measuring inducatnce & capacitance a high frequencies

hi,

can anyone help me please.

im trying to measure inductor & capcitor values at rf up to around 150 mhz (up to srf). im using a 50ohm output rf sig gen capable of covering these frequencies.im also using a cro with sufficient bandwidth.

can anyone devise a test circuit for inductors and caps separately. ive been trying to use either a resonant peek parallel LC or series LC resonant dip circuit. although this works, you can only determine L with known C or vice versa.


The results are awful at high frequency.

e.g.

resonant dip for C=22pf & L=1uH should be 33.9mhz, when measured its 32.7mhz. its quite good.

resonant dip for C=4.7pf & L=1uH should be 73.4mhz, when measured its 58.6mhz. its not good.

any help devising a test cct please, thanks

rick

The results are more or less expectable. You can only expect better results by ignoring the real properties of components, particularly inductor's winding capacitance respectively self-resonance frequency (another way to specify parallel capacitance).

Parasitic capacitance of your measurement setup might add. For a specific discussion, the involved inductor type should be mentioned, you should also describe the measurement circuit.

Re: measuring inducatnce & capacitance a high frequencies

I saw a website tell how to apply a sine wave to the coil through a known resistance, and measure the volts across the coil (or is it across the resistor?). The amount of voltage drop will let you calculate the impedance. From that you can use the time constant formula to derive the Henry value.

The known resistance should be 10 to 100 times the DC resistance of the coil, so that coil response is chiefly due to reactance, not ohmic resistance.

--------------

Edited to add this link. The author discusses many factors that are involved.

https://users.monash.edu.au/~ralphk/inductance-measuring.html

Re: measuring inducatnce & capacitance a high frequencies

hi,

thanks for your attention FvM & thnks for the link Brad, im looking at it. i'll get back to you both tomorrow cant say much today.

rick

Re: measuring inducatnce & capacitance a high frequencies

hi,

thanks FvM & Brad. inductors im using in range 0.3 to 2.2uh. there mainly through hole chokes or air coils.i havent used smd but, probably will. will they be any better.cant find much data on caps or coils but, id like some.

Brad' ive tried a circuit like that link. i find an LC series circuit more reliable as a resonant dip method but, still cant get any dips above say 90mhz. i wish i could find a circuit measuring L in terms of resistors but, these all fail.

i think id need an RF LCR meter that works up to 100s of mhz. Cant find any used ones anywhere at low price.

Also on return loss bridges. ive seen many made with claims of 30db directivity. when i had one made it only gave 21db. Also, do these measure real part or magnitude of impedance.

any advice on these things please.
rick.

My guess is 4.7pF is very low capacitance, input capacitance is going to add onto in.

My wild guess is this:

Use a coax to connect from generator to the copper board you use for making the test fixture. Terminate with 50ohm. build your parallel LC circuit close to the termination point. Connect a 1K resistor from the coax to the junction of the LC.

You need to put a high frequency amp to buffer the LC. Use an opamp in non inverting mode to get high output impedance. Then connect the LC to the +ve input of the opamp by a 1K resistor. The amp should have some gain. The output of the opamp should drive a coax to the scope through a 50ohm resistor.

My point is the isolate the LC with high resistance (1K) to isolate all the parasitic capacitance from reaching the LC. The impedance of the LC is low in all frequency except at resonance frequency. So even with 500ohm ( two 1K) loading, you should see it peaks at the resonance frequency. You might want to experiment with even higher resistance than 1K.

I am not too worry about the frequency flatness of the opamp as I am looking for peak at the calculated frequency. You have to be very careful in building the circuit as layout is very important at this frequency.

1.

Suppose you were to send pulses through the coil, and observe the coil's time constant behavior on a scope?
It would be the customary curve, across a resistor to ground.



The plateau voltage indicates the proportion to the coil's ohmic resistance.

2.

A method also comes to mind that involves a length of wire. See how long is the coil-induced delay, and match it to a length of wire.

At 150 MHz, one wavelength is 6.5 Ft. You wouldn't need that many feet, just maybe one foot, to roughly match the delay in the coil.

I'm not sure I would use a clock generator in this case. The fewer components involved, the better.
I would scratch a wire against a battery terminal to make definite on/off contact.

