Page 2. Femto Capacitance Meter
A/D Converters.

(Nowdays, digital panel meters are available, which eliminates the need to interface a display to a voltmeter chip.)

The prototypes use ADD3701 for 4,000 counts. Any of the variety of display driving A/D converters could be used. The 3701 will drive either LCD or LED displays. LCDs are easier to see under varied lighting conditions; and the low current is advantageous. But if 4,000 counts are used, LCDs are more difficult to construct, requiring 6 additional chips. If 2,000 counts are used with an LCD, an ICL7106 can be used; and it drives the display directly. The prototype LCD meter uses 4,000 counts. The circuits for both 3701 and 7106 LCDs are shown.  

Figure 2. shows the analog circuits for the 3701. The chip measures to ground with a 5V supply and has a gain of 2; so 2V input produces 4,000 counts. The reference voltage, going into pin 18, is adjusted to about 2.010V. The adjustment range shown is the same as the specified range (+29mV, -13mV, relative to 2V); but it is too touchy; so two trimmers might be used. The 50K trimmer is for offset adjust.

Each ground connection around the A/D is supposed to go directly to the ground hub to prevent ground currents and noises from producing errors.  

Figure 3. shows the digital circuits for the 3701 with an LCD. A square wave of about 44Hz is needed on the display. It is created by a 74C86 exclusive OR chip and is distributed by 4543 drivers. The A/D determines which driver functions through pins 21, 22, 23 and 24, which go to the latch disable pins.

Two phases are used with the square wave. One phase (pin 3) goes to the back plane of the LCD (pin 1); and the opposite phase goes to each segment which shows. The 4543s pick up the square wave in the same phase as the back plane (entering at pins 6).

The decimals are driven by the 10 pin of the 74C86, which is always out of phase with the back plane. However, there are 100K resistors which connect the decimals to the back plane; and they hold the decimals off, unless the rotary switch overrides the resistors with the opposite phase.

All of the segments which are continuously blanked are connected to the back plane. Segments will not blank if left unconnected. They must be in phase with the back plane to blank.

Overflow blanks 4 digits and shows a 1 at the left. The decimal stays on during overflow. The overflow signal from pin 7 of the A/D goes to the blanking inputs (7 pins) of the 4543s and to the XOR for showing the 1 by changing the phase of pin 3 of the LCD.
The 4,000 count LCD is an earlier Digi-Key part number LCD004, which is no longer extant but should be the same as more recent ones. The pinout is shown in Figure 4. along with the connections, which mostly go to the 4543s.
The LCD for ICL7106 is an earlier Digi-Key part number LCD002. It is identical in pinout to the one above except without the MSD (most significant digit); and the corresponding pins are not connected. The pinout for LCD002 is shown in Figure 5.  

Figure 6. shows the analog circuits for the ICL7106 with an LCD. The chip needs a negative voltage of at least -2.5V; so it uses the available -8. 5V. It has no offset voltage; and there is no device variation on the reference voltage. So the reference voltage is set at exactly 0.500V. Twice that amount, or 1V, is the amount of input voltage that reads full scale. Since full scale is 2,000 counts, there is an effective voltage gain of 2, as required.

Either an external or internal reference voltage can be used with the 7106, as shown. The external reference would be easier to use, since it is already available. The internal reference is based upon a zener diode which sets the 32 pin at 2.8V below V+. That reference is low in drift; but it has a wide tolerance; so it should be measured first and the resistor values tailored to the measurement.

The large capacitors can all be metal film except the one for integrating (pin 27), which should be polypropylene.
Figure 7. shows the digital circuits for 3701 with an LED display. The LED is 4 digit multiplexed. The digit select is buffered and inverted with an interface chip. The segment select pins have 18 ohm resistors to limit the current to 4OmA. Since each segment is on 25% of the time, the average current for each is 10mA. The cathode of each decimal point is switched to ground, while a 300 ohm resistor at the anode limits the current to 10mA.

Overflow is handled internally showing OFL; so the 7 pin of the 3701 is not connected.
Measuring Reference Capacitors.

Reference capacitors are the two on op amp B (100nF and 1.00nF).

Figure 8. shows a circuit for measuring reference capacitors by using a timer, which a frequency counter should have. One microsecond of resolution is desirable. When calibrated, the circuit is a complete capacitance meter. But there is no need to calibrate it for this purpose. The critical values are measured and calculated as odd numbers.

The first part of the circuit is the same precision triangle wave generator that the femto capacitance meter uses. What would be the reference capacitor is in this case the test capacitor. The resistor which determines the current flowing into the capacitor, Ra, is soldered on from the top; so it can be changed. If it is 1M, it produces about 4,000 ÁS for the 1nF capacitor. The 100nF cap produces 400mS; so Ra could be smaller.

Two comparators are used for upper and lower thresholds, with about 10V between them. That voltage needs to be precisely measured to determine the volts per second.

There is about 5mV of hysteresis on each comparator for stability at low frequencies. To calculate it, there is 5V across the 270K feedback resistor before the change of state. The resulting current through the 270 ohm resistor produces 5mV. For the lower comparator, there is 10V across the feedback resistor prior to the change of state. The hysteresis for the upper comparator is subtracted from the reference voltage; and for the lower one it is added. So the effects cancel; and the quantities can be ignored.

The AND gates at the output have the purpose of preventing an occurrence during the unused half of the cycle. The voltage going in is limited to 5V with a diode network.

