OpenRAMAN LD & TEC Drivers

Today’s post has been long-awaited and so I’m extremely glad to share it with you. In fact, it took me two full years to complete and it is a real cornerstone in the OpenRAMAN project!

Today’s post is about the low-cost LD & TEC driver that will drive the laser of the Performance Edition! A picture of the circuit board is given in Figure 1. All the files required to reproduce this board are given [∞] here. The circuit can be reproduced under 500€ ex-VAT.

Before I dig into the details of the driver, I would like to digress a little bit and give a few words on how I handled this subtask at the project management level because I believe it is a good example on how to proceed with your own experiments.

When I started working on the Performance Edition, I knew I would have to change the laser source to achieve a higher signal and higher resolution because the small CPS532 4.5 mW multimode laser pointer from Thorlabs would not be sufficient for that. Of course, it was possible to get narrow-linewidth lasers as whole units (even fiber coupled) but a 100 mW laser with 0.1 nm linewidth will already cost you about 1,500€ ex-VAT at CNI Laser (it is the best quality/price laser company I know to this day) and a 10 mW single mode laser makes the price boom above 3500€! Some manufacturers will sell you laser diodes for Raman, wavelength stabilized, at about 750€ but this is still relatively expensive and you don’t have a driver at that price. I was therefore extremely pleased to learn that Thorlabs DJ532 laser series were both cheap (<200€) and had extremely high wavelength purity. They were kind enough to test the coherence length for me: the 10 mW version had a 23 meters coherence length and the 40 mW version had an 11 meters coherence length. Translated to linewidth, this is on the order of 0.0001 nm!

The only problem with these laser diodes is that they come “naked” and you have to manage the current feed yourself as there is no driver included (no 110/230V power plug nor even 12/24V one). Also, these DPSS lasers first generate a high power NIR beam and therefore requires relatively high currents, up to 400 mA for the DJ532 40 mW version. The consequence is that they dissipate a large amount of heat, up to ~1 Watt, which makes them heat pretty quickly. The consequence of the heat-up is two-fold: first, the diode stops emitting green light when it goes out of the 20-25°C range (± some margin), second, the diode will fry if you keep powering it. So in addition to a current driver, you also have to handle the heat issue via a TEC driver.

Historically, I first tried to adapt my [»] 1 Amp LED driver to power up the laser, with some safeties added to it. I was confident it would work and ordered enough to make 5 boards but the first trial fried my LD in about 5 seconds! Laser diodes are unfortunately much more sensitive than LEDs to current spikes and overvoltage and the safeties I added were clearly not enough.

At that point I was a bit stuck between trying to fix the board or go to a different solution but none of my colleagues from the electronic department at the office were able to spot the issue which meant I would probably fry dozens of lasers before (maybe) finding the issue. I therefore chose to go slower and safer with some off the shelve approach.

I therefore bought the expensive laser driver kit from Thorlabs (>3,000€) that included everything required to operate the DJ532-40 laser safely. Everything in this kit was clearly overkill for the application but it had the advantage of offering a solution that would tell me if there was any chance that the spectrometer would achieve the required performances with the chosen LD. After all, I did not test the performances at that time and everything was based on the assumption that the DJ532-40 would be a good fit for the spectrometer.

This process is called risk mitigation in project management. It allows you to gain information about your system by assessing each risk one at a time. Usually risk mitigation trade risk against time and money because you move slower but with less risk of having the whole project failing. An important consequence that is often overlooked about all this is that risk is a leverage of effort. When you accept risk, you can skip some validation tasks and save both time and money. This concept is crucial in project management and even if it sounds obvious when stated like that, I have never heard anyone in the project management sphere discussing the fundamental nature of risk like this. In fact, I was first exposed to these notions through the reading of military textbooks where people have a much better understanding on the concept of risk, especially as a leverage of manpower (or how a clever general can lead a troop of 1,000 men to defeat a 10,000 men army).

Ultimately, I was able to validate the Performance Edition breadboard using the laser kit driver, as you already know. I then started replacing one part at a time, still in this idea of going step by step. I first replaced the large LDM56 cooler block with the tiny LDM21. Thorlabs did not recommend using the LDM21 to drive the DJ532-40 (they even confirmed me that in an e-mail) but it actually work fine if you machine some clamp for the LD diode. Be aware that it drives the LDM21 at the edge of what it can do and I had to later add an auxiliary active cooling on top of it to evacuate the heat of the hot side of the TEC element. I then swapped the expensive TEC driver by the tiny MTD415TE chip with its educational board which I posted [»] here and I also tested Thorlabs low-power current driver LD1255R which was a complete failure as explained [»] here. I was however relatively confident that the higher power version of the circuit, the LD3000R, would do the job – maybe with some additional cooling on top of it.

