iiinova — where your digital self keeps evolving

Andrey @spirit.re · 10/08/2026

Sneak peek of a small side project: The world's highest resolution, highest frame rate thermal imaging phone-attach camera: 640x512 at 60Hz(!), in a tiny 25x25x38mm package. Radiometric w/ recording, focuses from ~8mm macro to infinity. Works on Android&iOS + Windows/MacOS/Linux.

Render of a 640x512 thermal imaging phone dongle

Real photographs of the 640x512 phone dongle prototype. No label engraving. Real lens visible.

A superresolution capture(real thermal data, higher than 640x512 native pixel size) from the camera prototype, produced in-software, then upscaled another 2x for pleasant viewing.

040

Andrey @spirit.re · 29/07/2026

During the process of cleaning a capillary used to feed and bond the wire onto the chips, I had a small intrusive thought... and it may have won...

010

Andrey @spirit.re · 28/07/2026

I'm working on retrofitting a wire bonder with custom electronics as part of a small side project(yes, to also make them cheaper). I think it's neat how the electric flame-off(EFO) gold ball forming system works when viewed at high magnification. Here's a short video about it.

010

Andrey @spirit.re · 01/07/2026

A little sneak peek of another project I'm working on. Can you guess what this is? :)

An uncommon object that involves ruby, sapphire, stainless steel and gold.

000

Andrey @spirit.re · 29/06/2026

I got one of my high speed cameras to try filming an ozone generator's dielectric barrier discharge, but the camera wasn't sensitive enough. Can't put a high speed camera away without filming something, them's the rules, so have a can of brake cleaner spraying at 895fps instead.

130

Andrey @spirit.re · 06/05/2026

Decided to try a little experiment - zooming in from 15x to 100,000x in smooth progression. The SEM isn't particularly well calibrated so the result is only ~30-40nm resolution, but I think it's still pretty neat. Working on making SEMs so affordable you could have one at home :) 🧪 HQ link in reply.

110

Andrey @spirit.re · 12/04/2026

Tungsten & oxides deposit on the Wehnelt and anode of SEMs as filament evaporate and gunk in the chamber hits it. Abrasives are used, but I found an alternative: 1:1 solution of 10% ammonia to 3% H2O2! If aluminium is present, oxides grow and a 5% phosphoric wash is needed after. 🧪

A heavy oxide and tungsten metal film is visible deposited in a rainbow pattern on the filament side of the wehnelt cup of an electron microscope.

The same Wehnelt cup, after cleaning with a 50:50 mix of 10% ammonia and 3% hydrogen peroxide solution, as well as a 5% phosphoric acid rinse due to the presence of aluminium.

020

Andrey @spirit.re · 24/02/2026

Tinkering with electron optics simulation this evening, came up with a little test software for boundary element method simulation and particle tracing. So far so good, and I think it has potential as an educational & low-complexity pro design tool. Would you want to use this? 🧪

Screenshot of a work-in-progress charged particle optics simulation software, showing traces of electrons being emitted from a narrow bar's surface, traveling between and being deflected by two charged squares, accelerating into a cylinder at high potential.

030

Andrey @spirit.re · 05/01/2026

Neat paper I missed in 2024 - these guys took an ultrasonic wedge wire bonder - the kind that's used to attach wires to chips - scaled it up by 15x, and used it to 3D print aluminium! Pretty cheap to implement, not much to it. Pretty cool. Open access, DOI: 10.3390/ma17102188 🧪

020

Andrey @spirit.re · 29/11/2025

It's not really collimated, it's just a basic flat panel backlight for machine vision. It wouldn't work as well without a telecentric lens.

020

Andrey @spirit.re · 15/11/2025

One of the more unusual workplaces for an engineer to be. Plugging something you designed and built into an obsolete avionics system that would cost more than your entire net worth to replace is... one of the feelings of all time, for sure. The jet is a Cessna Citation Bravo.

An overview of the cockpit of a Cessna Citation Bravo in the middle of an overhaul, with a custom diagnostics instrument attached to the avionics.

020

Andrey @spirit.re · 04/11/2025

Built a small optical coordinate measurement machine and wrote a basic but mostly feature-complete software for it. Telecentric optics, so no perspective error. Resolution ended up at ~4µm for this ~30mm FOV. Would you want a nicer version of this(housing, lights) for ~2.5k? 🧪

An optical CMM based on a telecentric lens and a high resolution camera, mounted on a large and high precision vertical translation stage. Flat backlight visible below.

