LabVIEW DAQ: Stop on Speed Threshold, Not Sample Count

Reject the fixed-length quick fixes

Do not shorten the record by guessing how long the shaft engagement will take. The current call collects 30,000 samples from three channels at a sample rate of 20k. Any shorter fixed count can cut off a slow engagement, while the present count leaves unwanted constant-speed data after a fast engagement.

For a quick production recovery, acquire in manageable blocks, detect the target crossing, and retain data only through that sample. After production is stable, add noise qualification and explicit timeout handling.

Use the speed waveform as the stop condition

The shaft-speed channel already contains the event that defines the useful record. Replace the fixed sample count as the primary endpoint with a comparison between each acquired speed sample and the known target.

For acceleration toward a positive target, the usual crossing test is measured speed greater than or equal to target speed. For a decreasing-speed event, reverse the comparison. If the waveform represents a voltage proportional to speed, compare against the voltage corresponding to the target only after confirming the scaling. A modified LabVIEW example used an input rising above 4.0 V; that value demonstrates the comparison pattern and is not a shaft-speed setting.

Stopping a software loop happens after a read returns. The returned block can therefore extend past the physical crossing. The exact analytical endpoint is not the loop-stop time; it is the first qualifying sample inside the speed waveform. Preserve its index and use it to trim every channel.

Replace the one-shot read with a block loop

Modify the Acquire Waveforms VI arrangement so acquisition proceeds through repeated reads rather than one request for the entire 30,000-sample record. Keep all three channels in the same acquisition so their sample positions remain aligned.

1. Start the acquisition and initialize storage for the three waveform channels.
2. Read the next block from all channels.
3. Search the speed channel from its first new sample to its last new sample.
4. Append the block if no sample meets the target condition, then read again.
5. If a crossing is found, append only the samples through that crossing index for all three channels.
6. Stop and clear the acquisition after the retained data has been copied.
7. Pass the trimmed waveforms to the waveform-analysis VIs.

start acquisition
repeat
    read the next block from all three channels
    search the speed samples in chronological order

    if the first target crossing is found
        retain every channel through the same crossing index
        exit the acquisition loop
    else
        retain the complete block
until stopped or the allowed record length is reached

stop acquisition
send retained waveforms to analysis

Keep a maximum-record exit as a fault path. Without it, a broken speed sensor, an unreachable target, or a stalled shaft can leave the acquisition waiting indefinitely. The existing 30,000-sample limit can serve as that ceiling if it is long enough for the slowest valid engagement.

Trim every waveform at one crossing index

Search for the first qualifying speed sample, not the maximum speed and not the final sample above the target. If the crossing occurs at position N in the current speed block, retain positions from the start of that block through N from each of the three channel blocks. Include the crossing sample when the analysis needs the measured point that satisfied the endpoint condition.

When previous blocks have already been accumulated, the endpoint in the complete record is the number of retained samples from prior blocks plus the local crossing position. Keep the waveform timing information intact when taking the subset. Removing waveform metadata or rebuilding arrays without their time origin and sample interval can make otherwise aligned channels appear to have unrelated time bases.

No conversion from one channel's position to another is required when all channels return equal-length, simultaneous blocks. If channel lengths differ, or the channels came from separate acquisition tasks, stop and correct the acquisition architecture before trimming. A shared numeric index is valid only when the samples represent the same acquisition timeline.

Qualify noisy threshold crossings

A single greater-than comparison is enough for a clean, monotonic speed signal. Noise near the target can produce an early crossing, however. Choose the qualification method from the physical requirement rather than hiding the problem by raising the threshold.

Do not choose a consecutive-sample count until the permitted detection delay and signal bandwidth are known. Each added sample delays the declared crossing by one sample interval. Record that rule with the analysis configuration so a later maintenance change does not alter engagement duration silently.

Verify the captured engagement before release

1. Run with a target that the shaft reaches normally.
2. Confirm the retained speed waveform ends at the first sample that satisfies the selected comparison.
3. Confirm all three waveforms have the same sample count, time origin, and sample interval.
4. Check that the final portion no longer contains an extended constant-speed segment.
5. Run a slower engagement and verify that acquisition continues until the target rather than stopping at an assumed time.
6. Test an unreachable target and confirm the maximum-record or operator-stop path ends the acquisition cleanly.
7. Feed the trimmed record to each required waveform-analysis VI and check that it processes only the engagement interval.

If a complete block still appears after the crossing, the application is stopping only between reads. Keep the loop behavior, but trim the final block at its internal crossing index. If samples disappear before the event, inspect block accumulation and the maximum-record exit rather than lowering the target.

FAQ