Why a bench test at all
A receiver can pass a desk test — sensor next to it, screen showing numbers — and still fail on a
machine with a 12-metre chassis. The difference is not luck: it is the radio link running out of
margin. So we test the two things that decide whether the margin exists:
1. Centre frequency. 433.92 MHz is the band we work in; a unit that drifts off centre loses
range without ever showing a fault.
2. Transmit power and receive sensitivity, which together give the link budget that the machine
geometry then eats into (steel, distance, orientation).
Every receiver we ship goes through the bench before packing:
What the bench actually measures
| Test | What is measured | Why it matters on a machine |
|---|---|---|
| Centre frequency | Carrier at 433.92 MHz, tolerance checked against the band | A drifted unit loses range silently — it still "works" at 1 m |
| Transmit level | Sensor output power (we quote up to 20 dBm at the sensor) | Determines how much steel and distance the link can absorb |
| Sensitivity | Receiver sensitivity floor (we design around −95 dBm class parts) | Headroom for the vehicle's own electrical noise |
| Packet count over time | Frames received vs frames sent in a fixed window | The number that actually decides whether a wheel is monitored |
| Interface loop-back | CAN / RS485 / RS232 wrap to the test rig | Proves the interface you ordered is the interface we fitted |
The last two are the ones customers never ask about and always suffer from.
The number that matters is the packet count
A receiver's specification sheet can say 256 wheel positions; what decides the installation is
whether each of those positions is heard often enough. We measure it the same way every time: the
sensor increments a counter in each packet, cycling `0x00` to `0x28`, and any jump in that sequence
is a lost packet.
In our published field tests this produced loss rates in the 5–14 % range on port machinery —
and, importantly, showed that adding a repeater did not lower the total rate (it improved
*confirmed* loss from 55 to 41 packets). Those numbers come from counting, not from a marketing
claim.
Why "works at the desk" is not evidence
Three failure modes that a desk test cannot catch:
1. Metal in the path. A steel chassis between sensor and receiver costs more than any antenna
gain can recover. This is why our installation advice is about position, not power.
2. Collision under load. With many sensors reporting on the same channel, transmit windows
overlap. More repeaters and more sensors both increase the collision rate.
3. Standing on a bench, not moving. Vibration and temperature change the crystal load; a unit
that is on-frequency cold can drift once the machine is hot. Burn-in before the frequency check
is how we screen for it.
What we will not claim
We do not publish a single "range in metres" figure. Range is a property of an installation — a
number in a brochure that ignores steel, orientation and receiver position would be invented. What
we publish instead is the test method, the measured loss rates, and the installation rules that
follow from them.
FAQ
Do you test every receiver or sample them? Every receiver that ships goes through the bench
(frequency, power, packet count, interface loop-back).
What is the centre frequency tolerance? The unit is checked at 433.92 MHz against the band; a
unit outside tolerance does not ship.
Can you test with our sensor layout? Yes. Send the wheel map and receiver position; we can run
the same counter-based measurement on a bench rig configured to match.
Why publish the loss numbers at all? Because a buyer can reproduce them from the CAN log. Numbers
that can be checked are worth more than numbers that sound good.