#35 🚀 The AI Road Map: Measuring Losses

SuperPowers for Electrical Engineers

May 22, 2022

👋 Hello Supergineers! Dr. Molina here! 👨‍🔧

Last week, we recorded a Training session about Artificial Intelligence at Frenetic for our High Freq Magnetics Training (by the way, the next batch starts in June, and I have a 25% of discount for you 🥳 using the code NL25) and that motives me to talk about one of the main problems building the AI in magnetics. The data.

The big question in building an AI is, how you get the data and how many data points are you able to get in a specific space.

In this edition, I will talk about a method we are using for generating data of the core and winding losses and some learnings in the process. I have to thank you my colleague Juan Aguaron, who is the Project Engineer who really has the hands on the hardware, for explaining to me the details 🤗.

Core losses Method

The University of Bristol published work for measuring core losses and winding losses in an accurate way and we have arranged a collaboration with them, for using this technology for generating data for our AI. The method is called the Triple Pulse Method (TPT).

The Triple Pulse Method consists of exciting a magnetic component winding with a triple voltage pulse, starting in a state of magnetization known. The waveforms are represented in Figure 1.

Figure 1. A triple pulse applied to a winding

The circuit used for applying the TPT is represented in Figure 2. As you can see, there are some capacitors, which are really giving the energy during the pulses.

Measuring the Core Losses

Capturing the measurement It’s needed an additional winding, where we will measure the voltage induced (Vmea) in the winding of the device under test (DUT).

In the main winding, we will measure the current (Ip).

The integral of Vmea x Ip during a complete cycle will be the losses in the core.

Some of the advantages of this method:

Rectangular voltage excitation
One-shot measurement: independent of temperature, large currents are possible
Measurements under DC-biased current

One of the most interesting advantages in my opinion is the impact of the DC BIAS on the losses, which are taken into account.

The following graph shows the volumetric losses for a TX25 core of 3C90 calculated with the measurements Juan did in comparison with data from Ferroxcube (only two points in the datasheet) and from MagNet (the database of Princeton University).

As you can see, the results are very similar.

Figure 3. Volumetric losses calculated with the measurements of the TX25

Some of the interesting results of these measurements are the different volumetric curves of losses obtained depending on the core used for doing the calculation.

For example, in the graph below. The volumetric losses for 3C97 are different if we use for the characterization a PQ20/20, a PQ35/35, or the datasheet values provided by Ferroxcube, which use a TX25 core.

At B=0,2T the differences are around 20%. This result shows something we have covered in other editions, the core shape and size affect the losses, therefore, we can´t use exactly the same coefficients for a PQ65 as for a TX25 with the same material.

Figure 4, Volumetric losses calculated with different core shapes

Measuring the winding losses

One of the main advantages of this method is, that we can measure the winding losses.

In the first step, we have measured the voltage in an additional winding and doing the integral, we have the core losses (blue line in the graph).

In a second step, we measure the voltage directly in the main winding and multiply by the main current, doing the integral, the result is the total losses (red line in the graph).

Therefore, the difference between both lines (core and total losses) are the winding losses.

Figure 5. Losses calculated with the voltage and current measurement

If you want to read more, the article of reference from Bristol University is:

[1] J. Wang, X. Yuan, and N. Rasekh, "Triple Pulse Test (TPT) for Characterizing Power Loss in Magnetic Components in Analogous to Double Pulse Test (DPT) for Power Electronics Devices," IECON 2020 The 46th Annual Conference of the IEEE Industrial Electronics Society, 2020, pp. 4717-4724, doi: 10.1109/IECON43393.2020.9255039.

🎁 SPECIAL GIFT FOR FRENETIC NEWSLETTER READERS 🎁

The next session of our High-Frequency Magnetics Training program will start on June 1st. We have still some available seats and we want to give the chance to some of our newsletter readers to participate under special conditions: a 25% discount 😱

All the info about the Training Program is here.

When you check out, use the coupon "NL25" and you´ll get a 25% off. This promotion is valid up to Sunday 22nd of May, so hurry up!

You can get the Frenetic SuperPowers and save some money, don´t miss this opportunity...

If you want to get SuperPowers in power electronics, join our training.

Some of the instructors are: