Specialisation

• Matlab and Mathcad prototyping.
• Windows programs developed in Visual Basic, C++, C.
• Rapid C and C++ code generation from Matlab prototypes.
• DSP development for both low-cost intelligent sensors and high-speed processing in high added-value systems.
• Advanced signal processing methods used in ultrasound inspection - similar to airborne radar.
• Chirp signals and optimum correlation detection methods.
• Time-frequency interpretation methods.
• Neural networks for decision making and data fusion.
• Rapid turn-around of custom signal processing.
• Hardware control and interfacing.

Cambridge Software for signal processing

• Cambridge Software has twelve years experience in developing signal processing and software. We also have extensive hardware development experience.
• We use Matlab or Mathcad for rapid prototyping and testing.
• We use Matlab for generating ANSI C source code that can be used on the final target DSP/PC.
• Our staff are trained to a high level in this field (PhD typically) and offer an intelligent and creative approach to problem solving.
• Our experience results in lower development costs for our clients.

Application - quality inspection using ultrasonic resonance spectroscopy

Cambridge Software was the lead partner in a collaborative European research project in which our technical tasks were to develop a new lock-in amplifier circuit on a PC plug-in board and to write controlling Windows software. Both tasks were completed on time and within budget.

The Spectrum software had two main tasks: to control the lock-in amplifier for collection of ultrasonic resonance spectra and to perform signal processing of the spectra. An interface to an embedded artificial neural network was also included.

Some of the signal processing functions included in Spectrum are:

• Normalisation of the signal level.
• Peak detection.
• Histogram analysis of the distribution of peaks across frequency.
• De-convolution for the removal of one spectrum from another.
• Averaging of spectra.
• Taking the difference between two spctra.
• High-pass and low-pass FIR digital filtering.
• Fourier transformation.
• Re-scaling of frequency results - to correct for small temperature fluctuations.
• Macro commands for cascading signal processing functions - to automate custom processing.

The system has application to testing the quality of a wide range of industrial components, to find faults and defects. We are now marketing systems in collaboration with a US corporation.

Application - finding defects in concrete

We were approached by the Institut fur Massivbau, Darmstadt (a leading German university specializing in civil engineering), and asked to design and build a novel inspection system to probe concrete. The Institut fur Massivbau was sponsored by the German Institute of Standards and the German Concrete Association. The system built for them was based upon several years of research we had already done in this area. The project was successful and the performance of the system has set new standards for inspecting concrete.

Novel elements in the signal processing are:

• An array of transducers is used to allow spatial compounding of received signals.
• Linear sweep frequency chirp signals are transmitted, instead of the customary single pulse.
• Matched filtering is used to detect the chirp signals giving improved signal to (random) noise, approximately 10:1 pulse compression and some improvement in signal to (stochastic) noise.
• Spatial compounding is performed in two ways: using phase-sensitive and phase-insensitive processes.
• Useful information is obtained by comparing phase-sensitive and phase-insensitive signals, such as: the quality of the concrete and how well targets are located in the centre of the ultrasonic beam.
• Target size and distance can be measured from the processed signals. The system gives good performance to 0.5 m and fair performance out to 1.0 m.
• Images of the interior of concrete have been developed using our recommendations for processing - the Weibull distribution to suppress stochastic clutter.

Application - time/frequency domain

We have used the Wigner transformation to investigate ultrasonic dispersion in concrete. Dispersion causes sharp pulses to be lengthened, resulting in a loss of resolution. The Wigner transformation converts a signal in the time domain into a time-frequency graph. False colours are used to represent amplitude.

Swept frequency chirp signals are useful in ultrasonic inspection when working with difficult materials such as concrete, austenitic steels, cast iron or any material where the wavelength of ultrasound and grain size are close in value. The sweeping frequency can be seen in the time-frequency representation as a sloping feature. By detecting the time of arrival of the slope of the chirp and ignoring energy scattered and delayed to later times it is possible to reduce the effect of dispersion. By measuring the slope of the chirp it is possible to apply an anti-dispersion correction.

Application - inspection systems for oil wells

Ultrasound is used for inspecting oil wells. We have been active in this area combining our specialisations of visualizing ultrasound with signal processing to create improved inspection systems. Visualization allows us to see the ultrasonic pulse travelling from the transducer through water, for example. We can see it reflect from the casing and return to the transducer (or not return!). We can perform real-time visualization and signal processing, allowing the impact of any changes on the system to be monitored. This is a unique commercial service.

Acknowledgement

Two images of targets in concrete - with and without enhancement using the Weibull distribution - are reproduced by permission of Institut fur Massivbau, Darmstadt. Data by M.Schickert processing by R.Jahnson.