|Home | Articles | Forum | Glossary | Books|
As discussed in Section 1, the most important thing you need for effective troubleshooting is your wits. In addition to those, however, you'll normally want to have some equipment. This section itemizes the equipment that is necessary for most general troubleshooting tasks; some you can buy off the shelf, and some you can build yourself.
Before you begin your troubleshooting task, you should know that the equipment you use has a direct bearing on the time and effort you must spend to get the job done. Also know that the equipment you need to do a good job depends on the kind of circuit or product you are working on. For example, a DVM may be unnecessary for troubleshooting some problems in digital logic. And, the availability and accessibility of equipment may present certain obstacles. If you only have a mediocre oscilloscope and your company can't go out and buy or rent or borrow a fancy full-featured scope, then you will have to make do.
If you lack a piece of equipment, be aware that you are going into battle with inadequate tools; certain clues may take you much longer than necessary to spot. In many cases when you spent too much time finding one small problem, the time was wasted simply because you were foolish or were unaware of a particular troubleshooting technique; but, in other cases, the time was wasted because of the lack of a particular piece of equipment. It's important for you to recognize this last-mentioned situation.
Learning when you're wasting time because you lack the proper equipment is part of your education as a troubleshooter.
In addition to the proper tools, you also need to have a full understanding of how both your circuit and your equipment are supposed to work. I'm sure you've seen engineers or technicians work for many fruitless hours on a problem and then, when they finally find the solution, say, "Oh, I didn't know it was supposed to work that way." You can avoid this scenario by using equipment that you are comfortable and familiar with.
The following equipment is essential for most analog-circuit troubleshooting tasks.
This list can serve as a guide to both those setting up a lab and those who just want to make sure that they have everything they need-that they aren't missing any tricks.
1. A dual-trace oscilloscope. It's best to have one with a sensitivity of 1 or 2 mV/cm and a bandwidth of at least 100 MHz. Even when you are working with slow op amps, a wide-bandwidth scope is important because some transistors in "slow" applications can oscillate in the range of 80 or 160 MHz, and you should be able to see these little screams. Of course, when working with fast circuits, you may need to commandeer the lab's fastest scope to look for glitches. Sometimes a peak-to-peak automatic triggering mode is helpful and time-saving. Be sure you know how all the controls work, so you don't waste much time with setup and false-triggering problems.
2. Two or three scope probes. They should be in good condition and have suitable hooks or points. Switchable 1 X/10X probes are useful for looking at both large and very small signals. You should be aware that 1 X probes only have a 16 or 20-MHz bandwidth, even when used with a 100-MHz scope. When you use 10X probes, be sure to adjust the capacitive compensation of the probe by using the square-wave calibrator per FIG. 1. Failure to do so can be a terrible time-wasting source of trouble.
Ideally, you'll want three probes at your disposal, so that you can have one for the trigger input and one for each channel. For general-purpose troubleshooting, the probes should have a long ground wire, but for high-speed waveforms you'll need to change to a short ground wire ( FIG. 2) The shorter ground wires not only give you better frequency response and step response for your signal, but also better rejection of other noises around your circuit.
In some high-impedance circuits, even a 10X probe's capacitance, which is typically 9 to 15 pF, may be unacceptable. For these circuits, you can buy an active probe with a lower input capacitance of 1.5 to 3 pF ($395 to $1800), or you can build your own ( FIG. 3). When you have to work with switching regulators, you should have a couple of current probes, so you can tell what those current signals are doing. Some current probes go down to DC; others are inherently AC coupled (and are much less expensive).
3. An analog-storage oscilloscope. Such a scope can be extremely useful, especially when you are searching for an intermittent or evanescent signal. The scope can trigger off an event that may occur only rarely and can store that event and the events that follow it. Some storage scopes are balky or tricky to apply, but it's often worthwhile to expend the effort to learn how to use them. Digital-storage oscilloscopes (DSOs) let you do the same type of triggering and event storage as do the analog type, and some can display events that precede the trigger. They are sampled-data systems, however, so you must be sure to apply them correctly (Ref. 1). Once you learn how to use them, though, you'll appreciate the special features they offer, such as bright CRT displays, automatic pulse-parameter measurements, and the ability to obtain plots of waveforms.
