|Home | Articles | Forum | Glossary | Books|
After the first five sections about troubleshooting passive parts, this section shifts over to troubleshooting active components. We begin with the simple stuff--diodes and rectifiers, optically coupled devices, solar cells, and batteries.
Even the simplest active devices harbor the potential for causing baffling troubleshooting problems. Consider the lowly diode. The task of a diode sounds simple:
To conduct current when forward biased, and to block current when reverse biased, while allowing negligible leakage. That task sounds easy, but no diode is perfect. and diodes' imperfections are fascinating. Even these two-terminal devices are quite complex! up. An ideal diode may have an exponential characteristic with a slope delta_V/delta_I of All diodes start conducting current exponentially at low levels, nanoamperes and
g = (38.6 mS/mA) X I_F,
where mS = millisiemens = millimhos, and I_F = forward current. And indeed transistors do have this slope of 38.6 mS/mA at their emitters, at room temperature. This corresponds to 60 mV per decade of current. But the slopes of the exponential curves of different real (two-terminal) diodes vary considerably. Some, like a 1N645, have a slope as good as 70 or 75 mV per decade. Others like 1N9 14s have a slope as poor as 113 mV per decade. Others have intermediate values such as 90 mv/decade. When you go shopping for a diode, the data sheets never tell you about this. To tell the truth, I didn't even really recognize this. When I wrote the first version of this, as published in EDN, I assumed that the slopes started out from 60 mV per decade and then got worse-shifted over to 120 mV per decade at higher current levels, and I said so. But I was wrong. And nobody ever contradicted me. Such a strange world! Please refer to FIG. 1, which shows just a few of the different curves you may get when you buy a diode. None of these slopes are characterized or guaranteed; if you change vendors, all bets are off. So, qualify each vendor of diodes carefully for each application. (Note: For a detailed explanation of this graph and a table detailing exactly which diode is which.)
As the current level continues to increase, the conductance per milliampere gets even worse due to series resistance and high-level injection and other nonlinear factors.
Therefore, at a large forward current, a diode's forward voltage, V_F, will be considerably larger than predicted by simple theory-and larger than desired. Of course, some rectifiers--depending on their ratings--can handle large currents from amperes to kiloamperes; but the VFs of all diodes, no matter what their ratings, err from the theoretical at high current levels.
These days, Schottky diodes have smaller VFS than ordinary pn diodes. However, even germanium diodes and rectifiers still have their following because their low V_Fs are similar to the Schottky's. Just the other day I read about some new Germanium Schottky diodes that have even lower VFs.
All the other diodes you can buy have inferior conductances, and they are just about all different...surprise.
High-speed and ultra-high-speed (sometimes also called high-efficiency) silicon rectifiers are also available; they have been designed for fast switching-regulator and other high-frequency applications. They don't have quite as low VFS as Schottky diodes and are not quite as fast, but they are available with high reverse-voltage ratings and thus are useful for certain switch-mode circuit topologies that impress large flyback voltages on diodes.
When you reverse-bias these various diodes, ah, that is where you start to see even more wild dissimilarities. For example, the guaranteed reverse-current specification, I_REV for many types of diodes is 25 nA max at 25 C. When you measure them, many of these devices actually have merely 50 or 100 pA of leakage. But the popular 1N914 and its close cousin, the 1N4148, actually do have about 10 or 15nA of leakage at room temperature because of their gold doping. So although these diodes are inexpensive and popular, it's wrong to use them in low-leakage circuits since they're much leakier than other diodes with the same leakage specs.
Why, then, do some low-leakage diodes have the same mediocre 25-nA leakage spec as the 1N914? Diode manufacturers set the test and price at the level most people want to pay, because automatic test equipment can test at the 25-nA level-but no lower--without slowing down. If you want a diode characterized and tested for 100pA or better, you have to pay extra for the slow-speed testing. Of course, high-conductance diodes such as Schottkys, germaniums, and large rectifiers have much larger reverse leakage currents than do signal diodes, but that's not normally a problem.
If you want a very-low-leakage diode, use a transistor's collector-base junction instead of a discrete diode (Ref. 1). The popular 2N930 or 2N3707 have low leakage, typically. Some 2N3904s do, too, but some of these are gold-doped and are leakier.
The plastic-packaged parts are at least as good as the TO-18 hermetic ones. You can easily find such "diodes" having less than 1 pA leakage even at 7V, or 10 pA at 50V. Although this low leakage is not guaranteed, it's usually quite consistent.
