Making Meters With Chips

Lcd Monitor Repair Made Easy

How To Repair Lcd Monitors

Get Instant Access

By Sam Ulbing, N4UAU 5305 NW 57th Lane Gainesville, FL 32606

Have you ever built a project that needed a meter? Did you spend a lot of time and money to meet that need? Maybe you should consider using ICs to replace meters—it's cheaper and easier. When you see what these chips can do, I think you'll agree with me: Building a project with an analog meter is like building a QRP rig with vacuum tubes!

I'll discuss two chips that are ideal for making meters. Each chip is an analog-to-digitai (A/D) converter designed for direct visual display. As you'll see, the chips are different from one another. But what they have in common is that they are:

• low cost—much less than analog meters

• highly accurate

• easily adaptable to measure different voltages or currents

• able to include alarms or special indicators

The ICL7106 All-Purpose Voltmeter

The ICL7106 is a voltmeter chip that drives a 31/2 digit display, meaning it will display values up to ±1999, providing a resolution of better than 0.1%. With this kind of available resolution, the accuracy of the readout depends on the accuracy of the external resistors you use to build the meter. Fig 1 shows the basic components of the IC.

Fig 2 shows the circuit of an ICL7106 voltmeter. The

7106 Block Diagram

Fig 1—Block diagram of the ICL7106 voltmeter IC.

chip requires +5 and -5 volts. In this circuit, the +5 Vdc is provided by a 78L05 regulator IC, allowing a supply voltage of from 7 to 20 volts. -5 Vdc is developed using an ICL7660 voltage converter. The ICL7106 AID sampling rate is set by R1 and C1 to approximately 48 kHz. This rate results in about three readings per second and rejects 60-Hz noise. C2 is a reference capacitor which, in conjunction with C4, C5 and R2, allows the ICL7106 to calculate the input voltage. The reference voltage is set by R3 and R4. Ideally, C5 should be a polypropylene capacitor because of the low dielectric absorption of such capacitors. I've used ceramic and mylar capacitors here, though, with good results.

The meter in Fig 2 reads ± 20 Vdc, but you can select the range of your meter circuit by choosing the appropriate value for R5. The chip accepts a maximum input voltage of 200 mV, so select R5 to make the maximum input voltage correspond to 200 mV. Since the readout at the maximum input voltage will be 1999 (with a decimal point displayed at the appropriate spot), you probably will want to set the range to one of the voltages in Table 1, which shows the values of R5 needed.

The digital section of the circuit in Fig 2 connects the chip to an LCD display module. The LCD display is activated when an ac voltage, provided by the ICL7106, is applied to a given segment of the display, relative to the display backplane. The ICL7106 applies a continuous square-wave signal to the backplane and drives each of the display segments either in phase with this backplane signal (segment is off) or out of phase with the backplane

Table 1—Values of R5 in the ICL7106 circuit

R5 values

Full-scale V Exact 1% value

none 111.1 kQ 10.10 kQ 1.001 kQ

none 110 kQ 10.0 kQ 1.00 kQ

Using the nearest standard 1% resistor value introduces a small error. An exact resistor or trimmer pot can be used if necessary.

Fig 1—Block diagram of the ICL7106 voltmeter IC.

Fig 2—Schematic diagram of a general-purpose voltmeter using the ICL7106. The range of the voltmeter is set by selecting the value of R5 (see text).

C5—0.22 ^F capacitor, polypropylene, ceramic or mylar (see text). Q1—ZVN110A or equivalent N-channel MOSFET.

U1— ICL7106 LCD A/D converter IC. U2—LCD002 LCD display module U3—78L05 5-volt voltage regulator IC. U4—ICL7660 voltage converter IC.

All parts are available from Digi-Key Corp, 701 Brooks Ave S, PO Box 677, Thief River Falls, MN 56701-0677, tel: 800 344-4539.

For a given range, you may want to illuminate one of the decimal points on the display module shown in Fig 2. The unused decimal points are wired to the backplane to ensure they stay off, while the needed decimal point is wired to the drain of Q1. Fig 2 shows this arrangement for a 20-volt display. If using an LCD display module other than the one referenced in Fig 2, check the pin connections—they may differ.

