MAT 1975 HEWLETT-PACKARDJOURNAL © Copr. 1949-1998 Hewlett-Packard Co. An Understandable Test Set for Making Basic Measurements on Telephone Lines This new portable test set uses a digital processor to pre sent direct-reading, autoranged measurements of level and frequency, substantially reducing operator errors. Other conveniences simplify set-up and operation. by Michael B. Aken and David K. Deaver CONTINUING EXPANSION OF THE world's tele phone network intensifies the need for means of testing telephone lines expediently. Now more than ever, with the growth of data communications, oper ating companies and end users need test equipment that can check telephone lines quickly without re quiring a lot of personnel training. The HP Models 3551A (Fig. 1) and 3552A (Fig. 2) are dedicated test sets designed to fill this need. These instruments make basic measurements on voice-grade lines according to North American (Model 3551A) and CCITT (Model 3552A) standards. They measure tone level, noise level, and frequency, from which they obtain measurements of loss, atten uation distortion, message-circuit noise, noise with tone, noise to ground, and single frequency inter ference. Both test sets include an oscillator, a frequency meter, a level meter, and the various filters required for voice-channel measurements. They can send a test signal while simultaneously measuring it in a loop-back set-up, or two can be used as a pair for one way measurements. Combining the send and receive functions in one box makes the test set easier to carry and it also speeds measurements by making it pos sible to switch between send and receive without re connecting anything and while maintaining the telephone line in an "off-hook" condition. Design goals achieved with the realization of these test sets were: Reduction of operator set-up time because of an easily-understood front panel; Reduction of operator errors because of the autoranged, direct reading digital display; Increase in accuracy because of the low-distortion test signal with settability within 1 Hz; Increase in convenience because of the compact ness, light weight (13 Ibs), and choice of battery or line power. Front Panel Clarity A careful look at the front panel (Fig. 1) discloses how the telephone craftsman can tell the status of the instrument at a glance, and modify that status quickly with a minimum likelihood of making a wrong move. The right portion of the panel has the controls for the send unit. Frequency is selectable within a 40-Hz- Cover: Telephone channels need to be tested quickly, especially when restoring service on thousands of lines following a major disruption. Our thanks to the New York Telephone Company for the background photo of crafts men cleaning 16,000,000 relay contacts contaminated by the recent fire. The instrument is Model 3551 A Telephone Test Set, a number of which were used in restoring service. Also described in this issue is Model 5453 A, a programmable, computerized tele phone test system that uses digital signal analysis. In this Issue: An Understandable Test Set for Making Basic Measurements on Tele phone Lines, by Michael B. Aken and David K. Deaver â€” page 2 A Computer System for Analog Measurements on Voiceband Data Channels, by Stephan G. Cline, Robert H. Perdriau, and Roger F. Rauskolb . page 10 A Precision Spectrum Analyzer for the 10-Hz-to-1 3-MHz Range, by Jerry W. Daniels and Robert L. Atchley page 18 e Hewlett-Packard Company. 1975 Printed in U.S. A © Copr. 1949-1998 Hewlett-Packard Co. Fig. 1. Model 3551 A Transmis sion Test Set makes basic mea surements on telephone lines according to North American standards. The front panel is arranged to clarify operation of the instrument. The digital display is autoranged and automatically compensated for the selected impedance. to-60-kHz range and the amplitude is adjustable from -60 to +10 dBm. None of the controls is calibrated because the built-in measuring circuits enable read out of frequency and level with better resolution and accuracy than dial markings could provide. The fre quency range switch also includes a position for quick selection of the fixed holding tone used in noise-with-tone measurements. The central portion of the front panel is concerned with the type of measurement to be made on the re ceived signal. One switch selects the measurement to be made (tone, noise with tone, message circuit noise, noise to ground) while the other switch selects one of four weighting networks. Measurements can be made on tones that have amplitudes ranging from â€” 70 dBm to +15 dBm or on noise that ranges from 0 to +125 dBrn. The left portion is concerned with the physical con nection to the telephone lines. Two sets of terminals are provided to permit simultaneous send/receive measurements, each set consisting of a standard phone jack in parallel with a pair of 5-way binding posts. Either set may be used for measurements on a single pair. The FUNCTION switch selects the role of each set of terminals (send or receive) and it enables the roles to be interchanged without requiring any disconnecting and reconnecting of the lines. It also determines whether the receive line is to be bridged or terminated. A concentric switch selects the load impedance