Best practices & troubleshooting

A successful PIV measurement depends far more on experimental design and careful validation than on finding a "magic" combination of software settings. This page distills the recurring advice from years of PIVlab support questions — most problems trace back to the images, not the analysis parameters.

The five-stage mindset

Every PIV measurement goes through roughly the same five stages. Working through them in order — rather than jumping straight to parameter tuning — is the single biggest predictor of a smooth analysis:

Stage	What it covers
1. Acquire good images	Illumination, seeding, optics — see Particle seeding & image quality below, and Capturing images from a camera.
2. Choose acquisition parameters	Time separation (Δt) and expected particle displacement — see Choosing Δt together with the interrogation area.
3. Configure the analysis	Interrogation area, passes, algorithm — see Tips & tricks: choosing a suitable interrogation area.
4. Validate & interpret	Filters, interpolated vectors, VDP, correlation quality — see Velocity-based validation.
5. Troubleshoot systematically	Recognise a symptom, trace it back to acquisition, seeding or processing — see Troubleshooting by symptom below.

Particle seeding & image quality

Good particle images outperform clever parameter tuning every time. Aim for:

Particles with good optical contrast against the background.
A homogeneous distribution — avoid dense clustering in some areas and empty gaps in others.
Moderate density — too sparse starves the correlation, too dense causes particle images to merge.
Seeding across the entire field of view you intend to analyse.

Software can't fix bad images Poor seeding, compression artifacts, and uneven illumination cannot be reliably compensated for by adjusting interrogation area, preprocessing, or any other analysis setting. If results look wrong, improving the experiment usually has a far bigger impact than reworking parameters.

Diagnose before you tune

Before changing any software setting, work through this checklist — it takes a minute and often points straight at the real cause:

Are individual particles clearly visible in the raw images? The human eye is extremely good at tracking particles or textures — if you can't visually see groups of particles moving between the two images, it's unfortunately likely that the PIV software won't be able to find that motion either.
Is illumination uniform across the field of view?
Are there reflections off nearby surfaces or objects?
Is the laser sheet thin and well-aligned, or thick and diffuse?
Is seeding density sufficient — and consistent — across the whole image?
Could image compression be introducing artifacts?

Only once these are ruled out is it time to look at analysis settings.

Troubleshooting by symptom

A quick map from what you're seeing to what's likely causing it:

Symptom	Likely cause
Very noisy vectors	Interrogation area too small, poor seeding, or low signal-to-noise — see choosing an interrogation area.
Large regions of interpolated (orange) vectors	Poor image quality, reflections, or insufficient particles in that region — see Image pre-processing.
Random/spurious vectors near walls or objects	Missing or incomplete masks, or reflections off the object.
Unrealistically high velocities	Incorrect time step (Δt) or calibration.
Displacement measured close to zero where it shouldn't be	A strong, stationary background signal can correlate more strongly than the actual particle motion, biasing results toward "no motion." Subtract the background in Background subtraction — usually the better fix. Disable auto-correlation is an alternative for special situations where that isn't enough.
Correlation fails outright	Excessive particle displacement, compression artifacts, or poor seeding — try increasing the Pass 1 interrogation area first.
Different results between the GUI and a batch run	Different pre-processing settings applied in each case — double-check they match.

Where this advice comes from This page distills recurring, practical guidance from years of PIVlab user-support questions — it complements the rest of this manual, which documents what each control does, with advice on what values and workflow tend to work well in practice.