I was therefore feeling safe enough to make a custom circuit to use both the MTD415TE and LD3000R in conjunction with the modified LDM21 block to operate the laser. I could have decided to go one step at a time again but some simulations in Proteus and breadboarding test led me to think that I could directly go for a more final version of the board with all the safeties included. “Final Design” is by the way another interesting concept in terms of Project Management that is developed in-depth in Frederick P. Brooks Jr. book’s “The Design of Design”. In a few words, the author states that every design is an attempt at making the definitive version of the product but that it is limited by your current understanding of the project and its boundaries, therefore leading to the concept of revisions. This is off-topic but also a very interesting thing to be aware of in the context of project development cycles.

Now that I have covered the historical basis of the circuit, let’s dig into all its features!

As I previously mentioned, the circuit is based on Thorlabs’s MTD415TE TEC and LD3000R LD drivers. The TEC driver is used in its simplest form with no communication ports (they are grounded through 100K resistors). All configuration is done using the educational board as explained [»] here with a temperature window set to ±250 mK which means that the status pin will go to high state when the temperature has settled to the setpoint (22.50°C) within 0.25°C. Based on previous experimentation with the IC, once at steady state you can expect at least 10 mK stability.

A schematic of the TEC part of the circuit is given in Figure 2.

The status pin output is then mixed with a key switch, an interlock switch and a 5V remote disable input using a quad AND gate CD4082. Note that both the remote disable signal and the MTD415TE status are isolated using a dual ILD1 optocoupler. The status pin optocoupler is set as non-inverting while the remote disable pin is set as inverting which means that when a +5V signal is applied the signal will go to low state and the laser will be disabled. I’m using optocouplers to protect the laser driving part from any noise that could be injected into the power input of the laser driver. The power supply uses provided isolated +5V and ±12V supplies so I’m taking profit of this to avoid any noise from the TEC driver or the cooling fans that are powered up by the +5V as well (remember they use DC motors which are very noisy – electrically speaking).

Also, please note that this design does not include any protection on overvoltage or reverse bias of the remote disable so you should double-check that you send a +5V signal only.

The LD3000R is used in its external current configuration and a switch allows to toggle between two current values set by trimmers. Typically, I set the first one to ~330 mA for normal operation and the second one to ~170 mA for alignment. I did not had time to measure the actual optical power output yet because I do not own a powermeter and purchasing one would not be a smart expense at the moment as they are all above 1k€. In practice, I tune down the alignment setting until the laser is dimmed enough to allow comfortable alignment without safety glasses.

Warning: Never look directly into the laser, even at low power. Always wear laser safety goggles when power exceeds 1 mW. Proceed at your own risk. Purchasing a Class 3b laser might be illegal in your country without a license. Always check your local regulations first.

A schematic of the laser driving part of the circuit is given in Figure 3.

One small note concerning the LD3000R is that it requires a very clean power input and Thorlabs recommends using NO chopping power supplies. I achieved this using linear regulators 78L08 and 79L08 to create ±8V from the ±12V input. Note that -8V is actually the bare minimum required by the LD3000R and LDOs come with a 10% uncertainty on their output level. Here I was able to operate the system with a -7.6V power supply but there is no guarantee that it will always work so I would recommend using 78L09 and 78L09 ICs to produce ±9V instead of ±8V. For some reason, I was unable to procure the 9V versions whereas the 8V versions were in stock at DigiKey. Also, the LD3000R works with grounded anode laser (which is the case of the DJ532-40) so all the power is taken from the negative supply. At 0.4 Amp output (max recommended for this circuit although you should technically be able to push it to 500 mA), this means that both the LD3000R and the 79L08 will dissipate a fair amount of heat and they must be protected using heatsinks. All the tests I performed were done using active cooling on the LD3000R but experience shows that it might not be necessary. The heatsink for that element is only about 15€ so it’s up to you to decide to put it or not. The heatsink on the 79L08 is, on the other, not optional!

You may also notice that I included two diodes (D2 and D3) for surge and reverse bias protection. Although they should do their job, I did not check this so always be gentle with your laser. Also, Thorlabs recommends a maximum current of 400 mA and a maximum output power of 40 mW for the DJ532-40 diode. This means you should stop increasing current as soon as one of the two conditions are met. Since I do not own a power meter, I stay safe with a current of 330 mA which is given as typical drive current in Thorlabs datasheet. Note that for some diode this may already break the 40 mW limit!