4. A digital voltmeter (DVM). Choose one with at least five digits of resolution, such as the HP3455, the HP3456, the Fluke 8810A, or the Fluke 8842A. Be sure you can lock out the auto-range feature, so that the unit can achieve its highest accuracy and speed. Otherwise, you'll be wasting time while the DVM auto-ranges. For many analog circuits, it's important to have a high-impedance (>>10,000 M-ohm ) input that stays at high impedance up to 15 or 20 V; the four DVMs mentioned above have this feature. There are many other fine DVMs that have 10 MR inputs above 2 or 3V; and, if a 10 M-ohm input impedance is not a problem, they are acceptable. The most important reason to use a high-input-impedance DVM is because sometimes it's necessary to put 33 k-ohm or 100 k-ohm resistors in series with the probe, right near the circuit-under-test, to prevent the DVM's input capacitance from causing the circuit to oscillate. If you're using a DVM with a 10 M-ohm input impedance and you have a 100 k-ohm resistor in series with the probe, the DVM's measurements would lose 1 % of their accuracy. Fortunately, many good DVMs have less than 500 pA of input current, which would cause less than 50 pV of error in the case of 100 k-ohm source resistance. A high-resolution DVM lets you detect 100 to 200-pV deviations in an 11-V signal. You can best spot many semiconductor problems by finding these minor changes. A 4-digit DVM is a relatively poor tool; however, if you are desperate. you can detect small voltage changes if you refer the DVM's "low," or ground, side to a stable reference, such as a 10-V bus. Then, with the DVM in the 1-V range, you can spot small deviations in an 11-V signal. This measurement is more awkward and inconvenient than a ground-referenced measurement with a higher resolution meter would be, and this method can cause other problems as well. For instance, you can end up injecting noise generated by the DVM's A/D converter into the sensitive 10-V reference, thereby adversely affecting the performance of other circuits. In some cases, a little RC filtering may minimize this problem (see FIG. 4) but you never can be sure how easy it will be to get the noise to an acceptable level. . . .
5. Auxiliary meters. It may look silly to see a test setup consisting of two good DVM's, two little 3-digit DVMs monitoring a couple of voltage supplies. a couple more 3-digit DVMs monitoring current drain, and an analog meter monitoring something else. But, if you don't know exactly what you're looking for and you can borrow equipment, using lots of meters is an excellent way to attack a problem--even if you do have to wait until 5:15 pm to borrow all that equipment. When is an analog voltmeter better than a DVM? Well, the analog voltmeter usually has inferior accuracy and resolution, but when you watch an ordinary analog voltmeter your eye can detect a trend or rate-of-change that may be hard to spot on a DVM, especially in the presence of noise or jitter. As an example, if you suddenly connect an ordinary analog volt-ohmmeter across a 1 pF capacitor, your eye can resolve if the capacitor's value is 0.1 times or 10 times as large as it should be. You can't perform this kind of test with a DVM. Another advantage of analog meters is that they are passive devices: They don't inject noise into your circuit as digital meters can--even battery-powered ones. And they can have a lot less capacitance to ground.
6. A general-purpose function generator. While sine and square waves are popular test signals, I have often found triangular waveforms to be invaluable when searching for nonlinearities. Sometimes you need two function generators, one to sweep the operating point of the DUT, slowly, back and forth over its operating range, while you watch the response of the output to a small quick square wave-watching for oscillation or ringing or trouble.
7. Power supplies with stable outputs. They should have coarse and fine adjustment controls and adjustable current limits. Digital controls are usually not suitable; they don't let you continuously sweep the voltage up and down while you monitor the scope and watch for trends. In cases when the power supply's output capacitor causes problems, you may want a power supply whose output circuit, like that of an opamp, includes no output capacitor. You can buy such a supply, or you can make it with an op amp and a few transistors. The advantage of the supply shown in FIG. 5 is that you can design it to slew fast when you want it to.
(For speed, use a quick LF356 rather than a slow LM741). Also, if a circuit latches and pulls its power supply down, the circuit won't destroy itself by discharging a big capacitor.
While we are on the topic of power, another useful troubleshooting tool is a set of batteries. You can use a stack of one, two, or four 9-V batteries, ni-cads, gel cells, or whatever is suitable and convenient. Batteries are useful as an alternate power supply for low-noise preamplifiers: If the preamp's output doesn't get quieter when the batteries are substituted for the ordinary power supply, don't blame your circuit troubles on the power supply. You can also use these batteries to power low-noise circuits, such as those sealed in a metal box, without contaminating their signals with any external noise sources.
8. A few RC substitution boxes. You can purchase the VIZ Model WC-412A, which I refer to affectionately as a "Twiddle-box" ( FIG. 6) from R & D Electronics, 1432 South Main Street, Milpitas CA 95035, (408) 262-7 144. Or, inquire at VIZ, 175 Commerce Drive, Fort Washington, PA 19034, (800) 523-3696. You can set the unit in the following modes: R, C, series RC, parallel RC, open circuit, or short circuit.
