Gepard Enhancement Proposal (GEP1): reducing measurement time based on particle groups instead of statistical splitting
During a recent measurement we realised that our current practise of measuring samples with too many particle may lead to overestimation of large MP.
How is the current practise?
- particle detection in challenging samples (many particles, large particle size differences, partial overlays) is set to over-fragment particles in order to ensure detection of all small particles and get multiple spectra on larger ones
- to reduce spectra measurement time, a random subset of all detected particles are measured (e.g. 50%)
- the measured spectra are evaluated
- particles are evaluated and over-fragmentation is removed by combining sub-particles which visually appear to be one particle
- total numbers are multiplied with the splitting factor to extrapolate to the full sample
What is the resulting problem?
Any particle that was detected by more than 1 sub-particles due to over-fragmentation has a higher chance to end up in the final result. The larger the particle, the higher the likelihood that it includes one or more sub-particles that got measured.
What is proposed?
Gepard may identify groups of particles at the end of the particle detection. A group comprises of all particles that share a border (within some margin). If reduced measurement time is desired, instead of statistical splitting, Gepard should reduce the particles selected for measurement down to a certain minimum of all particles per group (e.g. 1 or a percentage).
What would be the benefit?
The problem of unequal detection of particles of varying size classes would be avoided, when reducing the measurement time by excluding detected particles from spectra scan. The necessity to extrapolate to a full sample would become obsolete.
Are there any drawbacks?
The possible reduction of measurement may be substantially less than in statistical splitting, depending on the composition of the sample (especially if the share of large particles is negligible).