Last but not least, the circuit also has a slow-start-fast-shutdown system independent of the current driver. A schematic is shown in Figure 4.

When the enable signal from the CB4082 is low (meaning the laser must be shut down), the current flows through the BC557 PNP transistor Q2 through the 10k resistor R8, enabling the CE junction of the transistor which fills the 22 µF capacitor C7 through the 470Ω resistor R4 relatively quickly to a voltage of about 3V which is enough to activate the N-MOSFET transistor. Since the R_on of the MOSFET is fairly low, all the current of the LD goes through the MOSFET which basically shunts the laser. This process is very fast since the characteristic time of RC is τ=10 ms. On the other hand, when the enable signal from the CB4082 goes high, the CE junction of the PNP transistor is not active anymore and the capacitor can discharge through the 33k resistor R7 until it reaches the voltage set by the ratio between the 220k and the 150k resistors R6 and R5, which is about -4.7V. The characteristic time of the RC is however this time much larger (τ=726 ms). The consequence of this mechanism is that the laser switches off pretty fast but takes a long time to switch back on. This behavior ensures that in case of a problem (e.g. interlock opens) the laser switches off as fast as possible but that it also warm up slowly as it is turned on which is a good thing when driving LD to increase their lifetime.

Note that you cannot directly use the characteristic time of the RCs to have a precise figure on the turn on/off time because the MOSFET will be activated before it reaches 3V and will be deactivated before it reaches -4.7V. The transition region should be between 0V and -3V depending on the current driven through the LD and the actual properties of the specific MOSFET you soldered. Also, you will notice a small lag between the moment you enable the laser and the moment current starts flowing into it because there is some room between the bias used in the circuit and the one that is actually required to be at the edge between the two states. You can try to reduce this headroom by changing the values of the resistor but be aware that it was designed like that to cope with change in the actual voltage of both the power planes and the LD.

During a test, I monitored the laser output using a DET36A/M photodiode from Thorlabs with my oscilloscope and triggered the acquisition as I switched off the laser using the safety key.

The results are given in Figure 5. There was an overhead of about 20 ms before the laser power started changing but the drop occurred in less than 5 ms. The total time to switch off the laser was, in this case, less than 25 ms. Although you may notice slightly different figures in your circuit, you should get roughly the same values.

Finally, there are two diodes associated to the circuit. I had to add the diode D6 in front of the PNP transistor because I noticed that my CB4082 had a non-CMOS compliant behavior and did not go close enough to V_cc to prevent the transistor from being biased (about ~0.7V). The diode adds a safety margin of about 2.2V which should leave 3V tolerance for the CMOS voltages. The red diode D1 simply indicate when the laser is active. Note that this might not be CE compliant because I believe regulations say that if the LED breaks it should prevent the laser from turning on. I will do a revision for this in the future.

I had the occasion to test the circuit for a few days/weeks now and it performs as expected. I let it ran for more than 1h at 600 mA (my mistake, I completely screwed up the laser settings) and it was still working fine despite I was not supposed to operate it at more than 400 mA! I probably strongly reduced the lifetime expectancy of my laser though… The safety mechanism works perfectly well as long as you stay within the recommended current range.

During my experimentation with the photodiode, I also noticed that the laser takes some time to stabilize. After switching it on, it typically required about 2 minutes to settle to a steady state. However, on other occasions, I had to stop the laser because it was not stabilized even after 10 minutes. I will continue experimenting with this to understand the conditions required to produce a stable output. Note that I was perfectly able to record very clean spectra of acetone even with an unstabilized diode so it should not be an issue for identifying compounds.

Typical strong laser noise you may observe is shown in Figure 6. It should become perfectly stable after a few minutes of warm-up. I’m continuing experimentation to understand this behavior.

Concerning the thermal management, I took a lot of precautions to evacuate the heat as much as possible to be able to operate the system for long periods of time. In addition to an active cooling system on the LDM21 block, I also added active cooling on the LD3000R. The 79L08 linear regulator uses passive cooling and was found to be the hottest (but still way acceptable) element in the circuit. I would strongly recommend that you pay extra care to the cooling of this element. I designed a special heatsink block for the MOSFET just to be safe even if it should not dissipate more than 300 mW of heat in normal conditions. The heatsink consists of an aluminum block with a setscrew to maintain contact between the TO92 socket and the block. A regular heatsink is then glued on top of that block but experience suggests that it might not be mandatory. Finally, I also added a small heatsink to the MTD415TE chip.

