Thomas

I measure things that break. 20yr T&M · Germany · Signal Integrity / Power / RF / EMC. Real field notes from real labs →

[Thomas @SignalByThomas]

This is what happens when you leave a service engineer alone with a function generator.
Our service engineer Hammer was supposed to test the waveform generator.
Instead, he used it to draw a dancing stick figure on the oscilloscope.
Then he decided to join the dance himself.
Engineering.
Not always serious.
Sometimes this happens.
🤣
#Oscilloscope
#WaveformGenerator
#EngineeringLife
#Electronics
#TestAndMeasurement

[Thomas @SignalByThomas]

As an engineer, I think repairability and usability are often two different discussions.
A product can be repairable and still fail commercially.
A product can be difficult to repair and still become the preferred tool for getting real work done.
Engineers appreciate repairability.
Customers often optimize for productivity.

[Thomas @SignalByThomas]

Sector-level indication, not precision tracking. That constraint is a design decision, not a limitation.
The engineers who understand the difference build things that actually ship.
#EdgeAI #ESP32 #SignalProcessing #Acoustics #I2MTC #EmbeddedSystems #MEMS

[Thomas @SignalByThomas]

https://t.co/pqmbLlLPs5

[Thomas @SignalByThomas]

SoundSL v1.0 — ESP32-WROOM-32U + 4× MEMS via I2S + GCC-PHAT TDOA pipeline.
8-sector azimuth output. 128 samples/window at 16kHz. 200–400kB RAM. Update rate 26.7Hz.
Best quadrant: 84.8%. Worst: 70.7%. Honest failure mode documentation included.
University of Salerno + Mid Sweden University. IEEE I2MTC 2026.

[Thomas @SignalByThomas]

A 1976 algorithm. A microcontroller that costs less than €10. Four MEMS microphones.
80.1% direction accuracy. 7.5ms latency. Runs locally. No cloud. No neural network.
I2MTC 2026, Nancy. A poster that asked the question most AI roadmaps avoid.
How much compute do you actually need to make a useful decision?

[Thomas @SignalByThomas]

https://t.co/pqmbLlLPs5

[Thomas @SignalByThomas]

I almost spent this weekend in Paris.

After four days at IEEE I2MTC 2026 in Nancy, I planned to take a train to Paris, meet a few friends from the industry, and enjoy a quiet weekend.

A French friend gave me a simple piece of advice:

"Don't come this weekend."

PSG was playing the Champions League Final.

He told me that whether PSG wins or loses, parts of Paris often become chaotic afterwards.

So instead, I went to Versailles.

What struck me was how different the experiences were.

The researchers I met in Nancy were thoughtful, polite, and passionate about measurement science.

Versailles was calm, elegant, and peaceful.

Yet only a short train ride away, another France was making headlines around the world.

After more than 20 years in Europe, I've learned something:

The country you see on television and the country you experience in person are often two very different places.

Perhaps that's part of Paris' charm.

Or perhaps that's simply modern Europe.

[Luc Auffret @LucAuffret]

La situation dégénère près du Parc des Princes après la victoire du PSG.

Plusieurs départs de feux. Charge de la police pour faire intervenir les pompiers.

#PSG #PSGARS #Arsenal #UCLfinal #Paris

[Thomas @SignalByThomas]

@Em_Nomadic @tnkrdotai Interesting.

Coming from the test & measurement industry, I see a similar problem.

Building hardware is often easier than understanding the ecosystem around it.

Documentation, workflows, integration and community usually matter more than people expect.

[Thomas @SignalByThomas]

Proud to support IEEE I2MTC 2026 as a Gold Sponsor together with so many respected names in the instrumentation and measurement industry.

Events like this are important not only for showcasing technologies, but also for connecting engineers, researchers, universities, and industry together.

The future of measurement is becoming increasingly interdisciplinary:
AI, power electronics, edge computing, RF, automotive, energy systems…

And behind all of them, measurement remains the foundation.

Looking forward to more conversations in Nancy this week.

[Thomas @SignalByThomas]

At I2MTC 2026 in Nancy, our RF demo looked almost too simple.

A signal generator.
A spectrum analyzer.
One coax cable between them.
But after standing there for a while, I realized the real question wasn’t about RF performance.
It was philosophical.

The signal generator defines what the analyzer should see.
The analyzer tells you what the generator actually produced.

So who is measuring whom?

The setup:
UNI-T USG5014M → UTS3084A
One creates the signal.
One validates it.
A closed RF loop.

And honestly — this is how most real engineering decisions are made.

Not inside national metrology institutes.
Not inside perfect calibration chains.
But on benches like this.

A stable spectrum trace tells you what exists now.
The waterfall tells you whether it stays true over time.
That difference matters more than many engineers realize.
#RF #SpectrumAnalyzer #SignalGenerator #Metrology #SignalIntegrity #Engineering #EmbeddedSystems #Electronics #I2MTC #TestAndMeasurement #UNI_T

[Thomas @SignalByThomas]

https://t.co/PP4Tdjc7JG

[Thomas @SignalByThomas]

https://t.co/PP4Tdjc7JG

[Thomas @SignalByThomas]

This was the Lattice booth I photographed at Embedded World in Nuremberg.

