A Two-Millisecond Exposure Can Span Five Milliseconds Across an Image

Exposure duration and row timing describe different things

An exposure accumulates a light-dependent signal over an interval. Its duration says how long that interval lasts. The shutter method determines how exposure intervals relate across the image. Basler's historical BAS1401 white paper describes global exposure as simultaneous across rows and rolling exposure as successive, potentially overlapping row intervals.

Basler's Electronic Shutter Types documentation makes a useful additional distinction: global exposure can be followed by row-by-row data readout. Sequential readout therefore does not by itself establish sequential exposure. Its rolling-shutter description identifies a temporal offset between successive row exposures, with the offset varying by camera model.

For a buyer evaluating a moving scene, the useful question is which part of the scene was integrated over which interval. Matching exposure-duration labels leave that question open. The following invented schedule isolates it without selecting a camera or operating mode.

References: [RS48][BS2]

RS-48: four 2 ms exposures occupy a 5 ms window

RS-48 is a hypothetical four-row image on one ideal local timeline. Stipulate that rows 1, 2, 3 and 4 begin exposing at 0, 1, 2 and 3 milliseconds respectively. Every row integrates for exactly 2 ms. These are invented teaching inputs, not camera specifications or measured results.

Row 1 therefore exposes from 0 to 2 ms. Row 2 exposes from 1 to 3 ms. Row 3 exposes from 2 to 4 ms, and row 4 from 3 to 5 ms. Adjacent row intervals overlap for 1 ms, but their starts and ends differ.

The earliest start is 0 ms and the latest end is 5 ms. The whole exposure window spans 5 - 0 = 5 ms. Each row still has a 2 ms exposure; no individual row has integrated for 5 ms. Summing the four durations to obtain 8 ms would count overlapping intervals repeatedly and would not describe the elapsed window either.

The first-to-last start delay is a different quantity: 3 - 0 = 3 ms. For this equal-duration schedule, adding that 3 ms delay to the 2 ms row exposure gives the 5 ms window. This arithmetic describes the stipulated exposure intervals, leaving any subsequent data transfer or processing time unspecified.

The row midpoints differ by 3 ms, corresponding to 3 mm in this example

The temporal midpoints of the four intervals are 1, 2, 3 and 4 ms. The first and last midpoints differ by 4 - 1 = 3 ms. A midpoint is a coordinate used to compare these intervals; each row still integrates across its full 2 ms rather than taking an instantaneous sample at its center.

Now stipulate a target translating uniformly at exactly 1 metre per second relative to a fixed camera. Between the first and last row midpoint times, its world-space travel is 1 m/s x 0.003 s = 0.003 m, or 3 mm. This explains why assigning one common capture instant to all four rows would conceal a relevant temporal difference.

The 3 mm belongs to the target's stipulated motion in the world. It is neither 3 pixels nor a calculated image distortion. Mapping motion into image coordinates would need the actual viewing geometry and image scale, among other applicable evidence. RS-48 supplies no pixel displacement, skew angle, detection error or task-quality result.

A separate global case shares the same 2 ms exposure label

For comparison, stipulate a different four-row acquisition in which all rows begin exposing at 0 ms and end at 2 ms. Every row again has a 2 ms exposure, but now every interval has the same 1 ms midpoint. The full exposure window is 2 ms and the first-to-last midpoint difference is zero.

Both examples can truthfully carry a 2 ms exposure label. That common label leaves their different within-frame timelines unresolved. In RS-48, the last row begins after the first row's exposure has finished. In the global case, all four rows cover the same interval. Those premises support different interpretations of the scene represented by the assembled image.

Motion blur concerns movement during an individual exposure. Rolling-shutter geometry concerns the temporal differences across rows. Basler's BAS1401 section 5 distinguishes the effects and notes that blur can also occur with global shutter. In the comparison here, aligning the four intervals does not make their 2 ms duration disappear. Global exposure therefore does not establish a blur-free image.

References: [RS48]

Frame rate and a frame timestamp leave other timing questions

A headline frames-per-second value does not supply the row schedule. BAS1401 section 4 specifically cautions against directly inferring rolling-shutter distortion from frame rate. RS-48 stipulates no repeat interval between images, so its 5 ms exposure window establishes no camera frame rate.

A single frame timestamp also needs a defined event. If an offer assigns a time to an image, establish whether it refers to an exposure boundary or another named event, and how the row intervals relate to it. Merely retaining that number cannot turn RS-48's four different intervals into one shared interval.

The related video presentation-timestamp guide addresses the spacing of delivered frames on a media timeline. The hardware-timestamp guide addresses event boundaries and relationships between clocks. Here, even a perfectly agreed local clock leaves the within-frame exposure schedule to establish.

References: [RS48]

Compare the timeline each camera-input offer actually establishes

Consider two conceptual offers for an industrial vision input. Both describe an image with a 2 ms exposure and the same advertised frame rate. Offer A stops at those labels. Offer B also identifies the proposed camera and mode, the delivered row ordering and the applicable exposure starts and ends relative to a named reference event.

Offer A leaves the temporal relationship between rows unresolved. Its labels alone are consistent with both of the example schedules. Offer B supplies the information needed to distinguish those schedules, provided its documentation actually applies to the offered configuration and delivered image. More explicit timing evidence makes the comparison interpretable; it does not establish a task outcome or make B the selected system.

Ask for the exposure duration, the relationship between row starts and ends, the image region to which that relationship applies, and the meaning of any associated frame timestamp. The evidence should connect the source rows to the image the consumer receives. If the supplied image changes row order, a timing diagram without that mapping leaves the association unresolved.

Use that record to evaluate the temporal premise behind the proposed vision result: which intervals contribute to the image, and does the intended interpretation require them to be simultaneous? A static-scene use and an application interpreting moving geometry can place different demands on that premise. The appointed vision team must establish the actual application requirement and evidence of suitability; this comparison chooses no shutter mode, trigger arrangement, lighting change or quality threshold.

Connect a separate equipment requirement to physical deployment

If the selected vision system also needs rack-mounted compute installed or IT cabling connected, Leviathan's Complete Physical Deployment service and related project inquiry provide a route to discuss those physical stages. Share a nonconfidential equipment, site and connection outline to confirm scope, requirements and availability.

Camera selection, exposure timing, optical setup and application evaluation belong with the appointed vision specialists. The physical equipment inquiry concerns the agreed installation and cabling work.

References: [LS1]