Why Does My Varactor Diode Test Circuit Behave Differently on a Breadboard?

[piepermd]

TL;DR Summary

I have an oscillator circuit for finding the resonant frequency of LC circuits. It works in LTSPICE incorporating Varactor diodes, but does not behave the same way on the breadboard.

In the circuit simulator there is no voltage at the Varactor cathodes when Vbias is zero, but depending on the Varactor I use on the breadboard it can be as high as 7.4 volts! There is minimal current through the series resistor when Vbias is applied in LTSPICE, but I always see a significant voltage drop across the series resistor for Vbias on the breadboard, indicating there is indeed current flowing. This is not supposed to happen unless the breakdown voltage is exceeded. Any ideas why the circuit does not behave as expected when built on the breadboard? I have tried this in different iterations with different breadboards and I always see the same result. I tried building a similar circuit on a perf board and still got a voltage drop across the series Vbias resistor.
Here is a JPG but I’ll try to send the file as well.

 

[Baluncore]

Welcome to PF.

Here is a JPG but I’ll try to send the file as well.

We need the "circuit.asc" file, that is ascii text on the inside, to get the details which are not shown on the screen, or in the "file.jpg".
Rename the "circuit.asc" file to "circuit.txt" or preferably to "circuit.asc.txt" so it can be attached to a post. We can then download it, rename it, and run it.

 

[piepermd]

Thanks, I was wondering why the .asc file was not allowed. Here's the file.

 

[piepermd]

Just in case I was not clear, the behavior in LTSPICE is the "correct" behavior, and the behavior on the breadboard is not.

 

[Baluncore]

Less space between components, with shorter wires, makes for a more readable image.
Use menu; Tools, CopyBitmapToClipboard rather than screen copy.

The J310 model is missing from my version. Link to the model you used?
The Osc_out label was not connected.
1MEG in spice is 1000000, so it saves counting zeros. 1m0 or 1M0 is 1.0 milli.
It is clearer if you can use k, m, u, n or p as a decimal point. 2k7, 0u1 = 100n

There is minimal current through the series resistor when Vbias is applied in LTSPICE, but I always see a significant voltage drop across the series resistor for Vbias on the breadboard, indicating there is indeed current flowing.

How are you measuring the varicap bias voltage on the circuit? The 1 Meg, R1 resistor could be reduced to 100k which would reduce the effect of leakage currents on a dirty or damp breadboard. A breadboard also has high capacitance between strips, which reduces the range of small varicaps. Maybe try connecting the varicaps and bias isolation resistor in the air, above the breadboard.
Maybe your FET has a leaky gate ?
Maybe a varicap has died, become leaky, or is backward ?

 

[piepermd]

For some reason, I get an error message when I use MEG. I’m measuring the voltage with a multimeter. The bias voltage is produced by an Arduino with the MCP4728 device from Adafruit. I have tried lower values on the isolation resistor but no joy. I have tried three different Varactors with the same results. I am able to see the expected increase in frequency with increasing bias voltage when using the NTE618 varicap because the cathode voltage is low with no bias applied. With the MV209 varicap the unbiased voltage is so high than when bias is applied the frequency goes down! I can only supply 4V of bias, but this is all I need for my ultimate application. I have checked the orientation of the varicaps innumerable times. I can try swapping out the J310s. Thanks for your suggestions thus far!

 

[tech99]

The varactors must not conduct on RF peaks. so the reverse DC bias must be more that the peak RF voltage. Maybe tap the varactors down the tuned circuit.

 

[Baluncore]

I do not think the FET is the problem, it is not connected to the bias node, which is where the problem appears.

The current through R1 must be flowing somewhere. Don't forget that your meter measuring the voltage across R1 will change things. (It probably looks like 10meg in parallel).

Funny things can happen when an RF oscillator is running nearby. You are looking for DC leakage first, so kill the oscillator.