150 MHz is pretty low. A 1 MHz or 10 MHz LCR meter would do that job very nicely.

If the exact model of the reactance is needed (including interturn winding capacitance, etc) then measuring it on a network analyzer is the way to go.

thanks brad , allan & biff.

brad, i was looking mainly for techiniques to get C or L as direct functions of
frequency. i think im interpreting correctly.

Allan, i like that idea. i can see what your saying and i hadnt thought of buffering
via amps, its good, i havent done every last thing yet. can you draw me the circuit
to completely clarify it please. also id probably try attenuating from source to LC
parallel & from amp output to cro as precaution. also could we use a 2n3904
considering its high input & output impedance. hmmm ...bit lost here.

biff id probably look to use a network analyser way off in future rather than spend
anything for only 10mhz right now as i am looking for exact behavoiur of l & c.

also everybody,

on filters with non ideal components. about 30mhz fundamental id probably be ok with
l & c measurements i can make now, but obviously what bout the harmonics say 4th.
,can anyone just point me in right direction here...most alogrithms rely on constant
l & c values. i know how to create non ideal models but how do we compensate for this in filters.

thanks all..

rick

The circuit should be soldered onto a copper plane board as shown. Amp use has to be at least 200MHz. R2 and R3 is used to isolate the LC tank, you can experiment with higher value. Use 1/4W leaded resistors for these two because longer body lower the capacitance between the two terminals of the resistor. R1 is 50ohm termination. You set the gain of the opamp by R4 and R5. Don't make it too high as you can roll off the frequency response of the opamp. Use current feedback opamp.

I forgot to put the value of the resistor from the output of the opamp to the coax. It should be 50ohm.

The generator sweep through the frequency. Any frequency other than the resonance frequency, the impedance of the LC is low and very little pass to the opamp. At resonance, the impedance of the LC goes up and frequency will go from generator to the opamp. You should see a peak at the resonance frequency. 150MHz is low, you should see the desired result.

i know how to create non ideal models but how do we compensate for this in filters.

Click to expand...

All L/C manufactures I know about characterize the components by two port VNA measurement. It results in S-parameters useable to calculate real behavior in filters, also beyond SFR. Most simulation tools can import these parameters for calculating a complex filter behavior over a wide frequency span.
Typical fixtures for dual port measurement: **broken link removed**
Measurement method from Murata inductor product catalog p.86:
It is also possible to measure L/C impedance with a single port: http://edocs.soco.agilent.com/display/iccapmhb/Time+Domain+Modeling
A dynamic impedance matching tool which takes non ideal component behavior including SFR and ESR in account when calculating optimized filters: http://www.antune.net
With this tool is it easy to make a real filter that results in similar curves as a theoretical calculated filter. It covers up to 6 GHz for most common component types from Murata (20 GHz for some special cases). Main function is impedance matching. Takes input data either directly from a VNA over GPIB or by importing a Touchstone file.
For frequencies below a few hundred MHz is it not that critical with complex losses in measurement fixture setup. Stray capacitance, serial inductance in measurement cables and losses in PCB dielectrics can be assumed ideal in a reasonable well designed measurement fixture. Then can a Wheatstone bridge connected to a signal generator be used to measure complex impedance as a function of frequency. That is basically how many cheap LCR measurement tools works, but self balancing. http://www.sciencedirect.com/science/article/pii/S0263224102000404

i was looking mainly for techiniques to get C or L as direct functions of frequency.

Click to expand...

You don't actually measure L or C as a function of frequency. You measure the complex impedance of a component with parasitic elements, particularly capacitor series inductance and inductor parallel capacitance, along with loss elements. The complex impedance appears as an apparent C or L variation over frequency. In addition to this effect, capacitance and inductance itself may be frequency dependend, but the effect is ussually small compared to parasitic elements.

Your measurement setup adds more parasitic elements and thus additional apparent frequency dependend variation. The parasitic elements can be identified and eleminated in the calculation. An important advantage of LCR meters as well as VNA is that they can perform this corrections automatically.

yeepox_m said:

hi,

can anyone help me please.

im trying to measure inductor & capcitor values at rf up to around 150 mhz (up to srf). im using a 50ohm output rf sig gen capable of covering these frequencies.im also using a cro with sufficient bandwidth.

rick

Click to expand...