That's it. You now have an accurate digital voltmeter for your project. To convert it to an ammeter, remove R5 and connect Vjn to a shunt resistance that has one end connected to ground. The shunt resistance should be of a value that produces 200 mV across the shunt at full-scale current. Table 2 shows typical shunt resistances.

The LM3914

While the ICL7106 provides a numeric display this often isn't the most effective visual aid. Sometimes, the "moving pointer" nature of an analog meter provides the desired visual indication. A digital display that provides this kind of animation, as well as actual measurement capability, is produced by the LM3914. Along with this difference, LM3914-based displays are less expensive, smaller and simpler than ICL7106-based displays.

The LM3914, an 18-pin chip from National Semiconductor, drives 10 LEDs in a bar-graph display. Two reference voltages, the high reference and the low reference, set the range of the voltage measured by the chip.

Fig 3 shows the basic LM3914 voltmeter circuit. The supply voltage applied to pin 3 of the LM3914 can be between 3 and 20 Vdc, although it should be at least 1.5 volts higher than the voltage to be measured. In this basic circuit the low reference (pin 4) is at 0 volts, while

Table 2—Shunt resistances for the ICL7106 ammeter

Full-scale I Shunt resistance

200.0 nA

the high reference (pin 6) is set by the reference voltage supplied by the chip on pin 7.

While an external reference voltage can be used, use of the internal reference source is usually easier. A 1,25-volt reference is supplied between pins 7 and 8. Placing a resistance (1 kQ in our circuits) across these terminals causes a current to flow through the 1 -k£2 resistance: 1=1.25/1000=1.25 mA. This current also must flow through R2, developing a voltage across that resistor that is proportional to the resistance: V=lxR2. This voltage is added to the 1.25 V across R1 to form the high reference voltage, setting the upper limit of the measurement range.

Actually, an exact measurement requires a bit of fine tuning of the R2 value. This is because some of the current from pin 7 flows to pin 6 and through the internal voltage divider, causing the current through R2 to be slightly less than theoretical. The small bias current from pin 8, along with chip-to-chip variations, are also factors. I do a rough calculation of R2 and then use a small pot to trim the reference voltage to the exact value.

in this circuit, as the measured voltage increases more LEDs light up. This provides a bar-graph style display. You also may elect to have a "moving dot" display. Which display you get is controlled by the voltage at pin 9. If this pin is left floating, you get the moving-dot display. If the pin is connected to the supply voltage, the bar-graph display results.

Lm3914 Voltmeter

Fig 3—The basic LM3914 voltmeter circuit.

Fig 4—Use of the LM3914 as an over-voltage alarm.

R1 «,1 kfi *See text for fin tuning adjustments

Fig 3—The basic LM3914 voltmeter circuit.

Fig 4—Use of the LM3914 as an over-voltage alarm.

Lm3914 Amateur Radio

Fig 5—Cascading of LM3914s to achieve a higher-resolution display. Use as a moving-dot display is shown. For use as a bar-graph display, remove RB, disconnect U1 pin 9 from U2 pin 1, and connect both U1-9 and U2-9 to Vcc. Select Rl to adjust LED brightness (see text). Reference voltages are set as follows:






Fig 5—Cascading of LM3914s to achieve a higher-resolution display. Use as a moving-dot display is shown. For use as a bar-graph display, remove RB, disconnect U1 pin 9 from U2 pin 1, and connect both U1-9 and U2-9 to Vcc. Select Rl to adjust LED brightness (see text). Reference voltages are set as follows:

For example, for a 10-volt meter, R1=100 kil, R2=14.3 kD.

over-voltage alarm by sensing when the uppermost LED comes on. The alarm circuit shown drives a piezoelectric buzzer. I put this circuit on a battery in my shack once, and when the alarm went off my unsuspecting wife ran from the house screaming, "fire!" Try getting that kind of reaction from an analog meter!

