Articles

Let's go on a journey and look at different types of circuit protection!

The resistors on my PCB perplexed me. They demanded "better access to the fields" and "a cleaner environment", how could I resist?

To design a good PCB, you've got to understand this!

Remember when you didn't need to care about return currents? I don't, I was born after!

Silicon must watch out, GaN's in town and it's coming for the crown!

Well... The performance and miniaturization in power electronics crown, but who's counting. Last year as part of my bachelors I had to write a review paper. There was a big marketing boom in tiny USB-C "chargers" attributing their size to GaN (Gallium Nitride), which I wanted to investigate. Now, I think that in 5-10 years, all of our devices will have a GaN powerstage. But first, what even is it? The MOSFET we all know and love is built with silicon using P/N junctions. GaN is also a semiconductor, but it has a wide-bandgap which allows operation at higher voltages and smaller devices. Not only that, but an AlGaN layer induces a highly conductive layer of electrons (called 2DEG) which reduces resistance. But the star feature of GaN is the low switching losses, allowing cool operation at very high frequencies. All in all, this means that GaN transistors can be made relatively tiny with very low Rds(on), high current carrying capacity, and low parasitic capacitance. But there are mature Silicon parts with even lower Rds(on) than the current young GaN market, so why go with GaN? In a Silicon MOSFET, low Rds(on) necessitates a large gate and thus high parasitic capacitance, resulting in high switching losses at high frequencies. But why would I want a high switching frequency? A power converter relies on storage elements: capacitors and inductors. A high switching frequency means that each cycle is shorter, thus less energy needs to be stored. Less energy means smaller L and C, thus smaller passives. 1+kW GaN power converters are demonstrated in credit card sizes with MLCC only capacitors! Here's the exciting part. During my search, I found many examples for power converters with GaN transistors being much smaller. For example, I compared two AC/DC forced air cooled power converters. The 4kW GaN model was 1.9x heavier and 1.1x bigger than a 1kW super-junction silicon model. More exciting is this module under development at EPC - Efficient Power Conversion (GaN transistor company) that has a 1 kW capacity with a footprint SMALLER than 1x1"! One distinction to make is that GaN transistors aren't MOSFETs. Most GaN structures don't use P/N junctions. The nearest device to a Si MOSFET is a GaN HEMT (High-electron-mobility transistor) but it is normally-on and isnt doped. There are many structures that solve the normally-on issue such as cascode, GIT, finFET, pillar, and other commercial and research grade transistors. So, will GaN replace Si? No. GaN is more expensive to fabricate (especially vertically, without a different substrate like Si or SiC). It is unlikely to be cheaper for applications that dont benefit from a faster switching or a small footprint. I believe that most future power converters will be GaN based, but silicon will still dominate the power marketshare. Have you used GaN? How has your experience been?

Sometimes old is gold, but it's 𝘴𝘩𝘰𝘤𝘬𝘪𝘯𝘨𝘭𝘺 quite pink in this presentation.

Picture this: you ask an electronics engineer to add 3 differential pairs onto a PCB, and they deliver the following, what do you do?

Can the cheapest component on a circuit board be made cheaper?

Last week I posted a photo of computer RAM, in which a special component caught my attention. On the blue PCB (printed circuit board), you can see a resistor array containing four resistors of the same value (15 ohm). This is less common than single resistors, such as the four seen on the red board (marked 221, which means 220 ohm). But why choose either one? I had previously heard from an experienced engineer that a resistor array can be cheaper, but I wanted to check this claim. I used digikey (American) and LCSC (Chineese) websites to compare, first finding an array, then comparing it to buying four of the closest compareable single resistor. The result of this can be seen in photo 2, in which we can see that in come cases, especially at a higher order amount, the array is indeed cheaper. If a board needs 100 resistors of the same value, and a million units are made over its lifetime, $37,500 can be saved (using 25 stackpole arrays instead of 100 stackpole resistors). One more reason resistor arrays can be cheaper is the fact that they only require one placement in the pick and palce machine, in low production runs this doesnt mean much, but if the machine is working 24/7 on one single design, it can quickly add up. When I posted last week, I was expecting cost to be the primary driver for using arrays, but Eng. Eyas Alsuhaibani pointed out that having the resistors be in an array means that they are better matched. Indeed, as seen in the last photo, arrays can be purchased with specific matching tolerances. Why does this matter? Because even a 0.1% resistor can mean that there is a 0.2% difference, and if the difference between the parts is more important than the absolute resistance, then an array is the better choice. One reason not to use resistor arrays is that some parameters may not be avaiable, such as low temperature coeffeiceint (200 ppm seems to be the most common, with some 100 ppm parts available, with only one vendor supplying 0.1-1 ppm). This should not matter in most cases, but calculation is necessary to ensure proper device operation. I have seen many parts such as diodes, mosfets, LEDs, etc in which multiple ones are packaged together, but this seems much less prevalent in capacitors (a capacitor array could be used for things like PCIe, where every device needs DC blocking capacitors). I think it may have to do with crosstalk, what do you think the reason may be?

Today I was preparing a report, and I found this beautiful photo of computer RAM. There are atleast a dozen of hidden details in here that help this board do its job, but one of them used to always catch my eye. Two of the wires coming from the bottom get close together and then far away, why do you think this is? Why do other wires not do this? Comment below and let's see who gets it right. Image from pxhere.com, CC0

Recently, my hotel room's mini fridge had a flickering light and a glass door. So instead of going to sleep, I discovered something awesome 📟