What matters
- CSV row rate is not necessarily the acquisition or update rate of every channel.
- A brief event can fall between samples; a repeating fast signal can also appear as a false slower pattern through aliasing.
- Timestamp spacing, channel selection, sensor bandwidth, ECU filtering, and logger configuration must travel with any timing claim.
Why this question matters
A datalog is a time-aligned record of channels, not a diagnosis. Useful interpretation begins with channel definitions, targets, operating context, and repeatability.
What the evidence supports
COBB warns that requesting more monitors can slow the logging rate on some platforms and tells users to watch the time between samples. NI defines sample rate as the conversion frequency, distinguishes it from measurement bandwidth, and explains that signal content above half the sample rate can alias into false lower-frequency content unless acquisition bandwidth is limited before sampling. These principles show why a CSV plot cannot reveal events the acquisition chain never preserved.
Separate row timestamps from channel freshness
A logger may write one row containing values obtained through several layers: sensor conversion, ECU task scheduling, bus transmission, diagnostic requests, device buffering, and CSV export. The file can receive a new timestamp while one channel repeats its previous value. Counting rows therefore measures file cadence, not automatically the age or true update rate of every signal.
Inspect consecutive timestamp differences instead of dividing row count by total duration alone. Record the median, minimum, maximum, and long gaps; count duplicate or reversed timestamps; and examine runs of identical values in channels that should be moving. Uneven spacing is timestamp jitter or dropped delivery, while repeated values may reflect quantization, a slow source, filtering, or a held value. The CSV alone may not distinguish those causes.
A missed transient and aliasing are different failures
At 10 rows per second, nominal timestamps are 100 milliseconds apart. A one-time 60-millisecond throttle closure can occur between adjacent observations and leave no captured minimum. If one sample lands inside it, the plot can make the closure appear aligned to that timestamp or as long as the drawing between points. This is temporal undersampling of a transient; interpolation adds pixels, not measurements.
Aliasing is more specific. NI explains that when repeating signal content exceeds half the sampling rate, it can appear in the sampled record as a false lower-frequency pattern. The familiar “sample at more than twice the highest frequency” condition assumes a band-limited signal and suitable acquisition. Automotive diagnostic channels may already be averaged, filtered, calculated, or asynchronously requested, so a row-rate Nyquist calculation cannot prove the original sensor waveform was captured.
Design the log around one test question
Choose the smallest group of channels that preserves cause, control, and consequence for the question. For a suspected throttle intervention, that may include accelerator request, throttle target and actual, engine speed, load, torque or protection state, plus a stable time base. COBB notes that monitor count can slow logging on some vehicles, so save the exact channel list, device and firmware version, vehicle software, units, and test conditions with the file.
Repeat the event under a safe, controlled procedure and look for the same ordering across runs. When event timing matters beyond the logger’s demonstrated resolution, use a faster source designed for that measurement—such as an appropriate CAN capture, oscilloscope, or dedicated data-acquisition channel—with correct bandwidth, probes, synchronization, and anti-alias filtering. Never infer sensor accuracy from a high CSV row count.
A 60-millisecond throttle closure can occur entirely between two 10 Hz samples. The chart may show no closure, one partial value, or a point that makes the event look longer than it was. Interpolation improves drawing, not underlying evidence.
Illustrative example, not a measured vehicle test.
A repeatable workflow
- Validate timestamp units, monotonic order, resolution, median spacing, maximum gap, and duplicates.
- Check each important channel for repeated values, quantization, filtering, and documented source rate.
- Save the channel list, logger and firmware version, vehicle software, and test conditions.
- Reduce the channel set, repeat safely, and escalate to a suitable faster synchronized instrument when the timing claim requires it.
Where the shortcut breaks
A high reported row rate does not establish sensor bandwidth, ADC rate, ECU task rate, channel synchronization, transport latency, accuracy, or freedom from filtering. A CSV is not evidence of the unsampled waveform between rows.
Sources and update method
TunerBench prefers government, standards-body, and component-manufacturer documentation. This guide is reviewed against the sources below and should be revisited when regulations, product data, or vehicle technology changes.