Blog

Posted on

Why Great Engineering Doesn’t Happen in Meetings (And Why I Still Vacuum My Own Lab)

When you build test benches for electric motors, you’re not just connecting wires. You’re orchestrating a complex symphony of high-current power, deterministic software, and mechanical precision. Over the last 20+ years, I’ve learned that the best way to ensure this symphony doesn’t hit a sour note is to eliminate the “walls” between departments.

The “5-in-1” Engineering Advantage

In large corporations, a single test bench project might pass through five different departments. Information gets lost, responsibilities are shifted, and the “Core” of the project loses its focus.

Since 2002, I have taken a different path. I handle the entire technological stack myself—not because I have to, but because it’s the only way to guarantee the level of quality my clients expect for their rotabench systems.

  • Power Electronics & EDA: I design the PCBs for high-current applications up to 150A AC. Because I draw the traces myself, I know exactly how the hardware will behave under load.
  • Embedded Software Architecture: I write the C-code for the microcontrollers from scratch—no bloated libraries, just high-performance, deterministic execution (including custom AES-256 encryption for secure firmware updates).
  • System Integration: As a (former – 2007) certified LabVIEW Architect, I bridge the gap between low-level hardware and high-level user interfaces, creating a seamless data flow from the sensor to the report.
  • Mechanical Construction: I design the physical housing and fixtures. Knowing exactly where a solder joint sits or how a high-current cable must be routed changes how you approach mechanical design.

Ownership Down to the Last Detail
Running my own company means I am the CEO, the Lead Architect, the Sales Guy, and—quite literally—the cleaning crew. If you visit my lab, you’ll find it spotless because I vacuum it myself.

Why does this matter to you? Because this level of “Ownership” is baked into every product I ship. If I’m willing to personally ensure my workspace is clean, imagine the level of care I put into your 50kHz control loop or your safety-critical shutdown logic.

No Middleware. No Excuses.

When you work with me, you’re talking directly to the guy who designed the circuit, wrote the code, and tested the 150A power stage. There are no “project managers” to translate your requirements into something the engineers don’t understand.

Just hardcore engineering, 20+ years of experience, and a deep respect for the physics of electric motors.

Posted on

Stop Solving Integration Problems – Start Delivering Solved Problems

The Engineering Reality

In the last 20 years I was building test benches, I often had to spend way too much time for my taste with low-level integration of test bench components. The lesson I learned was: you don’t get paid for solving problems, you get paid for solved problems! If that sounds like a stupid marketing claim to you, just ask yourself: does your customer pay you if you spend hundreds of hours to work on a problem you could not solve? Most probably not. Your customer expects you to deliver a working solution, not a bucket full of unsolved issues.

Lesson #2: When you encounter technical issues—especially with low-level integration—it will always use more resources to solve them than you’d like. Your best and most efficient option is to mitigate such potential issues and use a solution you can buy on the market.

This is why I created rotabench EPS: it lets me (and you) leapfrog over most of the low-level integration issues and focus on functionality instead.

The following clip shows the startup sequence of a system where the “low-level heavy lifting” is already done. It’s about getting to the point where you can actually do your job—testing.

Why this matters for your workflow:

  • Risk Mitigation: By using a pre-integrated stack, you eliminate the “unknown unknowns” of hardware communication.
  • Resource Allocation: Stop wasting senior engineering hours on driver conflicts or bus timing.
  • Predictable Timelines: A deterministic startup means a predictable project schedule.
Posted on

The Fast Track to Professional Test Bench Integration

Building a rotating test bench from scratch is a high-risk investment of engineering hours. Often, the “Entry-Level Burdens” of low-level hardware integration consume the majority of your project budget before the first measurement is even taken. With the rotabench® EPS Mod. 5, we are changing the game for system integrators.

The Challenge: Bridging the “Low-Level” Gap

rotating test bench with rotabench EPS test bench control system
rotating test bench with rotabench EPS test bench control system

The major challenge in system integration is bridging the gap between tedious hardware-level tasks and high-level application functionality. A standard test bench requires at least an industrial drive, torque sensors, safety controllers, and an automation PC. The time spent programming inverters or debugging sensor signals is time lost for developing your unique application logic.

The Solution: rotabench® EPS as your Unified Link

rotabench EPS as abstraction layer
rotabench EPS as abstraction layer

The rotabench® EPS acts as an intelligent link between your high-level application and the physical hardware. It provides a pre-engineered core that standardizes the interaction between all components.

  • Drive Control: Direct speed or torque control without the need for deep-level inverter programming.
  • Torque & Speed: 1:1 plug-in for professional sensors (e.g., Kistler 4503) with integrated high-speed signal processing.
  • Integrated Safety: Built-in interface to bridge the emergency stop circuit between hardware and software.
  • Clean Interface: All data is pre-processed and delivered via one single TCP/IP or CAN interface.

