# System Technical documentation for the current Modular Rack Power design. Canonical: https://mrps.karlquinsland.com/system/ Updated: 2026-09-19 Source: https://github.com/kquinsland/modular-rack-mount-power-system/blob/f4855f0c2ca8b85eb16c4d821fcad9c6971e39fe/site/content/system/_index.md --- The core design principle of MRPS is to separate shared infrastructure from replaceable output modules in a scalable and modular manner. 1. At least one high current DC supply is wired to a [backplane](https://mrps.karlquinsland.com/system/hardware/backplane/). 2. The backplane distributes the DC bus to six slots, measures aggregate current, powers local control electronics and fans. 3. Each [module](https://mrps.karlquinsland.com/system/hardware/modules/) mates with the backplane and occupies a slot. Each hosts a power-output module or accessory. 4. The backplane and every carrier have their own microcontroller to manage local peripherals and a CAN-FD transceiver to communicate with the rest of the system. CAN-FD is the only inter-board communication bus. ```text DC input │ ├── Backplane controller, current monitor, CAN-FD, and cooling │ └── Shared VIN_BUS ├── Carrier 1 ── USB-C PD module ├── Carrier 2 ── Accessory or USB-C PD module ├── ... └── Carrier 6 ── USB-C PD module ``` This repo is going "live" with the second-generation backplane, which consolidates functions that previously lived on separate backplane and controller boards. At least for _now_ there is only one "supported" carrier/module but there are plans to expand the range of compatible modules in the future. ## Current design targets To make a very long story short, the first iteration of the design is targeting a DC input of _at most_ 30V, nominally somewhere around 24V. This is because the [current/only carrier module](https://mrps.karlquinsland.com/system/hardware/modules/carrier/#sw3538-usb-c-pd-module) has a maximum input voltage of 30V. Furthermore, the `SW3538` based module can't do more than 100 W without using non-standard configurations so the effective 'ceiling' here is going to be 100 W per port until a new module revision is introduced. The only way to achieve 100W with USB-C/PD is to use [5A/20V (100W) configuration](https://en.wikipedia.org/wiki/USB_hardware#USB_Power_Delivery). Just for perspective, a _good quality_ power supply that can do ~ 27v and ~ 30A will cost somewhere in the range of 200-400 USD so there's also a cost reason to keep things at or below the 100 W per port ceiling at this time. Given all that: | Property | Current target | | --- | --- | | Initial system input | Nominal 24-28 V DC | | Module slots | Six | | Initial USB-C policy ceiling | 20 V, 5 A, 100 W per port | | Working aggregate input budget | Approximately 30 A | | Cooling | Two independently switched 3-wire fans per backplane | And so so so many additional properties that are "theoretical, untested" until I get the first physical articles in hand; they [have been ordered](/worklogs/2026/09/10-first-physical-prototypes/) and are expected to arrive soon. Very early prototypes did put per-slot switching on the backplane but that was later abandoned in favor of a simpler, shared input bus design; the current design does not presently switch or limit each backplane slot. Slot power is present whenever the shared input bus is energized due to the [limited number of pins](#carrier-slots) available on the connector. For the underlying engineering record, see the repository's [system overview](https://github.com/kquinsland/modular-rack-mount-power-system/blob/main/docs/system-overview.md). ## Interfaces ### System DC input The backplane accepts ground and `VIN_RAW` through a user-installed two-position terminal. The initial population targets a nominal 24 V source. After the aggregate current shunt, the rail becomes `VIN_BUS` and feeds the backplane converters and all six slots. ### External CAN-FD The external connection exposes ground, CAN low, and CAN high. Connector-side ESD protection and a common-mode choke protect this boundary. A normally-open 120 Ω termination option is intended only for installations where the backplane sits at a physical end of the CAN bus. ### Carrier slots Each combined power-and-signal connector carries: | Contact | Signal | Purpose | | ---: | --- | --- | | 1 | `VIN_BUS` | Unswitched power from the backplane | | 2 | `GND` | Power and signal return | | 3 | `CANH` | Shared CAN-FD high | | 4 | `CANL` | Shared CAN-FD low | Each carrier includes its own connector-side CAN protection. The backplane does not duplicate an ESD network at every slot. ### Backplane-local interfaces - I²C connects only the backplane STM32 and aggregate INA237 current monitor. - Two 3-wire fan headers provide ground, switched 12 V, and tachometer input. - The external DS18B20 header provides ground, 1-Wire data, and 3.3 V. - A Tag-Connect SWD footprint for programming and debugging the STM32. Consult the engineering [interface tables](https://github.com/kquinsland/modular-rack-mount-power-system/blob/main/docs/interfaces.md) before relying on connector reference designators or fabricating hardware. ## Power and cooling > [!WARNING] > Everything below is theoretical. > I have done some simulations and calculations to get a ballpark estimate but these numbers should not be taken as guaranteed performance. > Once I have physical prototypes, I'll do some tests to validate these theoretical numbers. ### Initial operating budget | Item | Working value | | --- | ---: | | Supply input | Nominal 24 V | | USB-C output policy | Up to 20 V, 5 A, 100 W | | Per-slot input budget | Approximately 4.7 A | | Aggregate backplane target | Approximately 30 A | | Aggregate current shunt | 1 mΩ | At 30 A, the aggregate shunt develops about 30 mV and dissipates about 0.9 W. The biggest thermal issue will be the copper traces on the backplane and the general topology - all the DC is supplied from one end of the backplane so there's a voltage drop and heating gradient along its length. There is a [simple tool](https://github.com/kquinsland/modular-rack-mount-power-system/blob/main/scripts/pcb_power_capacity.py) to estimate voltage drop and heating along the backplane traces. [![Backplane copper-capacity comparison showing layer geometry, current capacity along the bus, and estimated temperature rise for two copper stackups at 33 A]()]() IPC-2221 screening estimate at 33 A total load, comparing 2 oz outer / 1 oz inner copper with 1 oz outer / 0.5 oz inner copper. These are calculated estimates, not measured results. > [!NOTE] > The figure above is a visual representation of the estimated voltage drop and heating along the backplane traces based on the initial operating budget. > It is a _crude_ simulation and is NOT meant to be taken as precise or guaranteed. ### Local rails and cooling The estimated 11 W loss from each fully loaded initial USB-C module makes forced-air cooling a design requirement rather than an optional feature. To facilitate this, each backplane has a dedicated LM5164 converter for a 12V 1A rail meant to power the cooling fans. The STM32 is meant to monitor temperature from a few different locations and drive the two fans as needed. See the engineering [power budget](https://github.com/kquinsland/modular-rack-mount-power-system/blob/main/docs/power-budget.md) for conductor calculations and open protection questions.