Cascading LM3914s for Higher Resolution

The resolution of a 10-segment, 10-volt display is only about 1 volt. While this is useful in many circumstances, greater resolution is often needed. It can be obtained by cascading LM3914 chips to provide a display using more LEDs. Displays of up to 100 LEDs can be obtained in this manner. Fig 5 shows a 10-volt meter using two LM3914s, for 20 LEDs. With twice the resolution of the 10-LED meter, it still can be built for only about $6.

Several changes are needed to the basic circuit of Fig 3 in order to cascade the ICs. Each chip needs its own low and high reference voltages. My approach to providing these voltages is shown in Fig 5. The on-board reference voltage is fed to an op amp, and the op-amp gain is set to give the desired high reference voltage. The other reference voltages are developed by a voltage divider. Since the high and low reference pins of the LM3914 are at opposite ends of a resistive voltage divider, successive reference voltages can be obtained

12 QEX






Cascading Lm3914 Chips

Fig 6—Two circuits to measure voltage on an expanded scale. Use the circuit at A when the difference between the high and low reference voltages is less than 1.25 volts. Adjust R1 to set the range, the difference between the voltages on pis 6 and 4, and use R2 to set the absolute voltage on pin 4. The circuit at B is used when the reference difference is more than 1.25 volts. The equations for the resistances are:


Fig 6—Two circuits to measure voltage on an expanded scale. Use the circuit at A when the difference between the high and low reference voltages is less than 1.25 volts. Adjust R1 to set the range, the difference between the voltages on pis 6 and 4, and use R2 to set the absolute voltage on pin 4. The circuit at B is used when the reference difference is more than 1.25 volts. The equations for the resistances are:



Select Rl to set LED brightness (see text).

by simply cascading the references: Connect the low reference of a chip to the high reference of the next lower chip in the cascade.

These changes are effective for the bar-graph display mode. For the moving-dot mode, a few other changes are needed: Pin 11 of each chip except the highest one should be connected through a 22-kO resistor, and each chip's pin 9 should be tied to pin 1 of the next highest chip. These changes force the highest LED of a chip to turn off when the lowest LED of the next higher chip lights.

An R1 value of 1 kll isn't totally arbitrary. The current drawn from pin 7 adjusts the current through the LEDs, controlling their brightness. This is why no current-limiting resistors are used with the LEDs, The current through each LED is about 10 times the current drawn from pin 7.

Expanded-Scale LM3914 Meters

So far, all of the LM3914 circuits have measured from 0 to some higher voltage. In applications where only a narrow voltage range is of interest, the LM3914 really shines, since it becomes a much higher-resolution device. I've used this design for several years to keep track of my shack's 12-volt battery voltage. By setting the low reference to 11.8 volts and the high reference to 13.8 volts, the meter will monitor the voltage with 0.2-volt resolution.

Fig 6 shows two variations on the expanded-scale meter. Only a few changes from the basic circuit are needed. In these examples, the meter is measuring its own supply voltage. To keep the measured voltage well below the supply voltage, a resistive divider is used. In Fig 8A, the difference between the low and high reference voltages is less than 1.25 volts, which can be obtained directly from the reference circuit. While this technique can be used for larger reference-voltage differences by tapping down on R2, interactions with the internal voltage divider begin to make calibration difficult. In Fig 8B, an op amp is used to avoid this problem.

The resolution of the expanded-scale meter can be as good as 1 %. Not bad for a $3 meter!




MODEL RA4-50 RA7-50 50-51 MHz

RA8-2UWB 144-148MHZ F012-144 144-146.5MHz 145-148MHz

Freq. #_Be Length Meas. Gain 50-51 MHz 4el 12.3 ft

F012-147 F015-144 F016-222 F022-432 F022-ATV F025-432 F033-432 F011-440

144-145MHZ 222-225MHz 432-438MHZ 420-450MHZ 432-438MHZ 432-438MHz 440-450MHz

7el 8e! 12el ^2e\ 15el 16el 22el 22el 25el 33el 11 el

26.5ft 11.8ft 17.3ft 17.3ft 25.1ft 17.3ft 14ft 14ft 17.1ft 24.3ft 6ft oo


Cost 134.95 279.95 91.75 142.50 142.50 192.50 129.95 114.95 114.95 134.95 223.95 69.95




