All the power of digital signal processing
Linear filtering is a very important function to reduce unwanted noise (white or coloured). It is usually carried out in the analog domain, with first, second or (seldom) third order filters. These circuits use several capacitors, inductors and selection logic, which require a large amount of real estate and components of tight tolerances. Furthermore, only a small set of fixed low-pass and highpass frequencies can be chosen by the user. Analog filters show a non-flat frequency response in the band-pass, large transition bands and non-linear phase.
On the other hand, filtering in the digital domain nearly removes these limitations. FIR filters are linear in phase and provide arbitrarily defined cut-off frequencies within the Nyquist band. With a high order enough, response in the passband can be done flat and transition bands narrow, with a high attenuation in the band-stop. Furthermore, linear filtering is applied after the impulsive noise has been removed.
All DASEL ultrasound equipments incorporate a programmable 63-coefficient FIR filter that works in real-time at any sampling frequency. Coefficients and data are 10-bits wide; full precision results are 25-bits wide. Figure 1 shows the filter frequency response for two filters: narrowband (4 to 6 MHz) and wideband (2 to 8 MHz) for a sampling frequency fS = 80 MHz. The ratio of the pass-band to the stop-band attenuation is about 50 dB or more. The filter output is conveniently scaled to keep the response constant independently of the filter bandwidth, providing an extra gain that improves defects detection.
Fig. 1 - Filter response. Left: narrowband (4 to 6 MHz), right: wideband (2 to 8 MHz). Blue line was obtained with full precision coefficients and green line after truncating to 10 bits.
Read more at:
- A Full Featured Ultrasound NDE System in a Standard FPGA