inductance measurement

[ezpcb]

Hi all,<p>Currently I have a project, it's a inductor meter that needs to measure inductance from several hundreds of nH to H level. Could your guys tell me any inductance measurement method and circuit? thanks!<p>Mike

[cato]

1)Hit it with a pulse, measure the decay time.<p>2)Place it in series with a known resistance. Sweep frequency signal, measure 3db point.

[philba]

3) use it in an LC oscillator - measure frequency.

[Robert Reed]

These are all good ways to measure the inductance, however you will have to take into consideration the units ditributed capacitance. In each of these tests, the inductance will always be lower than test results would indicate. All inductors have a self resoneant frequency because of its distributed capacitance. Experience tells you how much capacity to add (as in LC resonance tests) so that added capacity will swamp out expected distributed capacity and therefore make the results more accurate. If you work with LC circuits enough, you just get a feel for what frequency range and associating resonating capacitance should be used for any particular inductor. Accurate inductance measurements as related to their intended circuitry is not the easiest parameter to measure, as a lot of things will change your measured value as 'tested'. Not the least of these would be DC current orDC offset in the intended circuit (such as normal amplifier bias current) or your test setup. You have an extremely wide range of inductances that you want to check and one method may be better than another for any particular inductor. If you plan on doing this on a regular basis, you may be better off by a low priced service grade 'L' meter. Even these readings will have to be taken with a "grain of salt". BTW did you catch the L-meter construction project a few issues back in N/V ?<p>[ September 26, 2005: Message edited by: ROBERT REED ]<p>[ September 28, 2005: Message edited by: ROBERT REED ]</p>

[ezpcb]

Thank you all. Robert is right, for different inductor, different measurement method should be applied. There's no one "good" way for 1uH to 1H. <p>I wish I can check the construction of "L" meter on the back issure of N/V but I have no way to find it. Could anybody find which issue is it?<p>BTW, another question is how to measure "Q" of a inductor?

[Robert Reed]

The L-meter is in N/V August 2005 issue and is titled 'What the L is it". It looks like a decent design,too. One way I have done quick and dirty tests for inductance is with a scope and rf signal generator.Parellel the coil with an appropiate capacitor( The trick here is to pick the 'C' value, and a rule of thumb would be 10 pf for nano henry range; 100 pf for lo micro henry range; 1000 pf for hi micro henry range and progressivly larger as you work up into millihenry range and beyond. This will insure that you swamp internal dist. capacitances and still not drag the Q of the LC circuit down to a point where it becomes unreadable.Feed the rf signal into the inductor thru a 100K resistor and bomb the hell out of it (I have various generators of 0 Dbm to +25 DBm out put).From this point connect your scope thru a 10K resistor, then manually sweep the RF while watching the scope for a sharp resonant peak. Stop at the peak and read the frequency of your generator. You can plug these Fo and C values into any number of formulas to get your answer. Personally I just use a L/C slide rule for this,but you can do it long hand.Remember that the reactance is identical for L and C at resonance. If generator out put isnt producing a hi enough level you can reduce the 10K &100K values, but usually an output of 100 mv rms is sufficient with a quality scope.
I know you are thinking -how do I pick a suitable resonating capacitor if I have no idea of the inductor value? Again all I can tell you is that it comes with experience. I have rough checked so many inductors over the years,that I can usually come up with a broad ballpark figure just by looking at it. But I still get surprised sometimes and have to resort to a lot of playing around to gets its value. All I can tell you is bigger is more and smaller is less. You wont do too much testing be fore you develop a handle on this. Speaking of L-meters, this is one design project I have had on my mind for sometime, and would like to build one when I get the opportunity.
As to 'Q' measurements- they are even more difficult as a lot of things come into play here that will affect test results. I have some info on building a Q-meter somewhere. If I can find it I will dig it out for you.

[rshayes]

Coils should be tested at a frequency close to that at which they will be used. This avoids suprises, such as iron coil coils whose inductances decrease with frequency (probably due to eddy currents). For large scale production, coils should probably be tested on equipment selected for each individual case.<p>For coils in the .1 microhenry to possibly the millihenry range, I prefer a Q meter. It is a little tedious to use, but it can measure the coil at the operating frequency. A second measurement below the operating frequency allows calculating the shunt capacitance and correcting the inductance value from the measured value to the actual value. Knowing that and the shunt capacitance, you can calculate the self resonant frequency of the coil and confirm that it is well above the operating frequency of the coil.<p>Once you are above the millihenry range, a low frequency bridge becomes a better choice. The Q of these coils is typically too low to get a well defined peak on the Q meter and extra capacitance may be required to get resonance. The frequency is usually fixed at 1000 Hz, so it isn't possible to separate the shunt capacitance by making a measurement at a different frequency. Laminated cores can be tricky, I got caught once with a coil that measured one value at 1000 Hz and another value substantially lower at 2200 Hz, the actual operating frequency.<p>Hewlett-Packard used to make a network analyzer that was great for this type of thing. It measured the impedance of the part over a swept frequency range and displayed the result on a CRT display. It then calculated the values of an equivalent circuit that would match that response. It was particularly good on crystals, since it gave parts values for all four parts in the euuivalent circuit. The price was about $20,000 in the early nineties.<p>I don't really trust the LCR meters. Each range has a different test frequency that is not under your control. As a result, you can wind up trying to make a measurement too close to the self resonant frequency.<p>[ September 27, 2005: Message edited by: stephen ]</p>

