开云体育

Hi All

SEESII H4 + 1.2.40 measuring a Ferrit inductance gives a little different values depend on stimulus span selected :

50Khz..30Mhz (calibrated with same load 401 point) ---> 0.100 Ohm + j 26.88 Ohm ==> Q = 268
13.5Mhz .. 14.5Mhz (calibrated with same load 401 point) ---> 0.200 Ohm + j 26.88 Ohm ==> Q= 134

NanoVNA-F gives same for both stimulus spans but ---> 0.400 Ohm + j 26.88 Ohm ==> Q= 67

can you explaine that , and which is the more accurate measurement ??

it's important to decide which inductance have the better Q coeff

Thanks
73's Nizar

Not much change in the resistance and why Q can be difficult to determine. The old Q meters (HP 4342, BRC 160, 260 etc. using a tuned-circuit technique) often did not agree very well with the modern instruments because of this. As instruments improved there was better and better agreement. I lived thru the transition and often worked with suppliers because of the disagreements. In some cases Q had to be ignored. We specified the Q with the ancient Q meters and the supplier used some other instrument and the Q reported was much lower and no matter what the tried they could not reach the specified Q.
Would be interesting to know the specifics of the inductor you are testing. What core material, frequencies, etc.. Inductors wound on ferrite often have Q's < 1 such as a common mode choke of or the isolation impedance from the input terminals to the output terminals.

QUOTE: 50Khz..30Mhz (calibrated with same load 401 point) ---> 0.100
Ohm + j 26.88 Ohm ==> Q = 268
13.5Mhz .. 14.5Mhz (calibrated with same load 401 point) ---> 0.200 Ohm +
j 26.88 Ohm ==> Q= 134

NanoVNA-F gives same for both stimulus spans but ---> 0.400 Ohm + j 26.88
Ohm ==> Q= 67

can you explaine that , and which is the more accurate measurement ??
****
****

This is an easy one .......

Q = Reactance / Resistance

Hopefully the resistance is measured at the frequency of interest!

So, for the first case: Q = 26.88 / 0.100 = 268 (unitless)
For the second case: Q = 26.88 / 0.200 = 134 (unitless)
For the final case: Q = 26.88 / 0.400 = 67.2 (unitless)

They are all accurate. What you are measuring is just the difference in
pure resistance as a function of frequency. As the pure resistance
increases (skin effect), the Q decreases. No surprise.

Dave - W?LEV



On Fri, Jan 31, 2025 at 7:23?PM Team-SIM SIM-Mode via groups.io <sim31_team=
[email protected]> wrote:

Hi All

SEESII H4 + 1.2.40 measuring a Ferrit inductance gives a little different
values depend on stimulus span selected :

50Khz..30Mhz (calibrated with same load 401 point) ---> 0.100 Ohm + j
26.88 Ohm ==> Q = 268
13.5Mhz .. 14.5Mhz (calibrated with same load 401 point) ---> 0.200 Ohm
+ j 26.88 Ohm ==> Q= 134

NanoVNA-F gives same for both stimulus spans but ---> 0.400 Ohm + j 26.88
Ohm ==> Q= 67

can you explaine that , and which is the more accurate measurement ??

it's important to decide which inductance have the better Q coeff

Thanks
73's Nizar

--

*Dave - W?LEV*


--
Dave - W?LEV

Hi WB2UAQ
The inductance is a 8 turns around T37-6 Core at 14.124 Khz.
The measurement focused on loss resistor to select the one having the lowest loss for filtre application, this require the comparaison to be done with the same stimulus span , same calibration and all things are the same , otherwise the comparaison does not have sens .
73s Nizar

On Fri, Jan 31, 2025 at 11:23 AM, Team-SIM SIM-Mode wrote:

can you explaine that , and which is the more accurate measurement ??

Nizar,

I suggest none of these measurements are very accurate. You are trying to measure a very low resistance (under 1 ohm) in series with an inductance. The magnitude of the reflection coefficient is very, very close to 1. You can see this as you are plotting on the perimeter of the Smith Chart circle. This results in a poor estimate of the R component of the complex impedance.

As others have pointed out the Q meter is a better instrument for the job. However there is one method that may work for you if you only have a NanoVNA.. That is to use the "S21 shunt method" with a variable capacitor in series with the inductor. The capacitor is tuned until the series reactance is close to 0 and then a reasonable estimate of R can be made.

