|
To the mathematically inclined from the mathematically challenged...
I understand the obsession with eliminating measurement errors. But how much error are we really talking about here? Suppose I take a short piece (say 6 inches) of 50 ohm coax cable, crimp a male sma connector on one end, and a couple of alligator clips on the other end. I "calibrate" with the alligator clips dangling (open), then with a short piece of wire (short), and, finally, with a non-wirewound, 50 ohm 1% resistor (load). Then I go to measure inductance of the coil I hand wound for my crystal radio. Are we talking 1% here? 10%? 100%? 1000%?
I understand that the error in such as setup is going to increase with frequency. But, say, between 3 MHz and 30 MHz...? Below that? Above that?
I also understand that there will be many factors affecting this setup: resistance between the alligator clips and whatever you have clipped into them; random capacitance and inductance from nearby stuff; temperature, humidity, spilling your coffee on it... If the effect of all possible factors affecting measurement using such a setup are just too great to make any kind of ball-park assessment of the range of error, you could just say that.
Thanks, ahead of time, for insight into this question.
|
I'm sometimes (often) also mathematically challenged so I perform a sanity check when doing these type of measurements. Instead of the coil connect a 25 ohm 1% resistor (or some other value different from 50 ohm but not too much) and see if this is indeed being measured as pure 25ohm. There should be no capacitance or no induction. Any deviation will probably also happen when measuring your coil. This will tell you at least if your calibration using your homemade fixture makes sense. -- NanoVNA Wiki: /g/nanovna-users/wiki/homeNanoVNA Files: /g/nanovna-users/filesErik, PD0EK
|
The errors scale proporionally with wavelength. I use my VNA mainly around 900 MHz in RFID development and applications. Here a single mm path length is on the threshold of an acceptable error for me. So scaling up in wavelength 30 times, we arrive at 30 MHz and 30 mm will be the length of wires that may start to be a source of errors.
Wavelength at 30 MHz is 10 meters, 30mm is just 0.3% of 10 meters. So you're fine if you keep the wires a bit short, or lower the frequency to 3 MHz. Then, it hardly matters anymore. But you can easily measure yourself the return loss of a 50 Ohm resistor connected with alligator clips, and measure short and open. This is the easiest if you calibrate the VNA at the end of the (SMA) cable. Solder the alligator clips with short wires to a female SMA connector.
Op 6-1-2020 om 17:29 schreef entilleser via Groups.Io:
toggle quoted message
Show quoted text
To the mathematically inclined from the mathematically challenged...
I understand the obsession with eliminating measurement errors. But how much error are we really talking about here? Suppose I take a short piece (say 6 inches) of 50 ohm coax cable, crimp a male sma connector on one end, and a couple of alligator clips on the other end. I "calibrate" with the alligator clips dangling (open), then with a short piece of wire (short), and, finally, with a non-wirewound, 50 ohm 1% resistor (load). Then I go to measure inductance of the coil I hand wound for my crystal radio. Are we talking 1% here? 10%? 100%? 1000%?
I understand that the error in such as setup is going to increase with frequency. But, say, between 3 MHz and 30 MHz...? Below that? Above that?
I also understand that there will be many factors affecting this setup: resistance between the alligator clips and whatever you have clipped into them; random capacitance and inductance from nearby stuff; temperature, humidity, spilling your coffee on it... If the effect of all possible factors affecting measurement using such a setup are just too great to make any kind of ball-park assessment of the range of error, you could just say that.
Thanks, ahead of time, for insight into this question.
|
I have also been looking to answer how accurate should the NanoVNA be at HF. Most components I want to measure are wire ended so my test jig is a 60cm RG316 with SMA on each end cut in half. Each cut end has 1cm of inner exposed and an alligator clip put on. This is calibrated with open, short and a 50 ohm made from 2 100ohm resistors with short wires soldered on. A 100pF silver mica then gives the attached which indicates 105pF at 7MHz, 109pF at 14MHz and 126pF at 28MHz. I remove the alligator clips and repeat soldering the connections (inconvenient but worth trying). This gives the same results. Should the NanoVNA do better than this? Are there firmware versions which will produce more accurate results? Is it my VNA hardware(made in China somewhere) or test jig or test method?
I have found the VNA gives good results using S21 on filters but using a single cable and S11, capacitors and inductors are difficult to measure. I expect ferrite cored inductors to show a frequency variation but a 100pF capacitor should still be 100pF at 30MHz. I've repeated calibration, tried different leads but I don't have any confidence in the accuracy of reactance measurement. From what I have read, I could manually add "fudge factors" to the calibration and remove what looks like a systematic error but that is not so easy for the mathematically challenged (like me).
My conclusion is that the accuracy depends on the accuracy of the standards you have and the amount of effort you put into "improving" the calibration. Some test jigs might be easier to get a good calibration but up to 30MHz the test jig is not the most important factor.
