An evaluation sample exists to answer one question: does this device meet my requirement under my conditions. Most programmes do not answer it. They power the sample up, point it at something convenient, observe that it reports plausible numbers, and move to the next line of the schedule. Months later, during integration, a performance problem appears that the sample evaluation would have caught in an afternoon, and the cause is traced to a test that was never written down. This article sets out a bench protocol for evaluating a laser rangefinder sample so that the result survives scrutiny: what to define before the sample arrives, what to measure and how, which artefacts to expect, and how to record the data so that a rejection can be defended and a fix can be verified.
WHAT A SAMPLE IS AND IS NOT FOR
A single sample cannot characterise a production population. It can establish that the claimed capability is physically real, that the interface behaves as documented, that the unit survives your environmental envelope, and that the artefacts a specification always contains are at the magnitude described. That is a substantial amount of risk to retire for the cost of one unit.
A sample cannot establish unit-to-unit spread, long-term reliability, or manufacturing consistency. Those require either a larger sample or process evidence from the supplier, which is a separate exercise covered in our guide to qualifying a module supplier. Expecting a single unit to prove consistency is the most common way sample evaluation programmes overreach, and it usually results in conclusions that are either too optimistic or unfairly harsh.
DEFINE PASS AND FAIL BEFORE THE BOX ARRIVES
The value of an evaluation collapses if the acceptance criteria are written after the measurements are known. Write them first, with numbers, and have them agreed by whoever will sign the design review. A test with a criterion added afterwards is a test with no power to reject anything.
Three categories of criterion are worth defining separately. Hard criteria are absolute: if the unit fails, the unit is rejected regardless of explanation. Typical examples are minimum range at the close distances the product must serve, interface behaviour under the documented protocol, and survival of the storage and operating temperature extremes. Soft criteria are performance targets where a marginal miss is a conversation rather than a rejection, such as maximum range being slightly below the restated figure. Informational criteria are measurements taken to feed the integrator's models, such as actual current draw at several duty cycles, which no one passes or fails but which are needed downstream.
Write each criterion as a sentence containing a parameter, a value, a condition and a method. "Minimum range shall be no more than fifteen metres, measured against a diffuse panel of reflectivity 0.2 to 0.3, with three consecutive measurements reported." That sentence can be executed by someone who was not present when it was written, which is the whole point.
BENCH SETUP: TARGETS, ATMOSPHERE AND A REFERENCE
Three things must be controlled or recorded for any range measurement to mean anything: the target, the atmosphere and the reference against which the rangefinder is judged.
The target should be a flat diffuse panel, not a wall, not a vehicle and not a convenient metal surface. Reflectivity should be known at the operating wavelength or at least characterised as a diffuse grey panel in the range 0.2 to 0.3, which matches painted military surfaces and is the assumption used in most published conditions. Panel dimensions must be large enough that the beam footprint falls entirely inside it at the test distance; if the footprint overruns the edge, the measurement has silently become a point-target test and the result will be pessimistic and unrepeatable.
The reference measurement is the most commonly skipped item. You need an independent distance to the target, accurate to better than a fifth of the accuracy you are trying to verify, from a device with a calibration trace. A total station, a surveying instrument, or a calibrated laser distance meter all serve. Without a reference you can only confirm repeatability, not accuracy, and repeatability of a biased instrument looks perfect while being wrong.
The atmosphere should be recorded, not controlled. Log visibility, relative humidity and temperature at the time of each measurement. Indoor range work is far more repeatable than outdoor work because the atmosphere is stable and the target is fixed; if the requirement involves long ranges that exceed indoor facilities, plan outdoor measurements and accept that you will need to normalise them using the extinction model described in our range performance analysis.
WHAT TO MEASURE AND HOW
A complete evaluation covers seven measurement groups. Each one is listed with the practical method that a two-person team can execute in a day.
