Why to choose AD5933, what are the alternatives ?
I stumbled upon this part few years ago when I was looking for articles about Discrete Fourier Transformation. I noticed the part and became interested in impedance measurement. The extra search did not show any impedance measurement chips as integrated and opened as this device. There is one well known Taiwanese chipset for handheld LCR/ESR meters, but it has no opened documentation. If anyone today will start building LCR/ESR meter using commonly available parts, then chips list will include at least following:
All these 7 blocks are included in AD5933. The part has limitation in A-to-D and D-to-A resolution, short sampling window (only 1024 samples) and very weak analog front-end. But the simplicity of having everything integrated in single chip outweighs the resolution limitations. The speed of 1MHz conversion is decent for measurements up to 100-200KHz. And external clocking allows to lower the range of AC frequencies down to single Hz.
This 7 blocks can be identified in most of classic LCR meters. Perhaps the design of AD5933 is not very original in theory of operation. The originality of this chip is high level of integration, everything is packed in a single chip.
The LCR meters theory of operation is, if expressed in few words, the phase and amplitude sensing for a pure sinusoidal signal with known frequency. More or less the method of sensing the phase and amplitude boils down to so named synchronous detector. The synchronous detector can be looked at as phenomenally sensitive radio receiver. If synchronous detector has a local oscillator with accurate 90 degrees quadrature output, the detector will produce so named quadrature components at the output.
What is quadrature synchronous detection when looked at as Fourier Transform ?
Quadrature modulators and detectors were invented in world of radio design to harness the informational capacity of frequency spectrum. If you are not familiar with I-Q Detectors, and do not want to dive into math, you only will need to understand next paragraph.
To describe any pure sinusoidal AC signal it is enough to know the amplitude. Sinus is a function with zero Y at zero X. But saying "sinusoidal", not always means that signal has started at zero time with zero amplitude. It can be shifted left or right few seconds, microseconds etc. Shift in degrees is starting phase. So to completely describe pure AC signal with known f the observer needs 2 numbers: Amplitude and Phase.
I-Q detectors detect this 2 numbers. Pure AC comes at input, and 2 DC signals are at output (in those variations where quadrature components are post-processed to produce DC): Amplitude and Phase.
The Single Bin Fourier Transform as I-Q Demodulator - Detector is no different, except that multiplications are happening in digital domain, not in analog domain. In addition to multiplication to carrier signal, AD5933 also does Hamming windowing, which improves the frequency band selectivity of detector. In theory this windowing has no much effect if source signal is narrow banded pure AC, but in practical measurements, every blocking of undesired noise helps.
Another great property of synchronous detectors (either digital or analog) is that they are completely immune to DC component. This is very good property, because DC is much more non-trivial when is involved in high accuracy measurements. The extra advantage of Discrete FFT that results are digital and can not be corrupted later down the chain. The legacy analog detectors have DC outputs, which are yet to be carefully preserved/scaled/converted to Digital domain.
When AD5933 is used, there is no single place in schematics, where DC offsets, temperature induced voltages are affecting measurements. Whole analog part of UA315 is DC immune.