Easy to Use Software
The challenge when writing any software is to design the interface to be easy to
use for all levels of user, from the first time user to the expert. Good software
focuses on user’s goals. The first step in creating a usable product is understanding
those goals in the context of the user’s environment, task or work flow. Usability
relies on user-feedback through evaluation rather than simply trusting the experience
and expertise of the designer.
Low noise electronics
All measurement systems will be limited by noise, and the goal of the electronics
designer is to ensure that the noise contribution of the electronics is negligible.
For some systems this isn't too hard to achieve, but instruments like FTIR
spectrometers and Raman spectrometers need ultra low noise amplifiers. All detectors
generate thermal noise, which is proportional to the square root of the resistance
and the absolute temperature. MCT detectors and InGaAs detectors are commonly cooled
to liquid nitrogen temperatures (77 kelvin), calling for amplifiers with sub nanovolt
or femtoamp per root hertz noise density. These levels of performance also require
the highest levels of EMI rejection.
Low distortion electronics
Spectroscopic techniques like FTIR and xray analysis are dependant on system linearity.
Any non-linearity moves energy from one part of the spectrum to another, and this
is particularly apparent in FTIR where most of the information is determined from
Assuming the instrument electronics don't contribute significantly to the noise
floor, bandwidth narrowing is key to performance. For best performance, stability
and flexibility, this is now mostly performed in the digital domain. The role of
analogue electronics is limited to performing equalisation, gain switching and anti-alias
filtering. Digital filtering may be performed by a DSP or by an FPGA. Any instrument
that scans a sample with some form of excitation can usually be set up to perform
an average of repeated scans, and this technique is widespread in all spectrometry
and techniques like, confocal microscopy, xray microanalysis and electron backscatter
Photomultiplier tubes (PMTs) are still unbeatable if you need high speed, large
area, low noise photodetectors. Although they are big, fragile and inefficient when
compared to solid state detectors like CCDs, their strengths make them popular in
confocal microscopy, flow cytometry, multi-photon microscopy and fluorescence lifetime
imaging. A PMT converts a photon in to an electron, which is then amplified by a
series of electron multipliers. With up to a dozen stages, gains of 10,000,000 are
easily achieved, and response times of a few nanoseconds equate to bandwidths of
hundreds of MHz. The traditional photomultiplier has a response curve which peaks
at the blue end of the visible spectrum, but new photocathode materials such as
GaAlAs give much improved performance towards the red end of the visible spectrum.
At low levels of light intensity, photomultipliers can distinguish individual photons.
This allows light intensity to be measured by counting the detected photons. Photon
counting lowers the system noise floor because leakage currents, and their associated
shot noise, can be rejected. Accurate measurements can also be made of the
arrival time of a photon, with optimised tubes introducing an uncertainty of less
than 0.1 nanoseconds.
Lasers are the preferred sources of optical excitation for confocal microscopy,
multiphoton microscopy, Raman spectroscopy and flow cytometry, as well as a wavelength
reference FTIR spectrometry. The choice of wavelengths is gradually increasing,
but huge variety of package sizes makes the design of adaptable or modular systems
a problem. As an example, a green 532 nm laser is available in a small, solid state
package, but if you must have 543 nm, then you are going to have to find space for
a 500mm long gas laser. Gas lasers have dreadfully low efficiencies, and as an example
an argon laser with a 50mW output power will need 1 kW of input power. A kilowatt
of waste heat can only be handled with forced air or water cooling, which adds to
the system engineer's headaches.
Controlling the power of the light is also a problem, as the power of many lasers
cannot be modulated directly. Some lasers use non-linear elements to create harmonics,
effectively doubling the output frequency. For these non-linear crystals to work
efficiently, they must be pumped by high intensity light, which means that they
cannot easily be modulated. There are a number of ways to modulate laser light,
such as ND filters, liquid crystals, acousto-optic tuneable filters and pockels
cells. Each method has its own drawbacks in terms of speed, dispersion and cost.
Multiphoton imaging and fluorescence livetime imaging use pulsed lasers with pulse
widths of only about 100 femtoseconds, so an optical attenuator must not introduce
much dispersion. If speed is not a requirement, ND filters can be used, but for
high speed work a pockels cell can be used.
The technique of fluorescence labelling provides an immensely powerful way of imaging
proteins, DNA or other cell structures. Better instrument sensitivity allows users
to either use lower label concentrations, shorter exposure times or lower excitation
powers. Excitation power is an important factor when dealing with live cells, as
short wavelength light causes photodamage, and may kill cells.
