parametric pumping of L's and C's

[F6FLT]

Diodes like 1N4007 and zener diodes can be used as varactors, they have quite big range

For example in this video test setup presented, capacitance change from 24pf to 2000pf

https://www.youtube.com/watch?v=zTKShnW4w-8

Regards,
Vasik

Thanks for the information, Vasik.
Impressive variation that I didn't expect!

[ion]

Thanks for the information, Vasik.
Impressive variation that I didn't expect!

Neither did I, as this exceeds most experimenter available diodes that are only around 20pF. But then it is curious that the upper limit is close to the value of the DC blocking cap, so I have to wonder if he is actually measuring the 2200 pF DC blocking cap as the impedance at the anode of the diode goes to a low value with the forward bias. If he changes the DC blocking cap to 4700pF what will be the new upper limit? Will it still be around 2000pF?

I will repeat this test as is, and with a small change to the test setup as time permits. Of course different cap meters can give variable results, so all is good.

Regards

[partzman]

Here is a sim test of the capacitance change for an On Semi 1N4007 model with a reverse current of 1ma.

As is seen, at ~10ns the capacitance is 16.04pfd and at 450ns the capacitance is 599ffd.

This technique will not work for measuring the forward capacitance of this same diode due to forward conduction.

Regards,
Pm 

[partzman]

This is a sim of the forward capacitance of the same 1N4007 diode at .1v, .2v, .3v, and .4v step levels.  The test is a simple series resonance of the diode's capacitance with the 200uH L2 inductor. 

The first resonance at .1v is initiated by the closure of S1 for 200ns and the following step changes produce the resulting resonance voltages.

As can be seen, the capacitance values are considerably less than the video tests indicate for this diode.

Regards,
Pm

Edit: Replaced sim to show the method used to calculate the capacitance of each step.

[partzman]

And yet another forward conduction test method sim that takes advantage of LtSpice to set certain initial conditions to nodes.  In this case, we set an initial voltage level across D1 prior to the start of the simulation and then we immediately discharge any energy in D1 through L1 and measure the peak current in L1 when the voltage across D1 is zero.  We can then calculate the apparent capacitance of the diode under any initial forward voltage bias.

The results here are more accurate and differ from the resonance tests because their voltage variations change the capacitance as can be seen in the non-linearity of the waveforms.

IMO, any attempt to utilize this apparent forward capacitance for OU will be in vain due to the amount of energy required to forward bias the diode as compared to the energy available from the capacitance.

Regards,
Pm

[Orthofield]

Hi All,

Still poking around on varactors. Since the C change is completely due to V, no bias current is really necessary-- it's an artifact--so I was looking for some sort of pure voltage source that could drive varactors. F6FLT mentioned electrets and that got me thinking. Generally these are charged to too high a V, but it seems like a weak electret could be switched into a set of varactors to drive parameter change.

I also looked at electret microphones which also have the high output impedance. Since sound is always present in the environment, it should be possible to make a 'sonic energy harvester' using electret mike + varactors. As in other capacitive energy harvesters, a battery could charge the varactors in portions of the cycle where C is dropping, then discharge the higher voltage into a load while C is rising. The battery could be kept charged, and excess energy used to power a load. This would be cheap to build and test.

This circuit for a small FM transmitter uses an electret to drive dual varactors:

http://electronics-diy.com/electronic_schematic.php?id=1066

I'm not sure if the HF dynamic behavior of the varactor would be relevant in this case, so it seems possible that the output power could be higher than the acoustic input power.

F.

[F6FLT]

A current is needed in any case where C or U change: i = d(C.U)/dt = C.dU/dt + U.dC/dt

No gain. The extra energy we expect comes from the increase in U following the decrease in C. But since we supply U to decrease C, we directly supply the "extra" energy!
With an electret, perhaps...

[Orthofield]

Hi F6LT,

You're right, certainly current must flow, as the V increases in the bias. Electrons have to get to the bias juncture. But the question is, how low does the initial static current need to be? I see no indication at all that current is required to maintain a static C in the varactor. Current exists in the static case because the varactor is not a perfect diode.
Say the bias is 2 V and 1 pA, from some electrostatic V source. The bias is raised to 10 V and 5 pA (assuming a linear relation between the two).  In this case, as far as I can see, the C change will happen in the same way as if the bias began at 2 V and 1 nA, etc. Sure, a current flows, but the power consumed is not the same. That's why I latched onto the electret idea.
I got out my old book "Varactor Applications" and there is not one word about energy supplied by the pump, except to say that all the energy is supplied by the pump! No discussion of conversion efficiency at all. In these old para. amplifiers, used mostly for radio astronomy and the like, they were concerned about noise, not power, so a lot of this information is not too useful when it comes to power balance. Like I said, aside from your two formulae, there is no discussion in the literature of what actually happens in terms of power. I still consider it an open question.