Build just the R1 with the two varicaps and L1, without the breadboard or oscillator. Control the bias and find the few microamps of current leakage. Maybe change R1 to 100k or even 10k, then look again for a bigger leak. Maybe you used 1 MEG with a space, which will flag an error with LTspice, 1.MEG or 1MEG will not. 1MEG = 1000k = 1e6 = 0G001

Some versions of spice have differentiated between 1M0 and 1m0, but that leads to confusion. MEG is the spice standard.

 

[tech99]

There might be an issue in measuring and applying the bias. If you apply the multimeter where it says"bias" then it requires a high series resistor and a bypass cap. The bias supply shown will require bypassing to ground. be careful about RF on the bias and multimeter circuitry.

 

[Baluncore]

The varactors must not conduct on RF peaks. so the reverse DC bias must be more that the peak RF voltage.

As presented, there is too much energy circulating in this oscillator, ±7 Vpeak. I changed R4 to 2k2 to reduce the gain. I reduced the supply from 18 V to 5 V. Also, I am using 2N5484 FETs from the LTspice standard library. All that has reduced the RF across the varicaps to below ±1.5 Vpeak.

The bias supply shown will require bypassing to ground.

The 1 meg resistor should isolate the RF on the bias supply, but yes, a capacitor between ground and the bias control voltage source will reduce radiated RF, and control voltage noise.

 

[piepermd]

The varactors must not conduct on RF peaks. so the reverse DC bias must be more that the peak RF voltage. Maybe tap the varactors down the tuned circuit.

My understanding is that the “back to back” configuration of the Varactors cancels the effect of RF peaks, but I could be wrong.

 

[berkeman]

Can you post a picture of your circuit that you built? Since the varactor diode capacitance is so small (low pF probably), any stray capacitance in that part of the circuit dulls the tuning capability of the varactor diodes. If you have 10pF of stray capacitance in parallel with the varactors, the circuit will likely not work at all.

 

[Baluncore]

My understanding is that the “back to back” configuration of the Varactors cancels the effect of RF peaks, but I could be wrong.

The two reverse biased diodes are capacitors. The capacitors are in series, in a tuned circuit. But the voltage across the capacitors is changing throughout the oscillator cycle. The capacitance parameter is therefore also changing throughout the cycle. By using two diodes, back to back, the parametric distortion of the oscillation is made symmetrical about zero. That reduces the even harmonic distortion of the oscillator.

The capacitance change dc/dv is a maximum at low reverse bias voltages. The parametric distortion is therefore a maximum for low bias voltages near the transition to forward bias. If the diodes become alternately forward biased, charge will be pumped from the tuned circuit into the control voltage network, that results in loss of frequency control. Avoid forward conduction of the diodes.

 

[piepermd]

Thank you very much, that is a more thorough explanation than I have ever read! Is it possible that forward biasing by RF at zero DC bias is causing a voltage to develop at the common cathode that could be shown by the multimeter even on its DC voltage setting?

 

[berkeman]

Thank you very much, that is a more thorough explanation than I have ever read! Is it possible that forward biasing by RF at zero DC bias is causing a voltage to develop at the common cathode that could be shown by the multimeter even on its DC voltage setting?

Do you know how much capacitance your multimeter adds to the varactors when you are making measurements...?

 

[Baluncore]

Is it possible that forward biasing by RF at zero DC bias is causing a voltage to develop at the common cathode that could be shown by the multimeter even on its DC voltage setting?

I believe that forward bias of the varactor diodes by the oscillation, would pump the frequency control voltage up, not down as you are seeing.

 

[piepermd]

I believe that forward bias of the varactor diodes by the oscillation, would pump the frequency control voltage up, not down as you are seeing.

I’m seeing a voltage at the cathodes of the varactors when the DC bias is zero.

 

[Baluncore]

I’m seeing a voltage at the cathodes of the varactors when the DC bias is zero.

That is due to forward conduction of the varactors. Any AC at the cathodes that falls below -0.6 volts will make the cathodes appear more positive.
You should not be operating that circuit with zero voltage, or with such a large AC signal on the tuned circuit.