The circuit could be set up on a breadboard; but for high precision, it should be soldered into an etched board.

The circuit picks up 60 cycle hum easily at the high impedance level; so it may need to be shielded. To do that, use a piece of tin or aluminum with four sides bent up to form an open box. Put paper in the bottom to insulate; ground it; and put the board in. Run the wires over the top. The top does not have to be covered.

The capacitors could be soldered onto the board; but I use insertion pins and solder them onto the board. Shielding between them is needed, if small capacitors are measured.

Add 3pF to the value of the 1nF capacitor, because the cross capacitance of op-amp B will be about that much.
Alternate Reference Circuit.

If you don't have a frequency counter with a timer, you can use a crystal oscillator for measuring the reference capacitors, as shown in Figure 9. This procedure is not as precise as the previous one. A breadboard can be used.

The capacitor being measured is in the feedback loop of an op-amp. A reference voltage is pulsed into the noninverting input, which causes the output voltage to increase at a fixed rate based upon the current flowing through the 1.00M resistor. The resistance can be an odd value; but it needs to be precisely measured. The reference voltage is the voltage across that resistor. After the pulse, the output voltage stays high and is measured with a meter. The output voltage divided by the time of the pulse determines the volts per second.

The crystal oscillator will have enough accuracy without measurement. Its signal is reduced to the appropriate frequency and goes into the clock input (3 pin) of a flip flop which functions as a pulser. The output pulse will have a time duration of exactly one duty cycle of the input frequency. For the 100nF capacitor, the input frequency is 16Hz producing a pulse of 0.0625 seconds. For the 1nF capacitor, the frequency is 256Hz producing a pulse of 3.906mS.

The 6 pin of the flip flop produces an up pulse which opens the line to the reference voltage through a CMOS switch (4066). The 5 pin of the flip flop produces a down pulse which closes the line to ground.

The offset voltage must be thoroughly trimmed out to prevent a slide in output voltage due to current flowing through the input resistor. For the 100nF capacitor, that is not difficult to do; but for the 1nF capacitor, it becomes a little problematic; and more refined offset trimming might be needed.

A momentary switch removes the voltage across the capacitor for repeating the measurement. When the 1nF capacitor is being measure, a 1M resistor is used with the reset switch to remove contact bounce. It is omitted for the 100nF capacitor.

With the 100nF capacitor, the output voltage will be about 1.5V. That measurement will have a potential accuracy of about 0.2%. The 1nF capacitor will produce an output voltage of about 9.6V; and its measurement is consistently high by about 1% or 10pF. I don't know why. Theoretically, the reason would seem to be input capacitance, since there is a common mode voltage increase; but adding capacitance makes little difference indicating otherwise. At any rate, subtracting 1% from the measurement of the 1nF capacitor should produce a useable quantity.

Momentary switches that work good on a breadboard are the flat switches or post switches. Twenty two gage wires are soldered onto the pins for breadboard use.
The Box.

The box is designed for convenience of use. That means an upward sloping front panel with everything on it, as shown in Figure 10. Two boards are used, attached to the top panel, so the switches can run across the center, while insertion pins are at the bottom, and the display is at the top. Everything but the switches is attached to the boards. The power supply, however, is on the bottom of the box.

Of course you don't buy a box shaped like that; it is bent out of aluminum sheet metal, which is available from heating and cooling shops or metal supply shops. The preferred thickness is 0.040 inches (1.0 mm). A detailed box pattern is in Figure 11. Markings are shown in Figure 12.

Toggles are the preferred switches. Toward the upper left is the on-off switch for ac line voltage. There are 20cm wires going to it from the bottom of the box. They are allowed to coil up in the box. If they touch the reference capacitors, they can induce a few parts per thousand error, so an aluminum plate jutting out from the side is placed over them. If it is desirable not to have ac wires coiled in the box, the ac switch could go on the back panel, which is attached to the bottom. If LCD is used, the on-off switch could be eliminated leaving the meter on most of the time. But if LED is use, the meter should be turned off except when used, because heavy currents in the display will heat the box producing about 3ppt drift in a half hour.
Switch-2 is a three pole toggle located near the center between the rotary for stages and the potentiometer for zeroing.

The insertion pins
(see description on page 3.) are soldered onto the copper side of the board; and they project through holes in the aluminum. For the upper three ranges, two pins are used on the T-wave side, because some of the large capacitors have wide spaced leads.

If capacitors have short leads which will not fit in the pins, alligator clips can be used by soldering 20 gage wires onto the ends of flexible wires for insertion into the pins.

If banana jacks are desired for test leads, they could be attached to the board near the pins. A suitable jack for that is Radio Shack's 274-661. The plastic top can be removed leaving a small diameter jack which can be attached to the board without soldering. An extra attachment screw for the board should be located near the jacks.

If wires are used for remotely located jacks, they will be antennas for picking up transition noise from the comparator, which will require the comparator and its output wiring to be covered with grounded tin for shielding.

The momentary switch which pulls down over-voltage (called the set switch) is located on the lower board towards the right. The Panasonic post switch is good for that. It solders to the copper side of the board, which is upward during use.

On each side of the box (inside) is a small aluminum bracket with a hole for gathering wires. One is 2cm below the on-off switch for the ac wires; and the other is directly across from it for four power supply wires (▒8.5V, +5V, and ground). Each of those wires is about 20 to 30cm long, so the top of the box can be separated from the bottom, which has the power supply on it.

to page three | to page one