Teec, The Electronics Engineer's Computer. PXI like, but low cost

[bruchmueller]

Dear electronics engineers,

this is a new concept for a electronics engineers computer: modular Test and Measurement Unit, in a 19 inch desktop or rack mountable form factor, with up to 8 modules, and a architecture a bit like the very expensive PXIe chassis and modules. But with a NUC (or other mini computer inside). We call it Teec (The Electronic Engineer's Computer).



For modular Test and Measurement, PXIe is the high end standard. But its expensive, complicated, and use dated concepts. On a lower cost side, there is the BB3, a open source solution from EEZ, but this not addresses features found in PXIe like star-trigger, trigger-Bus, up to 300W on the mainboard-module connectors, limited cooling capacities and implemented as enclosure up to now only for 3 modules. While PXI chassis are mainly made for rack-mount systems, the BB3 is exclusive made for desktop use. The Teec combines the best of both worlds.

Here some highlights:
19 inch 3U rack or desktop-mount
Near-PXI performance thanks to star-trigger, a per module 8in, 8out trigger lines, and a programmable trigger controller
300W main power supply for 48V, 12V, 5V and 3.3V on the bus, and up to 8 additional 300W module PSU, for up to 2.400W total power
cooling capacity of around 50W per module.
Backward compatible to the DIB v1.2. (but DIB firmware not yet ported).
Either a NUC, or a mini computer with max. 117 x 117 x 54mm. Intel or AMD, Windows 10 or 11, or Linux.  Or a Raspberry Pi.
The NUC is connected via USB to the mainboard, and Teec modules can operate directly with the NUC. Ideally, the Teec module use the Triggerbus and star-trigger for real time syncronisation, USB only for data transmission.
The NUC serve the build-in 4 inch TFT touchscreen, via HDMI. On the second HDMI port of the NUC, a external big monitor can be used.
The Programmable Trigger Controller(PTG) is build with the RP2354, offer 12 state machines and FPGA-like PIO.  The PTG is programmed from the NUC via USB.
PIO Trigger have a input-to-output latency from 10ns - 160ns, depending on the complexity of used trigger logic.
The module-to-module bus is implemented with a CAN-FD bus, what connects all 8 modules in the chassis, as well as external devices.
The mainboard contains a BIOS, what also translates the CAN-FD bus messages to the NUC and back. The BIOS is also used as a USB to GPIO expander, real time clock and fault monitor.
For DIB v1.2 backward compatibiity, a MCU can be installed. The MCU module is build with an other RP2354, offer 7 SPI busses, I2C and UART are switched, not shared. A new firmware has to be written from scratch, because a different MCU (RP2350 vs. STM32F7) and architectural differences

What is unique in this architecture is the safe routing of AC line voltage (110 or 220/240V) to the modules. We enclosed these Meanwell 300W AC/DC pcb-psu's, having only 101x51x30mm (4'' x 2'') size, with a safe plastic enclosure with ventilation holes, so that it can be mounted on a Teec module. This allows up to 8 x 300W in this 19 inch 3U enclosure, what is quit remarkable.

Teec modules can offer a USB 3.2 connector, for data transfer rates up to 10 GBit. Depending on the NUC model, an additional USB 3 hub may be required (not shown in the concept diagram). All USB traffic on the mainboard is USB2, max. 480 MBit/s.

On the software side, standard software like LabView, GNU radio, EEZ Studio, Sigrok, BenchVue, ngscopeclient or other SCPI based T&M software can be used.

Introducing USB to the instrument bus, offers great possibilities: Open Source projects, like the Haascope, LibreVNA or the Adalm2k or Pluto, or many others, are all based on USB. They just need to be adapted to the new Form-Faktor of the Teec, and can be used with the existing software. If they are attached to the Teec trigger system, adaptions in hardware and firmware will be required.