General macro overview of the scene being measured by the CMM. 0.5mm pitch connector, M3 screw, glass calibration scale.

Screenshot of the user interface with feature recognition enabled and a few measurements added.

Second screenshot of the user interface, measuring thread parameters.

120

Andrey @spirit.re · 13/10/2025

They did warn me that with Windows the USB3 driver is a little unstable and can't quite reach 400MB/s(stops at 396MB/s), but so far it's been... fine? For a price of $69(nice) it's very good either way, and it should be stable on Linux/Mac.

000

Andrey @spirit.re · 13/10/2025

Sipeed sent me their new SLogic16U3 USB3 logic analyzer to review(free, no obligation). Inside is just a tiny Gowin GW5AT-LV15 FPGA doing softcore USB3(!!!), which is how they made it so cheap. 800MHz@4ch is quite impressive. My verdict? Definitely going in the tool box. 🧪

Inside the logic analyzer. Gowin GW5AT-LV15 FPGA.

Underside. Power supply and miscellaneous.

Sigrok PulseView software showing 800MHz sample rate on 4 channels.

130

Andrey @spirit.re · 29/09/2025

Have you ever wanted to see a 90 second abridged explanation of how diode lasers work? Here it is. I took some pictures of a laser under my SEM, and it kind of just scope crept into a whole video. Not perfectly accurate, but it's hard to squeeze 100 books of info into 90 seconds.

031

Andrey @spirit.re · 13/09/2025

Ever wondered how infrared thermometers work? By connecting over 100 thermocouples in series, all on a MEMS chip, floating in a vacuum! The cold side is connected to the base, thermal radiation is focused onto the hot side, and the Seebeck effect gives us a proportional voltage. 🧪

This image shows an infrared thermometer, also known as a pyrometer. Inside is a small hermetically packaged metal can containing a thermopile. There are two images showing the thermopile as a general overview and up close. A reference thermistor that measures the base temperature is visible on top.

Two scanning electron microscope images of the device, much closer in magnification. Top shows the positive and negative wires, and the bottom shows the structure of multiple thermocouple junctions around an absorber.

Two scanning electron microscope images, showing a closeup of how the thermocouples are connected in series around the cold side, the substrate.

020

Andrey @spirit.re · 29/08/2025

First handheld laser welds, cuts, some cleaning. What the internet says is true this time - compared to regular welding(MMA/MIG), this is, no exaggeration, a point&click tool. 10 minute learning curve. Some nuances like penetration and fillet size, but overall just spectacular.

First handheld laser weld between 2.5mm thick square tube and 5mm thick plate. Both carbon steel.

Triple pass weld with the same settings to fill out the joint. Needs more filler wire or a slower run.

Aluminium tube cut freehand at 1.5kW power with a ~250µm wide kerf and near zero dross. Incredible maneuverability, like butter.

Closeup of the cut pieces, seen from the inside. In the background, a laser cleaned section of the plate, shiny finish.

010

Andrey @spirit.re · 21/08/2025

And of course, we need a little lens mount to test with.

Custom C&CS lens mount attached to the board

Lens mount with lens, nicely fitting directly on the sensor

000

Andrey @spirit.re · 21/08/2025

Design, reality. Looks pretty accurate :) High speed imaging will be affordable, and hopefully soon.

An image sensor test board, designed in KiCAD. 3D render.

The same board, but manufactured and assembled in real life.

140

Andrey @spirit.re · 06/08/2025

I've made a big step and finally touched an FPGA on a personal level. Fighting the development environment was 90% of the effort and only 10% was learning how this whole HDL/constraint thing works. Now to draw the rest of the owl. In my case, a high-speed camera shaped one.

040

Andrey @spirit.re · 02/08/2025

Who needs Boston Dynamics and Unitree when you can have Soviet Dynamics? This 45 YEAR OLD(!) film shows the exact thing techbros have been parading around for the past few years. LIDAR, SLAM, multi-core mainframes, live trajectory and path planning, IK, simulation, all from 1980 youtu.be/_klAzHoI7S0

youtu.be

WalkingMachinesNIITransmash1980

YouTube video by Spirit

010

Andrey @spirit.re · 31/07/2025

Did I *need* to create an extremely detailed 3D model of an image sensor, down to the bond wires, overlaid sensor structure and modeled pads, just to put it on a PCB design? No. But did I spend quality CAD time doing that and then rendering it out for fun? Absolutely.

