Almost every hardware project that slips does so on a decision that was deferred, not on a technology that failed. This is the checklist we run at kickoff — before the schematic starts, and certainly before the enclosure is tooled.
Bestom engineering note — written by our own design and BSP engineers from work on Rockchip platforms. Platform parameters quoted here are taken from our published datasheets; see Core Boards & SoM. For translated community notes, see the Tech Notes index.
Every external connector, with electrical requirement and connector part number. "We will add USB later" is how a re-spin happens. Include the debug interface — you will need it.
Not "fast enough". Numbers: frame rate, resolution, latency ceiling, boot time, sustained throughput. Sustained, not peak.
DDR density and eMMC size, plus the expected footprint growth over the product's life. State the assumption explicitly so it can be revisited.
If there is AI, name who owns accuracy after quantisation, who owns the calibration dataset, and who accepts the result. Unowned accuracy budgets are the most common cause of AI schedule slips.
With the ambient definition written down. Component grade follows from this single line — it is the cheapest constraint to settle and the most expensive to change.
Board outline, mounting holes, stack-up thickness, and every connector's insertion direction and keep-out. Give the mechanical drawing before layout, not during.
Input source (adapter, PoE, battery), voltage rails, sequencing requirements, and the worst-case current budget with margin. Battery products also need the charge and gauge decision here.
How heat leaves the SoC: heatsink, thermal pad to enclosure, airflow, or none of the above. Decide before the enclosure is tooled — a sealed plastic box with no thermal path derates the platform you chose.
Which OS, who maintains the BSP, how field updates are delivered and rolled back, and how long the BSP will be supported. Decide before the platform, not after.
Which markets, which marks, which test house, and who owns the sample submission. Certification lead time sits on the critical path and is frequently discovered late.
What is tested on the line, at which station, with what fixture. A functional test that needs a full boot into the OS does not scale; decide the test points while the schematic is open.
How long the product must stay in production, and which parts need an alternate. Long-life industrial products need this fixed at kickoff, not at last-time-buy.
Take it into the kickoff meeting and mark each item locked, assumed or open. Anything marked assumed should have a named owner and a date. Anything open is a schedule risk you have now made visible instead of discovering in layout.
This is the same list we work through with clients at project start — talk to us if you want to run it against your product spec.
Send us the product spec or the constraint list. We design and develop custom hardware on Rockchip platforms — and we will tell you if the platform you have in mind is the wrong one.
Start a project → All Tech Notes