No giant GPU demos.
No flashy marketing.
No giant AI claims.
Just three words:
Connectivity.
Security.
Vision.

Sometimes the quietest booths at engineering trade shows represent the biggest long-term shifts.

#EmbeddedWorld #Lattice #Engineering

[Thomas @SignalByThomas]

Everyone in AI is watching GPUs.

But after walking through Embedded World in Germany this year, I increasingly feel the market may be missing another layer entirely:

FPGA.

Not sexy.
Not consumer-facing.
Almost invisible to most investors.
But in real embedded AI systems —
industrial vision,
robotics,
sensor fusion,
edge inference,
real-time control —
these chips quietly sit between sensors and AI accelerators, doing the work nobody talks about.

As someone working around embedded systems and test equipment in Europe,
this doesn’t feel theoretical anymore.
AI is no longer staying inside datacenters.
It’s starting to enter the physical world.
And physical-world AI has very different hardware needs.
#AI #FPGA #EdgeAI #EmbeddedSystems #IndustrialAI

[Thomas @SignalByThomas]

While everyone debated NVIDIA vs AMD GPUs,
Lattice Semiconductor ($LSCC) quietly moved from ~$85 to ~$143 in less than 2 months.
That kind of move usually means:
the market is beginning to notice something structural.

Not every AI problem needs a 700W GPU.
Some need:
low latency,
deterministic behavior,
sensor synchronization,
real-time IO,
low power consumption,
industrial reliability.

That’s FPGA territory.

#LSCC #FPGA #EdgeComputing #IndustrialAI

[Thomas @SignalByThomas]

At IEEE #I2MTC2026 today.

Interesting contrast:

Everyone talks about AI.

But inside the conference halls,
engineers are still focused on:

signals,
measurement fidelity,
RF sensing,
real-time acquisition,
synchronization,
edge processing.

AI does not replace instrumentation.

It amplifies the value of good instrumentation.

That’s why high-quality signal acquisition still matters.

#RF
#Oscilloscope
#AI
#Engineering
#SignalIntegrity
#EmbeddedSystems
#UNI_T

[Thomas @SignalByThomas]

At IEEE #I2MTC2026 in France today.

One thing became very clear:

AI sensing does NOT start with AI.

It starts with:

reliable instrumentation
clean signal acquisition
synchronized data
real-world measurement fidelity

Many talks today focused on RF sensing, industrial AI, radar imaging, biomedical monitoring and edge systems.

But behind every “smart” AI system, there is still a very old engineering truth:

Bad signals create bad intelligence.

That’s why high-fidelity measurement systems still matter.

Real engineering.
Real waveforms.
Real systems.

#AI
#RF
#EdgeAI
#Oscilloscope
#SignalIntegrity
#Instrumentation
#EmbeddedSystems
#IEEE
#Sensors
#Measurement

[Thomas @SignalByThomas]

At IEEE #I2MTC2026 in France today.

One thing became very clear:

AI sensing does NOT start with AI.

It starts with:

reliable instrumentation
clean signal acquisition
synchronized data
real-world measurement fidelity

Many talks today focused on RF sensing, industrial AI, radar imaging, biomedical monitoring and edge systems.

But behind every “smart” AI system, there is still a very old engineering truth:

Bad signals create bad intelligence.

That’s why high-fidelity measurement systems still matter.

Real engineering.
Real waveforms.
Real systems.

#AI
#RF
#EdgeAI
#Oscilloscope
#SignalIntegrity
#Instrumentation
#EmbeddedSystems
#IEEE
#Sensors
#Measurement

[Thomas @SignalByThomas]

What actually drove AMD's stock to double in a month?I think I saw it coming — back in March, at the AMD booth in Nuremberg.Four demo zones: Robotics. Medical Ultrasound. Secure Industrial Control. Automotive.Not a single word about GPUs.Everyone's asking if AMD can beat NVIDIA. Wrong question.The real AI deployment battlefield is in factories, operating rooms, and vehicles — where you need FPGA and embedded SoC, not H100s.The market took two months to price #AMD #FPGA #PhysicalAI #EmbeddedAI

[Thomas @SignalByThomas]

One thing I still like about engineering culture in Europe:

students still build real things.

Not pitch decks.
Not AI wrappers.
Not growth hacks.
Actual flying systems.
And sometimes:
they fail,
they crash,
they overheat,
they oscillate.

That’s usually where the real learning begins.

[Thomas @SignalByThomas]

Most people see a student drone team.

Engineers see:
power systems,
thermal problems,
RF noise,
vibration,
sensor fusion,
firmware bugs,
battery instability,
and 2AM debugging sessions.
Real engineering education rarely starts with AI.
It starts when the drone suddenly falls out of the sky.

[Thomas @SignalByThomas]

One of the most underrated engineering skills:

finding the real problem.

The waveform may look fine.
The firmware may compile.
The motor may spin.
And the system still fails.
Because complex systems rarely fail in one place.