Simple circuit prevents processor latch-up

Michal Kobylecki and Wojda Wlodzimierz, MK Design, Warsaw, Poland

The circuit in Figure 1 uses a Dallas Semiconductor DS1820 one-wire digital thermometer in a multipoint temperature-measurement system. The DS1820 sensor allows distributed temperature measurement and uses only one wire for both data communication and the power supply. You can easily connect the one-wire interface to a µC (in this example, a PIC16C63). The application accommodates as many as 16 DS1820 sensors in a 100m-long network, dubbed MicroLAN by Dallas. Unfortunately, with such long wires near high-current power cables, inductively coupled high-voltage peaks can cause µC latch-up, because the transient-voltage suppressor, D₁, cannot limit the line voltage below 9.8V. The interface circuit in Figure 1 prevents such faults.

The sensor needs only one data line, but the interface circuit in Figure 1 needs two CPU-control signals. The first is the inverted-output data line, connected to transistor Q₁ through resistor R₃. If this line is at logic 1, the sensor's data line connects to ground. Otherwise, resistor R₄ pulls the sensor's data line to 5V. If the sensor transmits data, the µC captures the data via R₂. If an overvoltage peak occurs, R₂ prevents µC latch-up by limiting the current to the µC pin. During the temperature conversion, the DS1820 needs an effective pullup circuit to provide an accurate conversion. Transistors Q₁, Q₂, and R₁ provide an effective pullup function, which the µC activates by grounding the input signal to Q₃.

The pullup function also improves the rising edge of the transmitted data. If the sensor connects to the µC through a long wire, the capacitance of the line degrades the rising edge of the data signals, because of the line-capacitance-R₄ time constant. The short pulse from the pullup circuit improves the signal's rising edge. It's obvious that, when the pullup circuit turns on, Q₃ must be off. However, the circuit protects itself against unintended signal corruption. If Q₃ is open, the power voltage does not connect to the sensor's data line. If you replace R₄ with a 1-mA current source, you can obtain effective data transmission with lengths as great as 500m (empirically verified). (DI #2317).

Dual one-shot makes rising- and falling-edge detector

Santo Camonita, Catania, Italy

The design in Figure 1 is an upgraded version of the circuit in a previous Design Idea (" Edge detector runs off single supply ," EDN, Dec 4, 1997, pg 140). It has fewer components, draws less current, and has higher input impedance. The circuit uses a 4098 dual monostable multivibrator with both sections connected. The circuit generates a pulse on both the rising and the falling edges of a signal. The duration of the output pulse is T=0.5R₂C₂. R₁ and C₁ provide power-on reset. The circuit has another advantage: It provides two independent true outputs (Q1, Pin 6 and Q2, Pin 10) and two independent complementary outputs (, Pin 7 and , Pin 9). (DI #2316).

Simple circuit measures diesel's rotations per minute

David Magliocco, CDPI, Scientrier, France

You may find it useful to measure a diesel engine's rotations per minute to accurately adjust the idling or to compare the motor's speed under hot and cold conditions, for example. Not all cars or trucks come with a tachometer. The scheme in Figures 1 and 2 allows you to measure rotations per minute with a DMM or an oscilloscope. In Figure 1 , a piezoelectric sensor and an alligator clip, fastened directly to one of the four metal fuel pipes, detect the fuel-injection pulses. The piezo element generates a signal that connects to the signal-conditioning electronics in Figure 2 through a coaxial cable. The circuit uses a charge amplifier.

IC₁, a versatile ICL7611 chosen for its high input impedance, low bias current, and reasonable power consumption, acts as an inverting current-to-voltage converter. C₁ integrates the high-dV/dt sensor signal. R₁ ensures that the output of IC₂ is high in the absence of a signal. You must use high values for R₂, R₃, and R₄. The two diodes at the input protect the amplifier against overvoltage spikes. R₅ and R₆ create a virtual ground, so the circuit uses the full common-mode input range of IC₁. Figure 3 shows an oscilloscope trace of the amplifier's output. The three small peaks between the main pulses are parasitic and come from the injection pulses of the other cylinders. You can fine-tune the shape of the signal with C₁. A value near 1 nF yields a smooth signal with low amplitude; a value lower than 100 pF gives a noisy signal with narrow pulses. The value for the trace in Figure 3 is 100 pF.