1703 WARREN ST * NEW CUMBERLAND PA 17070 Orders 1-800-536-3268 Info. 1-717-774-3570 7pm-10pm EST


CC-1 Capacitor Kit contains 365 pieces. 5 ea. ot every 10% value Irom Ipt to 33|.l CR-1 Resistor Kit contains 1540pieces. lOea ol every 5% value Irom lOllto 10megll Sizes are 0805 and 1206 Each kit is ONLY $4995 and available tor Immediate One Day Delivery'

Order by toll-free phone, FAX, or mail We accept VISA. MC, COD, or Pre-paid orders. Company RO.'s accepted with approved credit. Call for free detailed brochure.


' 426 Wesl Taft Ave ■ Orange, CA 92665-4296 ' Local (714) 998-3021 • FAX (714) 974-3420

Entire USA 1-800-854-0547

WIFB's Help For New Hams

"Where do I begin? The license I worked so hard for is in my hands and I want to get on the air!" Let W1FB tell you how to setup and operate your first station. Watch your apprehension be replaced by confidence as you get on the air and make contact after contact. This book answers basic questions about selecting equipment, laying out your station, constructing and using antennas and of course, operating. It also describes station accessories, logging and safety concerns. In this 155-page soft cover book, Doug acts as your personal consultant— your "Elmer" in ham radio jargon. A must for the new ham! #2871 $10 plus $3 ($4 UPS) shipping and handling.


February 1993 13

Digital Communications v^^uksk

5949 Pudding Stone Lane Bethel Park, PA 15102 email: [email protected] (Internet) or 71635,1174 (CompuServe)

The first column of a new year is usually a time for reflection on previous columns. This makes it easy for the columnist to enjoy the holidays as he or she paraphrases the past. Alas, a rehash of the only other column in this series would be an obvious ploy. The only way to allow for time with friends and family is to present the following potpourri of Things You Should Know in 1993. First, some things that are so blindingly obvious, you've forgotten them.

Improving Network Throughput

Even advanced technical types may be able to use this hint. I noticed that the problem is still prevalent while listening to the Pittsburgh area packet frequencies by ear: many stations have their transmit keyup delay parameter set too high. Proper attention to this parameter can give a surprising increase in throughput. One local BBS has its delay set 400 ms too long. This adds to the turnaround time, and is especially painful when uploading to the BBS, when it sends only short acknowledgment frames. During file transfers with a window size of one frame, an extra 400 ms can add 20% or more to the transfer time at 1200 baud; it doubles the transfer time at 9600 baud.

The easiest way to select the proper value is to use trial and error. Each transmitter will be different, and you need to allow for differences in the time it takes for the receiving station's squelch (if any) to open. BBS stations will want to set their delay a little longer to allow for slow users, but 400 ms is too long. If you're using the default, get out your manual and read up on it. Many TNCs use different units for the time represented by the parameter. NET/NOS users also need to pay attention to this parameter. One user's guide gives an example of the TXDELAY parameter set to 500 ms. This is too long in almost all cases.

Inexpensive Elevation Rotor for PACSATS

Here's an idea for a very cheap ($0.00) elevation rotator for a PACSAT satellite station. You don't really need one for occasional use. If you have an antenna with a 20-degree beam width in the elevation plane (most smaller Yagis will have more), you can hear 70% or more of an average pass from most locations. Fig 1 shows the daily access times at NK6K for AO-16 or LO-19, UO-22 access tunes are similar. The graph shows that most of the time is spent at low elevations, 88% at less than 30 degrees. All low-earth-orbit satellites will exhibit sim ilar characteristics. You can borrow a 1200-baud PSK or 9600-baud FSK modem and get a taste of the action on these spacecraft without a major investment in tracking hardware. Since the software handles all the data work, you can steer the azimuth with one hand and the downlink with the other. You don't need a third hand for elevation.