for a terminated line (and also establishes the send source impedance). A set of clip posts for connecting a lineman's hand set can be switched in parallel with the left set of ter minals for dialing up a connection. A holding circuit is provided so a connection to a "wet" line can be held in an off-hook condition while the line is used for either send or receive measurements. The readout is an LED display that gives four-digit resolution in frequency, three-digit resolution in tone level, and two digit resolution in noise level measurerhents. The quantity displayed (receive level or frequency, or send level or frequency) is selected by pushbuttons. Both frequency and level measure ments are autoranging and automatically compen sated for the impedance selected to give fast, direct readout of the measured quantity. No mental calcula tions are required on the part of the operator. A monitor loudspeaker helps the craftsman iden tify single-frequency interference and, by the char acter of the sound, the source of other types of inter ference. Battery and AC Power Each of the instruments has internal rechargable Fig. 2. Model 3552/4 Transmission Test Set is fundamentally identical to the Model 3551 A but has connectors, filters, and impedance levels that conform to CCITT standards. © Copr. 1949-1998 Hewlett-Packard Co. AutoAttenuator 15 kHz Flat Fig. 3. processing block diagram of the 3551A/3552A Transmission Test Sets. Signal processing is analog up to the detector, then processing becomes digital. The hold circuit is a current source that appears as a broadband high impedance to the telephone line while supplying the current necessary to hold central office relays. The diode bridge, protected against high line transients by the gas-discharge tube, functions as an automatic polarity switch for the hold circuit. batteries that can power the instrument for four to six hours on one charge. The instruments can also operate on ac lines of 100, 120, 220, or 240 volts. In the new test sets, a level-sensing circuit moni tors the battery voltage and shuts off the instrument whenever the voltage falls below a useful level. This prevents erroneous readings and it also prevents cell reversal from deep discharge. This arrangement is considered preferable to meter monitoring because the NiCad batteries used maintain a fairly constant voltage during use and then lose voltage rather rapidly as they approach to tal discharge. With the usual meter monitoring, the operator would have to check the battery voltage quite frequently to avoid overlooking the onset of the rapid voltage fall. Internal Details The two instruments are fundamentally the same except for certain characteristics that conform to the telephone measurement standards where the instru ments are to be used. The Model 3551A has imped ances, weighting filters, and a hold tone that conform to standards established by the telephone industry in North America. In most of the rest of the world, stan dards are set by the International Telegraph and Tele phone Consultative Committee (CCITT) and the Model 3552A conforms to these standards. The block diagram shown in Fig. 3 applies to both instruments. The received signal is filtered and de tected in conventional analog fashion. The output of © Copr. 1949-1998 Hewlett-Packard Co. the detector, however, is converted to a proportionate time interval in the logarithmic converter. The digital circuits measure this time interval to get a digital indication of the input signal level. From here on, the measurement information is manipulated digitally, with appropriate factors added so the number dis played gives the measurement in the desired units. In a frequency measurement, the signal frequency is multiplied in a phase-lock loop so the counting circuits can accumulate enough counts in 50 millise conds for 4-digit resolution. This enables a 10-per-second sampling rate, even at low frequencies. One of the significant differences between these in struments and other telephone test sets is the use of a function generator as the send unit, rather than the traditional RC oscillator. The use of a function gener ator achieves significant cost and space savings. The basic function generator circuit was modified, how ever, to obtain a sine wave with very low distortion. Triangles First A schematic representation of the send unit's func tion generator is shown in the box on page 6. As in other function generators, a triangular waveform is generated by using constant currents to alternately charge and discharge a capacitor, shown as range ca pacitor CR in the diagram. The triangular wave is shaped into a sine wave by a nonlinear network. Diodes in this network are biased at progressively higher levels so more and more attenuation is switched in as the triangular waveform moves towards its posi tive or negative limit. By suitable choice of the attenu ation switched in by each diode, the triangle is rounded off to a sine wave. Maintaining high sine-wave purity under field conditions required some modifications to the basic waveform-generating mechanism. To begin with, the sine shaper is compensated for temperature changes by diodes that adjust the voltage on the bias ing networks to compensate for