开云体育

Concerning how to measure capacitance and inductance on the RF Demo board (or any capacitor or inductor), reference /g/nanovna-users/wiki/16592. I can get the curves shown in this demo just fine.. but how to read the actual values off of those curves eludes me. On the item 8 inductor example, if I vary the frequency, I can make that component read about any value I want. I must be missing something. Help, please.
--
Ed K9EK

The inductance will change with frequency.? I use the Smith chart and the resistance and inductance are displayed at the marker frequency.
Bob

From: "Bob Albert via groups.io" <bob91343@...>
Date: February 12, 2021 at 6:36:54 PM PST
To: [email protected]
Subject: Re: [nanovna-users] measuring Capacitance or Inductance
Reply-To: [email protected]

? The inductance will change with frequency. I use the Smith chart and the resistance and inductance are displayed at the marker frequency.
Bob

The inductance won’t change with frequency (to a first order approximation and especially over a somewhat narrow range).

What will change is the reactance with respect to frequency.

XL= 2*pi*F*L
XC= 1/(2*pi*F*C)

The L and C values shown by the marker is calculated by solving for those values at a specific frequency (where the marker is set to)

Ray WB6TPU

Thank you for the responses. I think I was unclear. What I am trying to understand is how one relates the marked value of a component to what nanovna shows. Using DL5FA’s item 8 inductor, my devices show 700 nH @50khz. But at 100mhz, the reading is 1uH. But the component is speced as 470nH, a value I can’t see anywhere, regardless fir frequency. How should I be reading this thing?
Thanks!
--
Ed K9EK

On 2/13/21 6:39 AM, Ed Krome wrote:

Thank you for the responses. I think I was unclear. What I am trying to understand is how one relates the marked value of a component to what nanovna shows. Using DL5FA’s item 8 inductor, my devices show 700 nH @50khz. But at 100mhz, the reading is 1uH. But the component is speced as 470nH, a value I can’t see anywhere, regardless fir frequency. How should I be reading this thing?
Thanks!

There is always some parasitic C around, which usually "reduces" (or cancels) some of the inductance - sort of the opposite of what you're seeing - but in any case, the amount of cancellation you get is frequency dependent (since the VNA is measuring X, and then converting that to nH)

Are you allowing for the inductance of the leads going to the component.? A handy thing is "about 1 nH/mm or 1 uH/meter" for a single wire.

HI,
I use a different approach when I need to know the exact value of inductance or capacitance, I build up a simple jig on a PCB mount connector with a series 50 ohm chip resistor then put the component in series to ground.? With the standard Xc or Xl formula determine the frequency where the value is between 25 to 100 ohms.? I use 1 or 10 MHz and using the formula in reverse extract the true value of the component.? Works very well.? It tells you the net value with all parasitic errors normalize in. So that the VNA reads Z50+/-j.
Mel, K6KBE

Hello Ed,

Curious, if you go to item13 on your card, take that reading at 50 kHz and SUBTRACT the L value from the reading at 50 kHz done at item8... Do you get 470 nH? A reading of 700 nH at 50 kHz points to a serious cal issue or a fixture parasitic that is begging to be removed as well understood.

I suspect that your L is not that far off or out of tolerance by nearly 100 %.

Alan

Ed,

What I am trying to
understand is how one relates the marked value of a component to what nanovna
shows.

The value marked on a component is supposed to be valid at a very low frequency. As soon as the frequency goes up, parasitic effects become increasingly important, so the actual value of the part varies with frequency. Also, of course, there is always a tolerance. With good capacitors and air-core inductors this tolerance might be only 5%, but with ferrite-cored inductors it can easily be 40%, and with some sorts of ceramic capacitors it can be even larger.

So, rule #1 is to measure on the lowest frequency possible, and rule #2 is to never forget that the marked value is subject to a tolerance.

But then the characteristics of the measuring instrument come into play. The NanoVNA is natively a 50? instrument. It should produce the best accuracy when measuring impedances reasonably close to 50?. When the impedance gets close to zero, or into the kiloohm range, the accuracy of the NanoVNA drops. So, if you are measuring small values of capacitance or inductance, measuring at the lowest frequency the NanovNA supports might produce poor measurement accuracy.

So, rule #3 is to measure at a frequency where the impedance of your part is at least close to the order of magnitude of 50?.