There will be at least three amateur PACSAT-class spacecraft launched in 1993. KITSAT-B will be a near copy of KO-23 (KITSAT-A) and should be sending pictures as good as the one on the cover of the October 1992 QEX. ITAMSAT and UNAMSAT, from Italy and Mexico, are AMSAT-NA style microsats. With these launches, amateurs will have 70 Mbytes of data storage in orbit, which is more than most of us had on the ground when PACSATs were first proposed in 1981!

TCP in '93

This may be the year that you give TCP/IP a try. TCP/IP is closer to "real networking" than an unembel-lished AX.25 link. TCP/IP, or more properly, the Internet Protocol Suite, is a generic name for many different protocols that are used for computer-to-computer communication. If you are in Amateur Radio to give yourself a hands-on technical education, you owe it to yourself to bring TCP/IP up on your system. Unfortunately, this is not as easy as you might think.

PACSAT Access Time vs. Elevation




Fig 1—Most of a PACSAT's time is spent at low elevations.

14 QEX

NOS (or its predecessor NET) is the software that almost all Amateur Radio TCP/IP packages are based on. .There are several tutorials, beginner's guides, help files, and FAQs (Frequently Asked Questions lists) out there to help you get started with NOS. A common thread in al! of them is this concept, quoted from Mike Gallaher's FAQ: "The best place to get NOS is from someone you know who is already using it." More generally, the best way to get on the air is to have someone help you. This in itself is somewhat of a problem, since getting at someone who is on the air usually requires you to already be on the air. There are several off-the-air sources for amateur TCP/IP information. The ones I've verified recently are:

• CompuServe's HAMNET Forum, data library 9. Contains some information files and older releases. The leading edge types seem to avoid the hoipolloion HAMNET. Someone needs to cross-fertilize. Network development is no fun without users to find errors, and HAMNET is fertile ground. As part of the research for this column, I placed a copy of the Mike Gallaher FAQ on DL9.

The best place to get TCP/IP information is via FTP off the Internet. Newcomers to TCP/IP aren't likely to have access, however. If you do, try: in hamradio/packet/tcpip. Almost everything of interest will be there somewhere. Information on what is where is sparse, however. At this writing, the FAQ, a good place to start, is in: hamradio/packet/tcpip/incoming/FAQ-1292.

The best place to get a pointer to the pointers is from someone who is already on the air. There are several IP address administrators; you probably have one in your area. Bob Hoffman, the coordinator for my area, tells me that some coordinators are listed in the ARRL Operating Manual. While these guys may not have time to help you out themselves, they should know someone who does. Implementations are available for DOS, OS/2 PM, Windows/NT, Macintosh and several UNIX platforms. Work is ongoing, so get involved.

TCP/IP and Innovation

The use of the internet protocols in amateur packet radio is a good thing in general. Along with giving people hands-on experience that they might not be able to get elsewhere, and the mom and apple pie aspects of standard, interoperable protocols, there is a down side to the movement toward the internet suite. I attended the 1992 ARRL Computer Networking Conference in New Jersey, where there were the usual number of people describing problems and their approach to solving them. The question/comment period after each paper showed a trend that has been getting worse recently. Rather than a discussion of the problem or solution, someone in the back would yell "That's covered by RFC xxxx," and that would be that. (RFCs are what the internet world calls its standards documents.) The problem is that, while the RFC does address the general problem, the solution sometimes assumes bigger, faster systems and data rates. While a solution involving flooding on a set of interlinked 10-Mbit token rings might be viable in some arenas, it is not likely to be optimum for a 9600-bps, or even 1200-bps, RF link. There is a tendency to view the very-low-bit rate networks as degenerate cases, or simply uninteresting problems. An existing standard should not be rejected out of hand, and minor tweaks to it may suffice. But we still need to allow for solutions that are tailored to our unique situation. Between the limitations imposed by physics, Part 97, and the wallet, we amateurs don't always fit into the general case.