However, this c-b diode generally doesn't turn ON or OFF very quickly.
Another source of ultra-low-leakage diodes are the 2N4117A and the PN4117A, -18 A, and -19 A. These devices are JFETS with very small junctions, so leakages well below 0.1 pA are standard with 1.0 pA max, guaranteed--not bad for a $0.40 part. The capacitances are small, too.
When a diode is carrying current, how long does it take to turn the current off? There's another wide-range phenomenon. Slow diodes can take dozens and hundreds of microseconds to turn off. For example, the collector-base junction of a 2N930 can take 30 us to recover from 10 mA to less than 1 mA, and even longer to the nanoampere level. This is largely due to the recombination time of the carriers stored in the collector region of the transistor. Other diodes, especially gold-doped ones, turn off much faster-down into the nanosecond region. Schottky diodes are even faster, much faster than 1 ns. However, one of my friends pointed out that when you have a Schottky diode that turns off pretty fast, it is still in parallel with a p-n junction that may still turn off slowly at a light current level. If a Schottky turns off from 4 mA in less than 1 ns, there may still be a few microamperes that do not turn off for a microsecond.
So if you want to use a Schottky as a precision clamp that will turn off very quickly, as in a settling detector (Ref. 2), don't be surprised if there is a small long "tail." Switching regulators all have a need for diodes and high-current rectifiers and transistors to turn off quickly. If the rep rate is high and the current large and the diode turns off slowly, it can fail due to overheating. You don't want to try a 1N4002 at 20 or 40 kHz, as it will work very badly, if at all. Sometimes if you need a moderate amount of current at high speed, you can use several 1N914s in parallel. I've done that in an emergency, and it seemed to work well, but I can't be sure I can recommend it as the right thing to do for long-term reliability. The right thing is to engineer the right amount of speed for your circuit. High-speed, fast-recovery, and ultrafast diodes are available. The Schottky rectifiers are even faster, but not available at high voltage breakdowns. When you start designing switching regulators at these speeds, you really must know what you are doing. Or at least, wear safety-goggles so you don't get hurt when the circuit blows up.
Turn 'Em Off-Turn 'Em On...
"Computer diodes" like the 1N914 are popular because they turn OFF quickly-in just a few nanoseconds-much faster than low-leakage diodes. What isn't well known is that these faster diodes not only turn OFF fast, they usually turn ON fast.
For example, when you feed a current of 1.0 mA toward the anode of a 1N914 in parallel with a 40 pF capacitance (20 pF of stray capacitance plus a scope probe or something similar), the 1N914 usually turns ON in less than 1 ns. Thus, the VF has only a few millivolts of overshoot.
But with some diodes-even 1N914s or 1N4148s from some manufacturers-the forward voltage may continue to ramp up past the expected DC level for 10 to 20 ns before the diode turns ON; this overshoot of 50 to 200 mV is quite surprising (FIG. 2). Even more astonishing, the VF overshoot may get worse at low repetition rates but can disappear at high repetition rates (FIG. 2b-d).
I spent several hours once discovering this particular peculiarity when a frequency-to-voltage converter suddenly developed a puzzling nonlinearity. The trickiest part of the problem with the circuit's diodes was that diodes from an earlier batch had not exhibited any slow-turn-on behavior. Further, some diodes in a batch of 100 from one manufacturer were as bad as the diodes in Figures 2b and 2v. Other parts in that batch and other manufacturers' parts had substantially no overshoot.
When I confronted the manufacturers of these nasty diodes, they at first tried to deny any differences, but at length they admitted that they had changed some diffusions to "improve" the product. One man's "improvement" is another man's poison.
Thus, you must always be alert for production changes that may cause problems.
When manufacturers change the diffusions or the process or the masks, they may think that the changes are minor, but these changes could have a major effect on your circuit.
Many circuits, obviously, require a diode that can turn ON and catch, or clamp, a voltage moving much faster than 20 V/ps. Therefore, if you want any consistency in a circuit with fast pulse detectors (for example), you'll need to qualify and approve only manufacturers whose diodes turn ON consistently. So, as with any other unspecified characteristic, be sure to protect yourself against "bad" parts by first evaluating and testing and then specifying the performance you need. Also if you want to see fast turn-on of a diode circuit, with low overshoot. you must keep the inductance of the layout small. It only takes a few inches of wire for the circuit's inductance to make even a good fast rectifier look bad, with bad overshoot.