Infinite Markets – One Core

Anything that rotates needs to be tested. The rotabench® EPS is the foundation that grows with your needs across various industries:

Automotive & E-MobilityHome & ProfessionalIndustrial & Green Tech
E-Bike & Micro-DrivesBattery Power ToolsHeat Pump Compressors
Chassis Actuators (EPS)Garden TechWind Energy Components
Active SuspensionPersonal CareConveyor & AGV Drives

Scalability Without Limits

We designed the rotabench® EPS to be the beginning, not the end. It makes complex hardware as easy to handle as a digital multimeter with an Ethernet port.

  • Software that Scales: Start with the included Basic Edition for commissioning and upgrade to the Professional Suite for full automation
  • Open for your Code: Seamless integration into your own LabVIEW, Python, or C# environments via Open API.
  • Form Factors: Available as a barebone PCB, in a DIN-rail housing (Plastic/Metal), or as a 19″ rack insert.
PDF icon

download the “system integrators cookbook”

Posted on

Small Device – big impact

From 19’’ racks to a single PCB: Streamlining Dyno Test Benches with the rotabench EPS.

rotabench EPS promo flyer page 1
rotabench EPS promo flyer page 1

I am not a marketing guy, I am an engineer trying to show the world the solutions I created. So please forgive me if this is not the perfect marketing pitch. On the other hand I can guarantee that my products are way better than my marketing 😉 … What I want to show you today is my new product: the rotabench EPS:

It is a test bench control PCB designed to combine and control all four major components of a dyno test bench — drive, torque sensor, encoder, and safety circuit — into one easy-to-use abstraction layer. Actually, it isn’t so new at all; the first versions go back to 2020. However, they were never intended as standard solutions for a wider market, but rather as custom builds for specific projects.

The main driver behind this project was my need for an extension board for my rotabench 6P lab-inverter for low-voltage electric motors. Without an active dyno test bench to brake the DUT (Device Under Test), an inverter is relatively useless. You can speed up the DUT and measure its idle behavior… yay, great. Not.

You need braking force to generate torque. Under these conditions, it’s a huge advantage if the inverter can also control the test bench hardware. It makes it much easier to measure, for example, characteristic curves when one device controls both engines: the drive and the DUT.

Since I built it myself, I designed it to be as convenient and efficient as possible:

  • TCP/IP Communication: I love it. You can use a $40 switch and a bunch of $5 cables to connect a complete test bench system to your control PC. With 10/100 Mbit Ethernet on the PCB, you can transfer more data than you’d typically need (the rtb EPS supports 5 kHz data streams).
  • CAN Bus: For situations where TCP/IP is overkill, I added CAN. It’s fast enough for up to 100 Hz streaming. Since coding low-level drivers isn’t the most exciting task, I kept the command footprint small: 10 commands are enough to start/stop drives, send setpoints, and query configurations, etc …
  • Integrated Signal Conditioning: I added the signal conditioning directly to the PCB. No external 19’’ racks filled with relays and connectors are necessary. It’s a simple wire from A to B.
  • WAGO 2091 Connectors: After crawling through confined spaces and ruining enough fingernails and screw heads, I chose push-in cage-clamp connectors. You can configure the wiring, click it in, and easily change it if needed.

The software for startup and maintenance (rotabench EPS basic) is included, featuring automatic speed and analog IO calibration (supporting DMMs like the Keithley DMM6500). There is also a “manual control” panel for hardware testing. (More on the software in next week’s post!)

Last but not least: I support the “Right to Repair”. The SWD connector is exposed, so if you want to, you could even write and run your own firmware on the board.

PDF icon

download promo flyer

Posted on

rotabench EPS Mod. 5 PCB

After re-inventing the wheel one too many times, I finally decided to design a standardized component for test bench control. The result is the rotabench EPS Mod. 5 Revision 1b

rotabench EPS Mod. 5 test bench control PCB
rotabench EPS Mod. 5 test bench control PCB
  • Drive Control: 24V Digital / ±10V Analog
  • Sensor Support: Kistler 4503B (RS232)
  • Connectivity: Ethernet & CAN-Bus, USB for Maintainance
  • Power-Supply: 24 Volt DC
  • Connectors: Wago 2091 Cage-Clamp

This PCB is engineered to control industrial drives via 24V digital I/O and ±10V analog I/O. It features native support for Kistler 4503B torque sensors (including integrated encoders) via RS232, and provides both Ethernet and CAN interfaces for seamless connection to a test bench control software.