I'm excited about CLOVER because it was designed to take the real world into account1. CLOVER is a combination of hardware and software that implements a data transfer system optimized for use on HF. It uses modern techniques, unencumbered by holdovers from the days when "hardware" meant motors and gears. HAL's PCI-4000 PC adapter card implements CLOVER with an 8-MHz MC68EC000 CPU and a 20-MHz DSP56001 DSP engine.2 This is serious processing power, and it is put to use to overcome the various limitations of HF, including the ionosphere as well as the audio stages of the common HF rig. CLOVER uses adaptive equalization, variable data rates, and Reed-Solomon coding to ensure some throughput in the worst of conditions, and faster data rates when conditions allow.

HAL has just started shipping the units and the first users are getting on the air. I've seen the user documentation; it is certainly comprehensive. At least two HF forwarding BBS implementors are adding CLOVER support, and HAL is interested in helping others to port software. With all that hardware on the card, plus software development to amortize, the cost of the PCI-4000 is a bit above the DSP-based TNCs, so I don't expect that everyone will have one. Still, if the card is as good in practice as it is on paper, I'd expect to see a substantial portion of HF forwarding traffic being sent by this mechanism by the end of 1993. Some of the RTTY/AMTOR big guns are already trying this new mode.

I hope to be able to report on the initial user experiences with CLOVER in my next column. As always, you are encouraged to talk back, either to the editor, or directly to my email address.

'Raymond C. Petit, "CLOVER-II: A Technical Overview,"

Proceedings of the 10th ARRL Computer Networking

Conference, ARRL 1991, pp 125-129. ' ZHAL Communications Corp, Box 365, Urbana, IL 61801.

By: Zack Lau, KH6CP 225 Main St. Newington, CT 06111 email: [email protected] (Internet)

Conjugate Matching of Nonlinear Amplifiers

The November 1991 issue of QST had a very interesting and somewhat controversial article by Warren Bruene, W50LY, the inventor of a popular SWR measuring circuit. The article, "RF Power Amplifiers and the Conjugate Match," presented a set of measurements that indicated that conjugate matching was not present in ¡he amplifiers Bruene measured. Not surprisingly, this generated so much controversy that an additional article in the May 1992 QST was published to further explain what was happening. Let's look at conjugate matching in an amplifier so simple that we can know exactly what is happening.

Consider Fig 1, a simple class-C amplifier. Vjn is large enough to produce a square-wave output. When Q1 is on, the drain-to-source resistance, Rds, is nearly zero. When Q1 is off, Rds is effectively infinite. Given that Q1 is on half the time and off half the time, what is the output impedance of the FET? (If it makes it any easier for you, you can make Rds(on) ^ and make Rds(off) 10 kO.) Does the output impedance equal Rl? How can it, since R|_ hasn't even been picked yet? You could look at page 24 of Solid State Design, which says that R|_=VCC72P0, where P0 is the output power, but this equation is wrong for our sample circuit, since it assumes a sine wave output rather than the unfiltered square wave we actually have. You have to be careful when pulling equations out of books—you really should make sure that the underlying assumptions still hold true. Modifying the equation for our case gives; R|_=VCC*VP0.

Note that this equation implies that the source resistance is zero. Otherwise, there is a maximum output power, which occurs when the load and source resis-

Fig 1

tances are equal. But solving the equation for output power gives P0=VCC2/R|_, which implies that the output power will increase as long as R|_ is decreased. I would like to be able to publish an accurate derivation of the source impedance, but I have yet to discover one.

Now, what happens when you hook up an SWR meter, transmission line, and load to this transmitter, as shown in Fig 2? According to transmission line theory, as long as Z0=R|_, any pulse sent down the transmission line will just disappear into the resistor, R|_. Practically, we know this isn't true. All you have to do is come up with a large enough pulse to destroy RL, and you'll get a reflection! But this does explain why the SWR meter reads 1:1 even though the load and source resistances are not equal. There is no reflected power measured because there is no reflected wave to measure—all the power traveling down the coax goes into the antenna. It

Homebrew Swr Meter

Fig 2

16 QEX

isn't always true that an SWR meter indicates that the load and source are conjugately matched.

Of course, this undoubtedly goes against truths heard long enough to be self-evident.

