September 24, 1998

Bipolars provide stable current source

Bill Morong, Morong's Harness, Dover-Foxcroft, ME

It's possible to implement a precise current source with a useful output at high frequencies, without using operational amplifiers. The circuit in Figure 1a suffers inaccuracies from both the V_BE drop and the finite base current of the transistor. The circuit in Figure 1b overcomes the base-current problem, but has two V_BE drops and does not perform well at high frequencies. The circuit in Figure 1c has no base-current problem and performs well at high frequencies, but is prone to inaccuracies from the V_GS of the FET. The circuit in Figure 2 largely overcomes these problems.

The V_BE of Q₂ cancels that of Q₁. Because the base current of Q₁ diverts (via Q₃) as shown around the current-setting resistor R₁, I_OUT is simply two-thirds of V_IN divided by R₁. Because the circuit provides error cancellation, the values and voltages are not critical, provided you match the upper and lower components. In this example, Q₁ has a beta of 50; the circuit self-adjusts as beta varies. Neon-driver transistors are a good choice for Q₁ because Q₂ and Q₃ need neither good high-frequency performance nor high output impedance. High-beta (400 in Figure 2) transistors are appropriate, as they minimize errors. Insofar as possible, it is desirable to have similar, high betas for Q₂ and Q₃. At high frequencies, it may be beneficial to bypass the base of Q₁ to ground. (DI #2257).

8-pin µC forms one-chip programmable VCO

Yongping Xia, Teldata Inc, Los Angeles, CA

The circuit in Figure 1 uses a Microchip 8-pin µC (PIC12C671) as a voltage-controlled oscillator (VCO). Because the PIC12C671 has an internal 4-MHz oscillator, four-channel 8-bit A/D converters, and built-in power-reset circuitry, you need no external components to configure the VCO. The µC reads two analog inputs through AN0 and AN1. The reference voltage for the A/D conversion is the µC's power supply V_DD. The converted 8-bit data determines the duration of output high and output low. Assume, for example, the digitized outputs from AN0 and AN1 are 43 and 87, respectively. Timer 0 loads the 43 after the µC sets output GP2 to logic one. Timer 0 receives its timing from the internal clock.

Once Timer 0 times out, the µC sets GP2 to logic zero before loading 87 into Timer 0. When Timer 0 times out again, the program starts the loop again. Thus, the voltage on AN0 determines the output-high duration, and the voltage on AN1 determines the output-low duration. If a simple 50%-duty-cycle output is satisfactory, you can tie AN0 and AN1 together and control them with one source. You can further program the VCO's output frequency by using D0 through D2. The µC has a prescaler between its clock and Timer 0. Each time the µC reads AN0 and AN1, it also reads D0 through D2 and loads the reading to the prescaler.

Table 1 shows some test results. The frequency range is 8 Hz to 8 kHz. Listing 1 is the assembly code for the µC. Because the voltage-to-frequency relationship is purely software-controlled, you can alter the program or use a look-up table to obtain a desired v-f curve. (DI #2259).

Cable tester is fast and cheap

Abel Raynus, Armatron International, Melrose, MA

The cable tester in Figure 1 uses a low-end 8-bit µC. The specific µC to use depends on the number of conductors in the cable you want to test. For the current application, two types of cables were under test, one with three conductors and another with seven. So, the Motorola 68HC705P9 µC was suitable. The program first determines which type of cable is under test by checking the cable-switch position (Listing 1). Then, the program checks each conductor line from A₀/C₀ to A₇/C₇ by putting a high-level voltage from output port A on one end of the wire and measuring the response on the other end, which is connected to the input port C. If all of the checks show conductivity, the green "pass" LED turns on. In the opposite case, the red "fail" LED turns on. The test checks not only for conductivity but also for the presence of a cross connection.

If you want to test a variety of cables, you can use more switches. If a cable has more than eight conductors, you can use a different type of µC or multiplex the inputs.

You can download Listing 1 and an assembly-language program by clicking here. (DI #2225)

Capacitive sensor "likes" parasitics

Boris Khaykin, Candid Logic Inc, Madison Heights, MI

Stray capacitance is a common problem with capacitive sensors. The capacitance changes within the measurement range are normally much smaller than the strays; the result is a loss of sensitivity. Various methods are available to increase the relative sensitivity (Deltaf/f₀): frequency subtraction, the use of bridges, and the use of a negatron to subtract the strays, for example. The idea here is not to do battle with the stray, but rather use it and turn its drawbacks to your advantage. This method uses frequency-dependent hysteresis in a classic op-amp multivibrator. Figure 1 shows a simple, flexible design for a capacitive sensor.

Without capacitor C₂, the design is a classic multivibrator based on comparator IC₁ with output buffer IC₂. If R₁=R₂, the frequency is

R₁, R₂, and R₃ define the hysteresis, 900 mV with the values shown. Frequency (f) is a function of capacitor C₁, as Figure 2 shows. Without C₂ and with C₁=60 pF and DeltaC=20 pF, f₀=159 kHz and the relative sensitivity Deltaf/f₀ is -18%. With C₂ connected in parallel with R₂, the hysteresis becomes frequency-dependent. The capacitive reactance (X_C=1/2pfC) in parallel with R₂ reduces the hysteresis in an inverse proportion to the frequency. As a result, the frequency increases. This increase reduces X_C, further reduces the hysteresis, and leads to a further increase in frequency. Thus the relative sensitivity Deltaf/f₀ increases significantly (see Figure 2 with C₂=40 pF).

With C₂=40 pF, C₁=60 pF, and DeltaC=20 pF, f₀=945.5 kHz and the relative sensitivity (Deltaf/f₀) is -82%. The sensitivity (Deltaf/f₀) (38.6) in this case is 26 times as high as the case without C₂ (Deltaf/f₀=1.45). You can obtain even more interesting results by replacing C₂ with a sensing capacitor. If C₁=200 pF, changing the value of C₂ from 0 to 200 pF changes the hysteresis from 900 to 28 mV, and changes the frequency from 30 to 1300 kHz. Figure 3 shows output frequency (f) as a function of capacitance C₂. With C₂=100 pF and DeltaC=20 pF, f₀=145.2 kHz and the relative sensitivity (Deltaf/f₀) is +393%. Thus, the frequency is directly proportional to the capacitance.

As Figure 3 demonstrates, you can adjust the desired initial frequency with R₃, and the sensitivity with R₅. Note that the higher sensitivity in this example occurs with a significant stray capacitance (100 pF, for example). If the real sensor has lower initial capacitance (50 pF, for example) the simple addition of a 50-pF capacitor in parallel with the sensor increases the sensitivity. The sensor "likes" the stray capacitance as it produces frequency-dependent hysteresis that, in turn, provides higher sensitivity. You could also use the added capacitor for temperature compensation.

If you use an extremely fast op amp or comparator in this design, there is a certain value of C₂ for which the output frequency jumps up a few kHz with a hysteresis of 5 to 7 pF (Figure 4). This quirk is particularly useful in the design of super-sensitive capacitive switches. You can adjust the switching point with R₃ and/or a capacitor in parallel with C₂. You can adjust the hysteresis by using a small resistance connected in series with C₂. On the other hand, the use of a slower comparator linearizes the frequency-versus-capacitance characteristic. For example, test results show that with an LM319 comparator, R₃=200 kOhm, R₅=200 Ohm, and C₁=200 pF, the output frequency follows the empirical equation f=140+3.327(C₂-100) kHz with 3% nonlinearity within the range C₂=100 to 400 pF. (DI #2258).

| EDN Access | Feedback | Table of Contents |