Thankfully, there isn't going to be a series of exhausting logs on the choice behind every component on the BoM. I just wanted to highlight how the system design has begun to create some freedoms and perhaps constraints on component selection. Hopefully someone will learn something too.

RdsOn

RdsOn is the equivalent resistance between the Drain and Source pins and it varies, mainly depending on the drive voltage applied to the Gate pin. When combined with a current flowing between Drain and Source, RdsOn leads to a voltage drop across the Drain and Source, defined by Ohm's Law, and this voltage drop varies according to current and the value of RdsOn, unlike in plain and Schottky diodes and transistors, where the voltage drop is pretty much fixed, regardless of the input to the Base or the current flowing between Collector and Emitter.

Having learned much of the limited amount I know about MOSFETs from a 3D printer heater control perspective, my instinct is to drive towards as low a value as possible for RdsOn, in order to reduce losses through heating.

Unlike in a simple 3D printer heater switching circuit, in this application I'm looking at an using an inductor with an equivalent series resistance in the order of an Ohm, which will be in series with the LED driver MOSFET. So here, pushing for 3mOhm over, say, 50mOhm RdsOn isn't a proportionate optimisation to make, as it will make barely any difference to the total series resistance. The other two MOSFETs are used in the battery charging circuit (at this point in my untested design) and they are likely to see a maximum current of an Amp, so this broadly equates to a milliVolt drop in the circuit per milliOhm RdsOn. However, as charging current drops, this voltage drop decreases, so the charging circuit MOSFETs aren't driving us towards an extremely low RdsOn either. This is good news as we can open up our RdsOn range and use the same part for all three N channel MOSFETs.

VGS(th)

VGS(th) (or other abbreviations) is the Threshold Voltage between Gate and Source. It is useful as it tells us how much the Gate drive voltage is required to be above the Source voltage before current can flow between Source and Drain. Beware that at this point the RdsOn will be significant compared to if the VGS was greater, so you aren't operating at an efficient part of the MOSFET's performance curve. So, we always aim to drive the Gate with a voltage higher than VGS(th). The MOSFET datasheets will often include the curves which show how RdsOn varies with VGS but we're operating below the current levels where the graphs usually show, so we'll just aim for around VGS(th) x2 for our likely selection of logic level MOSFETs.

Rise and Fall Time

The speed at which the MOSFET changes from allowing and preventing current flow is measured in low figures of nanoseconds! This is irrelevant for many applications but for high speed switching it can become relevant. We are going to be switching our LED driver MOSFET on and then off after about 3000 nanoseconds, so it would be good to keep Rise and Fall times (Ton and Toff) within a low percentage of that. This doesn't appear to significantly limit our range of options

Max Current and Power Dissipation

This isn't a challenge for this application like it is for my old 3D printer heater switch application. We're talking about ~4V and a few mA below 50% duty, so a few mW when in full lighting mode. All the signal MOSFETs seem capable of handling at least 100mA, so we're OK if we stick to 1W rated LEDs and below.

Max Voltages

With Yapolamp operating under 5V, we aren't ever likely to stress a MOSFET's ability to withstand voltage across the Drain and Source but you do need to check the value of VDS(max). We also need to check the maximum value that VGS can be, which is usally given as a +/- range. Again, especially in a high-side configuration where we're having to boost our Gate drive voltage, we are unlikely to rule out many MOSFETs based on VGS(max).

Package

With my old 3D printer heater, I was replacing a TO220 package so I opted for the same, especially as the IRLB8743PBF was recommended for this application and came in TO220. Here in the Yapolamp, we're less constrained. My hardest choice has been to decide whether to go surface mount or through hole for the components, where I can. This is becaus if other people wanted this torch, through hole hand soldering might make a reproduction more feasible. At this point I'll go with SMT as it allows for a smaller package, leading to smaller PCB and less cost to prototype the PCB at somewhere like @oshpark which charge by PCB area. I'm thinking SOT-23 or SOIC 8, rather than DP2AK / TO-263. Although a saving on PCB price might seem worth it on paper, if I find soldering a SOT-23 a nightmare, it might be worth splashing out on the extra PCB surface area to fit a D2PAK package in.

Price and Availability

There's no point finding the perfect combination of specs if the part isn't available. And for me when prototyping, that doesn't just mean available, it means available in the quantity where the smallest unit of purchase multiplied by the price at that quantity is acceptably low - I'm not going to buy a part reel of 500 units, even if each part costs a few pence. E.g. if I can buy one MOSFET at £1 or 10 MOSFETs at £0.10, I'm happy to prototype with either, although I'll try the £0.10 version first as I want at least three!

Conclusion

So, what part am I picking to start with? Well after a bit of trawling around, I'm going back to a MOSFET I used a few years ago, before I really knew what to look for. It's the FDN337N, from Fairchild Semi, now ON Semiconductor. Key specs:

• SOT-23
• VGS(th) Max 1V
• RdsOn 55-80mOhm, depending on VGS (2.5 - 4.5V for the stated range)
• 2.2A max continuous current, 10A pulsed (like we are using in Yapolamp)
• 30V max Drain-Source voltage, +/-8V max Gate-Source voltage
• Price ~£0.35 each in a pack of 5. Availability - on back order for delivery in 2 weeks. Hmmm, but given my PCBs will take longer, this is OK.

I will certainly keep a lookout for D2PAK and perhaps a through hole alternative. For breadboard prototyping, my trusty TO-220 IRLB8743PBFs will be fine for now.