DevHeads

Home
DevHeads

Contact information, map and directions, contact form, opening hours, services, ratings, photos, videos and announcements from DevHeads, Media/News Company, .

DevHeads is a community of professional IoT engineers who openly integrate technology building blocks into connected system PoCs—all while documenting their experience and sharing knowledge with peers.

14/07/2025

🐛 𝗪𝗵𝗲𝗻 𝗧𝗶𝗻𝘆𝗠𝗟 𝗠𝗼𝗱𝗲𝗹𝘀 𝗙𝗮𝗶𝗹 𝗦𝗶𝗹𝗲𝗻𝘁𝗹𝘆

You quantized the model.
You deployed it.
No crash. No errors.
Just... bad results.

Welcome to 𝗱𝗲𝗯𝘂𝗴𝗴𝗶𝗻𝗴 𝗧𝗶𝗻𝘆𝗠𝗟—where your float32 model runs perfectly on your laptop, but gives nonsense on-device. 🎯

Here are some 𝗿𝗲𝗮𝗹 𝗰𝘂𝗹𝗽𝗿𝗶𝘁𝘀 we’ve run into:

🔸 Silent overflows during inference (int8 woes 😵‍💫)
🔸 Preprocessing drift between Python + C versions
🔸 Missing ops in TFLite Micro
🔸 Unstable sensor data or out-of-distribution inputs
🔸 Tiny memory overflows (especially with circular buffers)

💡 Debugging edge ML isn’t just “check your logs.” It’s timing, memory, quantization, and signal integrity—all in one.

𝗙𝗲𝗹𝗹𝗼𝘄 𝗲𝗺𝗯𝗲𝗱𝗱𝗲𝗱 𝗠𝗟 𝗲𝗻𝗴𝗶𝗻𝗲𝗲𝗿𝘀:
🛠️ 𝗪𝗵𝗮𝘁’𝘀 𝘁𝗵𝗲 𝘄𝗲𝗶𝗿𝗱𝗲𝘀𝘁 𝗯𝘂𝗴 𝘆𝗼𝘂’𝘃𝗲 𝗵𝗶𝘁 𝘄𝗶𝘁𝗵 𝗧𝗶𝗻𝘆𝗠𝗟?
🎯 What’s your best trick for figuring out why the model misbehaves on-device?

👇 𝑫𝒓𝒐𝒑 𝒚𝒐𝒖𝒓 𝒔𝒕𝒐𝒓𝒊𝒆𝒔 𝒊𝒏 𝒕𝒉𝒆 𝒄𝒐𝒎𝒎𝒆𝒏𝒕𝒔 — 𝒘𝒆’𝒍𝒍 𝒇𝒆𝒂𝒕𝒖𝒓𝒆 𝒔𝒐𝒎𝒆 𝒐𝒇 𝒕𝒉𝒆 𝒃𝒆𝒔𝒕 𝒐𝒏𝒆𝒔 𝒏𝒆𝒙𝒕 𝒘𝒆𝒆𝒌.

09/07/2025

🧩 𝗠𝗼𝗱𝗲𝗹 𝗗𝗲𝘀𝗶𝗴𝗻 𝗦𝘁𝗿𝗮𝘁𝗲𝗴𝗶𝗲𝘀 𝗳𝗼𝗿 𝗧𝗶𝗻𝘆𝗠𝗟 𝗼𝗻 𝗠𝗶𝗰𝗿𝗼𝗰𝗼𝗻𝘁𝗿𝗼𝗹𝗹𝗲𝗿𝘀

Designing ML for microcontrollers isn’t about shrinking a model—it’s about 𝗮𝗿𝗰𝗵𝗶𝘁𝗲𝗰𝘁𝗶𝗻𝗴 𝗳𝗼𝗿 𝗰𝗼𝗻𝘀𝘁𝗿𝗮𝗶𝗻𝘁𝘀:

⚙️ 𝗠𝗖𝗨 𝗖𝗼𝗻𝘀𝘁𝗿𝗮𝗶𝗻𝘁𝘀:

07/07/2025

🔧 𝗝𝗼𝗶𝗻 𝘁𝗵𝗲 𝗗𝗲𝘃𝗛𝗲𝗮𝗱𝘀 𝗖𝗼𝗺𝗺𝘂𝗻𝗶𝘁𝘆
IoT • Embedded Systems • Edge AI • RTOS • Electronics • DSP • Firmware • Data Mining....