Accurate, realistic 3D render of a high speed CMOS image sensor in its package. Bond wires, pads, and clear chip outlines are visible, as well as some of the PCB layout.

010

Andrey @spirit.re · 10/07/2025

Some incredibly cool pre-production silicon just showed up in the mail: >2K resolution at >600 frames per second, >10kfps with windowing, global shutter, HDR. This may be the first image sensor to make high speed cameras truly affordable for *everyone*.

010

Andrey @spirit.re · 04/07/2025

Do we have anyone with contacts at Xilinx/AMD? I want to get some samples of the new Spartan UltraScale+ chips. Specifically XCSU65P in CMVB529. And ideally a dev/evaluation board(SU65P or higher, need PCIe). My DMs are open, email works too - see my site.

010

Andrey @spirit.re · 24/06/2025

Got some 1.5kW and 2kW air cooled laser welder+cutter systems in to try a distribution model(EU & US). Very clean builds inside - DC PSU, nameless laser source, OEM control panel, minor gas handling. Simple, but affordable and effective. Also going to mod cleaning mode into it ;)

A 1500W air-cooled laser welder, hand next to it for scale. Wire feeder in the background.

Inside the 1500W air cooled laser welder. Power supply visible in front, big laser source in the back, small power supply below and control electronics above. Also some gas handling hiding behind the small power supply.

Inside the 2000W unit. Same thing, except bigger, and slightly more advanced gas handling. Notable is the Huawei cell tower equipment power supply.

010

Andrey @spirit.re · 18/06/2025

Helped with a bit of repairs on a Mooney M20M(Bravo) recently, then went up in it! General aviation is really fun, and if you ever get the chance to go flying, you should definitely say yes!

A Mooney M20M(Bravo) single-engine, 4 seat, low-wing airplane.

Inside the Mooney M20M in flight. Modernized glass cockpit.

View over the right wing while flying low over a city on a clear, mostly cloudless, hot day.

030

Andrey @spirit.re · 11/06/2025

Anyone out there with access to a research fab at reasonable rates? I want to test a relatively simple and potentially groundbreaking semiconductor device, but with a very exotic stackup. I need ALD, PECVD, and likely DRIE. Low res maskless litho is fine. Preferably in the EU.

020

Andrey @spirit.re · 22/05/2025

Nope, it's a cool new-ish process, the only resins involved are solder mask and legend ink :)

000

Andrey @spirit.re · 22/05/2025

My process does not involve any 3D printing, the whole point is to produce boards that are identical to factory spec, using industry standard materials. Polymers you can plate are cool and all, but they're just not the same. Multi material resin is cool though.

200

Andrey @spirit.re · 21/05/2025

Standard FR4(maybe FR5) with regular copper :)

100

Andrey @spirit.re · 21/05/2025

Probably around $15k retail to start with.

100

Andrey @spirit.re · 20/05/2025

Any coating variation is possible with extra chemical steps, Ni/Au is relatively easy(but 10-50g of Au is $$$$). Copper thickness adjustable, but targeting regular 1oz/35um. Any composite substrate except PTFE, FR4 as standard. There is a path to 4L+, but 4x machine cost when jumping from 2 to 3+

100

Andrey @spirit.re · 18/05/2025

Maximum size depends on system cost, for something low-ish end I'd limit it to 100x100mm. Through holes are fully plated at the end of the process. Mask, silk, etc. More or less a factory quality board.

000

Andrey @spirit.re · 18/05/2025

Is there any commercial interest in a desktop-ish full-cycle PCB "printer"? I proved out all the processes for fully automated, mostly chemical-free production of 2-layer PCBs with trace/space/via as small as 75μm. It could be relatively affordable, but not quite hobbyist-priced.

Complete single-layer PCB without silkscreen, made by a fully automated system.

A relatively large plated copper via made by a fully automated process.

270

Andrey @spirit.re · 26/04/2025

Ever wondered what TIG welding with neon gas would be like? No? Anyway, here's TIG welding with pure neon instead of argon. Video from Zerg Labs' archive.