[Will]

This isn't really my subject but I'm interested so I shall probably learn something when I'm jumped on over all my mistakes - I think that (Using 'W' for Omega) at resonance frequency, W1 = sqrt(1/LC) so that, if you know it then LCi = 1/W1^2 - Such being the case then it's impossible ? to tell the inductance from just finding the resonance frequency because LC could have an infinite number of compound values. If you were to add a parallel capacitance (Ca) and find the resonance frequency of the combination then could 'L' not be determined from L = (1/W2 - 1/W1)/Ca ? - I think it could ?

[rshayes]

The idea behind a Q meter is to resonate the coil at a known frequency with a calibrated capacitor. From this you can calculate the inductance using the formula for resonant frequency.<p>Unfortunately, most coils have substantial amounts of stray capacitance. This is usually assumed to be a capacitor connected in parallel with the coil even though it is actually distributed capacitance between the turns of the coil.<p>The capacitance resonating the coil is thus the calibrated capacitance in the Q meter plus the stray capacitance of the coil. This means that the actual coil inductance is less than the measured coil inductance.<p>The inductance measurement can be corrected by maling a second measurement at a lower frequency. Usually, a frequency of half the original frequency is used. In this case, resonating the coil at the new frequency will take four times the capacitance originally required to resonate the coil. The additional capacitance is all added by increasing the calibrated capacitor. The change in capacitance is three time the capacitance required to resonate the coil at the higher frequency. One third of this capacitance will be larger than the first capacitance setting. The difference between the two is the stray capacitance of the coil. The revised capacitance can also be used with the original resonance frequency to calculate the actual coil inductance.<p>Basically, two measurement are needed, but they give you both the actual coil inductance and the coil's stray capacitance.

[MrAl]

Hi there Mike,<p>What you'd be measuring is basically the ac
inductance. That is, the inductance with no
dc current. For inductors that are to be used
in dc to dc converters for example, where there
can be any level of dc current flowing from zero
to max the only good way to measure inductance is
in the actual circuit. Looking at both current
and voltage on a scope you can deduce L from:
V=L*di/dt
The ac inductance changes with dc bias, so you
could measure a different inductance at no
load, half load, and full load. If the inductor
saturates, its inductance could change 100uH to
1uH. Even without saturating however it wouldnt
be too unusual to see it change from 100uH to
50uH, and some supplies are designed with an
inductor to change even more on purpose to
maximize efficiency (see: swinging choke).
The iron or ferrite core is the reason for this
change as dc current changes because the magnetic
properties change as the core reaches different
levels of magnetization.<p>Measuring the ac inductance with zero dc bias
might tell you if the coil is wound correctly
though...if you're winding your own inductors.<p>Let us know what you end up doing...<p>Take care,
Al

[Will]

Sorry guys,
The formula which I derived and offered up should have read
L = (1/(W2^2) - 1/(w1^2)/Ca
i.e. the reciprocal Omega values (Omega = 2.pi.f) should have been squared. I still believe this would work because I constructed an Excel speadsheet with several values and they all produced correct answers. Clearly the accuracy of the result would depend on the precision of the added capacitance (Ca) ?
Repeating (i) You find the resonance freq of the inductor alone (f = W/2.pi where W = 2.pi.f) then find the resonance freq with an added capacitor then apply the formula. Is this correct or not ?

[rshayes]

To have a resonance at all you need to have some capacitance. Without an added capacitor, this is the stray capacitance of the coil. Often the Q of the self resonant circuit is low and the frequency should be much higher than the operating frequency. And, as you pointed out, measuring the self resonant frequency alone does not give you actual values for inductance, since there is an infinite number of inductance-capacitance combinations for each self resonant frequency.<p>If the coil is resonated with an added capacitance that is larger than the stray capacitance, then the added capacitance value can be used to calculate the coil inductance. If the stray capacitance is 10 pF and the added capacitance is 100 pF, the error will be about 5 percent. This may be good enough and only requires one measurement.<p>If higher accuracy is needed, a sceond measurement can be made at half the frequency as I previously described, the stray capacitance and actual inductance calculated.<p>Most Q meters have variable capacitors that go up to about 400 pF. If the resonating capacitance (ie, the variable capacitor) is kept nera the upper end of its range, the error in measuring the inductance with a single measurement will be only a few percent.