Build at the intersection of hardware and intelligence.
From microcontrollers to ML models, sensors to signal processing — DevHeads is where deep-tech engineers, hackers, and tinkerers connect.

💬 Real builders. Smart tools. Zero fluff.

👉 https://hubs.ly/Q03w67fF0

04/07/2025

𝗛𝗮𝗽𝗽𝘆 𝗜𝗻𝗱𝗲𝗽𝗲𝗻𝗱𝗲𝗻𝗰𝗲 𝗗𝗮𝘆 𝗳𝗿𝗼𝗺 𝗗𝗲𝘃𝗛𝗲𝗮𝗱𝘀! 🇺🇸
This 4th of July, while the skies light up with fireworks, we're thinking about the real spark behind progress: 𝗲𝗻𝗴𝗶𝗻𝗲𝗲𝗿𝘀 𝘄𝗵𝗼 𝗯𝘂𝗶𝗹𝗱, 𝘁𝗲𝘀𝘁, 𝗮𝗻𝗱 𝗶𝗻𝘁𝗲𝗴𝗿𝗮𝘁𝗲 across every layer of hardware and software.

Whether you're wrangling sensor data, fine-tuning embedded code, or scaling cloud pipelines—your work fuels the systems that keep the world connected and moving forward.

Today’s a reminder that freedom means the power to create, experiment, and push boundaries—exactly what was built for.

💥 𝑩𝒖𝒊𝒍𝒅 𝒔𝒂𝒇𝒆𝒍𝒚. 𝑪𝒆𝒍𝒆𝒃𝒓𝒂𝒕𝒆 𝒔𝒎𝒂𝒓𝒕𝒍𝒚. 𝑾𝒆’𝒍𝒍 𝒔𝒆𝒆 𝒚𝒐𝒖 𝒐𝒏 𝒕𝒉𝒆 𝒏𝒆𝒙𝒕 𝒄𝒐𝒎𝒎𝒊𝒕.

01/07/2025

𝐓𝐢𝐧𝐲𝐌𝐋 𝐨𝐧 𝐌𝐢𝐜𝐫𝐨𝐜𝐨𝐧𝐭𝐫𝐨𝐥𝐥𝐞𝐫𝐬: 𝐒𝐭𝐚𝐭𝐢𝐜 𝐌𝐞𝐦𝐨𝐫𝐲 > 𝐦𝐚𝐥𝐥𝐨𝐜()
Following up on our post about real-time sensor preprocessing — let’s talk memory.

Most MCUs (STM32, ESP32, nRF52) give you

27/06/2025

📍 Ever wondered how CAN arbitration really works? In this clip from our session “𝗠𝘂𝗹𝘁𝗶𝗽𝗹𝗲 𝗖𝗔𝗡 𝗡𝗼𝗱𝗲 𝗙𝗶𝗿𝗺𝘄𝗮𝗿𝗲 𝘄𝗶𝘁𝗵 𝗦𝗧𝗠𝟯𝟮 𝗔𝗥𝗠 𝗖𝗼𝗿𝘁𝗲𝘅-𝗠𝟰 𝗠𝗖𝗨” we explore how different node priorities determine who transmits first — and why lower CAN IDs mean higher priority.

💡 This is just one of the hands-on demos from the event. Follow us on to stay updated on future sessions and more practical embedded systems content!

25/06/2025

🔬 𝗥𝗲𝗮𝗹-𝗧𝗶𝗺𝗲 𝗦𝗲𝗻𝘀𝗼𝗿 𝗣𝗿𝗲𝗽𝗿𝗼𝗰𝗲𝘀𝘀𝗶𝗻𝗴 𝗮𝘁 𝘁𝗵𝗲 𝗘𝗱𝗴𝗲
When you’re working with raw sensor data on a microcontroller — whether it’s an IMU, a microphone, or a vibration sensor — smart preprocessing is what enables real-time inference.

