An interesting application note about crystals (I know very little about them, and need to know more) and a paper about how best to solder QFN packages, plus some jellybeanish analogue ICs and a tiny tiny DC/DC converter.

LMV331 Single, LMV393 Dual, LMV339 Quad General-purpose Low-voltage Comparators (TI)

Jellybean comparators, working off 2.7-5V supply voltage, with open drain outputs. More or less all comparators seem to have those: makes sense, since you want to be able to drive whatever logic levels you're using from the outputs. In fact, because of the open drain outputs, one suggested application for these is level-shifting.

As usual with these things, the IC schematic is a complete mystery to me. I've started being able to identify some building blocks, but I'm lightyears from understanding these things.

A thing I learned: comparators switch faster if you overdrive them, i.e. if the difference between the inputs is larger than the input offset voltage.

28/41 pages in this datasheet are about packaging. TI, you're weird.

Quartz Crystals and Microchip ICs (Microchip ANTC207)

This one was very useful to me, because I know very little about how crystals are used, and it seems that more or less any circuit you design that uses an MCU will need one, and will need you to do load capacitance calculations, for which I feel inadequately prepared!

Anyway, crystals are passive components that use the piezoelectric properties of quartz to form a resonator: their electrical impedance drops at their resonant frequency, so a crystal can be used as a filter element.

Most are disk-shaped: a thinner disk means higher frequency. Tuned by (ion beam) etching of electrodes to reduce their mass. Also get "inverse mesa" shaped crystals, which are disk-shaped with a divot in the middle to make the middle thin. This gives a higher resonant frequency without the manufacturing hassle of having a disk that's thin everywhere.

An effective model for a crystal has ESR, ESL, series and parallel capacitance. The values depend mostly on crystal size, it seems (smaller capacitances for small SMD crystals, but higher ESR). The series capacitance and ESL form a series LR resonant circuit and the series capacitance determines the magnitude of "pulling" of the crystal's resonant frequency by load capacitance. On the one hand, that pulling can be a pain, if you want to solder down a crystal and a couple of capacitors and predict what the resonant frequency is (you need to think about parasitic capacitance in your PCB as well...). On the other hand, if you want to be able to tune the resonance of your crystal (e.g., in a VCXO), that "pulling" effect is what you want.

Away from resonance, a crystal acts like a normal capacitor with parallel capacitance. Crystals are calibrated to oscillate at their quoted frequency with a given load capacitance. It's important to get this right when using crystals!

The ideal crystal has large series capacitance (so it's tunable), small ESR (small energy losses), and small parallel capacitance. This is easier to achieve with larger crystals, which is a bit of a shame for making SMD crystals! Overtones to the rescue...

You can get crystals to oscillate at overtones by using a high-pass filter to reduce the gain at the fundamental frequency. ESR is higher for higher overtones, to it's harder to get them to work. Using an overtone means you can use a thicker crystal (with lower fundamental frequency), which is easier for manufacturing, but it does make the design of the oscillator more complex.

Sources of variations in crystal resonant frequency: calibration tolerance at whatever reference temperature, temperature drift, aging. Total stability of ±100 ppm is relatively easy to achieve; much less than this quickly gets difficult.

After reading this, I have to say that I still don't really understand how to set up crystals for driving MCU clocks, but I do have some more sources to read, and I know more than I did before! (The nRF52840 datasheet is the clearest I've seen so far, but it's still a bit fuzzy to me.)

LM358 et al. Industry-Standard Dual Operational Amplifiers (TI)

Not sure what "industry standard" actually means here!

This has the usual wide range of supply voltages (3-36V), but I'm not sure how useful these are in lower voltage applications because, if I'm reading the datasheet right, the output can only swing to within about 1.5V of the positive supply rail. Same sort of thing on the inputs too, I think.

Does "industry standard" mean "old part, don't use"?

The Impact of Via and Pad Design on QFN Assembly; Jennifer Nguyen, David Geiger and Anwar Mohammed (2016)

I'm planning to start doing some designs using QFN parts, which I've not used before, and the idea of soldering them is a little intimidating. However, I'd read in a couple of places that "hot plate + hot air pencil" soldering for these things isn't such a big deal, and is easier than fine-pitch QFPs, so that set my mind at rest a little.

Then I read this paper... Actually, it's not so bad, but it was interesting to see the range of options for board design for using these things. When working KiCad, you pick a footprint from a library and you can get the impression that that is the footprint for that package, but there are options.

The paper is mostly about thermal vias for QFN packages with exposed thermal pads, and the associated voids and solder protrusions you can get (a solder protrusion is where solder wicks through vias during reflow and makes little bumps peeking out on the back side of the board).

The have one interesting stencil design with no paste at all over thermal vias (they call this a "window pane" stencil).

In their results, solder protrusion is really common: the best way to fix it seems to be to use the window pane stencil and to include solder mask rings around the thermal vias in the thermal pad for the QFN.

This solder mask around thermal vias thing is not something I've seen before. I've never seen a footprint anywhere with it in. Maybe I need to try to find the IPC standard for these, like the one I read a few months ago.

TPS63036 High-Efficiency Buck-Boost Converter in Tiny Wafer Chip Scale Package (TI)

Fairly standard 1.8-5.5V input, 1.2-5.5V output, and generally looks like a fairly run-of-the-mill buck-boost converter, apart from the ridiculously small package (1.8 × 1 mm DSBGA with 0.4 mm ball pitch). My first though was: how does that compare with the possible sizes of inductors? You need 1.5 μH, and the part specified in the datasheet is 3 × 3 × 1.5 mm. That means that the layout only looks a tiny bit weird...

This has a sophisticated controller for switching between buck and boost. It's one of a whole menagerie of these very small DC/DC converters intended for Li-ion battery applications. They look easy to use, apart from the packaging!