08/30/2020 at 21:16 •
Having previously decided which, and how many, LEDS we are going to use. It is now time to design a driver to power them. This post goes through the calculations needed to design the circuit, the following post gives some recommendations on designing the PCB.
Here is the final product.
Choosing the main driver IC
According to our calculations, to achieve max brightness we need a driver capable of providing 1Amp @ 26.4V (the combined forward voltage of 6 ls). Also, since our power source is a 3S Lipo battery, we need a BOOST driver, as the voltage provided by the battery 11.1v will need to be stepped up to the voltage required by the LEDs.
We chose the AL8553 because it has an external MOSFET, It's small, inexpensive, has good integrated over-current and over-voltage protections, supports dimming control through PWM, and comes in a leaded IC package (important for ease of soldering).
Designing the Circuit
Designing switching converters is hard. Thankfully, the datasheet has all the instructions and equations required. Now, we only need to calculate the value of the components in the example schematics.
We are going to number the equations with the same numbers as the datasheet, so you can follow along if you want.
This is the main switching inductor of the circuit. Here's we need to calculate its Inductance and the Maximum Current it will handle. Here are the variables we know:
- Output Voltage: 26.4V
- Led Current: 900mA (To run the Leds at a little less than maximum power)
- Input Voltage: 11.1V (Nominal of a 3s Lipo battery)
- Efficiency η: 0.85 (Standard efficiency of BOOST drivers)
- Ripple Current Rate γ: 0.5 (Datasheet, p.10 suggest between 0.3 and 0.5)
- Switching Frequency f: 120Khz (According to datasheet p.4 table)
First we calculate the average current through the inductor:
Then the peak-to-peak variation of that current.
Now we know the maximum peak current that the inductor will need to withstand at any given point (And we can oversize it by 20% just to be sure):
Finally, we can calculate the inductance value like so:
Ok, so we know the inductor needs to be L1 = 43uH @ 3.8 Amps.
Looking through Mouser, I found this one that I liked SRP1265A-470M (Link to mouser listing) from Bourns it's the exact same inductance value, but 47uH is close enough..
This resistor controls how much current goes to the LEDs. And, assuming a LED current of 900mA can be calculated with the following formula:
This resistor is in the path of the led current, so it's mportant to double check his power dissipation.
Therefore R6 = 0.222ohm @ 180mW
or if you prefer to run the LEDs at 1Amp, R6 = 0.2ohm @ 200mW
This resistor controls the over-current protection trigger. The datasheet recommends setting this trigger point using 30% more than the maximum current expected by the circuit. Which happens to be the maximum peak current of L1, at the lowest expected Vin: 9v (The lowest safe voltage of a 3s lipo battery).
Using the equations from L1, we can calculate this peak current.
Finally, we over-size it by 30% to find the Over-Current Protection Current.
And now R3 is equal to:
We can also calculate the power consumed, just as we did with R6. Which means
R3 = 60m @ 0.54W
That's a lot of heat for an SMD resistor, So it is important to size it appropriatly. I found the WSLT2010R0600FEB18 to be adequate for the job.
R4 and R5:
This resistors are used in a voltage divider that sets the trigger for the over-voltage protection. Which is good to have in case the LEDs get accidentally disconnected during operation and leave the driver with an open-circuit. Mercifully, they have a fairly easy equation.
First, the datasheets suggests we define the Over-voltage threshold as 20% more than the nominal output voltage.
Now, the Relationship between R5 and R4 is given by.
Looking for a combination of standad value resistors that satisfies this ratio, I found a good pair with:
- R4 = 150K
- R5 = 10k
It is not an exact match, but it is close enough. These resistors don't need to dissipate a lot of heat, so it's critical which ones we choose. I liked these ones:
This is the main switching MOSFET. It only has a handful of recommendations on over specifying the voltage and current withstood by the transistor, namely
- Vds = 38.4V (20% more than over-voltage threshold)
- Ids = 4.63Amp (20% more than the nominal peak current)
- Vgsth < 13V (The datasheet p.4 says the max gate trigger voltage the chip can generate is 13V)
I searched around on Mouser and found the DMTH43M8LK3Q-13, which fits nicely with these recommendations.