In practice that means that you should look at what frequency the part has a reactance of 50?, and then measure at a frequency a few times lower. If the measured value is reasonably constant over the range between those two frequencies, then probably you have a valid measurement. If instead it varies all over the place, it probably means that there are too high parasitics even in that frequency range.

Of course I'm assuming that you properly calibrated the NanoVNA, putting the short and the load exactly at the same place where you then put the part to be measured.

In my experience it's best to use the shortest possible connection between the NanoVNA and the part under test. It seems that correct measurement of difficult impedances through a long piece of coax cable is harder for the NanoVNA, even when carefully calibrated through that long cable.

Always keep in mind that if you do all this and get a consistent result, then you are getting the low frequency value of the part you are testing, and that at higher frequency its actual value will change. At a high enough frequency a capacitor becomes a short circuit, further up it becomes an inductor. And what's an inductor at low frequency will become an open circuit at some high frequency, and a capacitor beyond that. Generally both inductors and capacitors will rise in value, when you start going up in frequency starting from a low frequency. But core materials tend to decrease their permeability beyond some frequency, and this effect can win over the other in some cases, so you might see cored inductors whose value goes down rather than up, when increasing the frequency. These are all real effects, not measurement errors! In RF work you often need to measure each part at the frequency you will be using it, rather than trusting the value printed on it, which is valid only at low frequencies.

Adding to my reply:

The reactance of 470nH at 50kHz is barely 0.15?. You cannot expect the NanoVNA to measure that accurately. You need at least 5MHz or so to bring the reactance up into a range where the NanoVNA is pretty accurate. A small SMD inductor like that one shouldn't have much parasitic capacitance effects yet at that frequency. 5MHz is "low frequency" in this case.

SMD inductors typically have pretty low Q. You need to make sure that you are not mixing up resistance with reactance. Use the RLC function.

Also SMD inductors typically have ±20% tolerance, so that can explain at least a part of the discrepancy you see.

CAPACITANCE AND INDUCTANCE ARE SPECIFIED AT A GIVEN FREQUENCY. See the
example.

[image: image.png]
*Clyde K. Spencer*



On Sun, Feb 14, 2021 at 9:51 AM Manfred Mornhinweg <manfred@...>
wrote:

Adding to my reply:

The reactance of 470nH at 50kHz is barely 0.15?. You cannot expect the
NanoVNA to measure that accurately. You need at least 5MHz or so to bring
the reactance up into a range where the NanoVNA is pretty accurate. A small
SMD inductor like that one shouldn't have much parasitic capacitance
effects yet at that frequency. 5MHz is "low frequency" in this case.

SMD inductors typically have pretty low Q. You need to make sure that you
are not mixing up resistance with reactance. Use the RLC function.

Also SMD inductors typically have ±20% tolerance, so that can explain at
least a part of the discrepancy you see.

Clyde Spencer wrote:

CAPACITANCE AND INDUCTANCE ARE SPECIFIED AT A GIVEN FREQUENCY. See the
example.

The chart says the frequency it is measured at.
Not how much inductance varies with measuring frequency.
I have carefully measured normal HF range inductors with a dip oscillator and SDR at working frequencies comparing with a meter that measures in the 100Khz range.
I noticed no significant difference.

Alan

EXAMPLE from DAYS PAST: In the TTL days of logic long past to newbies, it
was common practice to place a black, CK05 (0.1 ?F) capacitor at each end
of a row of logic chips. This was placed between Vcc and return. In those
days, we seldom considered self resonance of a passive device. The
intended use of them on the boards was to keep logic switching noise off
the DC rail. Turns out those CK05 capacitors which peppered our boards
became self resonant somewhere between 1 and 2 MHz, usually around 1.4 to
1.6 MHz. Therefore, above self resonance, they became DC-blocked
inductors. Now, ask yourself: "is an inductor good at bypassing rail noise
as the capacitor was intended to accomplish?" NO!

Several decades ago, I had the privilege of issuing a couple of new-hires
from Kent State into the real world of 'parasitic component' behavior (the
real world they would have to live in). I was given the task as they would
specify totally unrealistic component values for designs and had only an
understanding of the ideal behavior of electronic components. I asked
them to check out a small handful of those CK05 capacitors from engineering
stock. We warmed up the HP impedance meter of the time (the one that had a
tunable drum as a frequency indicator and topped out at 110 MHz). Sure
enough, *EVERY* CK05 capacitor went purely resistive between 1.4 and 1.6
MHz and inductive above that. Their eyes bugged out. They could not
understand or comprehend how a capacitor could possibly become resonant
(+jX = -jX) and ultimately become an inductor. I sent them back to their
test books and pointed them in the direction of our local building
library. It took them a week of digging, but they finally came back with
the classic capacitance to resistance to inductance curve with frequency.

Dave - W?LEV

On Sun, Feb 14, 2021 at 2:26 PM Manfred Mornhinweg <manfred@...>
wrote:

Ed,

What I am trying to
understand is how one relates the marked value of a component to what

nanovna

shows.

The value marked on a component is supposed to be valid at a very low
frequency. As soon as the frequency goes up, parasitic effects become
increasingly important, so the actual value of the part varies with
frequency. Also, of course, there is always a tolerance. With good
capacitors and air-core inductors this tolerance might be only 5%, but with
ferrite-cored inductors it can easily be 40%, and with some sorts of
ceramic capacitors it can be even larger.

So, rule #1 is to measure on the lowest frequency possible, and rule #2 is
to never forget that the marked value is subject to a tolerance.

But then the characteristics of the measuring instrument come into play.
The NanoVNA is natively a 50? instrument. It should produce the best
accuracy when measuring impedances reasonably close to 50?. When the
impedance gets close to zero, or into the kiloohm range, the accuracy of
the NanoVNA drops. So, if you are measuring small values of capacitance or
inductance, measuring at the lowest frequency the NanovNA supports might
produce poor measurement accuracy.

So, rule #3 is to measure at a frequency where the impedance of your part
is at least close to the order of magnitude of 50?.

In practice that means that you should look at what frequency the part has
a reactance of 50?, and then measure at a frequency a few times lower. If
the measured value is reasonably constant over the range between those two
frequencies, then probably you have a valid measurement. If instead it
varies all over the place, it probably means that there are too high
parasitics even in that frequency range.

Of course I'm assuming that you properly calibrated the NanoVNA, putting
the short and the load exactly at the same place where you then put the
part to be measured.

In my experience it's best to use the shortest possible connection between
the NanoVNA and the part under test. It seems that correct measurement of
difficult impedances through a long piece of coax cable is harder for the
NanoVNA, even when carefully calibrated through that long cable.

Always keep in mind that if you do all this and get a consistent result,
then you are getting the low frequency value of the part you are testing,
and that at higher frequency its actual value will change. At a high enough
frequency a capacitor becomes a short circuit, further up it becomes an
inductor. And what's an inductor at low frequency will become an open
circuit at some high frequency, and a capacitor beyond that. Generally both
inductors and capacitors will rise in value, when you start going up in
frequency starting from a low frequency. But core materials tend to
decrease their permeability beyond some frequency, and this effect can win
over the other in some cases, so you might see cored inductors whose value
goes down rather than up, when increasing the frequency. These are all real
effects, not measurement errors! In RF work you often need to measure each
part at the frequency you will be using it, rather than trusting the value
printed on it, which is valid only at low frequencies.

--
*Dave - W?LEV*
*Just Let Darwin Work*

On Fri, Feb 12, 2021 at 06:33 PM, Ed Krome wrote:

Concerning how to measure capacitance and inductance on the RF Demo board (or
any capacitor or inductor), reference
/g/nanovna-users/wiki/16592. I can get the curves shown in
this demo just fine.. but how to read the actual values off of those curves
eludes me. On the item 8 inductor example, if I vary the frequency, I can make
that component read about any value I want. I must be missing something. Help,
please.

Ed,

There have been a number of good posts by others describing the characteristics of inductors and capacitors versus frequency. Parasitic capacitance is a big factor and needs consideration.

This post will hopefully be a direct answer to your question of how to read the inductance value on the NanoVNA.

The first step is to make sure that you have calibrated correctly using the open, short and load that are on the test board. This is cumbersome because those little u.fl connectors are not easy to work with. Once you are sure that your calibration is OK you can measure the component. The Smith chart has a readout that will give you the resistance and inductance or capacitance for the marker frequency. As you move the marker you will see the resistance increase on the inductor due to the "skin effect" . The inductor value shown at the marker is calculated by dividing the reactance by 2*pi*frequency. You have to be careful interpreting the results because parasitic capacitance will affect the reactance measured and the estimate of L by this method will get worse as the frequency increases.