From one vehicle to another, the amplitude of the signal can vary over an approximately 2-to-1 ratio. You can't use a fixed threshold to shape the signal; either the threshold is too low, and you shape the parasitic pulses, or it's too high, and you obtain nothing. Thus, the signal-conditioning method compares the peak value to the average value: Only the main peaks are higher than the average value. C₂ and R₇ ignore the dc component of the charge amplifier's output signal, and transistors Q₁ and Q₂ charge C₃ to the peak value of that signal. You adjust R₈ and R₉ for a 90% ratio, which gives IC₂, a low-power CMOS comparator, a large enough signal to shape. IC₂ delivers a logiclike signal that you can measure with a DMM (using the frequency/period range, or the rotations-per-minute range on an automotive DMM), or with an oscilloscope. (DI #2318).

Fast, compact routine interfaces EEPROM to µC

Grzegorz Mazur, Institute of Computer Science, Nowowiejska, Poland

The code in Listing 1 provides the interface between any MCS-51 family of µCs and a 24C01a/2/4/8/16 I²C EEPROM. The interface is purely software-driven, so any µC port pins can control the EEPROM. This code is approximately two times smaller and approximately 20% faster than similar routines published by Atmel Corp ( www.atmel.com). Careful coding of low-level routines and structural optimization produce the performance improvements.

The code is tuned for a -51 running at 12 MHz, but you can also use it at lower clock speeds. For higher speeds, you must adjust low-level routines by inserting "nop" instructions to match I²C-specified timings. The maximum data rate while reading memory content as 16-byte blocks is as high as 8.1 kbytes/sec. With minor modifications, you can use the routines to interface any I²C slave device to a -51 µC (DI #2329).

Photodetector sorts objects

Alan Erzinger, Harris Semiconductor, Palm Bay, FL

Most object-sensing systemshave problems detecting the presence of an object. The system in Figure 1 uses an oscillator to ease detection problems and allow sorting. The oscillator reduces power dissipation in the photodiode by operating the diode at 50% duty cycle. The oscillator also enables a 50% increase in the noise-filter time constant, and it functions as a timebase to allow object sorting. The oscillator chops the photodiode's bias. The signal the photodetector receives is a square wave; thus, a filter can remove optical noise. You normally need filtering when light shines through fans onto the optical detector. When required, you can place a bandpass filter in series with Q₃'s output.

The oscillator frequency has two limits: the response time of the phototransistor and the accuracy of the object-sensing system. Q₄ and Q₃ are connected in a cascode configuration to minimize the Miller effect in Q₄. This connection reduces the pair's optical transient response to nanoseconds, thus allowing oscillator periods of submicroseconds. Objects on a conveyor belt travel at relatively low speeds. You can calculate their expected time in front of the photodetector, t, from t=d/s, where s is the conveyor speed in feet per second and d is the object width in feet.< /FONT>

An object 2 in. long with a belt speed of 3.5 ft/sec blocks the detector for 47.6 msec. If the oscillator period is 250 µsec, the object blocks the detector for approximately 200 periods, so each period equates to 0.5% length accuracy. The system senses two objects—one 2 in. long and one much longer—so 0.5% accuracy is more than adequate. When the detector is unblocked, the inverting input of IC₁ is dominant, and it keeps the output of IC₁ low. The low state prevents the oscillator's output from reaching the counter (IC₃). Blocking the detector allows C₃ to charge to 5V through R₆, and the noninverting input of IC₁ becomes dominant, starting the count. When the detector is unblocked, the one-shot comprising R₁₁, C₅, and IC_4A pulses IC_2B with an end-of-object pulse. The one-shot's trailing edge triggers the counter-reset one-shot comprising R₁₂, C₆, and IC_4C.

C₄ and R₆ form a nuisance filter that rejects short optical noise. The timing is such to enable a 2-in. object to clear the detector while both Q₆ and Q₇ are high. When the end-of-object pulse coincides with Q₆ high, Q₇ high, and Q₈ low, the output of IC_2B indicates a 2-in. object. This situation is the only time window that can indicate a 2-in. object. If the object is longer than 2 in., Q₈ goes high, indicating a large object. When the object clears the detector, the reset one-shot resets the counter for another cycle, and Q₃ quickly discharges C₄ in preparation for another cycle. With the component values shown, the system can sense and discriminate between objects as short as 2 in., separated by 0.1 in. (DI #2325).