Before your model ever sees a single value, a lot has to happen — and it has to happen fast:

✅ 𝗙𝗶𝗹𝘁𝗲𝗿𝗶𝗻𝗴 — Moving average, Butterworth, or high-pass filters to reduce noise and drift.
✅ 𝗡𝗼𝗿𝗺𝗮𝗹𝗶𝘇𝗮𝘁𝗶𝗼𝗻 — Zero-mean, unit-variance, or simple min–max scaling to match your model’s expectations.
✅ 𝗙𝗲𝗮𝘁𝘂𝗿𝗲 𝗘𝘅𝘁𝗿𝗮𝗰𝘁𝗶𝗼𝗻 — RMS, zero-crossing rate, spectral entropy — all give your model richer inputs.
✅ 𝗙𝗿𝗲𝗾𝘂𝗲𝗻𝗰𝘆-𝗱𝗼𝗺𝗮𝗶𝗻 𝗔𝗻𝗮𝗹𝘆𝘀𝗶𝘀 — An FFT can do wonders for periodic signals like audio or vibration.
✅ 𝗙𝗶𝘅𝗲𝗱-𝗣𝗼𝗶𝗻𝘁 𝗠𝗮𝘁𝗵 — Floats are costly on MCUs. Scaling and shifting efficiently is the secret sauce.

And here’s the kicker: all of this must happen in under 10 ms if you want to maintain real-time performance. Every cycle counts — this is edge AI at its most efficient.

💡 𝑺𝒎𝒂𝒓𝒕 𝒑𝒓𝒆𝒑𝒓𝒐𝒄𝒆𝒔𝒔𝒊𝒏𝒈 𝒊𝒔 𝒕𝒉𝒆 𝒓𝒆𝒂𝒍 𝒉𝒆𝒓𝒐 𝒐𝒇 𝒕𝒊𝒏𝒚𝑴𝑳 — 𝒍𝒆𝒂𝒏, 𝒇𝒂𝒔𝒕, 𝒂𝒏𝒅 𝒃𝒖𝒊𝒍𝒕 𝒇𝒐𝒓 𝒑𝒖𝒓𝒑𝒐𝒔𝒆. 💪

23/06/2025

That’s a quick demo of how these CAN nodes communicate with each other using the switches and LEDs.

If you’d like to see the full session — “𝗠𝘂𝗹𝘁𝗶𝗽𝗹𝗲 𝗖𝗔𝗡 𝗡𝗼𝗱𝗲 𝗙𝗶𝗿𝗺𝘄𝗮𝗿𝗲 𝘄𝗶𝘁𝗵 𝗦𝗧𝗠𝟯𝟮 𝗔𝗥𝗠 𝗖𝗼𝗿𝘁𝗲𝘅-𝗠𝟰 𝗠𝗖𝗨” — where we dive into multi-node CAN setups, STM32 firmware, and practical hands-on examples, check out the recording on our YouTube channel: https://hubs.ly/Q03tb-470

And if you want to connect with other engineers, ask questions, and share your projects, come join our Discord community! https://hubs.ly/Q03tbZr00

𝑻𝒉𝒂𝒏𝒌𝒔 𝒇𝒐𝒓 𝒘𝒂𝒕𝒄𝒉𝒊𝒏𝒈 — 𝒂𝒏𝒅 𝒔𝒆𝒆 𝒚𝒐𝒖 𝒕𝒉𝒆𝒓𝒆!

20/06/2025

Multi Mater Nodes
CAN Wiring and ranges
Setup Two CAN Node in Normal mode
Example Project: Communicate Two CAN Node using STM32 ARM Cortex-M MCU

18/06/2025

🧠𝗪𝗵𝗮𝘁 𝗜𝘀 𝘁𝗵𝗲 𝗖𝗔𝗡 𝗕𝘂𝘀 𝗣𝗵𝘆𝘀𝗶𝗰𝗮𝗹 𝗟𝗮𝘆𝗲𝗿?
If you’re building embedded systems with STM32 or other microcontrollers, 𝘂𝗻𝗱𝗲𝗿𝘀𝘁𝗮𝗻𝗱𝗶𝗻𝗴 𝘁𝗵𝗲 𝗽𝗵𝘆𝘀𝗶𝗰𝗮𝗹 𝗹𝗮𝘆𝗲𝗿 𝗼𝗳 𝗖𝗔𝗡 is essential.