I made some tests today on a small air-core inductor that measured 243 nH on a DE-5000 inductance meter at 100 kHz. I then connected it to a calibrated test jig and made some measurements. I attached a series of annotated screenshots showing the estimate of L by the NanoVNA at various frequencies up to 150 MHz. .

I then used the NanoVNA app by OneOfEleven to plot L, R and X versus frequency. You will note that at 50 kHz. the estimate of L is poor compared to my DE-5000 inductance meter. This is because the value of X is very small at this frequency and not in a reasonable measurement range for the NanoVNA. As the frequency increases to about 1 MHz. the accuracy of the estimate improves because the reactance is now a little over 1 ohm. Note the slight rise in calculated L with frequency due to the parasitic capacitance of the inductor.

Roger

Thank you to all who replied. But, since the values I was seeing didn't match prescribed values (and it was driving me a bit nuts), I tried the experimental approach. I took a single RF Demo Kit board with the same cable. I carefully calibrated my 3 nanoVNA's (nanoVNA 2.8", nanoVNA H4, nanoVNA SAA-2N), each 50kHz to 300MHz. Then I recorded the values of the components at position 7 (capacitor) and posn. 8 (inductor) at increments over the frequency range. Results showed remarkably good correlation over reduced frequency range on all three nano's. Pos. 7 showed the capacitor to be 100pF from 0.1 to 100 MHz. Posn 8 showed the inductor to be 700nH from 0.1 to 30 MHz. Things went rapidly askew at higher frequencies, although trends were similar. The capacitor lines followed each other reasonably well all the way to 300 MHz on all 3 nano's, but the inductor curves, while trending similarly, showed marked differences in values between the 3 different nano's. As a reality check, I measured the inductor on an AADE L/C meter IIB and read 700nH. This wasn't as precise a setup as with the nano's. Curves attached
I think I'm getting a handle on this now. Thanks again.

Ed K9EK

Clyde,

CAPACITANCE AND INDUCTANCE ARE SPECIFIED AT A GIVEN FREQUENCY. See the
example.

Yes, often the datasheet states the measurement frequency. But this is invariably a "low" frequency, low enough to keep parasitic effects negligible. The inductance or capacitance of a component won't vary significantly from that frequency down to DC.

The part value starts changing significantly at frequencies high enough to make parasitic effects significant.


Alan,

I have carefully measured normal HF range inductors with a dip oscillator and SDR at working frequencies comparing with a meter that measures
in the 100Khz range. I noticed no significant difference.

When the inductors are built in such a way that in the HF range the parasitics are low, then this is indeed the normal situation. But if you measure them at VHF, the values will very likely change. And at UHF they surely will. It's a simple matter of going high enough in frequency, to make the values of any part change dramatically.

At HF you can often get away with taking the low-frequency values and assuming they will hold true at your working frequency, But not always. And at VHF that becomes rarer, and at UHF it becomes very rare.


Dave:

Since I enjoy challenging you, I will do it again! ;-)

Now, ask yourself: "is an inductor good at bypassing rail noise
as the capacitor was intended to accomplish?" NO!

Well, it depends! A DC-blocked inductor can be a pretty good bypass element, no worse than a capacitor! It just depends on its impedance at the frequency in question. Such a "nasty" 100nF bypass capacitor is resonant at 1.5MHz when its equivalent series inductance is 112nH. A foil-wound capacitor might indeed be that bad. A ceramic capacitor only if mounted with very long leads. Anyway, assuming it has indeed 112nH and thus is resonant at 1.5MHz, how would it behave at 3MHz? Well, it would have a reactance of less than 2?! That's still a pretty good bypass, despite being inductive. At 100MHz it would be bad.

If you replace that nasty 100nF capacitor by a 10nF one, would it be better? NO, if you keep those long leads! It will be much worse bypassing low frequencies, it will be good at its resonant frequency near 5MHz, but at 100MHz it will be almost as bad as the 100nF one.

And what happens if you follow that old rule of putting the 100nF capacitor in parallel with a 1nF one? Well, at some frequency you get a might parallel resonant circuit, with the 100nF capacitor acting as the inductor, and at that frequency you get infinite impedance, and thus NO bypassing! Of course, only if the capacitors have high Q at that frequency. So the important point with bypass capacitors is: They should have enough capacitance for the low frequencies, low enough ESL for the high frequencies, and they should be bad! I mean, they should have a low Q. A high loss factor. That largely pevents getting unbypassed frequencies due to bypass caps happily parallel-resonating with each other.

