Frequency estimation

[infinitenothing]

I tried a few different ways of estimating frequency and I thought I'd compare them. If anyone has some other ideas or links to other techniques, I'd like to see how they perform.

Techniques used the VI:

• Zoom FFT
• PSD
• AutoCorrelation

Frequency Estimation.vi

[Francois Normandin]

If you need to track frequency in a noisy realtime environment, I've seen Kalman-based filtering being used. 
Unfortunately, I don't have an implementation example to provide. NI has this discrete filter in the Control Design and Simulation toolkit... which does not come cheap.

[ShaunR]

Wavelet?

Edited February 9, 2019 by ShaunR

[infinitenothing]

On 2/9/2019 at 7:45 AM, Francois Normandin said:

If you need to track frequency in a noisy realtime environment, I've seen Kalman-based filtering being used. 
Unfortunately, I don't have an implementation example to provide. NI has this discrete filter in the Control Design and Simulation toolkit... which does not come cheap.

Interesting. I've seen its use in things like GPS. Actually, in my particular application, I'm not tracking a frequency. There's some "dead air" which I simulated above with the zeros and then the signal which is unrelated to the previous signal. Does that rule out Kalman or am I misunderstanding how you'd use it. I have the platform suite so I think I have the control design toolkit.

On 2/9/2019 at 9:05 AM, ShaunR said:

Wavelet?

Mhmm..yeah... I know some of these words. 😉 I'll take a look at some of the algorithms they mention.

[ShaunR]

20 hours ago, infinitenothing said:

Interesting. I've seen its use in things like GPS. Actually, in my particular application, I'm not tracking a frequency. There's some "dead air" which I simulated above with the zeros and then the signal which is unrelated to the previous signal. Does that rule out Kalman or am I misunderstanding how you'd use it. I have the platform suite so I think I have the control design toolkit.

Mhmm..yeah... I know some of these words. 😉 I'll take a look at some of the algorithms they mention.

Wavelets are part of the Advanced Signal Processing Library.

[Tim_S]

Total Harmonic Distortion (appears to have been renamed Harmonic Distortion Analyzer from when last used it) can be used to find the fundamental frequency.

[infinitenothing]

Thank you for the tip. The performance seems similar to ZoomFFT.

[ShaunR]

18 hours ago, infinitenothing said:

Thank you for the tip. The performance seems similar to ZoomFFT.

I know you are looking for the most precise method for this type of waveform, but is there a reason you can't do a peak detect to remove the lead-in?  

[infinitenothing]

5 hours ago, ShaunR said:

I know you are looking for the most precise method for this type of waveform, but is there a reason you can't do a peak detect to remove the lead-in?  

At this point, AutoCorr is working well enough and I'm just curious about the advantages and disadvantages of different methods. My signals are rather noisy so it's not always super clear how far back the "start" is from the trigger.