Here’s a quick breakdown of the topology shown:
🔹 𝗠𝗮𝗶𝗻 𝗕𝗮𝗰𝗸𝗯𝗼𝗻𝗲 – A twisted pair (CANH & CANL) that forms the primary communication path.
🔹𝗦𝘁𝘂𝗯𝘀 – Short branch connections from each node to the backbone. Keep these short to avoid signal degradation.
🔹𝗧𝗲𝗿𝗺𝗶𝗻𝗮𝘁𝗶𝗼𝗻 𝗥𝗲𝘀𝗶𝘀𝘁𝗼𝗿𝘀 – Placed at both ends of the backbone to prevent reflections and ensure signal integrity.
🔹𝗖𝗔𝗡 𝗠𝗼𝗱𝘂𝗹𝗲𝘀 – Each module (or "node") communicates on the shared line in a 𝗱𝗲𝗰𝗲𝗻𝘁𝗿𝗮𝗹𝗶𝘇𝗲𝗱, 𝗺𝘂𝗹𝘁𝗶-𝗺𝗮𝘀𝘁𝗲𝗿 𝗻𝗲𝘁𝘄𝗼𝗿𝗸.

🎯 Want to 𝗯𝘂𝗶𝗹𝗱 𝗮𝗻𝗱 𝗳𝗹𝗮𝘀𝗵 𝗳𝗶𝗿𝗺𝘄𝗮𝗿𝗲 for your own multi-node CAN system?

Join us 𝗙𝗿𝗶𝗱𝗮𝘆, 𝗝𝘂𝗻𝗲 𝟮𝟬 𝗮𝘁 𝟭𝟱:𝟬𝟬 𝗨𝗧𝗖. for a hands-on session: "𝗠𝘂𝗹𝘁𝗶𝗽𝗹𝗲 𝗖𝗔𝗡 𝗡𝗼𝗱𝗲 𝗙𝗶𝗿𝗺𝘄𝗮𝗿𝗲 𝗼𝗻 𝗦𝗧𝗠𝟯𝟮 𝗔𝗥𝗠 𝗖𝗼𝗿𝘁𝗲𝘅-𝗠𝟰"
🔗 𝗝𝗼𝗶𝗻 𝗵𝗲𝗿𝗲: https://hubs.ly/Q03sFXPt0

16/06/2025

𝗠𝘂𝗹𝘁𝗶𝗽𝗹𝗲 𝗖𝗔𝗡 𝗡𝗼𝗱𝗲 𝗙𝗶𝗿𝗺𝘄𝗮𝗿𝗲 𝗼𝗻 𝗦𝗧𝗠𝟯𝟮 𝗔𝗥𝗠 𝗖𝗼𝗿𝘁𝗲𝘅-𝗠𝟰 🚀
Join us for an in-depth technical session where we’ll get hands-on with 𝗺𝘂𝗹𝘁𝗶-𝗻𝗼𝗱𝗲 𝗖𝗔𝗡 𝗯𝘂𝘀 𝗰𝗼𝗺𝗺𝘂𝗻𝗶𝗰𝗮𝘁𝗶𝗼𝗻 using the 𝗦𝗧𝗠𝟯𝟮 𝗔𝗥𝗠 𝗖𝗼𝗿𝘁𝗲𝘅-𝗠𝟰 𝗺𝗶𝗰𝗿𝗼𝗰𝗼𝗻𝘁𝗿𝗼𝗹𝗹𝗲𝗿. Whether you're building automotive systems or just love embedded development, this one’s for you!

