random thoughts of an engineer

 What is ModuGlow?  ModuGlow is modern, modular, organic lamp that can be anything between night lamp to art sculpture. It's built from 10x10cm panels containing 25 blinking bulbs. It can be stacked in 2 axis to provide arbitrary size lamp. Video showing 2x2 combination working. Neon behavior Neon bulbs exhibit S type dynamic negative resistance. If you understood last sentence feel free to skip this paragraph. Negative dynamic resistance is a behavior of some electric circuits when with increasing current current decreases or the other way around with increasing voltage current decreases. In neon bulbs it manifests itself between strike and maintain voltage. When bulb strikes the voltage starts to drop due to series resistor. Dropping voltage makes bulb want to draw more power further lowering voltage until negative resistance region ends near maintain voltage. That way bulb "turns on" at higher, striking voltage while "maintains on" at lower maintaining voltag...

The purpose of this blog post is to describe the process of connecting a MEMS microphone with PDM to a microcontroller, and the steps taken to optimize the PDM-to-microcontroller data processing for better performance. Introduction In my latest research I'm in need to connect MEMS microphone with Pulse Density Modulation (PDM) to micro-controller to evaluate their parameters. In contrary to more popular Pule Code Modulation (PCM) that sends data at reasonable sample rate (40-192khz) with wide words (8-24 bits) PDM sends one bit stream at Mhz sample rate straight from sigma-delta modulator. To do anything useful with this stream it needs to be decimated to lower sample rate while increasing word width and filtering out high frequency noise. My microphone generates PDM stream at around 3Mhz. I want to downsample it around 40khz to feed my "standard" software DSP to have enough time to process data. Why? Decimating signal is as simple as selecting one sample every N and d...

 As promised this is update to last post about our PCB holders. Technical difficulties It turns out that if I want tapped hole every 10mm that's a lot of tapping. Additionally tapping steel is quite hard, especially with such small tap. Most companies decided that they don't want to do such job, one that agreed is research institution so I guess they're used to nontrivial requirements. After some negotiations we got price of almost 5k euro for 300x300mm sheet. That's a bit too much for this project. Required changes My changing steel to aluminum and moving holes onto 15x15 grid we considerably reduced required engineering time. After initial testing it turned out that 15x15 hole spacing is more than sufficient, but swapping to aluminum means magnetic mounts are no longer an option. I guess it's low price to pay to reduce manufacturing cost to 740 euros. plates when they first arrived in our lab   Designed mounts We designed few mounts for plates to mount things on t...

 Today's post will be quick one. I got too irritated by all PCB clamps that don't all work all too wall so I decided to make my own! Inspiration came mostly from optical tables we use all around our labs. I decided to make thin, heavy base plate with M3 holes every 10mm and adapters to be able to mount support wherever I want. The design looks like this: quick render of design Before commiting to expensive and long process of getting them milled from steel plate I decided to make sure that it's as useful as I think. For that I 3d printed prototype out of PLA. I wanted to make it out of black PLA but part way through filament ended and I had to quickly switch to orange. This is how it looks assembled: PCB mounted for soldering everything prepared for mounting PCB It make look more complex than usual clamps to use but keep in mind that if board has screws on 10mm raster then spacers can be screwed directly into base plate. Additionally thanks to modular design boards can be m...

 So what this story is about? Qpass is a method of measuring weight of small mass by depositing it on quartz resonator and measuring changes in electrical properties. Most common are frequency and Q factor changes. Of course because it's sensitive method and requires disposable quartz resonators it's used mostly for very sensitive measurements, hence detected deviation values are very small too, often at the order of sub ppm. Typical solutions Most modern commercial solutions use frequency scanning, building frequency or frequency and phase response and searching for peaks and 3dB width. There're few problems with that solution. For once it requires very dense scanning to make sure you find basic mode and not spurious ones. Phase measurements help in finding "missed" resonances but requires more complex analog frontend. Big upside of this method is that one finds interesting mode they can just scan very small range to track Q and resonant frequency. Another someti...