In your case you will need a good test jig. From your screenshots your reactance is 26.8 ohms at 14.126 MHz. which is a small inductance of 300 nH. Stray inductance and capacitance need to be minimized.

On Fri, Jan 31, 2025 at 01:59 PM, Team-SIM SIM-Mode wrote:

The inductance is a 8 turns around T37-6 Core at 14.124 Khz.

It's odd that I happen to have the exact inductor in my junkbox. It's wound with #22 bare copper wire over 3/4 of the core, 0/5" leads. Q measures 195 at 14.1 MHz on my HP 4342A Q meter.

Brian

On Fri, Jan 31, 2025 at 01:59 PM, Team-SIM SIM-Mode wrote:

The inductance is a 8 turns around T37-6 Core at 14.124 Khz.

Nizar,

Attached is a simulation of a T37-6 Micrometals powdered iron core. Note how small a resistance is calculated at your operating frequency.

Roger

On Fri, Jan 31, 2025 at 02:38 PM, Roger Need wrote:

Attached is a simulation of a T37-6 Micrometals powdered iron core.

My results are somewhat different, especially Q. I don't know for sure that the core is type 6, but it's bright yellow and matches the current photos at Amidon where I got it years ago. I recall I was using type 6 at the time for a project. Inductance measured about 270 nH and Q 200 at 14.1 MHz after I spread the turns a bit. 0.5" leads.

Brian

On Fri, Jan 31, 2025 at 02:51 PM, Brian Beezley wrote:

My results are somewhat different, especially Q. I don't know for sure that
the core is type 6, but it's bright yellow and matches the current photos at
Amidon where I got it years ago. I recall I was using type 6 at the time for a
project. Inductance measured about 270 nH and Q 200 at 14.1 MHz after I spread
the turns a bit. 0.5" leads.

There can be some variation in production batches. Amidon does not make their own cores - they buy them from suppliers like Micrometals and Fair_rite.

The wire diameter and how you wind can affect inductance and resistance at a particular frequency. Subsequently Q is affected. Your inductance is pretty close to the calculated value for 8 turns. Here is a graph from Micrometals showing Q with 12 turns of #20 on a T37-6. Using smaller wire would lower Q.. Results are close to what you measured.

Roger, I think there is something wrong with your Q calculation. Using your X and R, Q = 17.1 / 0.0935 = 183, close to what I measure, although I measure a higher inductance.

Brian

On Fri, Jan 31, 2025 at 03:12 PM, Brian Beezley wrote:

Roger, I think there is something wrong with your Q calculation. Using your X
and R, Q = 17.1 / 0.0935 = 183, close to what I measure, although I measure a
higher inductance.

Brian,

You are indeed correct. The Q calculation is wrong. I ran the simulation at


I will let the author know...

Roger

Roger

On Fri, Jan 31, 2025 at 03:33 PM, Roger Need wrote:

The Q calculation is wrong.

OK, glad that's cleared up. I thought I must have done something really wrong to get 4 times the value you reported.

I hope Nizar got his answer. I think the best one was provided by Jim Lux. It didn't make it into this thread, but Nizar can find it. If I had only a NanoVNA to measure Q, I'd calibrate it over as narrow a frequency range as possible to minimize interpolation error. Then I'd take extra care to make low-resistance connections, which can really matter when Q is high.

Back into the woodwork...

Brian

Hi

Thanks Roger and all contributers for their interesting comments ,
i am agree of almost all , it is true that making measurements of the low resistances of the inductances with the 50 Ohm divider bridge is very tedious, it is certainly not the best arrangement for this kind of measurement even with S21 methode, but remains a question which may have to do with the details of the calculation model by the firmware : why is there so much change in result just by changing the stimulus Span as long as we have done the calibration with the same 50 Ohm load and same NanoVNA , there is only the stimulus span which has changed in the meantime, this phenomenon has not been observed with NanoDeeplec -F.

Brian , I dont think that it's an interpolation error, indeed both calibration's are done alone with each span with 401 sweep point's for each , so there is no need to do interpolation .

Excuse me , the inductance is a 8 turn within T30-6 ( not T37-6) .

73's Nizar.

Hi All

excuse me , i just doing the calibration again at the same time of the two different stimulus span , and remeasure the same inductor , and i have the same value of resistor loss , so the issue is resolved and explained , just a mismuch of calibration not done at the same time .

but as conclusion : it's rather to do comparaison with the same NanoVNA , same calibration , same stimulus span , nothing to change during comparaison .
Thanks to all .
73's Nizar

Hi All,

For those who have the excellent book 'Experimental Methods in RF Design'
(Wes Hayward W7ZO! et al.) look at
the following section.

7.9 Q measurement of LC resonators.

The test jig described would be ideal for S21 measurements with a nanoVNA
or one could use a signal generator
and tinySA.

I still use my ancient Advance T2 Q meter (A UK instrument using three
valves (tubes)) and can measure up to a
Q of 400 at any frequency from 100 kHz to 100 MHz. The meter is rather
coarse but who needs Q to a high
precision?

73
Phil G3SES

On Sat, 1 Feb 2025 at 12:10, Team-SIM SIM-Mode via groups.io <sim31_team=
[email protected]> wrote:

Hi All

excuse me , i just doing the calibration again at the same time of the two
different stimulus span , and remeasure the same inductor , and i have the
same value of resistor loss , so the issue is resolved and explained ,
just a mismuch of calibration not done at the same time .

but as conclusion : it's rather to do comparaison with the same NanoVNA ,
same calibration , same stimulus span , nothing to change during
comparaison .
Thanks to all .
73's Nizar

I recommend to you to use the resonance method, it’s more accurate. You are using a very simple VNA and there are errors of measurement bigger than the magnitude or real part of Z. I posted a method based on VNA that would be more accurate and repeatable. If you would like to measure the Q on several frequencies you can use a variable capacitor..
Figure that NanoVNA can measure resistor in a range of 50Ω there are no specifications for this instrument but for example for a R&S VNA model ZND ( I attached the accuracy plot) the S11 magnitude uncertainty is for Abs(S11) close to 1 is 0.02. Propagating errors it results for your specific case where the imaginary part of Z is 26,88 Ω in a real part error of 0,64 Ω * , and this is a mid range VNA!!
There are also a parasite capacitance on inductor that can affect the measurements.

You can find my post on : Step by step on measuring inductor and self resonance


* I used the appropriate formula :

?


Where X is the measured reactance , data gamma is the measurement uncertainty Zo=50Ω

Regards, Patricio.
?

Thanks Brian for steering me to this discussion. It looked like an opportunity for me to compare my measurements using this much discussed T37-6 with 8 turns. I have a bag of T37-6 that had come from Amidon (and almost for sure are Micro-Metals manufacture). So, I wound and measured one with #22 wire and one with #26 wire, all at 14.1 MHz.

#22 Wire, 260A Q-Meter L=250 nH, Q=164
#22 Wire, Insertion loss method (-46.0 dB) Q=176

#26 Wire, 260A Q-Meter L=248 nH, Q=138
#22 Wire, Insertion loss method (-44.25 dB) Q=140

A couple of notes. My 260A is really old, but through some miracle has all original tubes, including the very special detector tube. It also has the original thermocouple. It behaves just like it did 40 years ago. I have a box of "standards" that were measured in the 1960's on an almost new 260A (a different one). The checks with those are still close. Also, the 260A instruction manual notes a need for a correction due to the coupling resistor of 20 milliOhm, that appears in series with the inductor. This needs to be subtracted for low series resistance inductors. I did that for the numbers shown.

The attached shows how they look physically.

My Q numbers are enough lower than Brian's to be interesting to track down. Brian, can I mail you the coils?

Bob W7PUA

On Sun, Feb 2, 2025 at 02:27 PM, Bob Larkin wrote:

So, I wound and measured one with #22 wire and one with #26 wire, all at 14.1
MHz.

#22 Wire, 260A Q-Meter L=250 nH, Q=164
#22 Wire, Insertion loss method (-46.0 dB) Q=176

Bob,

Was the insertion loss method the one from your book Fig. 7.65? If so what value of capacitors did you use for Cin and Cout and the resonance cap?

Roger

A little on the pricey side, @ $115USD on amuzoon or higher on ebayism.

Mike C. Sand Mtn GA

Hi Roger - I use the method of figure 7.66 "Series TC." This is simple with a nanoVNA, or any VNA. For this particular coil, I found an old CD 470 pF silver mica cap and used that along with a 30 pF NP0 in parallel. I have kept a piece of PCB with coax connectors and an exposed microstrip line of an inch or so to solder coils and caps to. I use Eq 7.4 noting that A is the attenuation and thus a positive number. Bob