🔧 𝗪𝗵𝗮𝘁 𝗬𝗼𝘂’𝗹𝗹 𝗟𝗲𝗮𝗿𝗻:
✔️ Setting up 𝗺𝘂𝗹𝘁𝗶-𝗺𝗮𝘀𝘁𝗲𝗿 𝗖𝗔𝗡 𝗻𝗼𝗱𝗲𝘀 using STM32 Nucleo boards
✔️ Wiring, bus topology & termination for CAN networks
✔️ Building and deploying firmware for 𝘁𝘄𝗼 𝗰𝗼𝗺𝗺𝘂𝗻𝗶𝗰𝗮𝘁𝗶𝗻𝗴 𝗖𝗔𝗡 𝗻𝗼𝗱𝗲𝘀
✔️ Real-time demo: 𝗧𝗼𝗴𝗴𝗹𝗲 𝗟𝗘𝗗𝘀 𝗮𝗰𝗿𝗼𝘀𝘀 𝗻𝗼𝗱𝗲𝘀 𝘃𝗶𝗮 𝗖𝗔𝗡 𝗺𝗲𝘀𝘀𝗮𝗴𝗲𝘀
✔️ How to 𝗰𝗮𝗹𝗰𝘂𝗹𝗮𝘁𝗲 𝗖𝗔𝗡 𝗯𝘂𝘀 𝗹𝗼𝗮𝗱 to ensure efficient communication

🛠 𝗟𝗶𝘃𝗲 𝗣𝗿𝗼𝗷𝗲𝗰𝘁 𝗪𝗮𝗹𝗸𝘁𝗵𝗿𝗼𝘂𝗴𝗵
We'll 𝗰𝗼𝗱𝗲 𝗮𝗻𝗱 𝘁𝗲𝘀𝘁 a real-world example where two CAN nodes exchange button press signals and toggle LEDs using standard CAN IDs (0x322, 0x422) — all in 𝗻𝗼𝗿𝗺𝗮𝗹 𝗺𝗼𝗱𝗲 𝗰𝗼𝗺𝗺𝘂𝗻𝗶𝗰𝗮𝘁𝗶𝗼𝗻.

🗓 𝗪𝗵𝗲𝗻: 𝗙𝗿𝗶𝗱𝗮𝘆, 𝗝𝘂𝗻𝗲 𝟮𝟬 𝗮𝘁 𝟭𝟱:𝟬𝟬 𝗨𝗧𝗖
📍 𝗪𝗵𝗲𝗿𝗲: https://hubs.ly/Q03s915M0

🔁 𝑩𝒓𝒊𝒏𝒈 𝒚𝒐𝒖𝒓 𝒒𝒖𝒆𝒔𝒕𝒊𝒐𝒏𝒔, 𝒄𝒐𝒅𝒆 𝒂𝒍𝒐𝒏𝒈𝒔𝒊𝒅𝒆, 𝒂𝒏𝒅 𝒍𝒆𝒕’𝒔 𝒕𝒊𝒏𝒌𝒆𝒓 𝒕𝒐𝒈𝒆𝒕𝒉𝒆𝒓!

12/06/2025

The CAN transceiver is a key component that transforms real physical signals—CAN High and CAN Low—into digital logic that microcontrollers can understand.

It receives raw electrical data from the CAN bus, converts it into logic 0s and 1s, and hands it off to the CAN controller, which sets the baud rate and interprets the message before passing it on to the microcontroller.

This is how your embedded system stays in sync with the physical world. ⚙️💡
𝗪𝗮𝘁𝗰𝗵 𝘁𝗵𝗲 𝗳𝘂𝗹𝗹 𝘀𝗲𝘀𝘀𝗶𝗼𝗻 𝗼𝗻 𝗬𝗼𝘂𝗧𝘂𝗯𝗲: https://hubs.ly/Q03rWnLf0

𝗡𝗲𝘅𝘁 𝘂𝗽: June 20: 𝗠𝘂𝗹𝘁𝗶𝗽𝗹𝗲 𝗖𝗔𝗡 𝗡𝗼𝗱𝗲 𝗙𝗶𝗿𝗺𝘄𝗮𝗿𝗲 𝘄𝗶𝘁𝗵 𝗦𝗧𝗠𝟯𝟮 𝗔𝗥𝗠 𝗖𝗼𝗿𝘁𝗲𝘅-𝗠𝟰 𝗠
🔗 𝗥𝗲𝗴𝗶𝘀𝘁𝗲𝗿 𝗻𝗼𝘄: https://hubs.ly/Q03rWp_N0

Address

Website

https://linktr.ee/DevHeads

Alerts

Be the first to know and let us send you an email when DevHeads posts news and promotions. Your email address will not be used for any other purpose, and you can unsubscribe at any time.

Shortcuts

Address
Alerts
Claim ownership or report listing
Want your business to be the top-listed Media Company?

Country:

City: