the single-board computer and 24 V power converter here. That cylinder on the right is a PS/2 keyboard socket. We’ll resist the urge to try hacking it…
Reattach the front panel and connectors to the power switch/power supply, LCD and indicator diodes. The red switch part of the key lock pops out of the lock assembly.
Now put on the top panel, plug it in and fire it up. Yes, it is now working perfectly again!
Here is a quick rundown on how the delay stages are calibrated to relate the rotation angle to the delay time of the laser pulse.
An interferometer is built by inserting a mirror/beamsplitter combination after the delays to interfere pairs of the four beams. A Helium-Neon (HeNe) laser is used as the light source, as the atomic emission at 632.8 nm is very well defined — a requirement to make a nice interference pattern and for use in the calibration fit equation.
Here we show the vertical configuration of the beamsplitter/mirror combination, which interferes the top and bottom beams, directing them downwards towards the table. The beam is directed to a fibre-optic which feeds the spectrograph and camera.
I think this is the first look here at some of the spectrometer software! The intensity at the 632.8 nm laser line is collected as a function of one of the rotational stage angles. The rotational stage is swept between 0° (glass perpendicular to beam) to around 45°. An interference pattern is generated.
The grey trace is the raw intensity data as a function of rotational stage angle. The yellow trace is with a Savitzky–Golay noise filter applied. The red trace, and the yellow curve in the bottom panel show the theoretical angle-to-delay relationship. As the data is collected, or when the “Fit” button pressed, the fit is performed and the three fitting parameters (α, β, γ) are determined. The fit can also be adjusted in realtime as the parameters are changed manually.
The theoretical model of the glass angle vs delay match the data very well across the entire range of rotation angles. That’s good, as it’s pretty fundamental to the whole spectrometer design!
Here’s a nice picture showing an overview of the spectrometer layout. The NOPA is in the back-left, with the compressor on the left. The spectrometer arms run down along the bottom edge, through the delay stages in the bottom centre. The beams are then directed back up to the sample and detector on the right side of the image.
This is the back end of the spectrometer arms. The sample vial is visible, as well as the mirrors and lens used to direct light from the sample into the detector. The spectrograph is an Andor Shamrock, with an Andor Newton CCD camera as the detector. This is able to capture 500 spectra per second.
These are the three pairs of rotational stages and glass used to control the arrival times of the laser pulse sequence for the 2D spectrometer. The stages are computer controlled, and can rotate with an accuracy of one-thousandth of a degree. By rotating the glass, the effective thickness increases, delaying the laser pulse with control on the attosecond time scale. A pair of glass plates are used for each beam, rotated at opposite angles to correct for the path deviation.
Here’s the layout of the pulse compressor. It consists of both grating and prism stages.
Here’s the layout of the non-colinear optical parametric amplifier (NOPA). It’s a tunable light source which can produce laser light anywhere across the visible spectrum. It’s fed by the output of our titanium-sapphire regenerative amplifier (Spitfire, Spectra Physics), which provides 100 fs duration pulses with a 1 kHz repetition rate, centred at a wavelength of 800 nm. We split that to produce a white-light continuum seed and a 400 nm pump beam. The precise angle and timing of the overlap of these beams on the nonlinear crystal (beta barium borate, BBO) will change the wavelength of the NOPA output. Currently this is generating light centred around 600 nm.
The output is strongly affected by the quality of the white-light seed. The white light is produced by focusing a small amount of the 800 nm laser onto a pure sapphire crystal. The produced white light radiates out and needs to be collected and collimated using a precisely placed curved mirror.
Here’s my Ph.D thesis, entitled Theoretical and Spectroscopic Studies of Energy and Charge Transport in Organic Semiconductors. It was obtained at the University of Adelaide under the supervision of Tak W. Kee and David M. Huang.
Patrick Tapping – Doctoral Thesis (22 MB, PDF)