the change in diode forward voltage drop caused by temperature changes. For optimum operation of the sine shaper, the up slope of the triangle wave must be exactly equal to the down slope and the waveform must be centered on the zero level (equal positive and negative excur sions). Circuits for maintaining these conditions are designed into the function generator. These are de scribed in the box on page 6. As a result of these measures, total harmonic distor tion in the send unit output is more than 50 dB below the fundamental within the telephone voice band and at least 40 dB below outside that band. These specifications are held over an operating tempera ture range of 0Â°C to 55Â°C. Level Detection As noted in the block diagram of Fig. 3, the re ceived signal is appropriately filtered and adjusted Logarithmic Converter V, = dc Proportional to Signal Level Fig. 4. quasi-rms combines the outputs of peak and average detectors to derive a quasi-rms indication. The detector output is converted to a proportionate time interval by the logarithmic converter. © Copr. 1949-1998 Hewlett-Packard Co. A Function Generator with a Well-Defined Output The triangular waveform is shaped into a sine wave by the nonlinear network described in the text preceding. To assure proper operation of the sine shaper, the triangular waveform must be symmetrical. Equal positive and negative waveform ex cursions (x-axis symmetry) are assured by integrating the trian gular waveform and using the resulting dc level to modify the lower steady-state level in the bistable switch. The waveform's peak negative excursion is thus adjusted to equal the magni tude of the peak positive excursion. Maintaining equal up and down slopes (y-axis symmetry) is accomplished by sensing the symmetry of the square wave. Re ferring to the y-axis symmetry circuit in the diagram, current lb flows into the adjacent integrator when the square wave is at the upper level, reverse-biasing diode D1. When the square wave is at the lower level, the reverse bias is removed from diode D1, allowing the lower current source to draw a current equal to 2IS through diode D1 , with half of the current (ls) being drawn from the integrator. Thus, as long as the waveform is sym metrical, the average integrator output is zero. If the waveform were not symmetrical, a net charge would remain on the inte grator output, which would add to or subtract from the charging current in the triangle generator. The up slope of the triangle is thereby altered to make it match the down slope. Definite space-saving and cost advantages result from using a function generator as the send unit's oscillator in the Model 3551 A/3552A Transmission Test Set. Certain modifications, how ever, had to be made to the basic circuit to assure the wave form purity required for telephone tests. A simplified diagram of the function generator is shown be low. Positive and negative currents are switched by transistors Q1 and Q2 to alternately charge and discharge capacitor CR, thereby generating a triangular waveform. To minimize switch ing transients, steady current flow is maintained by operating transistors Q1 and Q2 in a bridge configuration with Q3 and Q4. This arrangement sinks one current to ground while the other is charging the capacitor and vice-versa. The transistors are turned on and off by a bistable switch that changes states when a comparator indicates that the triangle waveform has reached the same level as that being held by the switch. The amplitude of the triangular waveform is thus deter mined by the bistable switch, a precision circuit that maintains a well-defined level in either of its two states. The frequency is determined by the rate at which the capaci tor charges and discharges. The vernier frequency control changes the charge and discharge currents. Ranges are changed by switching in capacitors of different values. X-Axis Symmetry 3.3V 3.3V capacitor to the negative peak of the waveform. The average detector integrates negative half-cycles of one waveform to get an average value. The summing resistors at the input to the integrator determine the ratio of peak to average in the combined result. Dur ing tone measurements, only the average detector is used. in gain before being applied to the detector. For noise measurements, the detector functions as a quasi-rms type that derives the equivalent of an rms measure ment by combining the outputs of peak and average detectors in accordance with Bell System Technical Reference PUB 41009. The peak detector (Fig. 4) is a diode that charges a 6 © Copr. 1949-1998 Hewlett-Packard Co. To Digital Counter From Attenuator Amplifier Fig. 5. Frequency multiplier generates an output two times the input for inputs of W kHz and higher, and 20 times higher for inputs below 10 kHz. The detected dc level is then applied to the log con verter. This compares the detectors' dc level to the voltage on a discharging capacitor, generating a pulse when the two are equal. Ten times per second during tone level measurements, switch Si (Fig. 4) closes long enough to charge capacitor C1 to a fixed level. When Si opens, Ct discharges and when it falls to the level of the detector voltage, a pulse is generated. The time interval between