How fast can new results be retrieved / provided?
The timing is given by the detector array and electronics operating it. Furthermore, the processor capability may limit the transfer rate of the spectral respectively the crunching of the data in case of complex algorithm.
The time to acquire a spectrum is the combination of exposure time (integration time), readout time, transfer and computing time. The latter is heavily influenced by the used computer capacity.
The read(out) time is given by the number of pixels divided by the effective clock frequency, the frequency applied to the chip, controlling the switching from pixel to pixel.
tRead = NPixel/fCLK,
The overall measurement time includes the period of the exposure over which light is integrated:
tMeas. = tRead + tIntegr.,
During readout most sensors are sensitive, therefore, this is not a lost time. In this case the readout time defines the shortest possible integration time. If there is a separated exposure phase, then this integration time can be shorter, e.g. 0.1 ms. Back to top▲
If 1024 pixels are read at 1 MHz frequency, the readout takes about 1.1 ms, so every 1.1 ms a new spectrum could be provided, equaling a rate of about 900 spectra/s
To acquire real high data rate (NSpectra), the best is to apply a so-called Burst mode: a specific number of spectra is taken one after the other; the last read scan is the reset for the next spectra, no time is lost:
NSpectra = 1/tRead = fCLK/NPixel,
External triggering: The above is valid only if data are taking continuously. If the data collection has to be synchronized to an external process the picture changes. If the spectrometer is the Master, external events could be triggered by the spectrometer electronics, without any loss of spectra acquisition speed. The triggering is typically done at the end of readout process (related to EndOfScan signal generated by detector array). Back to top▲
However, if the data acquisition has to be triggered by the external event, the procedure differs.
To understand the complexity, it is important to recall that a typical detector array is sensitive permanently. If not read out frequently, charge builds up and will distort the spectral information. Also, the amount of signal is directly proportional the integration period (at constant light level). A slight change leads to a slightly different signal measured.
Various scenarios are possible, here three important ones:
a) Spectrometer controls integration time: first, the spectrometer has to be primed, causing delays of a few ms, then a trigger signal starts the first read process initiating the integration while a second scan reads the spectrum from the array.
tMeas. = tRead + tIntegr.+ tPrime,
The key demand is to provide a suitable trigger signal. Often light gates sensing the passing samples are a good means. The measurement can also be combined with a flash source, controlled by the spectrometer electronics.
b) External event controls integration period: to avoid any initial delay, the external event can be used to control the START pulse (read procedure). Any jitter of this external event will cause a variation in integration time.
c) If very high speed data acquisition is given (e.g. CMOS based technology) then the spectrometer may acquire spectra ongoing and adequate spectra are selected, e.g. based on signal level.
Timing Precision: is given by the stability of the master clock, of which the clock frequency as well as the START pulse generation is derived. Since the time distance between two consecutive START pulse define the exposure period, the accuracy of how much signal is detected depends heavily on this precision (in case of a constant light flux). Back to top▲
Example: to achieve a really 16 bit accuracy at 10 ms integration time, the integration period has to be accurate to 0.15 µs.
If the light is generated by a flash, then this flash has to be within the sensitive period of the detector only, the exact duration of the exposure period is no longer critical.