They are invaluable for running various "tests" that can lead to the right answer.
You may need component values beyond what the twiddle boxes offer; in our labs, we built a couple of home-brew versions (FIG. 7). The circuit shown in FIG. 7a provides variable low values of capacitance and is useful for fooling around with the damping of op amps and other delicate circuits. You can make your own calibration labels to mark the setting of the capacitance and resistance values. The circuit shown in FIG. 7b provides high capacitances of various types, for testing power supplies and damping various regulators.
9. An isolation transformer. If you are working on a line-operated switching regulator, this transformer helps you avoid lethal and illegal voltages on your test setup and on the body of your scope. If you have trouble obtaining an isolation transformer. you can use a pair of transformers (step-down, step-up) back-to-back ( FIG. 8). Or. if cost isn't an issue, you can use isolated probes. These probes let you display small signals that have common-mode voltages of hundreds of volts with respect to ground, and they won't require you to wear insulated gloves when adjusting your scope.
10. A variable autotransformer, often called a Variac. This instrument lets you change the line voltage and watch its effect on the circuit-a very useful trick. (Warning: A variable autotransformer is not normally an isolation transformer. You may need to cascade one of each, to get safe adjustable power.)
11. A curve tracer. A curve tracer can show you that two transistors may have the same saturation voltage under a given set of conditions even though the slope of one may be quite different from the slope of the other. If one of these transistors works well and the other badly, a curve tracer can help you understand why. A curve tracer can also be useful for spotting nonlinear resistances and conductances in diodes, capacitors, light bulbs, and resistors. A curve tracer can test a battery by loading it down or recharging it. It can check semiconductors for breakdown. And, when you buy the right adapters or cobble them up yourself, you can evaluate the shape of the gain, the CMRR, and the PSRR of op amps.
12. Spare repair parts for the circuit-under-test. You should have these parts readily available, so you can swap components to make sure they still work correctly.
13. A complete supply of resistors and capacitors. You should have resistors in the range from 0.1 R to 100 M-ohm and capacitors from 10 pF to 1 pF. Also, 10,100, and 1000 uF capacitors come in handy. Just because your circuit design doesn't include a 0.1 ohm or a 100 M-ohm resistor doesn't mean that these values won't be helpful in troubleshooting it. Similarly, you may not have a big capacitor in your circuit; but, if the circuit suddenly stops misbehaving when you put a 3800 pF capacitor across the power supply, you've seen a quick and dramatic demonstration that power-supply wobbles have a lot to do with the circuit's problems. Also, several feet of plastic-insulated solid wire (telephone wire) often come in handy. A few inches of this type of twisted-pair wire makes an excellent variable capacitor, sometimes called a "gimmick." Gimmicks are cheap and easy to vary by simply winding or unwinding them. Their capacitance is approximately one picofarad per inch.
14. Schematic diagrams. It's a good idea to have several copies of the schematic of the circuit-under-test. Mark up one copy with the normal voltages, currents, and waveforms to serve as a reference point. Use the others to record notes and waveform sketches that relate to the specific circuit-under-test. You'll also need a schematic of any homemade test circuit you plan to use. Sometimes, measurements made with your homemade test equipment may not agree with measurements made by purchased test equipment. The results from each tester may not really be "wrong": They might differ because of some design feature, such as signal filtering. If you have all the schematics for your test equipment, you can more easily explain these incompatibilities. And, finally, the data sheets and schematics of any ICs used in your circuit will also come in handy.
15. Access to any engineering or production test equipment, if possible. Use this equipment to be sure that when you fix one part of the circuit, you aren't adversely affecting another part. Other pieces of equipment and testers also fall under the category of specialized test equipment; their usefulness will depend on your circuit.
Three examples are a short-circuit-detector circuit, an AM transistor radio, and a grid-dip meter.
A short-circuit-detector circuit. This tool comes in handy when you have to repair a lot of large PC boards: It can help you find a short circuit between the ground bus and the power or signal busses. It's true that a sensitive DVM can also perform this function, but a short-circuit detector is much faster and more efficient. Also, this circuit turns itself off and draws no current when the probe is not connected. In the short-circuit-detector circuit shown in FIG. 9, the LM10 op amp amplifies the voltage drop and feeds it to the LM331 voltage-to-frequency converter, which you set up to emit its highest pitch when Vin = 0 mV. When using this circuit, use a 50- to 100-mA current-limited power supply. To calibrate the circuit, first ground the detector's two probes and trim the OFFSET ADJUST for a high pitch. Then, move the positive test probe to Vs at A and trim the GAIN ADJUST for a low pitch. FIG. 9 illustrates a case in which one of the five major power supply busses of the circuit-in-trouble has a solder short to ground.