Coherent CW Experimenters Need Help

Speaking of simple transmitters, one group of people having second thoughts is the Coherent CW, or CCW, experimenter community. Why second thoughts? Because the need to build a rig "from scratch" has inhibited the exploitation of CCW. Not everyone interested in CCW necessarily wants to build their entire station, preferring to spend more time on actual experimental issues It may make more sense for these people to modify existing transceivers, such as the Kenwood TS930 or TS450. In particular, these transceivers use a single reference oscillator, making it easy to substitute a highly accurate reference—needed for CCW—in place of the stock oscillator. Having had little success getting information from manufacturers, the two CCW experimenters listed below are interested in knowing what other commercial synthesized transceivers use a single reference, as opposed to multiple reference oscillators. Please help them if you can:

Cliff Buttschardt W6HDO 950 Pacific St Morro Bay CA 93442

Peter Lumb G3IRM 2 Briarwood Ave Bury St. Edmunds Suffolk IP33 3QF ENGLAND UK

Designing Cavity Preamplifiers

So, you want the ultimate in VHF/UHF low noise amplifiers? It's tough to compete with a properly designed cavity amplifier— except in terms of ease of construction and cost. With proper design, one can combine high selectivity with low noise, an ideal combination for preamplifiers.

You might ask, "why can't I just combine a high selectivity filter with a low noise preamplifier, simplifying the design into two easier designs?" You could, but you generally incur additional loss every time you have to make an impedance transformation. In designing the system this way, you might expect to make a transformation from the filter down to 50 ohms and another transformation from 50 ohms to whatever impedance the active device of the preamplifier wants to see. Not a problem if you can afford a dB or two of loss, but what if you need a receiver with a 21 -Kelvin noise temperature {0.3 dB NF) for low-power EME work?

Since many of our readers do like math, I'll take the liberty of highlighting some useful equations. For a single tuned circuit, the insertion loss is given by: 20log(1-Q|/Qu) dB where Q| is the loaded Q and Qu is the unloaded Q.

This equation tells us to make the ratio of the loaded Q to the unloaded Q as small as possible. Otherwise, the loss of the circuit may be too high. (Incidentally, many people get Q| and Qu confused. This has resulted in arguments whether high Q or low Q is best. You can get low losses with a high Qu, relative to Q|, or a low Q|, relative to Qu.)

Generally, Q|2>Ra/Rb-1, with Ra>Rb, where Ra and R^ are the two resistances being matched. You might want to derive the exact equations for your network, but this equation is close enough for our discussion. With real networks used for low noise preamplifiers, there is a minimum Q that depends on how different from the source impedance you allow the filter input impedance to be. This might mean that an intrinsically lower noise device, such as a PHEMT, might actually be worse than an ordinary GaAs FET at a given frequency because the required Q| is higher. It has been argued that one could just use an ideal transformer—perhaps a low-loss transmission line transformer, to match the impedances. But the dimensions for a 50-to-5000 ohm, 432 MHz (for example) transmission line transformer are a bit unreasonable!

There are also limits on how high Qu can get. While it is true that the loss of a cavity decreases as its volume—and Qu—increases, there are other considerations. Perhaps the most important is the possibility of multiple modes—different field distributions in the cavity. I think this sets a limit on the size of a cavity—the different field distributions will result in different coupling factors, which translates into a truly unreliable preamp.

Then there is the issue of silver solder. I have looked into this and found that there is little improvement in the resistivity of tin/lead solder when small amounts of silver are added. Perhaps high-silver-content solders (40+ percent) may be a significant improvement, but I don't have any data yet.

Coupling Into a Cavity

This is perhaps the biggest question concerning cavities. There are a variety of methods: loop coupling, probe coupling, etc. Unfortunately, many of the designs are empirically derived—determined by cut and try. The modern design method is to let a computer solve it via number crunching, but I've not seen an inexpensive program that will handle this for amateurs. For now, empirical designs are probably still the rule.