[Robert Reed]

Its called core saturation, however air core coils will not suffer from this malady. One reason I like the test I described is its simple- not highly accurate --but simple. It is also very easy to check the self resonance point while doing this. The coils distributed capacitance is fixed and we can do nothing about it,therefore it is of little consequence after knowing self resonance point. An inductor (at a given frequency) will look increasingly more inductive as we increase that freq, until we reach that self resonant point at which time it should look resistive. Increasing frequency beyond that point and it starts to look capacitive. Increasing frequency greatly beyond that point (several orders of magnitude) and it looks like a capacitor--albeit a poor one. This property change is common to all components (R,L,C) at Uhf and beyond, which is why work in this area becomes difficult.At high enough frequencys, the major determining factor of the circuit is the printed circuit board itself. Still as much as I like testing this way, when it comes to high accurracy or multiple part checking ,I use a calibrated variable capacitor or a 1% cap. decade box for large inductors (short test leads or stray box capacitance wont bother you much here due to the large LC values already employed). But you always need some sort of isolation for generator and scope to keep the 'Q' up for a suitable measurment. If you arrive at the LC values you will be actually using in your circuit (while still in testing mode) You can determine 'Q' by sweeping the generator thru a small range and noting the 3 DB corner points. Note that span and divide it into the center frequency and your result will be 'Q'.This calculation has some suspect to it,though. Just bear in mind that the cicuit it eventually goes into will affect that 'Q' by how much adjacent loading it introduces.
Stephan
In regards to your reply that self resonance point would be of low 'Q'--- by virtue of one of the many equations for determing 'Q' (and some of them contradict themselves) ,one comes to mind -- 'Q' = L/C which would indicate the highest possible 'Q' point for that inductor operating in its own LC configuration. Or could it be that inherant distributed capacity is of such a low 'Q', that it drags down the whole circuit when relying on this alone for resonance under those conditions ? <p>
[ September 28, 2005: Message edited by: ROBERT REED ]<p>[ September 28, 2005: Message edited by: ROBERT REED ]</p>

[rshayes]

The dielectric associated with the stray capacitance is usually the coil form and the wire insulation. The loss in the insulation tends to lower the Q. So, yes the distributed capacitance tends to lower the Q.<p>Also, the Q meter citcuit cannot reach the self resonant point directly. Unless there is some added capacity, there is no injection of energy into the resonant circuit. Using only the minimum capacity, usually around 30 pF results in an indicated Q lower than the actual Q. There is a correction for this, but life gets awkward when the indicated Q is less than 10 or so.<p>You can measure the self resonant frequency by using a working coil to establish the resonant circuit and then noting the change in frequency when you parallel the working inductor with the inductor being tested. If the frequency doesn't change, the working coil was tuned to the self resonant frequency of the coil being tested. Of course, you now have the losses of both coils to consider.<p>There are several definitions of Q. One is to assign all losses to a series resistance. In this case:<p>Q = (2 Pi f L) / Rs<p>Q = sqrt(L / C) / Rs<p>If the resistor is in parallel then:<p>Q = Rp / (2 Pi f L)<p>Q = Rp / sqrt(L / C)<p>Other definitions are based on 3 db bandwidth and on energy loss per cycle.

[Robert Reed]

Stephan
As I suspected that distributed capacitance was a "dirty" capacitance. I have had no problem measuring self resonant points with the method previously mentioned, however it does require heavy isolation, and since the 'Q' really comes down to resistance (either series or paralel) present in the LC circuit, I suppose it would be very difficult to get a high degree of accuracy, since the test setup itself would have to present a load no matter how small it might be. For my own use, 'Q' has basically boiled down to X/R or L/C. In designing hi-Q tank circuits for oscillators, I generally shoot for the highest L/C ratios feasable. This puts a limit on variable oscillators for any given range, due to very low L/C ratios when the variable capacitor is fully closed (highest C to L ratio at this point).
For this reason I usually limit tuning ranges to about 3.3 to 1,as when L/C ratios drop below 10 cannot sustain oscillator activity. I have had mixed success with the 3DB bandwidth method with an occasional puzzling result, but the main thing here was could I acheive the desired bandwidth as installed in the final circuit, and for that purpose the 3DB method has worked satisfactory.