 

[piepermd]

That is due to forward conduction of the varactors. Any AC at the cathodes that falls below -0.6 volts will make the cathodes appear more positive.
You should not be operating that circuit with zero voltage, or with such a large AC signal on the tuned circuit.

I wonder why LTSPICE shows no voltage at the cathodes at zero DC bias. Perhaps it is just one of the many limitations of the circuit simulator. Can an AC voltage show up on a multimeter on the DCV setting under certain circumstances?

 

[piepermd]

Do you know how much capacitance your multimeter adds to the varactors when you are making measurements...?

I’ll have to look it up

 

[berkeman]

I’ll have to look it up

Well it's way more than the varactor capacitance. So you can measure some of the DC voltages in the circuit with your DVM, but using it during RF circuit operation will likely change things a fair amount. BTW, were you going to upload a picture of your prototype circuit at some point? (use "Attach files" below the Edit window) Thanks.

 

[piepermd]

Well it's way more than the varactor capacitance. So you can measure some of the DC voltages in the circuit with your DVM, but using it during RF circuit operation will likely change things a fair amount. BTW, were you going to upload a picture of your prototype circuit at some point? (use "Attach files" below the Edit window) Thanks.

I can do it tonight, thanks!

 

[piepermd]

I wonder why LTSPICE shows no voltage at the cathodes at zero DC bias. Perhaps it is just one of the many limitations of the circuit simulator. Can an AC voltage show up on a multimeter on the DCV setting under certain circumstances?

Now that I think of it, the simulation on LTSPICE showing the voltage at all the nodes only shows DC voltage…

 

[Baluncore]

Now that I think of it, the simulation on LTSPICE showing the voltage at all the nodes only shows DC voltage…

When you place the cursor on a node, LTspice gives you the DC operating point, the voltages and currents used at the start of a simulation. Change the transient startup parameters "Skip Initial operating point solution:" or "Start external DC supply voltages at 0V:" to see the difference.

You need an RC LPF to measure the cathode voltage in the transient simulation. The LC oscillator amplitude is also attenuated by forward conduction of the diodes.

 

[piepermd]

Well it's way more than the varactor capacitance. So you can measure some of the DC voltages in the circuit with your DVM, but using it during RF circuit operation will likely change things a fair amount. BTW, were you going to upload a picture of your prototype circuit at some point? (use "Attach files" below the Edit window) Thanks.

Here’s the promised photo

 

[berkeman]

Here’s the promised photo

That is not a low-capacitance way to breadboard this varactor circuit.

The capacitance in parallel with the varactors must be very small in order for the circuit to work (preferably 0.1pF or less). That can only be achieved with a careful PCB design, or for first prototype you can use directly-soldered connections ("flying bug" style). Keep the leads short (trim those TO-92 leads), and think about where the parallel capacitance is coming from (and minimize it).

 

[piepermd]

That is not a low-capacitance way to breadboard this varactor circuit.

The capacitance in parallel with the varactors must be very small in order for the circuit to work (preferably 0.1pF or less). That can only be achieved with a careful PCB design, or for first prototype you can use directly-soldered connections ("flying bug" style). Keep the leads short (trim those TO-92 leads), and think about where the parallel capacitance is coming from (and minimize it).

Flying bug style with just the LC circuit still gives me a .035V drop across a 100K series resistor with 4V of reverse bias applied, indicating a .35 micro amp current is flowing. The maximum reverse voltage leakage current for the MV209 varicap is listed as .1 uA, and while the data sheet doesn’t say so, the leakage current is usually measured just shy of breakdown voltage which is 30V for this device. So I don’t think I should be seeing anywhere near this much current at 4V. I have seen unexpected leakage current with three different types of varicaps, and series resistance ranging from 10K to 1MEG. Are all my varicaps out of spec?

 

[Baluncore]

Flying bug style with just the LC circuit still gives me a .035V drop across a 100K series resistor with 4V of reverse bias applied, indicating a .35 micro amp current is flowing.

Is the oscillator running with C5 = 30 pF in place, while you are making that measurement ?