The best way to do this is to use a VNA, with an S-paramter test test. Make up a PCB with a some SMA connectors, and characteriste your board using opens, shorts and a load. The VNA will measure the S-paramters,. An 8753A/B/C etc will do you ok. If you switch to the "Smith Chart" view, you can directly read the capacitance or inductance as well as the resistance. It will give you an impedance R + j X, and also convert the imaginary part directly to capacitance or inductance.

This is covered in quite some detail in this book.

https://www.amazon.co.uk/Handbook-Microwave-Component-Measurements-Techniques/dp/1119979552

If you do buy a used HP VNA, I suggest you buy one with option 010, which is the TDR option. That can be quite useful for this sort of thing. If you buy an HP 8753B or later, you might want to consider puchasing option 006, which increases the frequeuncy range to 6 GHz. There's also an option 002, which does harmonic related things, but I don't believe it is much use.

Of course, this may mean buying more test equipment.

Dave

hi

thanks everyone for your useful replies.

Alan thanks for the circuit ill try it. op amps a bit expensive though at >= 200mhz bandwidth so can you suggest precisely which opamp to use please. ive not tryed this way of doing it before. usually ive put v small caps either side of LC if parallel.

FvM & E kafeman thanks. i understand trying to get at C or L by measuring at lower frequency. then measure esr & parastic by finding the srf at high frequency. i wasnt exactly trying to make a model, rather do obvious approach of actually measuring impedance over a frequency range or look at a resonance in LC parallel or series. on filters, can the antune software actually "derive" a filter network with characteristics that would be equivalent to the case if components maintained ideal behaviour. i.e. will it derive a compensatory network for each component if non ideal behaviour is significantly present.

just one extra question here does anyone know where to get smd inductors cheaply?


drkirby also thanks. Unfortunately this is the bad news with lots of £ involved.


once again thanks all.

In my view, I will simulate to choose the Op-Amp, that is very easy to simulate in ADS.

When I need good accuracy I do exactly what you are proposing. The trick is knowing what you are dealing with.

You need some accurate reference points. Two two main ones is accuracy frequency souce, the second is a known capacitance at frequency of measurment desires.

Assuming the known capacitance is usually the problem. One method is to use a transmission line stub which can be pretty accuracy predicted.

You want to create a resonate tank circuit with as little external loading as possible. As long as the loading resistance is much much less then the parallel equivalent Rp of the coil you have met the light loading requirement.

I routinely do this accurately at frequency above 1000 MHz for determining inductance of airwound coils.

Here is example of what I do.

Lets say you need a 12 nH coil at 1000 MHz. That is +j75.4 ohms. It would parallel resonate at 1000 MHz with a 2.11 pF capacitor. The best airwound coil you can make at this frequency would be less then a Q of 100. If I got to Q of 100 that would mean the coil Rp would be 100 x 75.4 or 7.5k ohms. So if I don't want to degrade the measured Q I should not load the parallel tank more then 30k ohms or greater. If I lightly couple onto the tank with series caps from a generator and take of a sense to some RF level measuring instrument (like a spectrum analyser or low level power meter) I would use two coupling caps of -j900 or about 0.2 pF from the generator and 0.2 pF to the power meter. I can ignore the Rp from loading but should add 0.4 pF to the capacitance loading on the coil. So I need a 1.7 pF cap instead of 2.1 pF cap to resonate at 1000 MHz. I stand up high tolerance chip caps on a copper clad test board and solder coil directly across chip cap to minimize lead inductance from chip cap. You can check with cap manf. to see if they list a series inductance on a chip cap. 0603 caps are in the order of less then 0.5 nH Ls. You can change the cap to adjust for cap Ls. At 100 MHz you will unlikely need to worry about it. If you have a good idea of a caps equivalent series inductance you can use a low frequency cap meter to measure a poor tolerance cap and then correct capactance effective value at desired frequency.

I crank up the generator at 1000 MHz until I get an indication of the output sensing instrument. Then sweep freq until output sensing peak. This is resonate freq. Move freq gen lower until signal drops 3 db. Do same on high side. This allows me to calculate Q of tank which is dominately coil.

You likely will not hit coil right on. So use actual measured frequency to calculate coil value.