Bias and random error across the envelope. Place the target at five distances spanning the operating range, ideally including one distance close to the specified maximum. At each distance, take twenty consecutive measurements against the reference and compute the mean offset (bias) and the standard deviation (random error). Report both. A module with a two metre bias and a ten centimetre spread is a very different proposition from one with zero bias and one metre spread, even if both are described as accurate to plus or minus one metre.
Minimum range. Walk the target in from twenty metres to the closest distance you can safely place it, taking readings at two metre intervals. Record the shortest distance at which ten consecutive readings agree with the reference. Expect a blind zone; the question is whether it is inside or outside your requirement.
Two-target resolution. Set two panels so that the nearer one partially occludes the farther one, then vary the separation along the line of sight. The separation at which the module begins to report the farther panel distinctly is the resolution. Combine this with a check of which return the module reports when both are inside the beam, which reveals the first-return or last-return logic more reliably than any datasheet statement.
Update rate and latency. Trigger a stopwatch function in software, or log the module's own timestamps, and measure the interval between successive measurements at the configured rate. Then measure end-to-end latency by moving the target abruptly and counting the number of stale readings before the reported value matches the new position. Both numbers feed the integrator's latency budget.
Detection probability at the marginal range. At a distance where the module ranges most of the time, take one hundred single-shot attempts without accumulation and count successes. That ratio is the practical detection probability at that range under today's atmosphere. It is the single most informative measurement in the whole protocol, because it converts a specification into an observation.
Boresight and optical axis. Mount the module alongside a collimated reference and measure the angular offset between the reported axis and the module's mechanical datum. Repeat after the thermal soak, because boresight drift with temperature is common and is much easier to discover on a bench than in a fielded payload.
Electrical behaviour. Record average current at the configured repetition rate, and capture the peak current and inrush waveform on a current probe at power-up. Also verify behaviour when the supply is deliberately made marginal, since a module that resets or misreports under a sagging rail is a module that will misbehave in a real vehicle.
THERMAL SOAK AND COLD START
Thermal behaviour is where sample evaluation earns its cost, because thermal problems are invisible on a bench at room temperature and expensive in the field.
Soak the module at the upper specified operating temperature for long enough that the baseplate reaches equilibrium, which typically means at least an hour and often two, then repeat the bias and boresight measurements. Compare against the cold measurements. Any change beyond the accuracy budget is a derating issue that must be understood before design freeze, and the magnitude should be added to the error budget of the whole system.
Then cool the unit to the lower specified temperature, soak again, and apply power. Measure time to first valid measurement, because optical and electronic subsystems sometimes need settling time at cold start. A module that requires a two minute warm-up before reporting accurate ranges is acceptable in some applications and disqualifying in others, and the datasheet will rarely tell you.
INTERFERENCE AND POWER QUALITY
A benchtop supply and a short cable are not representative of a payload. Two interference tests are cheap to run and find a large share of real integration problems.
First, place a brushless motor or a radio transmitter near the cable run and observe whether measurement throughput or accuracy changes. A pulsed laser driver is a fast high-current switching circuit and a rangefinder receiver is an extremely sensitive photodetector, so an enclosure containing both is a design problem, and a module that degrades near a motor on the bench will degrade further inside a vehicle.
Second, introduce controlled supply ripple at the pulse repetition frequency and at harmonics of it, and check whether range bias shifts. Ripple coupling into the receiver appears as a periodic noise signature that looks exactly like a marginal detection threshold, which is why it is often misdiagnosed as an optical problem later.
RECORDING THE DATA SO IT SURVIVES SCRUTINY
The output of an evaluation is not an opinion; it is a dataset. Record, for every measurement: the date, the operator, the target identity and reflectivity, the reference distance, the visibility and humidity, the module temperature, the raw readings, and the pass or fail against the pre-agreed criterion.
Keep the raw readings, not just the summary statistics. Disagreements about evaluations are almost always resolved by returning to the raw data, and reconstructions from memory are worthless. Photograph the setup including the target and the reference instrument display. It takes five minutes and ends arguments.
