The Tech Blog

Back To Basics: Eddy Current Sensors

Posted under Back to Basics, Technology on March 1st, 2026 by Amaldev

I was doing some research for a client project this week. The solution eventually needed some eddy current displacement sensors. It’s the kind of tool you forget until contact measurement starts to become a problem. Let’s discuss that this week.

An eddy current displacement sensor is basically a coil driven with a high-frequency AC signal. That alternating field induces circulating eddy currents in any nearby conductive metal. By Lenz’s law, the magnetic field produced by those eddy currents opposes the probe’s field. For the same drive current, the probe coil stores less magnetic energy because some of the field is being pushed back. From the coil’s point of view that shows up as a drop in effective inductance, so the inductive reactance term falls. At the same time, energy is dissipated as I²R heating in the target, which the coil feels as added loss, so the effective resistance rises.

The probe electronics drive the coil at a fixed frequency and measure the resulting voltage/current amplitude and phase, infer the complex impedance (R+jωL), and then use a stored calibration curve to convert that impedance change into a distance output.

These sensors are used when you need reliable distance or vibration data in harsh conditions. Like maintaining tip clearance, predictive maintenance on rotating equipment cases, measuring shaft runout etc. They’re popular because they’re non-contact, fast, and can work in all the worse places with oil mist, dust, coolant spray, and even high-temperature scenarios.

The biggest tradeoff is that they only see conductive targets, range is usually a few mm (to low cm), and readings shift with material, geometry, EMI pickup, and temperature, so mounting, shielding, and calibration matter in real machines. Most companies providing the sensors, also do give the matching drivers with temperature calibration as a full set.

If you plan on using multiple of these units, please space them out so that one sensor’s field doesn’t affect the field of the other. Some systems use different drive frequencies for adjacent units.

If you liked the post, Share it with your friends!

Tech Explained: Taalas AI Inference Chip

Posted under Tech Explained on February 22nd, 2026 by Amaldev

A couple of days back I heard about an interesting new startup called Taalas, and they are doing some interesting work in the LLM inference space. I thought we’ll try to deep dive on what I found about that today.

They take a trained model like Llama 3.1 8B and turn it into a fixed chip. Instead of a GPU pulling weights from High Bandwidth Memory(HBM) every token, the weights are baked into silicon in a big ROM like fabric. The model and the hardware are basically the same thing, so you are really buying a specific model in PCIe card form.

On their first part, HC1, they talk about roughly 17k tokens per second per user for that 8B model at around 200W power. The nearest competitor is the Cerebras chip which does around 2k tokens per sec. With every weight on the die and no HBM in the loop, most of the work becomes local switching instead of moving GBs back and forth. To visualize it, think of a grid where each weight is a tiny logic cell, not a number in memory. When an activation vector arrives, that grid lights up along fixed paths and each cell contributes its small multiply and add. The layer output appears with almost no indexing overhead. SRAM around it holds KV cache and adapters, so you can still add LoRA style tweaks. Try out at chatjimmy.ai It’s mind-blowingly fast, the moment you hit the Enter key.

Where could that be useful? Anywhere you are happy to standardise on a stable model and really care about latency and cost per query. Meaning robots, edge devices, superfast agentic frameworks etc.

Am I really convinced if this is the way ahead? I am not sure. Since it’s baked into the Silicon, you loose flexibility when you want to change things. You need big deployment volumes to make the economics work. By the time a tape out happens(they say 60days), the next generation model would be out there. Also not sure how it scales for large Trillion+ token models. A classic case is for stable popular older models like ChatGPT 4o if its weights are open sourced, its fanboys would love to bake it with Taalas and use for deployment since OpenAI sunsetted the model last week. There is potential lets see where this goes.

If you liked the post, Share it with your friends!