Opinion needed for measurement accuracy?

I have a P type inductor and I'm attempting to chart its impedance and inductance at freq. between 1KHz to 10KHz. I was using a PC based software and some simple hardware but now I got a hold of an AD5933 evaluation board from Analog Devices (this is some kind of a network analyzer), they call it an impedance converter. I am calibrating it using a 1KOhm 0.1% resistor and do the scan while I assure that feedback does not saturate the ADC.

This is the impedance chart (looks nice):


I find the reactive part using |Z| x sin(phase) and solve L = Xl / 2pif
Here is the chart:


And off course phase:


X axis is freq. in all charts, Y axis is Ohm on the 1st chart & Henry on the 2nd & degrees on the 3rd. This seems a little odd to me. The trend is visible, however the behavior is VERY non linear. What do you think?

however the behavior is VERY non linear

Click to expand...

No. The inductance measurement has a variance of a few % that can be possibly reduced by averaging multiple measurements. And there seems to be certain systematical frequency dependance of inductance that isn't unusual for a ferromagnetic core.

Yuden P type chokes are tested at 1MHz. Which P/N did you use?
You can open the attached zip files in any text reader to read the S parameters from 1 to 10MHz

There are many types of ferrite for different purposes. High permeability types for low frequency and low µ types for high frequency, but none work at constant value over a wide range. It is best to use near tested frequency. Also DC current will reduce the value and operating near SRF is not a good idea, but likely that is well above your test range.

I notice some monotonicity errors on your "nice" linear curve fig 1, so try to use shielded wire pair not coax for probe. But actually it appears to be digital value noise on analog signal ground or reference current. i.e. where data changes from x011111111 to x1111111 for X0 changing form x1 10 11 on most significant bits.

Also drive signal on shield as a "guard signal" method of reducing capacitance of cable if using one. Otherwise probe calibration method is needed and averaging N samples reduces noise variation by √N {as FvM indicated.}

I hope my zip file is attached.

This inductor is a no-brand part I'm afraid. I have no idea who made it or when.
The DCR is pretty high, around 30 ohms according to a DMM.

I'll try to get a shielded wire tomorrow... I imagine it will help a little with the accuracy.

One thing worth mentioning is that according to my PC software tested the inductance is very much different.

One thing worth mentioning is that according to my PC software tested the inductance is very much different.

Click to expand...

Different indeed, but not very much.

You didn't tell about the core material. Some materials have considerable level dependent effective permeability.

Why do you say they are not very different?
I agree they "start out" similar but the 1st one goes from about 0.6H @ 1KHz to 0.56H @ 10KHz, the 2nd one shows from about 0.61H @ 1KHz to 0.78H @ 10KHz.

I'm afraid I don't know anything about the core material. All I can tell is that I am testing at the operation freq.

It sure isn't a P type SMD RF choke, sounds like a big old chunk of laminated iron flat cores with a few hundred turns of AWG30
from my spereadsheet I get RL=2 PI L
The impedance is ~4KΩ~50KΩ from 1~10KHz or

[H] 0.63 0.79
[Hz] 1,000 10,000
[Ω] 3,900 49,600
Rs= 30 Ohm @ 0.334Ω/m assuming AWG30 thus 292 ft of wire (90m)

So it might be good for a high voltage low current filter for < 5W ( depending on thermal dissipation and size) =I^2R or I <5mA

Did this come from an old TV?
hmmm Oh Henry.. it is approaching 1 Henry.


Generally I use the self resonant approach and sweep the inductor with the circuit capacitance see what the resonant frequency is with the measured or a known capacitance in same ballpark.

Then calculate L using L= 1/(C*(2PI*F)^2) standard formula (re-arranged)

Have fun. You can always make sparks with it by switching 300mA from a fresh 9V battery across it then safely switching it off and get an arc around 1kV per mm and or get a real jolt if you are not careful .... one henry. It might generate 10kV or a 10mm arc.

SunnySkyguy - it's an old inductor. made from two ferrite materiel pieces, they are held together by a mechanical ring of sorts. How did you calc the wire dc resistance?

I actually did the SRF tank-thing using a cap, two transistors and a resistor. I got these figures:
freq. / mH
136.6 646.43
447.6 629.02
627.7 639.05
1456.9 765.48
1526.8 782.3
1978.1 942.3
2497.1 1187.8

As you can see they are more or less in the region but not exactly. From whatever reason the readings I got in low freq. using the AD5933 and the impedance meter (from kit) were identical and lower.

- - - Updated - - -

SunnySkyguy - it's an old inductor. made from two ferrite materiel pieces, they are held together by a mechanical ring of sorts. How did you calc the wire dc resistance?

I actually did the SRF tank-thing using a cap, two transistors and a resistor. I got these figures:
freq. / mH
136.6 646.43
447.6 629.02
627.7 639.05
1456.9 765.48
1526.8 782.3
1978.1 942.3
2497.1 1187.8

As you can see they are more or less in the region but not exactly. From whatever reason the readings I got in low freq. using the AD5933 and the impedance meter (from kit) were identical and lower.

If you had DC current net flow it may have affected the readings, or even different AC drive levels can saturate the core, or as FvM said activate the core and increase inductance with F. It must have a permeability sweet spot> 10Khz.

I just searched for "AWG resistance" and guessed the turns and chose the wire. Ballpark numbers only.

Why do you say they are not very different?

Click to expand...

The inductance of an inductor with ungapped core, either iron or high permeable ferrite will be flux and frequency dependent. Variations of at several 10 % are not surprizing.