I have done what Erik suggests and measured a 100 ohm resistor. It shows no significant reactance at 30MHz. I don't have a record of it but certainly not 26pF.
73 Brian.
|
For flying wires and alligator clips to work, everything must be in the *exact
same position* as when the calibration was done. I use a couple of BNC
female connectors mounted on a piece of double sided FT-4 board. I've then
soldered alligator clips to the BNC teat which I use to mount the component
under test. This keeps everything in the same position as the cal.
Dave - W?LEV
toggle quoted message
Show quoted text
On Tue, Jan 7, 2020 at 11:43 PM Brian <vk4bap@...> wrote: I have also been looking to answer how accurate should the NanoVNA be at
HF. Most components I want to measure are wire ended so my test jig is a
60cm RG316 with SMA on each end cut in half. Each cut end has 1cm of inner
exposed and an alligator clip put on. This is calibrated with open, short
and a 50 ohm made from 2 100ohm resistors with short wires soldered on. A
100pF silver mica then gives the attached which indicates 105pF at 7MHz,
109pF at 14MHz and 126pF at 28MHz. I remove the alligator clips and repeat
soldering the connections (inconvenient but worth trying). This gives the
same results. Should the NanoVNA do better than this? Are there firmware
versions which will produce more accurate results? Is it my VNA
hardware(made in China somewhere) or test jig or test method?
I have found the VNA gives good results using S21 on filters but using a
single cable and S11, capacitors and inductors are difficult to measure. I
expect ferrite cored inductors to show a frequency variation but a 100pF
capacitor should still be 100pF at 30MHz. I've repeated calibration, tried
different leads but I don't have any confidence in the accuracy of
reactance measurement. From what I have read, I could manually add "fudge
factors" to the calibration and remove what looks like a systematic error
but that is not so easy for the mathematically challenged (like me).
My conclusion is that the accuracy depends on the accuracy of the
standards you have and the amount of effort you put into "improving" the
calibration. Some test jigs might be easier to get a good calibration but
up to 30MHz the test jig is not the most important factor.
I have done what Erik suggests and measured a 100 ohm resistor. It shows
no significant reactance at 30MHz. I don't have a record of it but
certainly not 26pF.
73 Brian.
--
*Dave - W?LEV*
*Just Let Darwin Work*
*Just Think*
|
Brian,
How much different are your results if you clip to the capacitor 0.25" away from the cap body vs 0.5" away from the cap body? The inductance of the leads in series with the capacitor will form a resonator. The measured capacitance will increase as the resonance is approached from below.
--John Gord
toggle quoted message
Show quoted text
On Tue, Jan 7, 2020 at 03:43 PM, Brian wrote:
I have also been looking to answer how accurate should the NanoVNA be at HF.
Most components I want to measure are wire ended so my test jig is a 60cm
RG316 with SMA on each end cut in half. Each cut end has 1cm of inner exposed
and an alligator clip put on. This is calibrated with open, short and a 50 ohm
made from 2 100ohm resistors with short wires soldered on. A 100pF silver mica
then gives the attached which indicates 105pF at 7MHz, 109pF at 14MHz and
126pF at 28MHz. I remove the alligator clips and repeat soldering the
connections (inconvenient but worth trying). This gives the same results.
Should the NanoVNA do better than this? Are there firmware versions which will
produce more accurate results? Is it my VNA hardware(made in China somewhere)
or test jig or test method?
I have found the VNA gives good results using S21 on filters but using a
single cable and S11, capacitors and inductors are difficult to measure. I
expect ferrite cored inductors to show a frequency variation but a 100pF
capacitor should still be 100pF at 30MHz. I've repeated calibration, tried
different leads but I don't have any confidence in the accuracy of reactance
measurement. From what I have read, I could manually add "fudge factors" to
the calibration and remove what looks like a systematic error but that is not
so easy for the mathematically challenged (like me).
My conclusion is that the accuracy depends on the accuracy of the standards
you have and the amount of effort you put into "improving" the calibration.
Some test jigs might be easier to get a good calibration but up to 30MHz the
test jig is not the most important factor.
I have done what Erik suggests and measured a 100 ohm resistor. It shows no
significant reactance at 30MHz. I don't have a record of it but certainly not
26pF.
73 Brian.
|
Hi Brian,
Unfortunate but a 100 pF will NOT remain a 100 pF C independent of frequency. And in fact the C value will increase with an increase in frequency. In your case with the values you provided, the result you present is due to the presence of series inductance of ~ 67 nH. Part of this are leads on the mica C as well as L from the fixture. In any case ALL passive RLC components have a parasitic component and at the vary least, a C as well as L will show an increase in value with increased frequency as you have shown in your chart. Eventually the C will obtain series resonance and an L will obtain anti resonance or parallel resonate frequency.