010

Andrey @spirit.re · 03/04/2025

Everyone posts videos from these at infinity, so here's something you'll rarely see: thermal macro! The phone dongle will have manual focus, so you'll be able to do this out of the box. No denoising or post-processing applied, so it's a bit fuzzy.

010

Andrey @spirit.re · 03/04/2025

Got this very neat radiometric 640x512 thermal imaging core. Miniaturization is wonderful, it's just the WLVP image sensor directly wire bonded onto the board, with an ASIC on the back to crunch data. There are no affordable phone dongles with this resolution, so I'll make one.

Thermal image core with the lens and lens mount removed. The shutter assembly is visible in its retracted state, with the sensor visible through it.

The lens assembly on its own, showing the back element and design simplicity.

Back of the main thermal sensor board. An ASIC(supposedly) is visible in the middle, right below where the sensor is mounted on the other side. Some power supplies, flash memory, and other minor components.

110

Andrey @spirit.re · 12/03/2025

Silly computer peripheral idea: would you want an ultra-lightweight(world's lightest?), completely battery-free, permanently wireless 8kHz polling rate mouse, if it was confined to a special powered mousepad(~0.6mm thick, goes under your actual mousepad)?

120

Andrey @spirit.re · 18/02/2025

Is it safe to clean ball bearings in an ultrasonic cleaner? Turns out, it isn't! I investigate this with a scanning electron microscope in this video: www.youtube.com/watch?v=-g6g... 🧪

youtube.com

Does ultrasonic cleaning damage ball bearings? Electron microscope investigation!

YouTube video by Spirit

030

Andrey @spirit.re · 15/02/2025

Someone asked whether it's possible to build a DIY backscattered electron detector for a SEM. Yes, it's pretty easy! You just need bare PIN photodiodes, a trаnsimpedance amplifier, and a signal input. S=SED, R=BSED. Sample: rock section, ~nm of Au on top. Some charging visible.

A small collage showing the detector as-built, the photodiode, where it's installed, and a schematic for replication.

Secondary electron image. Some charging(full white spots) is visible due to the poor gold coating.

Backscattered electron image, better showing compositional contrast.

040

Andrey @spirit.re · 26/01/2025

Here's a soviet K573RF2 EPROM under the newly repaired SEM.

Soviet K573RF2 EPROM scanning electron microscope image

140

Andrey @spirit.re · 26/01/2025

After removing the cover and manually returning the stage back within the correct range of travel, replacing the stage controller with a 3D printer board running grbl, and messing around with software(upgrading the OS too), everything worked! See next post for imaging results!

040

Andrey @spirit.re · 26/01/2025

When I powered it up, something was very wrong. The motor controller caused the software to hang with a nonsensical error. Investigation revealed that the table was very hard crashed during transport, and that the controller board was fried.

Sample stage with visible signs of overtravel damage

130

Andrey @spirit.re · 26/01/2025

The electronics are quite modern for its time. Three FPGAs handle imaging, with a Hitachi SH2 processor handling the config, all connected via FireWire, except for the motor control board, which has a separate USB interface. And of course, Apex opamps for raster scanning.

Main control board with FPGAs, DACs, CPU, and everything else. Mostly digital, some analog.

110

Andrey @spirit.re · 26/01/2025

It's a fairly simple SEM in terms of construction. I would say a bit too simple. There's only 3 valves, none for isolation - one on the forepump, one to slow the turbo down, and one to quickly vent to the atmosphere. No vacuum gauges - it just guesses based on the turbo's speed!

Inside the microscope, with the top cover removed

Back side of the microscope, turbomolecular pump visible

Electron gun chamber open, cathode visible

110

Andrey @spirit.re · 26/01/2025

Oops! All electron microscopes. I have another patient in the lab, this time it's a Keyence VE-7800 from the early 2000s. Let's dig in and take a look at the insides to bring it back to life! Lots of pictures in the thread.

Keyence VE-7800 scanning electron microscope, face on.

Keyence VE-7800 scanning electron microscope, view from the side. Extra ports are visible.

130

Andrey @spirit.re · 25/01/2025

Here's a reconstructed CT scan of a SEM filament. The disk on top is the Wehnelt... disk, I suppose, that serves to create the crossover pinch and adjust emission spot size. The four screws are used to center the filament within the cartridge.

050

Andrey @spirit.re · 24/01/2025

Video of the pump silencer and vibration isolation box from the previous post. Simple, but works very well.

010