As well as showing the presence or absence of a target molecule, some labels are
sensitive to their local environment. Some labels exhibit a shift in emission
wavelength according to an ion concentration, whilst others exhibit a change in
their decay lifetime. The challenge of distinguishing between livetimes in the range
of 2 to 4 nanoseconds is considerable, but possible. Femtosecond pulsed lasers
have sufficiently accurate pulse repetition periods to allow repeated photon arrival
measurements to be made.
High performance electro-mechanical actuators
Many instruments need some form of precision movement for their operation. An FTIR
spectrometer must move mirrors in the interferometer to modulate the beam, and laser
scanning microscopes use mirrors on galvanometers to perform the X and Y scanning.
Confocal and multiphoton systems acquire 3D data sets, so they also need to scan
in the Z axis. Z axis microscope movements are usually performed by a stepper motor,
but piezo-actuators are also used. Closed loop operation calls for the design of
some challenging feedback loops as many systems are intrinsically prone to instability.
In addition to actuators which form part of the fundamental operation of a system,
there is a general need to motorise the configuration of an instrument. This is
not because users are too lazy to move something manually, but because mechanisms
that aren't motorised can't be automated. A lack of automation leads to the possibility
that an experiment will be performed when a part of the instrument is in the wrong
configuration, which leads to bad results. Preventing such problems is covered by
"Good Laboratory Practice".
Some instruments require precise control of the temperature of the sample or of
the detector, or use low temperatures to reduce thermal noise. Liquid nitrogen is
sometimes used to cool detectors, but keeping a system topped up is an inconvenience
to the users. For temperatures down to -50 C, peltier cooling may be a better solution.
Peltier heat pumps can heat as well as cool, so they can also be used to temperature
stabilise components. Bidirectional control presents significant problems in the
design of the control loop due to the different loop gain when heating as opposed
to cooling. Some systems avoid this problem by running the cooler all the time and
achieving through a heater element.
The move by Intel and AMD to dual and quad core processors follows the multiprocessing
path troden by the scientific computing community over 20 years ago. PC programmers
are now coming to terms with new classes of problem like the deadly embrace deadlock.
Partitioning problems efficiently between processors requires a clear understanding
of the problem domain and of the resources available.
The mass produced PC provides a stunning amount of capability at a bargain price.
PCs have become the standard user interface for scientific instruments, taking on
the user input, data storage, data analysis and data visualisation functions. Dedicated
keypads and displays have virtually disappeared, although some instruments embed
a complete PC. This approach is double edged, as the embedded PC provides instant
storage, processing and connectivity, but with the drawback of a short hardware
lifecycle and a constant need for the users to keep the operating system patched
and free from malware.
Connecting dedicated electronics to a PC used to be simply a matter of using RS232,
GPIB or the ISA bus. These interfaces are simple, and the PC operating system allowed
direct access to the hardware. The ISA bus disappeared many years ago, PCs may not
have RS232 ports any more, and direct access to hardware is not permitted in modern
operating systems, so new solutions are now used.
The PCI bus has proved to be a worthy successor to the ISA bus. As the performance
limits of PCI have now been reached, PCI-Express will replace PCI for ultra-high
throughput applications. PCI is not likely to disappear for a while though, because
it is well understood, cheap and capable of handling many applications. PCI slave
interfaces are available as soft cores for many FPGAs, which means that dedicated
slave interface chips may not be necessary. The drawbacks of using a dedicated PCI
board is that you need to write a device driver and you need to open up the PC in
manufacturing to plug the board in.
A high performance alternative to designing a PCI board is to use the IEEE1394 firewire
bus. With data rates of 400 Mbps (or 800 Mbps for the b specification), power management
and bandwidth management, it is an attractive solution to the problem of getting
data in to a PC. One drawback of firewire devices is that you need to provide a
device driver. Firewire is a peer to peer network, which allows devices to exchange
data without the intervention of a host.
Now that USB has a high speed (480 Mbps) capability, it is a rival to firewire.
It's ubiquitous nature is a benefit, and you don't necessarily have to write a device
driver, as you may be able to use one of the class drivers. Some developers are
migrating old RS232 designs to USB by adding a USB to RS232 virtual com port chip
inside their box. Whilst this circumvents the problems caused by the disappearance
of PC RS232 ports, it is not without problems, and not a suitable approach for new