One "diode" that does turn ON and OFF quickly is a diode-connected transistor. A typical 2N3904 emitter diode can turn ON or OFF in 0.1 nsec with negligible overshoot and less than 1 pA of leakage at 1 V, or less than 10 pA at 4 V. (This diode does, of course, have the base tied to the collector.) However, this diode can only withstand 5 or 6 V of reverse voltage, and most emitter-base junctions start to break down at 6 or 8V. Still, if you can arrange your circuits for just a few volts, these diode-connected transistors make nice, fast, low-leakage diodes. Their capacitance is somewhat more than the 1N914's 1pF.
Other Strange Things That Diodes Can Do to You...
If you keep LEDs in the dark, they make an impressive, low-leakage diode because of the high band-gap voltage of their materials. Such LEDs can exhibit less than 0.1 pA of leakage when forward biased by 100 mV or reverse biased by 1V.
Of course, you don't have to reverse-bias a diode a lot to get a leakage problem.
One time I was designing a hybrid op amp, and I specified that the diodes be connected in the normal parallel-opposing connection across the input of the second stage to avoid severe overdrive (FIG. 3). I thought nothing more of these diodes until we had the circuit running-the op amp's voltage gain was falling badly at 125 C. Why? Because the diodes were 1N914s, and their leakage currents were increasing from 10 nA at room temperature to about 8 FA at the high temperature. And remember that the conductance of a diode at zero voltage is approximately (20 to 30 mS/mA) x I_LEAKAGE.
That means each of the two diodes really measured only 6 k-Ohm.
Because the impedance at each input was only 6 k-Ohm, the op amp's gain fell by a factor of four, even though the diodes may have only been forward or reverse biased by a millivolt. When we substituted collector-base junctions of transistors for the diodes, the gain went back up where it belonged.
Thus you cannot safely assume that the impedance of a diode at zero bias is high if the junction's saturation current is large. For example, at 25 C a typical 1N914 will leak 200 to 400 pA even with only 1 mV across it. Therefore, a 1N914 can prove unsuitable as a clamp or protection diode--even at room temperature-despite having virtually no voltage biased across it, in even simple applications such as a clamp across the inputs of a FET-input op amp.
How can diodes fail? Well, if you were expecting a diode to turn ON and OFF, but instead it does something unexpected-of the sort I have been mentioning-that unexpected behavior may not be a failure, but it could sure cause trouble.
Further, you can kill a diode by applying excessive reverse voltage without limiting the current or by feeding it excessive forward current. When a diode fails, it tends to short out, becoming a small blob of muddy silicon rather than an open circuit. I did once see a batch of 1N4148s that acted like thermostats and went open circuit at 75 C, but such cases are rare these days.
One of the best ways to kill a diode is to ask it to charge up too big a capacitor during circuit turn-on. Most rectifiers have maximum ratings for how much current they can pass, on a repetitive and on a nonrecurring basis. I've always been favorably impressed by the big Motorola (Phoenix, AZ) books with all the curves of safe areas for forward current as a function of pulse time and repetition rate. These curves aren't easy to figure out at first, but after a while they're fairly handy tools.
Manufacturers can play tricks on you other than changing processes. If you expect a diode to have its arrow pointing toward the painted band (sometimes called the cathode by the snobbish) and the manufacturer put the painted band on the wrong end, your circuit won't work very well. Fortunately reverse-marked diodes are pretty rare these days. But just this morning, I heard an engineer call the "pointed" end of the diode an anode, which led to confusion and destruction. Sigh . . . .
Once I built a precision test box that worked right away and gave exactly the right readings until I picked up the box to look at some waveforms. Then the leakage test shifted way off zero. Every time I lifted up the box, the meter gave an indication; I thought I had designed an altimeter. After some study, I localized the problem to an FD300 diode, whose body is a clear glass DO-35 package covered with black paint.
This particular diode's paint had been scratched a little bit, so when I picked up the test box, the light shone under the fixture and onto the diode. Most of these diodes didn't exhibit this behavior; the paint wasn't scratched on most of them.
To minimize problems such as the ones I have listed, I recommend the following strategies:
Have each manufacturer's components specifically qualified for critical applications.
This is usually a full-time job for a components engineer, with help and advice from the design engineer and consultation with manufacturing engineers.
Establish a good relationship with each manufacturer.