Perhaps Partzman can simulate a very low current reverse bias, maybe using an FET, to see if I am right or wrong about using such a source to alter the C?

F.

[partzman]

Perhaps Partzman can simulate a very low current reverse bias, maybe using an FET, to see if I am right or wrong about using such a source to alter the C?

F.

Fred,

I have done many simulations with various non-linear caps including mosfets and as F6FLT stated, the energy required to charge a given device is always more than can be recovered.  At this point, I have found no exceptions but this is not to say it is impossible.

For example, there may be a possibility this can be done with BiCmos devices as described by Tsividis but the source and drain must be separate from the substrate and a connection to the substrate must be available.

His configuration allows a parametric amplifier to be built with such devices which would be equivalent to moving plates of a physical capacitor apart.  The question would be if his boost voltage would require more energy than would be available from the increase voltage at the gate.  I have not found a model to try in LtSpice nor have I tried any bench tests so I have no idea whether this is possible or not.  I can say that standard depletion type mosfets do not work!

I included one of his papers below.

Regards,
Pm

[Orthofield]

Hi Partzman,

Yes, we've certainly discussed this before. But you may have possibly misunderstood what I was asking, because of my use of the term 'FET'. I'm thinking of a hypothetical source that limits bias current to a varactor to less than the datasheet's leakage current. For instance, say the varactor bias has a leakage current of 20 nA. This hypothetical gadget would supply way less than this current, say 20 pA in the 1-10 V range. The question then becomes, does this very low current still allow the varactor C to be changed, as bias V is raised?

F.

[partzman]

Hi Partzman,

Yes, we've certainly discussed this before. But you may have possibly misunderstood what I was asking, because of my use of the term 'FET'. I'm thinking of a hypothetical source that limits bias current to a varactor to less than the datasheet's leakage current. For instance, say the varactor bias has a leakage current of 20 nA. This hypothetical gadget would supply way less than this current, say 20 pA in the 1-10 V range. The question then becomes, does this very low current still allow the varactor C to be changed, as bias V is raised?

F.

Ah yes, I misunderstood.  I wonder if you charge at a current below the leakage current if you would have any charging of the device but maybe I'm still missing something.  I have tried adiabatic charging of non-linear caps and at best it results in a COP ~1.

Regards,
Pm

[F6FLT]

Hi Partzman,

Yes, we've certainly discussed this before. But you may have possibly misunderstood what I was asking, because of my use of the term 'FET'. I'm thinking of a hypothetical source that limits bias current to a varactor to less than the datasheet's leakage current. For instance, say the varactor bias has a leakage current of 20 nA. This hypothetical gadget would supply way less than this current, say 20 pA in the 1-10 V range. The question then becomes, does this very low current still allow the varactor C to be changed, as bias V is raised?

F.

The polarization current is a question of a not perfect varactor, not a question of operating principle, it is a pure and simple loss of energy, in addition to that at the origin of the change of C and which will be due to the current already mentioned i(t) = dQ/dt = d(C(t).U(t))/dt. Indeed, the electrical charge being conserved, if the one retained in a capacitor changes, it is because it is input or output from the capacitor, so it is a current.
There is no dynamic current if and only if C and U are covariant in the opposite direction. This is the case when the plates of a capacitor are moved away from each other: U increases proportionally as C decreases. The energy is proportional to U² so it increases. In this case, it is the mechanical energy of separating the plates that is transformed into electrical energy, and the charge is conserved.
In the case where the variation of C is obtained by modifying U, we always have this work of separating the charges of the 2 plates, this follows from the Coulomb law. Here this work is obtained from U, we will consume a power U(t).i(t) = U(t) . d(C(t).U(t))/dt.

For this reason, even if the static bias current were zero, or limited, we would still have COP <= 1.
The trick of a parametric system is to tap a free energy to change the parameter (heat, ZPE, nuclear, electronic spin...). If we use the same energy as the one we want to produce, in this case electrical energy, and on the same port as where we want to recover it (a capacitor electrode), we go round in circles.

[Orthofield]

Hi Partzman, F6LT,

On reading both your replies, I suddenly got it. To admit my ignorance is hard, but it's the right way. I assumed that the back bias didn't charge the varactor. I forgot about the polarization or displacement current. Simple and obvious but somehow I missed it, through all the years of working on this stuff.

I've realized that my own overunity ideas are not particularly good, and taking people's time. On reviewing where I can do some good, it returns to simply notifying when I've found an interesting patent or paper, so I'm going to stick to that, and stop pitching my ideas.

I also want to broadcast the Dahlberg and some thermoelectric patents to other forums devoted to these subjects, since they deserve wider awareness. I need to write up some articles on them that will get wide distribution. So those are the two areas I will continue to move forward.

Fred