the opening of Si and the comparator pulse is thus inversely propor tional to the absolute value of the input voltage and, since the C^ discharge curve is exponential, to the logarithm of the input voltage. The digital circuits measure this time interval to derive the number for display. The 100-kHz clock frequency gives a mea surement resolution of 0.02 dB, which is truncated to give a display resolution of 0.1 dB. During noise measurements, an additional lowpass filter is switched in at the input to the log con verter to provide more averaging for the noise sig nal. Switch Si is then activated only two times per second. The digital circuits also use the output of the log converter to sense when the attenuator range should be changed. If the time between the opening of switch Si and the comparator pulse is less than 5 ms (sig nal too large), the attenuator is up-ranged. If the time is greater than 20 ms (signal too low), the attenuator is down-ranged. The 15-ms interval between these points is equivalent to 15 dB, giving a comfortable overlap of the 10-dB range on each attenuator step. The attenuator consists of resistive dividers with taps switched by an eight-channel analog multi plexer under control of the digital system. The con trol sequence is such that an amplitude measurement and range correction is always made before a fre quency measurement is made. The frequency mea suring circuits are thus assured of a suitable signal level. Fast-Responding Frequency Measurements To get 1-Hz resolution in a conventional frequencycounter measurement of an audio frequency, say 4 kHz, a one-second counting time is required. A sampling rate of 10 per second was desired for the Models 3551 A/3552A so the results of adjustments to the telephone line are immediately apparent to the craftsman. The counting time was thus made 50 milli seconds, which allows time for an amplitude mea surement within each 100-millisecond measurement interval. To achieve 1-Hz resolution in this time in terval, input frequencies lower than 10 kHz are mul tiplied by a factor of 20 before counting. Input fre quencies of 10 kHz and higher are multiplied by 2, giving 10-Hz resolution with the four-digit counting circuits. The frequency-multiplication circuit is shown in Fig. 5. A commercially-available integrated-circuit phase-lock loop is at the core of the circuit. It has a voltage-controlled oscillator controlled by a phase detector that compares the input signal to a divideddown version of the VCO output frequency. The VCQ is thus locked to a multiple of the input frequency. Digital Control The operating simplicity of these instruments re sults from the use of a digital processor to manipu late the raw measurement information. The proces sor monitors the signal level and frequency and the front-panel control settings, and uses this informa tion to derive control signals for the measurement routines and for the display. The digital processor (Fig. 6) is an algorithmic state machine (ASM) that uses MSICs (medium-scale integrated circuits) and 8000 bits of memory in lowcost ROMs. It is divided into two parts, a control sec tion and a display section, although functions over lap somewhat. A four-digit counter that is part of the display section performs measurements of both level and frequency. It counts 100-kHz pulses for the duration of the log converter's measurement interval to derive a digital number proportional to signal level. During the 50-ms frequency measuring interval, it counts the output of the frequency-multiplier circuit (Fig. 5). At the end of a frequency measurement interval, the counter's contents are latched into a register and held there for direct application to the display. The contents of the latch register are scanned digit-by-digit and applied to a seven-segment decoder that drives the display LEDs one at a time. Scans occur 2500 times per second. © Copr. 1949-1998 Hewlett-Packard Co. Clock Range Up Down Counter â€¢^^ From Log Converter Input Qualifier from Display and Receive Switch Controller Output Qualifier and Output Select Lines Digit Activation Lines V7 Measurement Interval Programmed for Digit Scanning Time Base and Clock Circuits Frequency Multiplier Frequency Underrange to Controller Qualifier Select Block From Frequency Multiplier â€¢ Range Block to Automatic Amplitude Ranging Circuit and Qualifier Select Block Display Control Lines Fig. 6. Simplified block diagram of Test Set's digital circuits. the qualifier select block, it monitors timing signals and determines the proper frequency range and the points at which autoranging is initiated. It partitions the 100-ms sampling interval into the 25-ms signal amplitude measurement interval and the 50-ms fre quency measurement interval. It issues steady-state command voltages for opening and closing switches, and pulsed commands for initiating timerelated functions. At the end of an amplitude measurement, the regis ter contents are used as an address to find a new number in the display ROM according to the mea surement being made. For example, in tone measure ments, a 0-dBm input signal results in a log converter time interval of 10 ms (1000 counts). This is dis played as 0 dBm. A higher count (longer time inter val, lower