Figure 7 shows the custom TO92 heatsink made of the custom part #2020-20/1 and the heatsink APF30-30-13CB by CTS Thermal Management. The TO92 is maintained in contact with the aluminum block using an M3×5 DIN913 set screw (do not force!). Thermal paste ensures the contact between the TO92 socket and the heatsink and a few points of Scotchweld 2216 epoxy glue maintain the heatsink in contact with the block. Notice that I also put some Kapton tape below the block to avoid short-circuiting the vias on the PCBs. This was mandatory here because I machined the block myself. Although it should not be required with anodized aluminum, I strongly recommend putting Kapton in that case too.

All thermal contact points were made using NT-H1 thermal paste from NOCTUA and elements secured with a few drops (or more…) of space-grade epoxy glue. You don’t need a lot of thermal paste (just one or two drop depending on the surface is enough) but you really want to have it everywhere as a homogeneous thin film that contact all the surface that have to exchange heat. High-quality thermal paste is strongly recommended to avoid having to replace it after a while. The one I used here is made for computer CPUs so it should be good for several years!

I captured a thermal image of the circuit which is shown in Figure 8 after 1h or operation at 600 mA. The -8V linear supply was the hottest element, reaching 40°C. All other elements were below that. I measured 35°C on the TEC driver and 30°C on the power supply. I measured ambient temperature on the LD3000R IC and the MOSFET. In the picture, you can compare the heat signature of the pipes running through the floor on the left-hand side and both the FLIR camera and motor of the fan mounted on the LDM21 heatsink. The power supply is located on the right-hand side of the picture.

Concerning the assembly, it should take you about 2h to complete and requires nothing more than a soldering gun. I personally use a Weller PU 81 and I recommend that you use one that has thermal management. Be sure to clean the tip of your soldering gun using a product such as Chip Quick and scrub the extra lead on your tip regularly. Always use an ESD mat and ground yourself when handling ESD components. You will also have to solder a jumper as shown in Figure 9 to set the LD3000R to external current mode. If you don’t use an interlock, put a jumper in the header block too (as seen in the bottom left of Figure 9).

Once the assembly is finished, do not plug the LD3000R yet nor the MTD415TE. Using a good quality voltmeter (or an oscilloscope if you have one), check all the voltage pins of the MTD415TE chip that should read +5V or 0V. The pins of the LD3000R should read +8V and -8V. Check the current control pin voltage and adjust the first trimmer to read 0.650V and 0.35V on the second (you will have to flip the switch to change mode). This should enable a current of 325 mA for the first position and 175 mA for the second position. You can then place the LD3000R (shut off power when doing this) and add a jumper wire in place of the MTD415TE +5V and status pin to simulate that status is OK. Use three 1N4001 diodes in series to simulate the LD. Check the voltage at the A-K header block as you turn on and off the safety key. You should see if the current flows or not through the diodes as you turn on the key. The red LED should also toggle on and off as you do so. You can then put the MTD415TE chip in place and connect all elements including fans.

All the required operations are noted in Figure 9. Please note that the LD3000R and MTD415TE use different grounds so use the +12V ground to test the LD3000R and the 5V GND to test the MTD415TE. All the pins labeled “GND” should read 0V because they are grounded at the circuit level. The green bond represents the wire required to bypass the MTD415TE IC and simulate a valid status pin output.

The LD and TEC must be connected using SUB-D9 sockets with the wiring of Figure 10. I use 18AWG cable for power and a shielded cable to connect the thermistor of the TEC element. Note that the circuit board offers a “shield” input to connect the cable to. About 20-30 cm of cabling is way enough to connect the elements. Secure the SUB-D9 sockets on the LDM21 using screws. You can now test your laser. It is recommended to always turn the safety key to “off” position as you switch on/off the main power supply to protect the LD from transient voltages. Note that you will have to program the MTD415TE chip using the evaluation board as explained in [»] this previous post. Use a temperature window of 250 mK. Note also that the laser will turn on only when the LDM21 has reached its temperature set point which should take about 20-30 seconds in a room close to 20°C.

That is all for today! I hope you enjoyed this post as much as I did :D See you very soon for more updates on the OpenRAMAN project! In the next step, I will update the baseplate of OpenRAMAN to attach the circuit to it and that will conclude the Performance Edition!

I would like to give a big thanks to James, Daniel, Mikhail, Naif and Lilith who have supported this post through [∞] Patreon. I also take the occasion to invite you to donate through Patreon, even as little as $1. I cannot stress it more, you can really help me to post more content and make more experiments!