If the module logs internally, archive the log together with your own notes and note the firmware version. A future performance question can often be answered by re-reading data you already have, provided you know which firmware produced it.
WHEN TO REJECT VERSUS WHEN TO ASK FOR A FIX
Not every failure is a rejection, and treating every failure as a rejection costs programmes good suppliers. A useful rule is to separate failures into three classes.
Class one failures are fundamental: the product's physical architecture cannot meet the requirement. A resolution of ten metres against a requirement of two, or a blind zone of two hundred metres against a requirement of fifty, will not be fixed by a firmware change or a rework instruction. These are rejections.
Class two failures are calibration or configuration issues: bias outside tolerance, boresight offset beyond a stated window, an interface parameter set incorrectly at the factory. These are fixable and should be reported as nonconformances with a request for corrected units and a statement of what changed.
Class three failures are documentation mismatches: the unit performs correctly but not as the document described. This is the most valuable class to find, because it is cheap to resolve and it identifies a supplier whose documentation process needs attention before you commit to volume.
ERDI Laser supplies evaluation samples across its 905 nm, 1535 nm and 1570 nm ranging module families with no minimum order quantity, dispatched within one week of order confirmation, so a protocol of this kind can be executed on real hardware before a production decision. Batch production is delivered within one month, and the technical team in Longquanyi district, Chengdu answers method questions directly rather than through a sales channel.
FREQUENTLY ASKED QUESTIONS
How many samples do I need for a meaningful evaluation?
One sample retires the question of whether the claimed capability is physically real, and that is most of the available value. Three samples begin to reveal unit-to-unit spread in boresight and accuracy, which is worth having when the parameter is critical. More than three is rarely justified at evaluation stage, because consistency is better established through the supplier's process evidence than through a larger bench programme.
Can I evaluate range performance indoors?
Indoors you can verify accuracy, resolution, minimum range, timing and thermal behaviour very well, and you should, because the atmosphere is stable and the target is fixed. Indoor distance is usually too short to verify maximum range. For the long-range figure, either use an outdoor range with recorded atmosphere, or treat the published figure as a supplier claim and verify it later in the end application.
What is the most commonly skipped measurement?
Minimum range. Evaluation teams naturally test the capability they are excited about, which is the long-range figure, and neglect the near field that the product will actually operate in. The second most commonly skipped is detection probability at marginal range, which is the measurement that tells you how the product will feel to an operator.
How long should the thermal soak be?
Long enough for the baseplate to reach equilibrium, which is typically one to two hours depending on mass and the chamber's air movement. Measuring baseplate temperature rather than chamber air temperature is essential, because the two differ substantially and it is the baseplate that determines optical behaviour. If the baseplate has not stabilised, the measurement describes a transient rather than the specified condition.
Should I evaluate against a retroreflector as well as a diffuse target?
A retroreflector test is useful for one purpose only: confirming that the optical alignment and receiver path work at all. It is not a performance measurement, because a retroreflector returns orders of magnitude more light than any real surface and will make almost any module appear to range a very long way. Include it as a diagnostic, never as evidence.
What should happen to the evaluation data afterwards?
It should become the reference baseline for integration. When a problem appears during payload testing, the first question is whether the module behaved correctly in isolation, and the evaluation dataset answers that immediately. Keep the raw readings, the criteria document, the photographs and the firmware version together as a single package, and treat it as a controlled document rather than a folder of notes.
RELATED READING
- Custom Laser Rangefinder Module: OEM Guide
- Laser Ranging and OEM Integration FAQs
- Laser Rangefinder Datasheet Red Flags
- Laser Rangefinder Module Integration Checklist
- Laser Ranging and Designation Glossary
RECOMMENDED PRODUCTS
Hardware configurations from the ERDI Laser catalogue that match the topics in this article: