Electroplating DIY PCB Vias At Home Without Chemical Baths

January 31, 2025 by Maya Posch 20 Comments

Although DIY PCB making has made great strides since the early days of chemical etching, there’s one fly in the ointment: vias. These connect individual layers of the board with a conductive tube, and are essential for dual-layer PCBs, never mind boards with a larger layer stack. The industry standard way of producing them is rather cumbersome and doesn’t scale well to a hobby or prototyping context. Might there be a better way? This is the question that [Levi Janssen] set out to answer with a new home PCB manufacturing project.

The goal here is to still electroplate the vias as with the commercial solution, just without having to use chemical baths. This way it should be suitable for an automated setup, with a tool head that performs the coating of the via with a high-resistance conductive ink before the electroplating step, all without submerging the entire PCB. After an initial experiment showed promising results, [Levi] committed to a full prototype.

This turned out to be a bridge too far, so the prototype was scaled down to a simpler machine. This is where the main issue with electroplating one via at a time became clear, as a standard 0.3 mm via takes easily 10 minutes to electroplate, even with an increase in voltage. At that point ordering a PCB from China becomes the faster option if you have enough vias in the design. Fortunately [Levi] figures he may have some solutions there, so we’ll have to wait and see what those are in the next installment. The video is below the break.

Continue reading “Electroplating DIY PCB Vias At Home Without Chemical Baths” →

Patching Up Failing Hearts With Engineered Muscle Tissue

January 30, 2025 by Maya Posch 3 Comments

As the most important muscle in our body, any serious issues with our heart are considered critical and reason for replacement with a donor heart. Unfortunately donor hearts are rather rare, making alternatives absolutely necessary, or at the very least a way to coax the old heart along for longer. A new method here seems to be literally patching up a patient’s heart with healthy heart tissue, per the first human study results by [Ahmad-Fawad Jebran] et al. as published in Nature (as well as a partially paywalled accompanying article).

Currently, simple artificial hearts are a popular bridging method, which provide a patient with effectively a supporting pump. This new method is more refined, in that it uses induced pluripotent stem cells (iPS) from an existing hiPSC cell line (TC1133) which are then coaxed into forming cardiomyocytes and stromal cells, effectively engineered heart muscle (EHM). After first testing this procedure on rhesus macaque monkeys, a human trial was started involving a 46-year old woman with heart failure after a heart attack a few years prior.

During an operation in 2021, 10 patches of EHMs containing about 400 million cells each were grafted onto the failing heart. When this patient received a donor heart three months later, the removed old heart was examined and the newly grafted sections found to be healthy, including the development of blood vessels.

Although currently purely intended to be a way to keep people alive until they can get a donor heart, this research opens the tantalizing possibility of repairing a patient’s heart using their own cells, which would be significantly easier than growing (or bioprinting) an entire heart from scratch, while providing the benefit of such tissue patches grown from one’s own iPS cells not evoking an immune response and thus mitigating the need for life-long immune system suppressant drugs.

Featured image: Explanted heart obtained 3 months after EHM implantation, showing the healthy grafts. (Credit: Jebran et al., 2025, Nature)

Comparing Adhesives For Gluing PETG Prints

January 30, 2025 by Maya Posch 31 Comments

Testing every kind of glue with PETG, including wood glue. (Credit: Cosel, YouTube)

PETG is a pretty great material to print 3D models with, but one issue with it is that gluing it can be a bit of a pain. In a recent video by [Cosel] (German language, with English auto-dub) he notes that he found that with many adhesives the adhesion between PETG parts would tend to fail over time, so he set out to do a large test with just about any adhesive he could get his hands on. This included everything from epoxy to wood glue and various adhesives for plastics

TL;DR: Some superglues seem to weaken PETG, and a construction polyurethane glue is the absolute winner.

For the test, two flat surfaces were printed in PETG for each test, glued together and allowed to fully dry over multiple days. After about a week each sample was put into a rig that tried to pull the two surfaces apart while measuring the force required to do so.

With e.g. two-part epoxy and super glue the parts would break rather than the glue layer, while with others the glue layer would give way first. All of these results are noted in the above graphic that has the force listed in Newton. The special notes and symbols stand for strong smell (‘Geruch’), the PETG itself breaking (‘Substrat gebrochen’) and high variability (‘hohe Streuung’) between the multiple samples tested per adhesive.

Interesting is that multiple superglues (‘Sekundenkleber’) show different results, while MMA (Methyl Methacrylate) and similar score the highest. The Bostik P580 is a polyurethane construction adhesive, usually used for gluing just about anything to anything in interior and exterior applications, so perhaps its high score isn’t so surprising. Trailing at the end are the wood glue in last place, with the UHU general adhesive also scoring rather poorly.

Clearly there are many options for gluing PETG parts, but some are definitely more sturdy than others.

Thanks to [Risu no Kairu] for the tip.

Continue reading “Comparing Adhesives For Gluing PETG Prints” →

The Guanella 1:1 balun. (Credit: Steve Arar)

Using Guanella Baluns As Impedance Transformers

January 29, 2025 by Maya Posch 2 Comments

Guanella Impedance Transformer. (Credit: FesZ Electronics)

Even before entering the mystical realms of UHF design, radio frequency (RF) circuits come with a whole range of fun design aspects as well. A case in point can be found in transmission line transformers, which are commonly used in RF power amplifiers, with the Guanella transformer (balun) being one example. Allowing balanced and unbalanced (hence ‘balun’) systems to interface without issues, they’re both very simple and very complex. This type of transformer and its various uses is explained in a video by [FesZ Electronics], and also the subject of an article by [Dr. Steve Arar] as part of a larger series, the latter of which is recommended to start with you’re not familiar with RF circuitry.

Transmission line transformers are similar to regular transformers, except that the former relies on transmission line action to transfer energy rather than magnetic flux and provides no DC isolation. The Guanella balun transformer was originally described by Gustav Guanella in 1944. Beyond the 1:1 balun other configurations are also possible, which [Dr. Arar] describes in a follow-up article, and which are also covered in the [FesZ] video, alongside the explanation of another use of Guanella transformers: as an impedance transformer. This shows just how flexible transformers are once you can wrap your mind around the theory.

We have previously covered RF amplifier builds as well as some rather interesting balun hacks.

Heading image: The Guanella 1:1 balun. (Credit: Steve Arar)

Continue reading “Using Guanella Baluns As Impedance Transformers” →

Crystal structure of a monolayer of transition metal dichalcogenide.(Credit: 3113Ian, Wikimedia)

Transition-Metal Dichalcogenides: Super-Conducting, Super-Capacitor Semiconductors

January 28, 2025 by Maya Posch 3 Comments

Transition-metal dichalcogenides (TMDs) are the subject of an emerging field in semiconductor research, with these materials offering a range of useful properties that include not only semiconductor applications, but also in superconducting material research and in supercapacitors. A recent number of papers have been published on these latter two applications, with [Rui] et al. demonstrating superconductivity in (InSe₂)_xNbSe₂. The superconducting transition occurred at 11.6 K with ambient pressure.

Two review papers on transition metal sulfide TMDs as supercapacitor electrodes were also recently published by [Mohammad Shariq] et al. and [Can Zhang] et al. showing it to be a highly promising material owing to strong redox properties. As usual there are plenty of challenges to bring something like TMDs from the laboratory to a production line, but TMDs (really TMD monolayers) have already seen structures like field effect transistors (FETs) made with them, and used in sensing applications.

TMDs consist of a transition-metal (M, e.g. molybdenum, tungsten) and a chalcogen atom (X, e.g. sulfur) in a monolayer with two X atoms (yellow in the above image) encapsulating a single M atom (black). Much like with other monolayers like graphene, molybdenene and goldene, it is this configuration that gives rise to unexpected properties. In the case of TMDs, some have a direct band gap, making them very suitable for transistors and perhaps most interestingly also for directly growing 3D semiconductor structures.

Heading image: Crystal structure of a monolayer of transition metal dichalcogenide.(Credit: 3113Ian, Wikimedia)

Google Open Sources PebbleOS: New Pebble Device In Development

January 28, 2025 by Maya Posch 35 Comments

The Pebble smartwatch was introduced in 2012 as part of a Kickstarter campaign and saw moderate success before the company behind it got bought out by Fitbit. Although a group of enthusiasts kept their Pebble devices alive, including via the alternate Rebble project for online services, it seemed that no new Pebble devices would grace this Earth. However, we now got a flurry of Pebble updates, with Google, the current owner of Fitbit, open sourcing the PebbleOS source, and [Eric Migicovsky] as the original Pebble founder announcing new Pebble watches.

These new Pebble watches would be very much like the original Pebble, though switching from a memory LCD to an e-paper screen but keeping compatibility with the original Pebble watch and its hackability. Currently there’s just a rePebble site where you can sign up for announcements. Over at the Rebble project people are understandably excited, with the PebbleOS source available on GitHub.

A lot of work still remains, of course. The Apache 2.0-licensed PebbleOS source was stripped of everything from fonts to the voice codec and Bluetooth stack, and of course bootstrapping whole new hardware production will require serious investment. Even so, for lovers of smart watches that work with modern-day smartphones, featuring an always-on display and amazing battery life the future has never been more bright.

Thanks to [Will0] for the tip.

New Open Source DeepSeek V3 Language Model Making Waves

January 27, 2025 by Maya Posch 82 Comments

In the world of large language models (LLMs) there tend to be relatively few upsets ever since OpenAI barged onto the scene with its transformer-based GPT models a few years ago, yet now it seems that Chinese company DeepSeek has upended the status quo. Its new DeepSeek-V3 model is not only open source, it also claims to have been trained for only a fraction of the effort required by competing models, while performing significantly better.

The full training of DeepSeek-V3’s 671B parameters is claimed to have only taken 2.788 M hours on NVidia H800 (Hopper-based) GPUs, which is almost a factor of ten less than others. Naturally this has the LLM industry somewhat up in a mild panic, but for those who are not investors in LLM companies or NVidia can partake in this new OSS model that has been released under the MIT license, along with the DeepSeek-R1 reasoning model.

Both of these models can be run locally, using both AMD and NVidia GPUs, as well as using the online APIs. If these models do indeed perform as efficiently as claimed, they stand to massively reduce the hardware and power required to not only train but also query LLMs.