Require that manufacturers notify you when, or preferably before, they make changes in their products.
Keep an alternate source qualified and running in production whenever possible.
My boss may gripe if I say this too loudly, but it is well known that having two good sources is better than having one. The argument that "One source is better than two" falls hollow on my ears. Two may be better than seven or eight, but one is not better than two.
Zener, Zener, Zener.. .
Just about all diodes will break down if you apply too much reverse voltage, but zener diodes are designed to break down in a predictable and well-behaved way. The most common way to have problems with a zener is to starve it. If you pass too little current through a zener, it may get too noisy. Many zeners have a clean and crisp knee at a small reverse-bias current, but this sharp knee is not guaranteed below the rated knee current.
Some zeners won't perform well no matter how carefully you apply them. In contrast to high-voltage zeners, low-voltage (3.3 to 4.7V) zeners are poor performers and have poor noise and impedance specs and bad temperature coefficients--even if you feed them a lot of current to get above the knee, which is very soft. This is because "zeners" at voltages above 6V are really avalanche-mode devices and employ a mechanism quite different from (and superior to) the low-voltage ones, which are real zener diodes. At low-voltage levels, band-gap references such as LM336s and LM385s are popular, because their performance is good compared with low-voltage zeners.
Zener references with low temperature coefficients, such as the 1N825, are only guaranteed to have low temperature coefficients when operated at their rated current, such as 7.5 mA. If you adjust the bias current up or down, you can sometimes tweak the temperature coefficient, but some zeners aren't happy if operated away from their specified bias. Also, don't test your 1N825 to see what its "forward-conduction voltage" is because in the "forward" direction, the device's temperature-compensating diode may break down at 70 or 80 V. This break-down damages the device's junction, degrades the device's performance and stability, and increases its noise.
The LM329 is popular as a 6.9-V reference because its TC is invariant of operating current, as it can run from any current from 1 to 10 mA. The LM399 is even more popular because of its built-in heater that holds the junction at +85 "C. Consequently it can hold 1/2 or 1 ppm per OC. The LM329 and LM399 types also have good long-term stability, such as 5 or 10 ppm per 1000 hours, typically. The buried zeners in the LM129/LM199/LM169 also have better stability than most discrete references (1N825 or similar) when the references are turned on and off.
And before you subject a zener to a surge of current, check its derating curves for current vs. time, which are similar to the rectifiers' curves mentioned earlier.
These curves will tell you that you can't bang an ampere into a 10-V, 1-W zener for very long.
In fact, most rectifiers are rated to be operated strictly within their voltage ratings, and if you insist on exceeding that reverse voltage rating and breaking them down, their reliability will be degraded. To avoid unreliability, you can redesign the circuit to avoid over-voltage, or you might add in an R-C-diode damper to soak up the energy; or you could shop for a controlled-avalanche rectifier. These rectifiers are rated to survive (safely and reliably) repetitive excursions into breakdown when you exceed their rated breakdown voltage. The manufacturers of these devices can give you a good explanation of how to keep out of trouble.
If you do need a zener to conduct a surge of current, check out the specially designed surge-rated zener devices--also called transient-voltage suppressors--from General Semiconductor Industries Inc. ( Tempe, AZ). You'll find that their 1-W devices, such as the 1N5629 through 1N5665A, can handle a surge of current better than most 10- or 50-W zeners. If you need a really high-current zener, a power transistor can help out (FIG. 4). As mentioned earlier, a diode tends to fail by becoming a short circuit when overpowered, and zeners cannot absorb as much power as you would expect from short pulses. How dreadful; but, can IC designers serendipitously take advantage of this situation? Yes! The Vos of an op amp usually depends on the ratio of its first-stage load resistors.
IC designers can connect several zeners across various small fractions of the load resistor. When they measure the Vos, they can decide which zener to short out--or zq-with a 5-ms, 0.3- to 1.8-A pulse. The zener quickly turns into a low-impedance (= 1 short), so that part of the resistive network shorts out, and the Vos is improved.
In its LM108, National Semiconductor first used zener zapping, although Precision Monolithics (Santa Clara, CA) wrote about zener zapping first and used it extensively later on. Although zener zapping is a useful technique, you have to be sure that nobody discharges a large electrostatic charge into any of the pins that are connected to the zener zaps. If you like to zap zeners for fun and profit, you probably know that they really do make a cute lightning flash in the dark when you zap them. Otherwise, be careful not to hit zeners hard, if you don't want them to zap and short out.