The main purpose of this device is to eliminate the engineering overhead of creating a custom hardware abstraction layer for every new project. Instead, it serves as a turn-key solution that provides all essential features out of the box. By moving away from bespoke setups, engineering teams can bypass the complexities of low-level integration and focus immediately on their core testing objectives.

rtb EPS Mod. 5 in the test bench control cabinet
rtb EPS Mod. 5 in the test bench control cabinet

And this is how it looks like: the rotabench EPS Mod. 5 PCB in a control cabinet on my electric motor test bench.

Posted on

rotabench ECO Rev 01

… a lightweight Ethernet to RS485 and CAN Converter

With the end of support for Windows 10 there came a difficult decision for me: upgrade all test bench + device control PCs to Windows 11, or leave them on Windows 10? I decided against the upgrade. It would have meant to spend money and lots of work/time to re-establish the status-quo ante, just to get the prior functionality back. Instead I followed the advice “never change a running system”, created a 2nd network without internet access for lab-use only and migrated most of the rarely used control machines into virtual machines.

rotabench ECO Rev 01
rotabench ECO Rev 01

While Hyper-V on Windows 11 is a convenient approach, it has a major drawback: no USB pass-through. The official statement from MS, why there is no USB pass-through, is due to security reasons. Well, yea, sure …

This means I can not use my USB-devices, like my NI USB-RS485 or NI USB-CAN adapters on a Hyper-V virtual machine, when my software needs eg. RS485 communication or a CAN Bus. As my software was written for Windows, I had to stick to it, and I needed a simple approach to give my software its hardware capabilities back. Porting all this software to Linux would cost me years, so that’s not an option. So I built this:

rotabench ECO R01a KiCAD screenshot
rotabench ECO R01a

An Ethernet & USB to RS485 Half Duplex and RS485 Full Duplex Converter, with 2 CAN Ports, based on an STM32H7 MCU. And as I could, I made the device driver-less. The TCP-IP Protocol is the “driver”. And because there were still pins free on the MCU after all the mandatory elements were placed, I gave it 8 MB QSPI-SRAM – just in case.

Full Feature List:

– RS485 Half Duplex Port
– RS485 Full Duplex Port
– 2 CAN Ports – CAN-FD (5 Mbit/s) capable but not yet implemented in the Firmware
– 10/100 Mbit Ethernet
– USB for configuration
– 8 MB QSPI SRAM, 16 MB Flash Memory, 8 kb EEPROM
– STM32H723 MCU @ 550 MHz

On the software side, I simply have to add the communication library to my software and re-route the communication over Ethernet, instead of using the drivers like NI-Serial. This also means the device is implicitly cross-platform, as the protocol is the same on every platform.

no AI was used in the engineering process. Only human brain cells were tortured.
no AI was used in the engineering process.

Posted on

rotabench 6P


After several years of R&D and testing, my lab-inverter “rotabench 6P” (rotabench Sixpack) is finally ready for the end customer. If you are producing or using EC-motors (PMSMs) with low voltages and high currents, like automotive electric motors, e-bike motors or motors for cordless power tools, this device is the right tool for your test bench(es) and your lab. Finally, you can get rid of DIY and makeshift solutions, if you want to. Because now you have the right tool, designed exactly for this purpose.


As a lab-inverter is combines features like (of course) motor control and data acquisition but also test bench control (Digital IO, PSU, brake engine drive), data visualization and automated measurement sequences. It even has a basic power analyzer with decent accuracy included, to measure the fundamental wave.

It can handle motors up to approx. 5 kWatt (48 Volt Motor @ 100 Amperes), a 150 Ampere version is in the queue.

If you need more information, please visit the product page or download the flyer!

If you need a “test drive”: just contact me!

#rotabench #rotabenchSixpack #labinverter

Posted on

rotabench 6P diode test

Stories from the rotabench lab: >>Let them (almost) burn!<<

rotabench 6P diodes test setup
rotabench 6P diodes test setup

The title picture is a photograph of the setup for testing the “freewheel” diodes on the rotabench 6P lab-inverter power stage. The purpose of this experiment was to find out, how much “reverse current” the power stage can handle in a fault condition. The experiment was motivated by a couple of destroyed power stages at the customer and in my lab.

One of the worst-case scenarios for an inverter power stage is, when suddenly the DC voltage drops while running a PMSM on a 4-quadrant test bench with a 2-quadrant PSU. When this happens, the brake engine delivers mechanical power to the DUT, the DUT becomes a generator, converts the mechanical energy into electrical energy and delivers current back to the inverter, which is converted in to heat (or fed back into the grid) by the load/PSU.

Another situation – with a similar outcome – would be if the PMSM is operating in the field weakening area and the DC-Bus voltage is lower than the theoretical Back-EMF voltage at a (theoretical) iD = 0 and for some reason the control fails, e.g. because the coupling slips. This could also lead to a situation where the DUT is forced into breaking operation and delivering current back to the inverter.

