Inaccuracy at frequencies in which there are background distortions

[Beno68]

Hi,

Thanks a lot for sharing the impedance scan tool. It is a very advanced tool but it has critical technical issues which I have been dealing with some time ago. Since the tool is going to have many users, please consider updating the code with the following description:

The impedance scan gives faulty results at frequencies in which there are already harmonic emissions. The issue comes from the FFT and impedance calculations methods as follows:

1. The initial emissions should be deducted from the current and voltage components, otherwise the impedance values would be faulty.
   (Ih = Ih(after injection) - Ih(before injection)) ,(Vh = Vh(after injection) - Vh(before injection))

2. The FFT for currents should be calculated at zero-crossing point of voltage, otherwise there will be an angle error for the current components.

To see the error, you can add an emission source to the example RLC circuit. You would see that the impedance shall be calculated differently with and without the emission.
RLC circuit with distortion at 250Hz:

Impedance without distortion:

Impedance with distortion at 250Hz:

For more information, please refer to this paper: https://ietresearch.onlinelibrary.wiley.com/doi/full/10.1049/rpg2.12245

Best Regards,
Behnam

[YLIenerginet]

Dear Behnam,

Thanks for your prompt feedback and testing our tool. This is a very good question. I will add it also into the tool's FAQ page.

Actually we tried to investigate the scan method with and without deducting background harmonics during the development, because we did find the inaccuracy at harmonic orders. But we found the two methods resulted in the same scan results at the harmonic order, and the inaccuracy still existed. We tried to analyze why, and think this is because in principle the impedance should be calculated by knowing both the PoC voltage and the harmonic emission source's voltage, i.e.

Z(fp) = [ Vpoc(fp) - Vemission(fp) ] / Ipoc(fp)
assuming fp at the harmonic order.

For actual testing, the device under test may have some harmonic source emission, but since they are black-boxed models, we could not measure Vemission(fp). Therefore it is not possible to solve the above equation. Of course one can measure Vpoc(fp) and Ipoc(fp) with and without adding perturbation, but under the two simulation runs, we cannot gurantee Vemission(fp) is unchanged, and we do not know what Vemission(fp) is, so we cannot measure the correct impedance.

Recommendations:

• So far, the released version of tool does not apply a deduction of background harmonics for the impedance calculation. And we recommend to use a frequency scan setting that skips the harmonic order or just ignore the response at the harmonic order.
• If a model developer has access to their internal parameters and is able to measure the harmonic emission source from their device, they can further develop the tool to measure more accurate result at the harmonic order. Our subscript "IMTB_calculate.py" has a global parameter "ENABLE_DEDUCTION_DFT_NONINJ = False", so user may enable this as "True" and further investigate how to do the right calculation.

I hope this addresses your concern and we welcome your further feedbacks or contributions.

Best regards,
Yicheng

[Beno68]

Dear Yicheng,

Thanks for the reply.
This is a common issue regarding the impedance calculation and it could lead to extremely wrong results unless we ignore those frequencies as you mentioned. However, the calculation can be done on a black-boxed model as well. We just need to calculate the voltage and current FFTs at two separate time window: one before perturbation and another one after perturbation. Another important consideration is that we should take the current FFT aligned with voltage, otherwise it will lead to angle error and faulty results. The current FFT window should be at zero-crossing voltage instance. Most probably, this is the reason you have gotten the same results with two methods.

Vh = [Vh(FFT after injection) - Vh(FFT before injection)], Ih = [Ih(FFT after injection) - Ih(FFT before injection)]

Then, ignoring the couplings, the impedance shall be: Zh=Vh/Ih

I hope this comment could improve the accuracy of the studies for all the stakeholders.

Best Regards,
Behnam