If stand alone without oscillation, disconnect a parallel diode or component to identify the path of the leakage current. The datasheet on the KV1471 specifies the maximum reverse current as less than 50 nA.

The oscillator power supply voltage was too high. Ignoring any resonance peak, high RF drive through C5 =30 pF, (±9 V = 18 Vpp) may have damaged the diodes.

 

[piepermd]

The circuit I just tested is just the inductor and the varicaps.

 

[piepermd]

And I get the exact same numbers with a pair of NTE618 varicaps fresh out of the box. The others were MV209. The KV1471 devices were just for the simulation, because there are only two choices for varicaps native to the LTSPICE library. If anything I would think the parallel resistance of the multimeter would make the voltage drop appear slightly less than it really is.

 

[piepermd]

With a 9V battery and the NTE618’s I get .8uA of current, or 800nA. Per the data sheet for this device, reverse current at 9V is only 100nA. Recall that the observed reverse current at 4V was 300nA. My ultimate use of the varicaps is for adapting a WWII aircraft radio to electronic tuning. It relies on delivering a very precise voltage to the circuit over a fairly narrow range. My findings are making it difficult to predict the voltage actually delivered at the bias point because of the unexpectedly high reverse current that is causing a significant drop in voltage across the bias resistor, a voltage drop that varies with the supplied voltage.

 

[Tom.G]

Referring to the photo in post #27, try twisting all the leads together. This will greatly reduce stray pickup from the environment.

Three additional possibilities I can think of off-hand:
Using the physical setup in the photo of post #27, check the following:

1. Meter is not Zeroed; does it read zero with the leads shorted?
2. Stray pickup from the environment; nearby fluorescent or LED lamps often use switching regulators that can be very noisy. Even the old style 4-foot long fluorescent ceiling lamps with a ballast will radiate spikes from the several-hundred volt arc ignition voltage.
3. There may be ripple or oscillation coming from the power supply you are using. Try putting a 0.1uF disc ceramic capacitor across its output terminals.

You can also put as much of the circuit as possible in an oven that has a window so you can still read the meter. A Microwave oven has better shielding, but a kitchen stove oven has a fair amount. (Don't turn on the oven though. )

A photo showing complete equipment layout of the setup from post 27 may give us some more clues.

Cheers,
Tom

 

[Baluncore]

My ultimate use of the varicaps is for adapting a WWII aircraft radio to electronic tuning.

That is quite a challenge.
What is the radio make and model number ?
What is the frequency range of the LO ?
Are you replacing the LO only, or are you also tuning some other RF circuits ?

 

[piepermd]

That is quite a challenge.
What is the radio make and model number ?
What is the frequency range of the LO ?
Are you replacing the LO only, or are you also tuning some other RF circuits ?

It’s the SCR-522 from the P-51 Mustang. The LO range will be 106-126 MHz for civil aviation. I am tuning 3 RF circuits as well, up to the mixer stage. The desired frequency is selected on a microcontroller which then calculates the bias voltage needed to create the requisite capacitance for resonance at that frequency minus the IF. The old ratchet style tuning with crystal oscillators for four channels is missing from my radio, no parts are available, and no other such radios are available in the world. Without the ratchet system it would not be possible to have four preset channels available for the pilot to select, but it can be done with electronic tuning.

 

[Baluncore]

It’s the SCR-522 from the P-51 Mustang. The LO range will be 106-126 MHz for civil aviation. I am tuning 3 RF circuits as well, up to the mixer stage.

It looks to me like it has RF tuning only on RX, but has two separate switched four channel crystal oscillators, one for TX, the other for RX. It should be possible to replace the crystal oscillator(s) with digital synthesizer board(s). The harmonic multipliers that generate the TX carrier, and the LO for RX, can still be used.

There should be no problem using varactors to tune the small RF signals in the RX path, but you can expect problems generating the higher amplitude RF needed to replace the crystal oscillators. For those, you should look at synthesizer ICs that include a current controlled CMOS oscillator inside a PLL.