Analogue vs digital timing for time-of-flight mass spectrometry
Relative merits and discussion
Here is a brief description of the two strategies.
- Preserves information on the pulse amplitudes, which means that up to a point one can distinguish between single and multiple ion arrival events. If enough ions are present, it may be possible to achieve a useful spectrum in a single spectrometer cycle.
- At least double the cost of the digital option for a comparable time resolution, even before considering that a higher performance detector/preamplifier may be required (see below). High speed digitisers tend to have a limited number of bits, often using 8 bit ADCs, with an "effective" number of bits that is rather less than the nominal ADC data size.
- Electronic noise is recorded and appears in the data, as does any ugliness in the pulse shape.
- Time resolution is limited by the pulse width from the detector/preamplifier combination. If good mass resolution is important, it may be necessary to invest in expensive detector arrangements, with conical anodes and very high performance pre-amplification. This is particularly difficult/expensive if the detector needs to be floated away from ground, for example, to measure negative ions.
- Absolute quantification is very difficult because the detector gain has a first order effect on spectrum peak heights. Detector analogue gain is a highly non-linear function of excitation voltage, subject to drift and difficult to measure unless extra hardware is present, such as a Faraday cup. In practice an internal standard is necessary (a mass peak to normalise against).
- As well as drift and uncertainly in average detector gain, there is also a large scatter in the gain for each ion. This means that ion bunches have to be quite large before the pulse peak height will give reasonable precision for the measurement of the pulse charge. As a rough guide, at least 100 ions would be needed for a precision of 10%.
- To achieve a better precision than is available in a single spectrometer cycle either due to limited digitiser performance or because of the ion arrival statistics, many traces are added together. However, this is much more computationally intensive than building the digital histogram. This tends to further limit the maximum cycle repetition rate or push up the cost .
- The analogue form of data is very inefficient; most of the memory is occupied by describing the baseline between peaks. The data-volume is proportional to the timing resolution. This can become a problem if trying to monitor a fast changing analyte. Either data will need to be processed on-the-fly, or information thrown away, or very large data-sets stored. The data size is large even if the signal level is low.
- Relatively inexpensive (less than half cost) compared to analogue methods of comparable time resolution. The advantage becomes ever bigger as higher performance is demanded from the spectrometer..
- Discriminates against electronic noise by using a threshold. Spectrum baseline is often simply zero.
- Timing resolution can be about 10 x better than the width of the pulse from the pre-amplifier.
- Spectrum peaks are free from pulse artefacts such as bounce or ringing.
- Quantification is reasonably straightforward except for the saturation issue (see below) - one timestamp per ion. The statistical fluctuation in multiplier gain is largely removed
- The detection system is operating in the most sensitive way that it can because it deals in single ion arrivals.
- Data is held efficiently; the data-volume is roughly proportional to the number of ions detected and independent of the timing resolution. This makes it much easier to hold all the information and extract a variety of "views" of the data in post-processing, for example chemical images, or single ion chromatograms in additional to conventional mass spectra.
- Care must be taken to operate the multiplier at the correct gain so that The majority of single ion arrivals create a pulse above the threshold and are thus recorded as a single timestamp.
- The detection circuits will always have a "dead-time", typically at least 5ns. If two or more ions arrive within this dead-time, only the first ion will be recorded. Given the tendency for ions of the same mass to arrive close together in time, the probability for multiple ions of a given mass to arrive in a single spectrometer cycle needs to be kept low for the spectrum peak to yield good quantitative information. The statistical properties of this "saturation" effect are well understood, so there are methods to make corrections, but the possibility for saturation has to be borne in mind whenever signal levels are high.
- Many cycles are required to yield a spectrum with good dynamic range. Put crudely, if you want to have a spectrum where you have quantitative information about two mass peaks where one peak is 1000 times the intensity of the second, it is clearly necessary to perform more than 1000 cycles, in practice more like 10,000. This may be fine if the spectrometer repetition rate is 10,000 Hz, a quite normal rate, but it may be rather limiting if something like a laser limits operation to 10Hz. The saturation effect, described above, limits the rate at which the more intense peak can be recorded, even if plenty of ions are available. Fortunately, in practice, nature provides us with a work-around. There is nearly always a minor isotopic peak associated with the intense peak that can be monitored instead. The major peak can then be allowed to saturate and an estimate of its true area derived from the minor isotopic peak.
In general, if you can use a counting solution you should! If there is a choice between having fewer cycles with lots of ions, or many cycles with a few ions, there is no contest, the counting solution is much better. Although the "one ion per peak per cycle" saturation limit sounds a little radical, typical spectrometer repetition rates range from 50kHz for small spectrometers to 10kHz for larger models. A few 10 - 20 milliseconds of data collection in counting mode will gave the same dynamic range as an 8 bit digitiser, at much higher resolution and much lower cost.
However, there are occasions where an analogue system is the best choice, for example, a situation where the spectrometer repetition rate is limited by a laser that will only fire at 1Hz, each "shot" gives thousands of ions at the detector, and unit mass resolution is sufficient. In this case a good digital oscilloscope with at least 300MHz bandwidth, a sampling rate around 1GHz and at least 50,000 point memory would be a reasonable choice.
As luck would have it, many lasers will fire at about 10Hz, but at this speed a regular digital oscilloscope won't be able to keep up with the averaging; some more specialist hardware will be required. This is the regime where the choice between analogue and digital can be finely balanced. The optimum choice will depend on the detailed experimental priorities and the available budget.
Last updated: 16:53 26/02/2014
© Kore Technology Limited 2014