To find the exact location of the short, you simply slide the positive input probe along the busses. In this example, if you slide it from A toward B or D, the pitch won't change because there is no change in voltage at these points-no current flowing along those busses. But, if you slide the probe along the path from A to C or from K to M, the pitch will change because the voltage drop is changing along those paths.
It's an easy and natural technique to learn to follow the shifting frequency signals.
An AM radio. What do you do when trouble is everywhere? A typical scenario starts out like this: You make a minor improvement on a linear circuit, and when you fire it up you notice a terrible oscillation riding on the circuit's output. You check everything about the circuit, but the oscillation remains. In fact, the oscillation is riding on the output, the inputs, on several internal nodes. and even on ground. You turn off the DVM, the function generator, the soldering iron, and finally even the power supplies, but the oscillation is still there.
Now you start looking around the lab to see who has started a new oscillator or switching regulator that is doubling as a medium-power transmitter. Aside from yelling, "Who has a new circuit oscillating at 87 kHz?" what can you do to solve the problem? One useful tool is an ordinary AM transistor radio. As we have all learned, FM radios reject many kinds of noise, but AM radios scoop up noise at repetition rates and frequencies that would surprise you.
How can a crummy little receiver with an audio bandwidth of perhaps 5 kHz detect noise in the kilohertz and megahertz regions? Of course, the answer is that many repetitive noise-pulse trains (whose repetition rates are higher than the audible spectrum but below the AM frequency band) have harmonics that extend into the vicinity of 600 kHz, where the AM receiver is quite sensitive. This sensitivity extends to signals with amplitudes of just a few microvolts per meter.
If you are skeptical about an AM radio's ability to detect these signals, tune its dial down to the low end, between stations. Then, hold it near a DVM or a computer or computer keyboard, and listen for the hash. Notice, too, that the ferrite stick antenna has definite directional sensitivity, so you can estimate where the noise is coming from by using either the null mode or pointing the antenna to get the strongest signal.
So, the humble AM radio may be able to help you as you hike around the lab and smile pleasantly at your comrades until you find the culprit whose new switching regulator isn't working quite right but which he neglected to turn off when he went out to get a cup of coffee.
The grid-dip meter. On other occasions, the frequency and repetition rate of the noise are so high that an AM receiver won't be helpful in detecting the problem.
What's the tool to use then? Back in the early days of radio, engineers found that if you ran a vacuum-tube oscillator and immersed it in a field of high-power oscillations at a comparable frequency, the tube's grid current would shift or dip when the frequencies matched. This tool became known as the "grid-dip meter." I can't say that I am an expert in the theory or usage of the grid-dip meter, but I do recall being impressed in the early days of monolithic ICs: A particular linear circuit was oscillating at 98 MHz, and the grid-dip meter could tickle the apparent rectified output error as I tuned the frequency dial back and forth.
That was 25 years ago, and, of course, Heathkit has discontinued their old Grid Dip and Tunnel Dip meters in favor of a more modem design. The new one, simply dubbed HD1250 "Dip Meter," uses transistors and tetrode FETs. At the bargain price of $89, every lab should have one. They'll help you ferret out the source of nasty oscillations as high as 250 MHz. The literature that comes with the HD1250 dip-meter kit also lists several troubleshooting tips.
When grid-dip meters first became popular, the fastest oscilloscope you could buy had a bandwidth of only a few dozen megahertz. These days, it is possible to buy a scope with a bandwidth of many hundreds of megahertz, so there are fewer occasions when you might need a grid-dip meter. Still, there are times when it is exactly the right tool. For example, you can use its oscillator to activate passive tuned circuits and detect their modes of resonance. Also, in a small company where you can't afford to shell out the many thousands of dollars for a fast scope, the dip meter is an inexpensive alternative.
16. A few working circuits, if available. By comparing a bad unit to a good one, you can 17. A sturdy, broad workbench. It should be equipped with a ground plane of metal that you can easily connect to the power ground. The purpose of this ground plane is to keep RF. 60-Hz, and all other noise from coupling into the circuit. Place insulating cardboard between the bench and the circuit-under-test. so that nothing tends to short to ground. Another way to prevent noise from interfering with the circuit is to use a broad sheet of single-sided copper-clad board. Placed copper-side down and with a ground wire soldered to the copper, it provides an alternate ground plane. To prevent electrostatic-discharge (ESD) damage to CMOS circuits, you'll need a wrist strap to ground your body through 1 M-ohm.
18. Safety equipment. When working on medium- or high-power circuits that might explode with considerable power in the case of a fault condition. You should be wearing safety goggles or glasses with safety lenses. Keep a fire extinguisher nearby, too.