MOSFET chosen for Q1, picture from the Mouser Listing
This diode also only has a handful of recommendations. It should withstand the same voltage as the MOSFET and as much current as the LEDs. Also, I personally recommend a Schottky diode, because the less the Forward voltage of D1, the less heat it will need to dissipate, and the more efficient the circuit.
Here is a good candidate: PMEG045V100EPDAZ, it has a nice Vf: 0.3V, which was about as low as I could find for this forward current rating.
This resistor controls the Slope Compensation features of the AL8853. However, we are not using this feature in this design, so this resistor is not necessary.
R2 = 0ohm.
This is the output capacitor of the driver, it helps stabilize the output current. There are actually no suggestions of how to choose it. So, I took some inspiraton from some Texas Instrument Application Notes of high power LED drivers, and set it to:
C3 = 330uF @ 50V
I suggest this one UCM1H331MNL1GS. It has a high enough capacitance, and can safely withstand the output voltage.
This is the Input decoupling capacitor, Also has no indications of how to select it. In this case we decide to use 2 capacitors in parallel:
- C1 = 4.7uF
- C4 = 100nF
Funnily enough there is no mention anywhere of why this capacitor exists, or what it does. So I just chose this value and hoped it would work.
C2 = 100nF
I suggest using the same as C4.
Complete circuit schematic
Finally, we can put all this together, to get the full schematic
08/26/2020 at 16:32 •
After a lot of tests and fails with UVA joint V1, we have developed a second version that solves the weaknesses of the first attempt. The main fail with the V1 was in the snap-fit cantilever connection in the legs ring (UVA-MJ-LV1_1b). This is the part of the joint that will be attached to the tool.
As you can see in the pictures it failed because of the lack of stiffness in the legs. The vertical force practised by the bending piece "1a" in the snaps ring at the moment of snapping also bent the legs in the legs ring. This part wasn't either design or prepare to bend. Because of the way that FDM 3d printed pieces work, the maximum flex and bend resistance is achieved parallel to the layers. In this case, the force applied to the leg was perpendicular to the layers and, after some attempts to connect both rings, the piece failed and break in the joint between the plastic layers.
In the UVA Joint V2 (UVA-MJ-V2) we increased the area of the legs to achieve the necessary stiffness to prevent bending and also to increase the surface of contact between the layers. With the new solution came some new problems, mainly related to the printing process. The leg parts both snap-fit and locks have to fit in the snap parts with not too much tolerance to guarantee a good joint. So the termination of those has to be very precise and it is necessary to avoid supports to print the legs. As you can see, we left some holes over the legs to guarantee a qualitative termination.
Furthermore, we already took the next step and designed the electrical connection in the legs ring, and we are really close to finished the battery pack design but it will be discussed in another log.
08/26/2020 at 13:07 •
Thanks to Hackaday and the community vote for awarding us with the bootstrap money. One of our investments in the project was to buy a 3d printer. The reason for this decision was based on the fact that we needed to experiment with the 3d printing process. It was necessary to take full control of all the steps of the 3d printing because we have to improve the design in its final stage but also in the fabrication process.
With the 3d printer, we can optimize the use of plastic, temperatures, the resistance of materials, and we have the tool to create a detailed manual of how to produce UVA.
We buy a Creality Ender 3, it's one of the best in relation to price and quality. If we are designing a product to be built in adverse situations, It is interesting to design with a low-cost machine. We know that if our pieces are printable with good quality in Ender 3, it will be possible to print it in better and more expensive machines.
We also made a lot of prints to understood the limits of the machine, the tolerances and the capability to print. Here some important conclusions about the printing process with an Ender 3.
