I'm working on a Ultrasonic Transducers resonance tracking system. Where phase shift between Voltage and Current is the feedback signal for my system when phase difference is Zero that means system is resonant other wise depending on the phase difference frequency will be changed proportionally. I went through lots of different techniques of phase detection which includes Several

Analog(Comparatively Simple) techniques such as:-

  1. Phase Locked Loop

  2. Zero Crossing Based Phase Measurement

  3. Simple Phase Detector ICs

Digital(Complex DSP Based) Techniques Using Mathematical Algorithms such as:-

  1. Cross Correlation

  2. FFT

  3. Goertzels Algorithm

  4. Sine Wave fit Methods

  5. VVV

So, My question is:- Which system will be best to use for a phase difference calculation analog or digital?

Considerations & Preferences:- Transducer is noisy (Gaussian white noise + harmonics of fundamental). Accurate phase difference calculation is crucial.

  • $\begingroup$ Have you looked for papers comparing the two? $\endgroup$ – Phonon Feb 22 '14 at 23:30
  • $\begingroup$ When there are a lot of competing approaches, you'll need to use your application's unique requirements and constraints to select one that is appropriate for your case. The only consideration you noted is that transducers are noisy, so you want accurate phase calculations. I'd say that's a pretty universal desire. Without additional detail, I think this question is too broad. $\endgroup$ – Jason R Feb 26 '14 at 0:15
  • $\begingroup$ What additional information should I include ? $\endgroup$ – rahul_b Feb 26 '14 at 5:52
  • $\begingroup$ Sorts of information we'd need: what frequencies are being used? (e.g. 40kHz? 100kHz? something else?) Can you digitize the signals at that rate? What accuracy do you need? As it is, the question is too broad to get a sensible answer. $\endgroup$ – Peter K. Mar 20 '14 at 23:57

Digital has one major challenge for control application: latency.

Any digital system needs to include A/D conversion, processing and D/A conversion. The minimum amount of latency is 1/SampleRate even if you don't do anything else but a straight wire. You also typically can't use off-the-shelf sigma-delta-converters since these have significant amount of conversion latency.

Most digital control systems tend to use heavy oversampling and fast one-shot converters to work around this problem. Once you have the basic infrastructure in place, the digital domain offers a large variety of algorithms that are highly configurable and effect. However the cost of entry is high.


Not the answer you're looking for? Browse other questions tagged or ask your own question.