NI offer more then 600 different PXIe modules, for a premium price. 900 more are offered by various manufacturers. With the Teec, now every individual or company can offer Teec modules with comparable functionality, for a much lower price point. While the Teec chassis and the Teec concept is intellectual property of RevoBrick LLC, Teec modules are free to be manufactured and sold by everybody. We will offer reference implementations and Kicad templated, to kickstart new projects.

We expect the MSPR of the 19 inch 3U Teec chassis, including Mainboard, PSU+power distribution, 4 inch TFT, in the ballpark of $450 - $550 (Project name Teec 19-8). To compare with PXIe, a NI PXIe-1088 (8+1 slot chassis), cost $5,166.
We plan a Kickstarter campaign later this year.

I just finished the PCB layout of the mainboard and the psu+power distribution board. My colleague will import the PCB design into the 3D model of the chassis tomorrow, if everything work smooth, we could have a first prototype in 4 weeks.

On what modules would you guys be most interested to be ported to Teec?
- An other programmable power supply (i have a 2 quadrant 40V, 8A PSU in mind),
- a function generator like this https://github.com/MatthiasElectronic/AWG_DHO8-900
- LibreVNA Spectrum Analyser 6GHz
- LTDZ Spectrum Analyser 35M-4.4GHz
- Haascope Pro - use USB3
- Adalm Pluto
- Adalm2000
(all those ported designs will be published under their original license)
- or something else?

Would be glad, hearing your feedback.
BR, Michael

[Arten]

Hey, I'm the dev behind Moducard / MiniModucard System (MMS) — figured I'd jump in since i see almost full overlap.

Just to be clear up front: yes, this architecture already exists and has for a while. We've got a backplane-based modular system in production use across several robots (Nomad 1/2/3 and others), fully open source — schematics, PCB layouts, KiCad templates, protocol docs, the works. Not trying to gatekeep the idea, just flagging prior art since a lot of the bones here line up closely with what we built.

What's the same:

Passive backplane, cards plug in perpendicular via DIN41612, backplane distributes power + a shared bus to every slot
A compute "brain" card that's just a commodity SBC/SOM (we use RPi CM5 / Jetson) rather than a bespoke CPU design — same move as your NUC
CAN as the module-to-module bus, with the brain card gatewaying out to the wider world (ours is CAN↔Ethernet for ROS2; yours is CAN-FD↔USB for the T&M software stack)
A dedicated PSU card handling power sequencing on command from the brain
Open module ecosystem — publish the backplane spec/connector pinout, let anyone design compatible cards without touching the standard
Timing, new backplane feachures 2x differential triggers.

Where Teec is genuinely different, not just a reskin:

Domain/purpose: we built this for robotics (sensors, motor drivers, ROS2 nodes). Teec is aimed at bench/rack T&M instrumentation competing with PXIe — different problem entirely.
Bus: plain CAN (2×) for us vs. CAN-FD for you — higher throughput, matters more for your acquisition side.
Mechanical: our backplane is sized for robot enclosures, yours is a proper 19" 3U rack/desktop form factor.
Software target: we're ROS2/YAML-defined CAN frames; you're SCPI-based T&M tooling (LabVIEW, EEZ Studio, etc.) — basically no overlap here.

So: same core pattern (backplane + swappable cards + external commodity brain + CAN + open ecosystem), but applied to a different domain with a genuinely harder power and timing problem on your side. Might be worth a look at how we handled card addressing/discovery and the connector pinout before you lock yours in — no point rediscovering the same footguns — but the trigger bus and power architecture are real engineering you'd still have to do either way. Happy to compare notes if useful.

https://github.com/Dominik-Workshop/moducard-cm5-module
https://github.com/ArthorH/ModuCard-backplane
https://github.com/ArthorH/ModuCard-OrangePie
https://youtu.be/_UtSGDRLs88?is=ysHf5-WkFyyHgAz4

Contact me on linkedin if you are interested, or reply to post with contact info 
https://www.linkedin.com/in/artem-horiunov/