- If you want to print a piece that has to fit really tight, between 0.2 and 0.3 mm of tolerance is ok
- If the fit has to pass through and be gentle and soft we recommend 0.4mm or more
- It is important to be familiar with the tool, try and fail a lot, try to find the limits
We also find these useful resources about the capability and configuration of the Ender 3 and Cura:
07/18/2020 at 12:33 •
To achieve a multi-tool kit it is necessary to understand the parts of it. This system is composed of two main elements: The battery pack and the Instrument or tool (in this case the ultraviolet light for curing adhesives).
It's a simple idea based on one challenge, the joint. This is also the most important mechanical part of the design because it is the starting point for the development of different tool-heads. The UVA joint has to achieve 4 principles:
- Simplicity (to encourage people to develop new tools)
- Durability (It has to resist the wear of daily connection/disconnection and also the falls and shocks)
- Reliability (the user should be confident that the electrical and mechanical joint wouldn't fail and the tool will work as it is expected)
- Replaceability (The modular design of the UVA system allow to replace the chassis where the joint is build in without affecting the electronics).
With these bases in mind, we design a rotational joint that is structured by three different connections: the snap-fit cantilever, the electrical contact and the Locks. It is necessary to leave some tolerance (in our case 0.4mm) because FDM prints are not 100% accurate. Calibration, environmental conditions and human interaction could affect the quality of the finished pieces. We need a tight fit because we have to prevent movement between the battery pack and the tool, but also we need to prevent the plastic to fail, crack or to not fit at all.
To fit the part it is necessary to make a 28 degrees rotation in the longitudinal axis. It will snap by this rotation and do not return to the starting point without some force. Every connection has its mission to constraint the parts together.
- Snap-fit cantilever: this part is the most complex to design and is the one that ensures that the UVA joint will keep together the battery pack and the tool. It makes a restriction in the rotation after the snap. This rotation restriction could be broken after applying some force to unlock the uva joint.
- Locks: This connection is responsible for avoiding the movement in the vertical axis. The locks will help to reduce or eliminate the forces over the snap-fit cantilever after the snap, increasing the quality of the joint and the durability of the connections.
- Electrical connection: This part is under development. It will be the mechanism responsible to transmit the electricity between the battery pack and the tool.
The joint was modeled with Autodesk Fusion 360, here the link to the model: https://a360.co/2OywktE
We are going to be using this software because it is free and tremendously good for mechanical design. It can also be use to simulate electrical circuits.
Here some information that we used to design the UVA joint:
Snap-fit design calculator:
friction coefficient: 0.492
Secant modulus: 3300 Mpa
Allowable material strain: 3%
07/05/2020 at 11:58 •
LEDs come in all manners of sizes, colors and brightness. To be able to choose the most appropriate ones, we need to know what are they going to be used for. For this purpose, we did some market research on UV curing glues. After checking the datasheets of 89 models of adhesives across 4 different companies, we came up with following graph:
The source of this information can be found in spreadsheet format, here.
Let’s highlight some important aspects of this dataset:
- Save for
a handful of glues specifically designed to be cured by visible
light. All studied adhesives can be cured by a wavelength of 365nm.
- The large majority of datasheets recommend an irradiance of around 100 mW/cm^2 or less for ideal curing conditions. (Though the documentation can be somewhat ambiguous on the absolute minimum irradiation required to still cure the product).
- Small portion of glues from Henkel recommend using a secondary wavelength of 250nm to improve curing on surfaces exposed to oxygen. We will not be addressing this, as it would prohibitively increase the cost of the project.
From this information we can conclude that we should ideally aim for a lamp that emits 100 mW/cm^2 @ 365nm, as this gives the project the best coverage of commercially available glues. Though as we will soon see, it is not trivial to emit light at such high intensity from a battery powered device.
