Internals Of Bode Plot Measurement

[hugocoolens]

When using the bode plot measurement, does the system internally measure the peak-to-peak output voltage and divide this by the peak-to-peak input voltage to determine the amplitude characteristic? And does it measure the delay between two corresponding zero crossings of the in- and output signal to determine the phase characteristic?

 

regards,

hugo

[attila]

Typical usage of Network Analyzer uses the AWG 1 output and Oscilloscope Channel 1 input of the device connected to the filter input, oscilloscope Channel 2 to the filter output. As shown in the picture below.

 

 

The analysis is performed from start to stop frequency in the specified number of steps. For each step the AWG is set to a constant frequency and the Oscilloscope performs an acquisition. Using the FFT result from the index corresponding to the current frequency step, the magnitude and phase value is calculated. The magnitude of Channel 1 (Filter input) is calculated relative to AWG Amplitude. The other channels magnitude and phase is calculated relative to Channel 1.

 

[hugocoolens]

 

Typical usage of Network Analyzer uses the AWG 1 output and Oscilloscope Channel 1 input of the device connected to the filter input, oscilloscope Channel 2 to the filter output. As shown in the picture below.

 

bode.connection.png

 

The analysis is performed from start to stop frequency in the specified number of steps. For each step the AWG is set to a constant frequency and the Oscilloscope performs an acquisition. Using the FFT result from the index corresponding to the current frequency step, the magnitude and phase value is calculated. The magnitude of Channel 1 (Filter input) is calculated relative to AWG Amplitude. The other channels magnitude and phase is calculated relative to Channel 1.

 

Thanks for the reply, I think this is really an interesting technique, I never saw it applied before (until now I always used sweep gen and log convertor). I wonder about some issues of it though:

1. How many periods of a certain frequency per step are applied?

2. What's the effect of a transient response on the result?

3. How sensitive is the result when distortion of the output signal occurs (of course ideally the output should be a pure sine-wave, but

a little distortion may be allowable), I assume you will only look at the spectral component corresponding with the applied frequency and neglect other spectral components (which ideally should not

be present)

 

To investigate the approach used by the Analog Discovery with the standard approach (measuring peak-to-peak values), it would be nice if I could write a script

which does the following:

1. apply a sinusoidal voltage of frequency f1 ,

measure the peak-to-peak voltage of input and output signal

write frequency f1, input-ptp-voltage, output-ptp-voltage in a text file

2. apply a sinusoidal voltage of frequency f2

measure the peak-to-peak voltage of input and output signal

write frequency f2, input-ptp-voltage, output-ptp-voltage in a text file

.

.

3. process the results with matlab or gnuplot and compare with the results obtained from the Analog Discovery module

 

Could you suggest how to accomplish steps 1, 2, ...? (3 is not a problem)

 

 

thanks in advance

Hugo

[attila]

1. The minimum number of periods can be specified in Network Analyzer Settings/Options/General.

The more periods produce more precise result, but at low frequencies the analysis will take more time. In this case the number of steps can be reduced.

 

2. The frequency step transition problem could occur when analyzing resonant circuit, like a speaker. This can be solved by increasing the Settle time in Options. This specifies the milliseconds between frequency change and acquisition.

 

3. This is the advantage of FFT over the Pk2Pk analysis. It is looking only at a specific frequency component, so it is less susceptible on distortion harmonics and other noise components.

 

Below you have a batch-script for pk2pk bode and excel plot comparing this with the FFT Bode.

 

https://drive.google.com/file/d/0B-YtGABZihr-b01rYTI1YWZraFk/view?usp=sharing

 

 

https://drive.google.com/file/d/0B-YtGABZihr-dnVRb3U0bWhHaTg/view?usp=sharing

[hugocoolens]

1. The minimum number of periods can be specified in Network Analyzer Settings/Options/General.

The more periods produce more precise result, but at low frequencies the analysis will take more time. In this case the number of steps can be reduced.

 

2. The frequency step transition problem could occur when analyzing resonant circuit, like a speaker. This can be solved by increasing the Settle time in Options. This specifies the milliseconds between frequency change and acquisition.

 

3. This is the advantage of FFT over the Pk2Pk analysis. It is looking only at a specific frequency component, so it is less susceptible on distortion harmonics and other noise components.

 

Below you have a batch-script for pk2pk bode and excel plot comparing this with the FFT Bode.

 

https://drive.google.com/file/d/0B-YtGABZihr-b01rYTI1YWZraFk/view?usp=sharing

 

pk2pk_vs_bode.png

Thanks a lot for the comments, this is real interesting material for investigating further, I'm going to check out your batch file asap. Is it possible, you forgot to attach the excel file?

 

kind regards,

Hugo

[attila]

I added a comparison in csv format:

https://drive.google.com/file/d/0B-YtGABZihr-dnVRb3U0bWhHaTg/view?usp=sharing

[hugocoolens]

I added a comparison in csv format:

https://drive.google.com/file/d/0B-YtGABZihr-dnVRb3U0bWhHaTg/view?usp=sharing

Thanks a lot!

hugo