Why does USL utilise hardware logarithmic amplifiers?

19th January 2023

In a time where converting linear amplifier signals is commonplace, why persist with hardware logarithmic amplifiers?

The core issue is the performance required from a digitiser to replace the log detector in the PM40 (and previous generations of USL pulser-receivers) and give a net result at least as good. The 2 key parameters for any digitiser are number of bits and sampling rate.

USL’s system comprises an amplifier chain plus filters which convert incoming sinusoidal RF signals to one whose instantaneous amplitude corresponds to the logarithm of the magnitude of the incoming RF. It does this with an accuracy better than 0.5dB over a signal range of 95dB and frequency range of 100kHz to 30MHz.  Our log detector is linear over a range of >95dB which is a ratio of 1:56,234.

This ratio implies the smallest RF signal at the input to the log detector is 1/56234 of the largest. If you want any sort of resolution when digitising the smallest signal, you need to be able to capture it with at least 2 bits resolution. This means a suitable replacement digitiser must have at least 56,234*4=224,936 quantisation levels over its input range and to achieve that it must be at least 18 bits resolution.

To capture an ultrasound signal with sufficient fidelity for scanning you need about 10 samples per cycle so at 30MHz, the sampling rate would be 300 MSPS.

While digitisers of 18 bits at a sampling rate of 300 MSPS do exist, they do not exist in a form (cost, size, power consumption) that make them viable for use in NDT inspection systems. Digitiser chips exceeding 250 MSPS are available but not with 18 bits of usable resolution.

While converting signals from linear to logarithmic or accepting a lower maximum usable bandwidth or reduced dynamic range is becoming common practice, the performance of our hardware logarithmic amplifiers shows why hardware still beats conversion methods hands-down.