Getting RF Out of the Device

The preceding discussion has been about efficiently getting RF into the device. Present common wisdom is that you don't worry about getting RF out of the device efficiently. While an efficient output network may result in the best amplifier, it usually results in a lousy oscillator instead. Even at 1.3 GHz, modern FETs have more than enough gain for most applications, allowing you to waste some of it. A 47-Q drain resistor and a coupling capacitor seem to work very well with the Avantek

ATF10135, giving 15 to 17 dB of gain. That's less than the device is capable of giving, but what you gain for this loss (excuse the pun) is ease of tuning. Typically, you need at least two adjustments for the input circuit—coupling and approximate resonance. If you had two output adjustments as well, you would probably wear out the adjustments before you found the optimum settings!

Now you know "Zack's secret to a good preamp": don't tune the preamp to death. Gold-plated parts are nice, but there really isn't much gold there, so don't wear it out. Tune the preamp just a few times, as further tuning is unlikely to improve performance. It also helps to know how tuning screws shift as you tighten the lock nuts. With typical screws, further clockwise motion forces the screw into the cavity. But tightening the lock nut pulls the screw back out of the cavity, so preset the tuning screw to account for this. (Of course, it takes experience to learn just how much to preset the tuning.) I find that copper tubing makes poor threaded holes—it may be necessary to solder a nut in place rather than relying on the tapped hole, though the tapped hole may be electrically superior.

Biasing the Device

So far, I've always source biased my cavity preamps. The rationale for this is that grounding the gate of the FET allows for minimal RF losses in the input circuitry. If


Support the Amateur Space Program gate biasing, where does one optimally add a blocking capacitor and RF choke to bring in negative bias? Most implementations I've seen put the RF choke at a high impedance point—not a good idea if you want minimal loss. Of course, the question really depends on the FET used: a high gain PHEMPT may not be stable with source biasing. Then again, that instability might just be due to the designer not working hard enough.

What Device is Best?

Ideally, I'd choose a device that could be made unconditionally stable with reasonable drain loading and source feedback. In other words, it wouldn't oscillate no matter how you adjusted the cavity, while still giving you good noise figure and gain. But the latest devices seem to be getting worse, not better, in this regard. The Mitsubishi power devices such as the MGF1801 and MGF2116 seem to do the best on 144/222 MHz. At 432 MHz, the inexpensive MGF 1302 seems to do as well as anything else, possibly because the advantage to be gained from a better device is small. At 903 and 1296 MHz, the choices seem to be the MGF 1302, Avantek ATF10135, and various PHEMTs. As I mentioned before, the loss-versus-impedance-matching issue may prevent PHEMTs from outperforming ordinary GaAs FETs at 0.9 and 1.2 GHz, but there hasn't yet been enough published on designs using the different devices to draw a conclusion.

Finally, cavity preamps might be worthwhile on the higher bands, now that we have devices that want to see such high impedances for optimum noise figure. It will be interesting to see the progress made in this area in the next few years.

AM SAT Has Established Amateur Radio As a Permanent Resident in Space!

From operating any of 12 Amateur satellites circling the globe today to participating In Amateur Radio activities from the Space Shuttle, the benefits of space based Amateur Radio are available to you by becoming an AMSAT member. Our volunteers design, build and launch state-of-the-art satelites for use by Radio Amateurs the world over. We provide educational programs that teach our young people about space and Amateur Radio. Most of all, we provide our members with an impressive array of member benefits including:

• Operating aides such as discounted tracking software and land line BBS

• An extensive network of volunteers to provide you local technical assistance.

• The AMSAT Journal, your bi-monthly periodical devoted to the Amateur Space program.

JOIN TODAY. For more information, call or write for your free information packet. Or send your dues now, check or charge: $30 U.S., $36 Canada ¡Mexico, $45 all else. ($15 towards the AMSAT journal.)


A collection of outstanding articles from RSGB's RADCOMM Magazine. Covers single and multielement horizontal and vertical antennas, very small transmitting and receiving antennas, feeders, tuners, modeling, mechanics and measuring instruments, 233 pages, paperbound. Order No. 3770 $18 plus $3 ($4 UPS) shipping and handling.

ARRL7 225 Main Street,

Newington, CT 06111

Was this article helpful?

0 0


  • welde robel
    8 years ago

Post a comment