Alan
|
On Tue, Jan 7, 2020 at 03:43 PM, Brian wrote:
" ..A 100pF silver mica then gives the attached which indicates 105pF at 7MHz, 109pF at 14MHz and 126pF at 28MHz .."
=======================================================
The above frequency dependent impedance effects are why manufacturers specify their component values at specific test frequencies. A dedicated LCR instrument, i.e. the well regarded DE-5000, has five test frequencies 100Hz/120Hz/1kHz/10kHz /100kHz one of which will generally match the manufacturers specified test frequency.
You can judge the accuracy of the NanoVNA at 100 kHz by comparing its results with those of a dedicated instrument like the DE-5000. At HF, you're a bit out of luck unless you have access to something like a high dollar HP 4294A Precision Impedance Analyzer, 40 Hz to 110 MHz. You might start a collection of "golden" components whose impedance values at HF you have determined using other instruments and use those as transfer standards to check the accuracy of the NanoVNA.
For your 100 pf capacitor example, I would expect at 100 kHz for the NanoVNA to measure 100 pf +/- the specified manufacturer's tolerance. At higher frequencies, especially for leaded components, I would expect values similar to what you measured.
- Herb
|
On 1/7/20 6:27 PM, hwalker wrote: " ..A 100pF silver mica then gives the attached which indicates 105pF at 7MHz, 109pF at 14MHz and 126pF at 28MHz .."
=======================================================
The above frequency dependent impedance effects are why manufacturers specify their component values at specific test frequencies. A dedicated LCR instrument, i.e. the well regarded DE-5000, has five test frequencies 100Hz/120Hz/1kHz/10kHz /100kHz one of which will generally match the manufacturers specified test frequency. ???? Also, they choose lower frequencies not only because it's easier to measure at low frequencies, but also because many components suck at higher frequencies, and by specifying low test frequencies, component manufacturers were able to hide the uselessness of their parts.? Several decades ago I had access to an HP network analyzer, and used it to evaluate components for use in the VHF/UHF radio equipment I was designing.? Among standard ceramic chip capacitors, I found that only Murata and Panasonic were suitable for RF designs.? Pretty much everybody else's "capacitors" were much closer to resistors at those frequencies. But at the specified test frequencies of a few kilohertz to **maybe** a few megahertz, they looked fine, and I assume they would work fine, IF you were only designing audio equipment.
|
You would need to have in the ballpark of 80 nH of series inductance due to series parasitics.. Seems a little high but not out of the question.
Measure caps a low freq first to get values with minimum parasitic effects. Don't go below 2 MHz to stay over about 500-800 ohms of reactance as the accuracy gets poorer at higher value of reactance.
If you have access to a good accuracy 10KHz or 100KHz LCR bridge it would give you a good base to work from.
|
Thanks all for the comments. I am not yet convinced that series inductance of the capacitor leads explains why I measure 26pF more at 28MHz. The 100pF was connected with minimal lead length so I wouldn't expect 60 to 80 nH stray inductance.
I agree that at low frequencies it measures the same as my LC meter.
I think Dave Eckhardt is correct and the answer to the original question is that the test fixture must be pretty rigid and stable even at 30MHz. At 300MHz you need something really good to accurately measure components. Ideally matched to 50 ohms all the way to the component.
Thanks Brian.
|
On Wed, Jan 8, 2020 at 03:22 PM, Brian wrote:
" .. I am not yet convinced that series inductance of the capacitor leads explains why I measure 26pF more at 28MHz. .."
=========================================
Even with a perfect fixture the capacitive reactance of a capacitor will decrease as the frequency across its plates increases. Therefore, the measured value is going to change in inverse to the frequency and not remain the same. How accurately the NanoVNA, or any other VNA, can measure the change will depend on the test fixture you use and how accurately you calibrate out its effects.
- Herb
|
For flying wires and alligator clips to work, everything must be in the *exact
same position* as when the calibration was done I evolved from Mickey Mouse "tails" to Donald Duck bill:
|
Looks *much* better! Now, for results?
Dave - W ?LEV
toggle quoted message
Show quoted text
On Thu, Jan 9, 2020 at 12:22 AM Oristo <ormpoa@...> wrote: For flying wires and alligator clips to work, everything must be in the *exact
same position* as when the calibration was done I evolved from Mickey Mouse "tails" to Donald Duck bill:
--
*Dave - W?LEV*
*Just Let Darwin Work*
*Just Think*
|
I evolved from Mickey Mouse "tails" to Donald Duck bill: Looks *much* better! Now, for results?
Oh, I'm lately working at < 10 MHz and am satisfied to get with 10% of expectations. I wish nanoVNA worked at < 10 kHz..
|
Jim Allyn - N7JA
Oristo