19. A suitable hot soldering iron. If you have to solder or unsolder heavy busses from broad PC-board traces, use a big-enough iron or gun. For small and delicate traces around ICs, a small tip is essential. And, be sure that the iron is hot enough. An easy way to delaminate a trace or pad, whether you want to or not, is to heat it for too long a time. which might happen if your iron weren't big enough or hot enough. (The old Heathkit warnings not to use a hot iron became obsolete along with the germanium transistor.) In some cases, a grounded soldering iron is required: in others. a portable (ungrounded or rechargeable) soldering iron is ideal. Make sure you know whether your iron is grounded or floating.
20. Tools for removing solder, such as solder wick or a solder sucker. You should be comfortable with whatever tools you are using; a well-practiced technique is sometimes critical for getting good results. If you are working on static-sensitive components, an antistatic solder-sucker is less likely to generate high voltages due to internal friction than is an ordinary solder-sucker. I have been cautioned that a large solder-sucker may cause problems when working on narrow PC traces: in that case, solder wick may be the better choice.
21. Hand tools. Among the tools you'll need are sharp diagonal nippers. suitable pliers. screwdrivers. large cutters. wrenches, wire strippers, and a jack knife or Exacto-knife.
22. Signal leads, connectors. cables, BNC adapters, wires, clip leads, ball hooks, and alligator clips-as needed. Scrimping and chintzing in this area can waste lots of time: shaky leads can fall off or short out.
23. Freeze mist and a hair dryer. The freeze mist available in aerosol cans lets you quickly cool individual components. A hair dryer lets you warm up a whole circuit.
You'll want to know the dryer's output air temperature because that's the temperature to which you'll be heating the components.
NOTE: Ideally we should not use cooling sprays based on chlorofluorocarbons (CFCs), which are detrimental to the environment. I have a few cans that some people would say I shouldn't use. But what else should I do--send the can to the dump? Then it will soon enter the atmosphere, without doing anybody any good. I will continue to use up any sprays with CFC-based propellants that I already have, but when it is time to buy more, I'll buy environmentally safe ones.
24. A magnifying glass or hand lens. These devices are useful for inspecting boards, wires, and components for cracks, flaws, hairline solder shorts, and cold-soldered joints.
25. An incandescent lamp or flashlight. You should be able to see clearly what you are doing, and bright lights also help you to inspect boards and components.
26. A thermocouple-based thermometer. If your thermometer is floating and battery powered, you can connect the thermocouple to any point in the circuit and measure the correct temperature with virtually no electrical or thermal effect on the circuit.
FIG. 10 shows a thermocouple amplifier with designed-in cold-junction compensation.
Some people have suggested that an LM35 temperature-sensor IC ( FIG. 11) is a simple way to measure temperature, and so it is. But, if you touch or solder an LM35 in its TO46 package to a resistor or a device in a TO-5 or TO-3 case, the LM35 will increase the thermal mass and its leads will conduct heat away from the device whose temperature you are trying to measure. Thus, your measurements will be less accurate than if you had used a tiny thermocouple with small wires.
27. Little filters in neat metal boxes, to facilitate getting a good signal-to-noise ratio when you want to feed a signal to a scope. They should be set up with switch-selectable cut-off frequencies, and neat connectors. If in your business you need sharp roll-offs, well, you can roll your own. Maybe even with op-amps and batteries. You figure out what you need. Usually I just need a couple simple Rs and Cs, with an alligator clip to select the right ones.
28. Line adapters-those 3-wire-to-2-wire adapters for your 3-prong power cords. You need several of them. You only need them because too many scopes and function generators have their ground tied to the line-cord's neutral. You need some of these to avoid ground-loops. You also need a few spares because your buddies will steal yours. For that matter, keep a few spare cube taps. When they rewired our benches a few years ago, the electricians tried to give us five outlets per bench. I stamped my feet and insisted on ten per bench, and that's just barely enough, most of the time.
You've come to the end of my list of essential equipment for ordinary analog-circuit troubleshooting. Depending on your circuit, you may not need all these items; and, of course, the list did not include a multitude of other equipment that you may find useful. Logic analyzers, impedance analyzers, spectrum analyzers, programmable current pumps, capacitance meters and testers, and pulse generators can all ease various troubleshooting tasks.
Each of you will have your own idea of what is essential and what is unnecessary for your special case, and I would be delighted to get feedback on this subject.
1. Collins, Jack, and David White, "Time-domain analysis of aliasing helps to alleviate DSO errors," EDN, September 15, 1988, p. 207.