- A signal is an output reading from an instrument that represents a sample
- All signals are either Gaussian or Lorentzian in distribution, meaning they are all perfectly symmetrical
- If not symmetrical, the signal will be made of constituent symmetrical contributors
- The process of reversing this is called deconvolution and is a computational process.
- Signals are always accompanied by noise, the general background uncertainty of the universe.
- The ratio of the signal:noise can be used as a measure of statistical usefulness.
- Arise from determinate (non random) errors
- Can be:
- Instrumental - the instrument is not functioning correctly. (operating temp, calibration error, etc.)
- Personal - Judgement errors (reading at the wrong angle, subjective determination of endpoint)
- Methodical - a result of poor experiment design (slow reactions, instability of reagents, concentration change from volatilisation)
- How well a technique is capable of detecting a change in signal
- How much does the signal change for a change in the measured variable
- Can be depicted by the slope of the calibration curve
- Dependent on the scale of both axis
- The smallest amount of analyte that can be reliably read
- Often considered to be \(3\times\)SNR
- Is the limit of what the instrument can be used to make quantitative determinations.
- Considered to be \(10\times\)SNR
- As samples are taken of increasing concentration, often, if the readings continue, a linear trend will disappear
- The range between the quantisation limit and linearity limit
- How capable is the technique of detecting the analyte without excessive interference
- This can be accounted for with a method blank, however all components cannot always be negated completely
- A selectivity coefficient can be produced to represent how much of the signal is actually from the analyte, compared to the rest of the matrix.