Choosing the right capacitor and inductor

[darunium]

Hi All,

To caveat this I am no EE, sorry if my question is silly! Any help is really appreciated!

I am trying to probe a load with both a DC bias via one instrument and an AC bias via another (I've tried putting them in series in a variety of configurations, but that hasn't worked out because of the way that the internal electronics on the instruments connect to grounds).

The method I would like to imploy is two loops, one for DC current and one for AC current, but to isolate them from one another I'd add an inductor to the DC loop, and a capacitor to the AC loop.

My load is in the Mohm-Gohm range. My DC bias 10-800mV. My AC bias 2-10mV at 500-1500Hz.


My question is: what qualities of the inductor and capacitor should I look for? Material? Maker? Parameters?

My understanding is that I would want a high inductance and high capacitance based on the idealized impedance formulae.

but in practice do I just want the highest L and C possible or should I be careful about that? To remind my load is in the 100kOhm-1Gohm region, so I want my capacitor to provide <1kOhm resistance to the 1kHz signal, and the inductor to provide <1kOhm resistance to the DC signal, as well as minimal noise in both.

Thanks for any advice!

 

[jiripolivka]

Hi All,

To caveat this I am no EE, sorry if my question is silly! Any help is really appreciated!

I am trying to probe a load with both a DC bias via one instrument and an AC bias via another (I've tried putting them in series in a variety of configurations, but that hasn't worked out because of the way that the internal electronics on the instruments connect to grounds).

The method I would like to imploy is two loops, one for DC current and one for AC current, but to isolate them from one another I'd add an inductor to the DC loop, and a capacitor to the AC loop.

My load is in the Mohm-Gohm range. My DC bias 10-800mV. My AC bias 2-10mV at 500-1500Hz.


My question is: what qualities of the inductor and capacitor should I look for? Material? Maker? Parameters?

My understanding is that I would want a high inductance and high capacitance based on the idealized impedance formulae.

but in practice do I just want the highest L and C possible or should I be careful about that? To remind my load is in the 100kOhm-1Gohm region, so I want my capacitor to provide <1kOhm resistance to the 1kHz signal, and the inductor to provide <1kOhm resistance to the DC signal, as well as minimal noise in both.


xxxxxxxxxxx
Thanks for any advice!

You are exactly right that the coils (chokes) you need at 1 kHz should have a substantially higher reactance that is the "standard" source/load impedance of your system, here shown as 100 kOhm.
Capacitors then should have the smaller reactance, to allow the 1 kHz signal to pass with a low loss.

As your "base" impedance is 100 kOhms, a typical coil or choke for use at 1 kHz can have an iron (good quality transformer core), ferrite core; for the best performance some use air core but it makes a coil or choke too large. For capacitors any good dielectric can be used; with ceramic capacitors, some materials are piezoelectric and could generate sounds or pickup noise.

Below ~10 Hz and above ~100 kHz, coil and capacitor materials should be chosen more carefully; at 1 kHz, you are in a "safe" region. RF-quality components should be used above 100 kHz.

 

[dave9000]

If you want to use only passive elements, you may consider a simple RC solution instead of a LC.
For example with a 1Kohm resistor in place of L1 you introduce a maximum error of 1% (due to the minimum load of 100K).
In this case you can use C1 > 300 nF (for example 1uF or so) and you are set.

Or you can build an active circuit with an opamp.
In this case you can achieve a nearly ideal setup.

 

[darunium]

Thanks for the great advice! I'm aiming for something then in the 200-2000uF range, which should give my 1kHz signal only 1-0.1Ohm of resistance (essentially zero). Unfortunately for the inductor, to get a 10^11Ohm resistance for my 1kHz signal, to force it over the load, I'd need a 1MH inductor, but inductors only go up to 10^2 Henry orderof magnitude. Any way to deal with this? The second post mentioned using a resistor instead of the inductor, but I'm afraid I don't see how that would impede the oscillating current any more than the DC.


On the capacitor material: Aluminum capacitors it seems have prety high leakage currents, in the uA regime, which is way too high since my currents will only be in the nA-pA orders of magnitude. Ceramic capacitors have lower leakages, but the noise issue was mentioned above. Is leakage current really an issue when there is no parallel resistor, or would it be an issue even here?

Thanks again!

 

[FvM]

I don't think, that the measurement problem has been yet clearly specified. To avoid jumping into conclusions, I want to ask for a clear specification first.

What's the exact point with separating DC and AC loops. As long as the DC and AC source are just low impedance sources, I don't see a reason to have e.g. a "substantially higher reactance" than 100 k load impedance. For higher load impedances, this objective would be simply illusional.

A simple understandable and not too difficult achievable requirement is not to introduce large errors of the AC respectively DC voltage. The situation would be different, if one or both sources are actually source-meters and you want them to measure load impedances.

Finally, the low AC and DC ranges suggest an active circuit adding both voltages and feeding the sum to the load, completely avoiding the problematical bias elements.

P.S.: The signal (sine) generator can be expected to expose 50 ohm output impedance. Thus requesting zero capacitor reactance doesn't sound justified.

 

[jiripolivka]

I don't think, that the measurement problem has been yet clearly specified. To avoid jumping into conclusions, I want to ask for a clear specification first.

What's the exact point with separating DC and AC loops. As long as the DC and AC source are just low impedance sources, I don't see a reason to have e.g. a "substantially higher reactance" than 100 k load impedance. For higher load impedances, this objective would be simply illusional.

A simple understandable and not too difficult achievable requirement is not to introduce large errors of the AC respectively DC voltage. The situation would be different, if one or both sources are actually source-meters and you want them to measure load impedances.

Finally, the low AC and DC ranges suggest an active circuit adding both voltages and feeding the sum to the load, completely avoiding the problematical bias elements.