SELECTING THE LEDS ON MOUSER/DIGIKEY
It turns out there are not that many high power UV LEDs available in Mouser or Digikey. Even less when you search for ones that:
- Emit at 365nm
- Have high power and efficiency.
- Are inexpensive.
- Are readily available and well stocked.
- Have small viewing angles (we need a concentrated beam light to reach those 100mW/cm^2)
We decided on the IN-C39ATOU2 from Inolux, link to the Mouser listing here.
Some highlight of the LED are:
- Forward Voltage: 4.4V
- Forward Current: 1Amp
- Radiant Flux: 1600mW
- Wavelength: 365 – 370 nm
- Viewing Angle: 30°
- Size: 4x4mm
- Cost: 13.62 $ per led
Note: Yeah, I know that they look really expensive. But the other were not much better. UV LEDs are just really expensive.
HOW MANY LEDS DO YOU USE ?
Ideally we don’t want to use more LEDs than absolutely needed, both because of the price, and to save on battery life. Thankfully the datasheet provides us with all the information needed to run a couple of simulations and figure this out. Namely, the total power emitted as light (1.6W) and the radiance pattern, which looks like this:
Image source: IN-C39ATOU2 datasheet
No we can program a small script in Python to calculate how this beam pattern would look projected over a surface, at different heights.
From here we can see that at its peak, 6 LEDs can indeed generate the 100mW/cm^2 we were looking for, even if it just in a small point. Now let’s look at those irradiance patterns a bit more closely:
At 7cm from the target, the center of the pattern surpasses 100 mW/cm^2, and there is a circle of about 2.4cm in diameter of light above 90 mW/cm^2. Not much in terms of area, but that’s quite an impressive power density.
Now, if you’re working with glue that cures at lower energy levels and prefer less power over a larger area. At 25cm away, the six LEDs can generate a circle of 14cm in diameter with over 20 mW/cm^2. Which is rather respectable for a handheld battery powered device.
For a total electrical power consumption of 26.4W and around 81.72$, these LEDs seem like the most sensible choice for the job. They cover both scenario, focused high intensity UV light from up close for those glues that require it. And, wide area low intensity UV light from far away for when area coverage is more important than high intensity.
EXTRA: WHY NOT USE STANDARD 5MM UV LEDS ?
Let’s explore a fair question that might be in your mind after reading this post. If you search on Amazon or Ebay right now for “UV LED” you might find about a dozen of listing like the following, selling 100 UV LEDs for just a few USD.
Image source: Listing of UV LEDs from Ebay.
If this is so, then why pay almost a hundred dollars for just 6 LEDs? Why not use these? After all, they are not only exponentially cheaper, but are even more readily available. Which makes them even more inline with our search criteria. Well,
TL;DR: sadly, they just don’t work for this application.
Two main reasons for this:
- Most of these listings are for ~400nm wavelength leds. Which as we’ve seen above, don’t coincide with the activation wavelength of many of the UV curing adhesives.
- More importantly, they are very weak. So weak in fact that they might not even be able to cure anything but the most sensitive of glues.
Now, let’s back up this claim by running some simulations. It’s hard to find datasheets for these cheap LEDs, so instead we are going to use the data from a similar well documented LED to feed the simulation. Namely, the UV5TZ-390-30 from Bivar (Link to the Mouser listing here)
According to its datasheet, at maximum power this led can produce 40mW of light with a viewing angle of 30°. As seen below.
To give it the best possible chance we are going to simulate a 7x7 grid of LEDs, for a total of 49 emitters working at maximum power.
There we can see that even at the point of peak irradiance, this array of LEDs is just barely able to reach 10 mW/cm^2.
Therefore, as attractive as their price seems, they are not suited for this application. As even if they had the correct wavelength for UV glue curing, they don’t output enough power to be effective at this task.
- Save for a handful of glues specifically designed to be cured by visible light. All studied adhesives can be cured by a wavelength of 365nm.