Such a sudden event – which happens typically within a timeframe < 1 second – is capable of wreaking havoc among the power electronic components, destroying the power stage within a blink of an eye and releasing the “magic dust”. The power stage takes the max. amount of damage possible in very little time.

Therefore such a situation needs to be addressed fast, fully automated and in the proper way. The most important countermeasure is to cut off the AC connections to the motor as fast as possible to stop the motor from feeding currents back into the inverter.

But: failure detection algorithms need time “to be sure”. No one want’s a device that shuts down into error mode at every little voltage spike, which simply could be a measurement error. Contactors need time for switching, you can’t stop a rotating machine within no time, and so on. So there is a short time span, where the power stage must be able to handle harsh overload conditions, without getting destroyed.

The power stage in the rotabench 6P lab-inverter has “freewheel” diodes arranged in a B6 configuration in parallel to the body-diodes of the MOSFETS. The main purpose of those (Schottky) diodes is to catch the (positive and negative) Back-EMF voltage spikes generated by the inductance of the DUT and the wiring (not to mistake with the Back-EMF of the motor!) during normal operation. In the scenarios described above those diodes become an involuntary B6 bridge rectifier, which rectify the (short circuit) current delivered the DUT and fed it into the DC Bus. There is not much you can do to prevent this, you just can design the PCB in a way that it is capable of handling this situation for the time necessary to separate the DUT from the inverter and shut everything down.

B6 bridge rectifier
B6 bridge rectifier

In order to check how much current the diodes can handle, I built a B6 rectifier bridge in the same configuration like on the rotabench 6P lab-inverter power stage but without the MOSFETs, short circuited the DC bus to allow the max. amount of current flow and used an EPS motor (Bosch, 4 pole-pairs, 12 Volt PMSM) driven by my test bench as generator. Then I measured the temperature on the diodes with a thermal camera, with the following results:

rotabench 6P diodes test - thermo image
rotabench 6P diodes test – thermo image

At a rectified DC current of slightly above 100 Ampere DC (200 Hz AC frequency), the temperature of the diodes increases from room temperature to ~80°C in one second and reaches ~120°C after two seconds, which is close to thermal destruction of the diodes. However, the encouraging result is: I have at least one second to detect an error condition in the motor control, switching off the AC contactor and shutting down the system. And in the microcontroller world, thinking in milliseconds, a second is a huge amount of time, even if the contactor is slow and needs 100 ms for a full separation of the electric circuit. So the conclusion is: 1 second is enough to shut down the test bench without major damage on the inverter.

Posted on

Lab setup for group delay measurement and phase accuracy calibration

Do I really need that much equipment? obviously yes.

lab setup for phase accuracy measurement
lab setup for phase accuracy measurement

With this setup, I can measure the group delay of the analog inputs for the voltage and current measurement of the rotabench 6P inverter.

The calibration process has two steps: in the first step, the outputs of a simulated PMSM are fed into the inverter. The device “thinks” it is driving a motor. The signals are generated by two Siglent Arbitrary Waveform generators, which are fed into a signal conditioning and amplifier PCB. The amplifier simulates the current output of the current sensors. A FPGA is used to simulate the encoder signals. With the known outputs of the signal generator the phase delay of the inverter measurement can be determined and compensated.

In the second step, I use a big choke to generate some currents around 100 A RMS. These currents are measured with a cRIO, a signal conditioning and amplifier device and Signaltec CT200 current sensors. With these data the D- and Q- currents are calculated, which can be compared with the inverter setpoint. As the first step does not use real currents and the current sensors of the inverter, some adjustmens for the group delay of the current sensors is necessary.

When anything is running smoothly, I can achieve a phase accuracy < 1° (electric) on a (simulated) PMSM with 4 pole pairs, running at 1000 Hz (electrical) frequency.

Posted on

Signal Conditioning PCB for LEM CT-200 Current Transducers

Hot on the bench: Signal Conditioning PCB for LEM CT-200 current sensors.

LEM CT-200 Signal Conditioning PCB
LEM CT-200 Signal Conditioning PCB

I just finished soldering the first samples of my signal conditioning PCB for LEM CT-200 current transducers. These devices converts the current output of up to 4 current CT-200 transducers into a voltage. For high precision I use Vishay Z-Foil burden resistors with an accuracy of 0.02% and a very low temperature coefficient. With some reduction in accuracy also CT-100 or CT-60 sensors could be used.

Format: 160 x 100 mm, 4 channels with Vishay Z-Foil 5 Ohm 0,02% burden resistors. Passive cooling. Power Supply: 2 x 18 Volt DC. The 5 Volt Rail for the Sensor Status output is generated by a DC-DC converter on the PCB.

Each Channel has a +-10 Volt (DC) Output and a 1 Volt RMS output on SMA connectors. The 1 Volt RMS outputs are intended to be used with my new-old Norma D6100, the 10 Volt Outputs go to a compact RIO.