For your circuits at 1 kHz and 100 kOhms, you can also try audio transformers. Some do not like DC current, so use caution.

 

[darunium]

I don't think, that the measurement problem has been yet clearly specified. To avoid jumping into conclusions, I want to ask for a clear specification first.

What's the exact point with separating DC and AC loops. As long as the DC and AC source are just low impedance sources, I don't see a reason to have e.g. a "substantially higher reactance" than 100 k load impedance. For higher load impedances, this objective would be simply illusional.

A simple understandable and not too difficult achievable requirement is not to introduce large errors of the AC respectively DC voltage. The situation would be different, if one or both sources are actually source-meters and you want them to measure load impedances.

Finally, the low AC and DC ranges suggest an active circuit adding both voltages and feeding the sum to the load, completely avoiding the problematical bias elements.

P.S.: The signal (sine) generator can be expected to expose 50 ohm output impedance. Thus requesting zero capacitor reactance doesn't sound justified.

Hopefully I can clarify a bit, if I'm understanding your comment correctly:

- The reason I think I need to isolate the two loops is because without doing so, almost all of the current flows on the outer loop, avoiding my sample, MOhm resistor. My aim is to measure the current flowing through the sample resistance at both DC and at my oscillation frequency (1kOhm) (using a lock-in amplifier to pick out the amplitude of current oscillations at 1kHz)
- you're correct in that in channels in which I want to *pass* current I don't need 1 Ohm resistances, anything around 100Ohm or less is totally fine, even kOhm is fine for the measurements I'm making as long as it's not fluctuating any more than 1 Ohm (which I imagine is not likely for a kOhm resistor). The issue I'm running into is in the loops where I want to forbid either AC or DC. In the DC case, it looks like a capacitor will do the trick perfectly. In the AC case however, it seems that I would need an impossibly large inductance to provide the 1kHz current with a resistance greater than my load.

does that get at what your asking?

 

[FvM]

There's is no provision for current measurement in your circuit. How do you intend to perform it?

it seems that I would need an impossibly large inductance to provide the 1kHz current with a resistance greater than my load

Yes, the idea simply doesn't work. That's why i would refer to an active source feeding the sum of both voltages. Measurement with GOhm load impedance will be still affected by a considerable error, but at least not additional by a bad bias circuit.

 

[darunium]

There's is no provision for current measurement in your circuit. How do you intend to perform it?


Yes, the idea simply doesn't work. That's why i would refer to an active source feeding the sum of both voltages. Measurement with GOhm load impedance will be still affected by a considerable error, but at least not additional by a bad bias circuit.

Thanks for the comment.

The circuit is simplified, there are current and voltage measurements in series and parallel with the DC power supply, and a lock-in amplifier with 1kOhm series resistance in series with the AC power source.

Given the limited inductance that it is possible to get out of an inductor, I will consider your suggestion of using an op-amp with feedback as a voltage summer. If I can get my hands on a sufficiently low noise op-amp, that should be fine. I may alternatively try to make an active low-pass filter in series with DC power supply, to try to minimize the amp noise in my 1kHz region. I will continue to try to work a series solution as well (V_DC and V_AC in series), but as I mentioned in my first post there are some loops internal to the equipment that have so far frustrated those efforts.

 

[varunkant2k]

Hi, For DC loop, instead of using large Inductor, You can use Low pass filter with good (active -good, passive-moderate) accuracy. Assume, low pass filter does not pass any current higher than 100Hz, ( required, 1.59nF and 1Mohm if load is 100kOhm). This filter resistance will carry, one tenth of load current, make gives 10% error. But if it can be accepted is the good way to do that.Offcource your active filter is the best way.

 

[FvM]

The circuit is simplified, there are current and voltage measurements in series and parallel with the DC power supply, and a lock-in amplifier with 1kOhm series resistance in series with the AC power source.

O.K., this sounds more plausible. I think, I had be a good idea to indicate the current measurement in the original post, because it gives sense to the circuit.

If you understand varunkant2k's post as suggestion to implement the DC path as current source with a low-pass voltage control loop, it shows another possible solution. But the GOhm order of magnitude suggests a single AC+DC source without bias T to minimize error terms, I think.

Your setup is very similar to a standard impedance measurement problem as implemented in high performance LCR meters. Studying the circuits used e.g. by HP/Agilent for this problem can give additional insights. Although their solutions mainly target to a wider frequency range up to MHz, you get similar problems when trying to determine the reactive impedance part of a GOhm load. There's a previous edaboard thread about the measurement circuits of these instruments.

 

[darunium]

O.K., this sounds more plausible. I think, I had be a good idea to indicate the current measurement in the original post, because it gives sense to the circuit.

If you understand varunkant2k's post as suggestion to implement the DC path as current source with a low-pass voltage control loop, it shows another possible solution. But the GOhm order of magnitude suggests a single AC+DC source without bias T to minimize error terms, I think.

Your setup is very similar to a standard impedance measurement problem as implemented in high performance LCR meters. Studying the circuits used e.g. by HP/Agilent for this problem can give additional insights. Although their solutions mainly target to a wider frequency range up to MHz, you get similar problems when trying to determine the reactive impedance part of a GOhm load. There's a previous edaboard thread about the measurement circuits of these instruments.

Thanks guys! That gives me a couple things to try, and it seems like the same op-amp would work for either configuration.

Yeah I took a look at Agilent's solutions for combining Network and semiconductor analyzers, I figured my problem would be a lot easier since I'm operating at lower freqs, but it turns out that they actually use the higher freqs to make wider use of inductors, something I only realize now :-( 1kHz makes my capacitor problem easier, but my low-pass filter problem trickier. Them's the breaks.