There is a long-standing myth about electrolytic caps needing a parallel-connected ceramic cap to provide bypassing over a wide frequency range. Using a parallel ceramic cap is indeed useful if this is a chip capacitor. But placing something like a an old-fashioned disc ceramic cap in parallel with an electrolytic of comparable path length doesn't help much, since both have roughly the same ESL.

In some equipment I often see real collections of 6 or more different capacitors in parallel, placed there by some designer who thinks that each frequency will then take the path it likes best. The only problem is that physics don't work like that. Those nice showcases of six different capacitors in parallel are mainly good for one thing: Getting a good laugh!

I also often laugh about that old rule of "one bypass cap per IC". When using slow ICs, often a single bypass cap is enough for the entire board, and in other cases one cap every so much distance is enough. One can save quite a bit of money in series production by leaving out unnecessary parts. Of course without overdoing that...

Dave,

We warmed up the HP impedance meter of the time (the one that had a
tunable drum as a frequency indicator and topped out at 110 MHz). Sure
enough, *EVERY* CK05 capacitor went purely resistive between 1.4 and 1.6
MHz and inductive above that.

What lead length did you use for that test?????

After my last post I grabbed my box of 100nF capacitors, fired up the NanoVNA (works well even without warming up!), and measured two dozen of them, with lead lengths typical for PCB mounting. Their resonant frequencies all fell in the range of 6 to 8.4MHz. As was to be expected, the smallest ones (leaded ceramic chips) had the highest resonance, and the largest foil capacitors had the lowest, inside that range.

I then measured with full length wires. My longest-legged one resonated at 1.85MHz. That one has 35mm long legs (each), of strongly magnetic material, which probably contributes to add lead inductance.

To get those low resonant frequencies, you must have had very long-legged capacitors, like 5cm, and you must have measured them with full lead lengths. Of course nobody would mount a bypass cap with full-length leads! So what you were getting on those boards must have been much better. Resonating around 7MHz, and producing acceptable bypassing to 30MHz or so, depending on the impedance requirements. In a great many situations that's good enough, specially with older electronics.

And since it's sunday and I have time for playing with the NanoVNA, I made measurements to show all of you the effect of placing bypass caps of different value in parallel. Attached are the impedance plots for 100nF alone, 1nF alone, and both in parallel. Whoa! Which option would you prefer?

Note that the 100nF cap alone provides a bypassing impedance below 3? from about 700kHz to about 70MHz. that's pretty good, I would say. The 1nF is bad on low frequencies, and the parallel combination is a total disaster! Yet that's what you will find in equipment designed by people who have never thought about this point, and are just following intuition, which often is wrong...

Here is the phenomenon I referred to in a previous post in this thread.
[image: image.png]

This paper is written by Cadence and can be accessed at:




Take only the 470 nF unit - the left in each plot. Below 15 MHz, the
device is capacitive, although decreasing in value as frequency is
increased. At resonance, it is purely resistive. Above resonance it
becomes inductive. The phase would also have to be shown to confirm that
last statement, but this is typical performance of a capacitor as it goes
through self resonance. Inductors act in a similar manner.

Dave - W?LEV

--
*Dave - W?LEV*
*Just Let Darwin Work*

And what's the best bypass cap? Well, a single, plain, cheap aluminium electrolytic!

Attached is the impedance plot for a 47?F, 25V electrolytic cap, measured with lead lengths compatible with mounting it snugly on a PCB. Their narrow pin spacing helps a lot in keeping their ESL low. I kept the same scale to make comparison easy.

YES, a single 47?F electrolytic is a much better bypass cap than a parallel combination of two ceramic caps of different values! Even in the low VHF range!

The problems with electrolytic caps is that their ESR rises with age, and rises much faster if they run hot, or if they have to carry large ripple current. So they can't be applied in every situation. But in situations that are kind to them, they are the cheapest and easiest way to get an excellent wideband bypass.

Here is the phenomenon I referred to in a previous post in this thread.
[image: image.png]

This paper is written by Cadence and can be accessed at:

Good to see that they show the parallel resonance between two different bypass caps! But too many circuit designers aren't aware of this.