These zener zaps are also becoming popular in digital ICs under the name of "vertical fuses" or, more correctly, "anti-fuses." If an IC designer uses platinum silicide instead of aluminum metallization for internal connections, the diode resists zapping.
Diodes That Glow in the Dark, Efficiently
Once I needed 100 LEDS, so I bought 200 LEDs from the cheapest supplier. I hoped to find some good ones and maybe just a few units that were weak or performed poorly, which I could use for worst-case testing. I lost out; every one of the 200 was of uniformly good intensity. In a variation on Murphy's Law, worst-case parts will typically appear only when you are depending on having uniform ones.
So long as you don't fry LEDs with your soldering iron or grossly excessive milliamps of current, LEDs are awfully reliable these days. I have a thermometer display on my wall, which has 650 inexpensive, plastic-packaged LEDS. These LEDs have amassed 40,000,000 device-hours with just one failure. The only problem I ever have with LEDs is trying to remember which lead is "plus"-I just measure the diode and re-derive it, every time.
An optoisolator, also called a photo-coupler or opto-coupler, usually consists of an infrared LED and a sensitive phototransistor to detect the LED's radiation. In the course of working with the cheaper 4N28s, I've found it necessary to add circuitry to achieve moderate speeds. For example, if you tailor the biases per FIG. 5, you can get a 4N28's response up toward 50 kHz; otherwise the devices can't make even 4 kHz reliably. The trick is decreasing the phototransistor's turn-off time by using a resistor from pin 4 to pin 6.
I've evaluated many different makes and lots of 4N28s and have found widely divergent responses. For example, the overall current gain at 8 mA can vary from 15 to 104%, even though the spec is simply 10% min. Further, the transfer efficiency from the LED to the photodiode varies over a range wider than 10:1, and the p of the transistor varies from 300 to 3000. Consequently, the transistor's speed of response, which is of course related to p and f_3dB, would vary over a 10:1 range.
If your circuit doesn't allow for gains and frequency responses that vary so wildly and widely, expect trouble. For example, two circuits, one an optoisolated switching regulator (Ref. 3) and the other a detector for 4- to 20-mA currents (Ref. 4), have enough degeneration so that any 4N28 you can buy will work. I used to have a group of several "worst-case" 4N28s from various manufacturers that I would try out in prototypes and problem circuits. Unfortunately, I don't have those marginal devices anymore, but they were pretty useful.
Also, the data sheets for optoelectronic components often don't have a clear VF curve or list any realistic typical values; the sheets list only the worst-case values.
Therefore, you may not realize that the VF of an LED in an opto-isolator is a couple hundred millivolts smaller than that of discrete red or infrared LEDs. Conversely, the VF of high-intensity, or high-efficiency, red LEDs tends to be 150 mV larger than that of ordinary red LEDs. And the VF of DEADs (a DEAD is a Darkness Emitting Arsenide Diode; that is, a defunct LED) is not even defined.
Once I was troubleshooting some interruptor modules. In these modules, a gap separated an infrared LED and a phototransistor. An interruptor--say a gear tooth in the gap can thus block the light. I tested one module with a piece of paper and nothing happened-the transistor stayed ON. What was that again? It turned out that the single sheet of paper could diffuse the infrared light but not completely attenuate it. A thin sheet of cardboard or two sheets of paper would indeed block the light.
Extraneous, unwanted light impinging on the pn junction of a semiconductor is only one of many tricky problems you can encounter when you try to design and operate precision amplifiers-specially high-impedance amplifiers. Just like a diode's pn junction, a transistor's collector-base junction makes a good photodiode, but a transistor's plastic or epoxy or metal package normally does a very good job of blocking out the light.
When light falls onto the pn junction of any diode, the light's energy is converted to electricity and the diode forward biases itself. If you connect a load across the diode's terminals, you can draw useful amounts of voltage and current from it. For example, you could stack a large number of large-area diodes in series and use them for recharging a battery. The most unreliable part of this system is the battery. Even if you never abuse them, batteries don't like to be discharged a large number of cycles, and your battery will eventually refuse to take a charge. These days one reads all sorts of marvelous hype about battery-powered cars, but the writers always ignore the terrible expense of replacing the batteries after just a few hundred cycles. They seem to be pretending that if they ignore that problem, it will go away...
So much for the charms of solar-recharged batteries. It's much better to use a solar-powered night-light. Remember that one? A solar-powered night-light doesn't need a battery; it simply needs a 12,000-mile extension cord. To be serious, the most critical problem with solar cells is their packaging; most semiconductors don't have to sit out in the sun and the rain as solar cells do. And it's hard to make a reliable package when low cost is-as it is for solar cells-a major requirement.
In addition to packaging, another major trouble area with solar cells is their temperature coefficients. Just like every other diode, the VF of a solar cell tends to decrease at 2 mV/T of temperature rise. Therefore, as more and more sunlight shines on the solar cell, it puts out more and more current, but its voltage could eventually drop below the battery's voltage, whereupon charging stops. Using a reflector to get even more light onto the cell contributes to this temperature-coefficient problem.
Cooling would help, but the attendant complications rapidly overpower the original advantage of solar cells' simplicity.
(image coming soon) FIG. 6. With a solar-cell array, you can make electricity when the sun shines.
Assault and Battery
Lastly, I want to say a few things about batteries. The only thing that batteries have in common with diodes is that they are both two-terminal devices. Batteries are complicated electrochemical systems, and large books have been written about the characteristics of each type (Refs. 5-10). I couldn't possibly give batteries a full and fair treatment here, but I will outline the basics of troubleshooting them.
First, always refer to the manufacturer's data sheet for advice on which loads and what charging cycles will yield optimal battery life. When you recharge a nickel-cadmium battery, charge it with a constant current, not constant voltage. And be sure that the poor little thing doesn't heat up after it is nearly fully charged. Heat is the enemy of batteries as it is for semiconductors. If you're subjecting your battery to deep-discharge cycles, refer to the data sheet or the manufacturer's specifications and usage manual for advice. Some authorities recommend that you do an occasional deep discharge, all the way to zero; others say that when you do a deep discharge, some cells in the battery discharge before the others and then get reversed, which is not good for them. I cannot tell you who's correct.
Sometimes a NiCad cell will short out. If this happens during a state of low charge, the cell may stay shorted until you ZAP it with a brief burst of high current. I find that discharging a 470 pF capacitor charged to 12 V into a battery does a good job of opening up a shorted cell. If 470 pF doesn't do it, I keep a 3800 pF to do the job.
When you recharge a lead-acid battery, charge it to a float voltage of 2.33 V per cell. At elevated temperatures, you should decrease this float voltage by about 6 mV/ C; again, refer to the manufacturer's recommendations. When a lead-acid battery is deeply discharged (below 1.8 V per cell), it should be recharged right away or its longevity will suffer due to sulfation.
Be careful when you draw excessive current from a lead-acid battery; the good strong ones can overheat or explode. Also be careful when charging them; beware of the accumulation of hydrogen or other gases that are potentially dangerous or explosive.
And, please dispose of all dead batteries in an environmentally sound way. Call your local solid-waste-disposal agency for their advice on when and where to dispose of batteries. Perhaps some can be recycled.
1. Pease, Robert A., "Bounding, clamping techniques improve on performance," EDN, November 10,1983, p. 277.
2. Pease, Bob, and Ed Maddox, "The Subtleties of Settling Time," The New Lightning Empiricist, Teledyne Philbrick, Dedham MA, June 1971.
3. Pease, Robert A., "Feedback provides regulator isolation," EDN, November 24, 1983, p. 195.
4. Pease, Robert A., "Simple circuit detects loss of 4-20 mA signal," Instruments & Control Systems, March 1982, p. 85.
5. Eveready Ni-Cad Battery Handbook, Eveready, Battery Products Div., 39 Old Ridgebury Rd., Danbury, CT. (203) 794-2000.
6. Battery Application Manual, Gates Energy Products, Box 861, Gainesville, FL 32602 (1-800-627-1700). (Note: A sealed lead-acid and NiCd battery manual.)
7. Perez, Richard, The Complete Battery Book, Tab Books, Blue Ridge Summit, PA, 1985.
8. Small, Charles H., "Backup batteries," EDN, October 30, 1986, p.123.
9. Linden, David, Editor-in-Chief, The Handbook of Batteries and Fuel Cells, McGraw-Hill Book Co., New York, NY, 1984. (Note: The battery industry's bible.)
10. Independent Battery Manufacturers Association, SLIG System Buyer's Guide, 100 Larchwood Dr., Largo, FL 33540. (813) 586-1408. (Note: Don't be put off by the title; this book is the best reference for lead-acid batteries.)