amplitude) is interpreted by the display ROM as a larger number (dB below 0 dBm) and it adds a minus sign to the display. A lower count (shorter time interval, higher amplitude) is also dis played as a larger number (dB above 0 dBm) but with a plus sign. Although the counter is unidirec tional, it appears to behave as a reversible counter be cause of the ROM processing. Noise measurements are processed in a similar manner so the correct number is displayed. In the case of measurements of noise-to-ground, the ROM adds 40 dB to the measurement number to correct the reading for the signal attenuation caused by the noise-to-ground input configuration (see Fig. 3). The craftsman is never required to add this correction mentally. Using firmware to process the numbers, rather than adders, registers, and other hardware, reduced the parts count significantly and hence reduced the cost. The control section of the digital processor uses a 4K ROM that contains all of the instructions for the level, noise, and frequency measurements. Through Test and Calibration The control section ROM also contains instruc tions for a progressive series of tests for troubleshoot ing both the digital components and analog subsys tems. These are internally-accessible tests, which means that they cannot be accessed by the operator from the front panel but they can be used as a trouble shooting tool by maintenance personnel, either in the shop or in the field. Tests may be conducted in either a fixed informational display mode, which indicates proper operation of the test routines and thus the instrument, or a dynamic response mode using an oscilloscope, which allows individual IC's, circuits, and subsystems to be examined at normal operating repetition rates. Some of the tests that can be performed are: â€¢ High-speed exercising of all transfers, resets, com mand outputs, and flip-flops to assure their proper operation. 8 © Copr. 1949-1998 Hewlett-Packard Co. points than would have been possible manually while at the same time reducing test time. The sys tem automatically sets a digital voltmeter and a synthesizer to the proper ranges, checks the frontpanel control settings of the instrument under test, tells the technician the adjustment to make, and checks that the calibration is within prescribed limits. â€¢ Monitoring the display ROM outputs to check proper operation of the ROM and its associated cir cuitry. â€¢ Fast or slow repetitive operation of the analog autoranging circuitry to check proper gains and tran sient responses. One of the cost-saving features of these instru ments is the inclusion of two test sockets that allow connection to a calculator-controlled test system. The inclusion of a digital controller in the Models 3551A/3552A made it relatively simple to make most of the pertinent information available to the HP Inter face Bus.1 It was thus possible to design a calculatorcontrolled test system that assures that the necessary calibrations are performed. It also checks more Acknowledgments The authors would like to acknowledge the contri butions of Ray Hanson for the design of the send oscil lator and as the group leader of the 3551A. Dick Huff man provided the mechanical design. S Reference 1. D.W. Ricci and P.S. Stone, "Putting Together Instru mentation Systems at Minimum Cost," Hewlett-Packard Journal, January 1975. David K. Deaver Dave Deaver joined HP in 1969 after earning his BSEE degree at Washington State University. Since being with HP, he con tributed to the designs of the Models 3575A Gain/Phase Meter and the 3570A Network Analyzer be?ori? jninina the 3551A project. He is working towards his MSEE degree at Colorado State Univer sity in the HP Honors Co-op pro gram. Dave plays basketball in the Loveland, Colorado, com munity intramural program and carries this activity over into community youth work, refereeing Saturday morning basket ball games. He and his wife have one daughter, 2. Michael Aken With HP since 1966, Mike Aken worked on voltmeters (3450A and 3480A) and the 3570A Network Analyzer before joining the 3551 A project. A graduate of the Univer sity of Wisconsin (BSEE), he earned his MSEE at Colorado State University in the HP Honors Co-op program. Married, and with three children, 10, 8 and 5, Mike enjoys gardening, both outdoor and indoor with fluores cent lights, and he plays some tennis. SPECIFICATIONS Model 3551 A 3552 A Transmission Test Sal Receive Section Noae-Mtn-tone -SOdBmlo 5 dBm (60011. 900Ã1) Noae-to-growna - SO dBm 10 * 35 dBm' Resolution 1 OB Sample rite 2 second Detecto) type RMS responding ACCURACY Message orcmt naÂ»e -1 dB (-70 dBm 10 -5 dBm). -2 dB (-90 dBm to - 70 dBm) None-mttvtone Â±idB( -70 dBm to -5 dBm} *2dB( -SOdBmto 70 dBm) Nose-to-grotnd â€¢ ' dB ( - 30 dBm to * 35 dBm) - 2 dB I - 50 dBm 10 - 30 dBm) WEIGHTING FILTERS Teiephon.(CClTTPÂ»ophometnc) ShHziat iSXHifiai programme (CCITT) FREQUENCY MEASUREMENTS FREQUENCY RANGE 40 Hi to 00 kHz DYNAMIC RANGE - 1 5 dBm to - 70 dBm RESOLUTION 1 Hz (40 Hz to 10 kHz}. 10 Hz ( 10 kHz to 60 kHz) SAMPLE RATE 10/eecond ACCURACY - 1 count LEVEL MEASUREMENTS FREQUENCY RANGE 40 Hi to 60 kHz DYNAMIC RANGE - 1 5 dBm lo - 70 dBm RESOLUTION 0 1 08 SAMPLE RATE 10>MCOnd DETECTOR TYPE â€¢vO'Â·gÂ· rMfxxxttig ACCURACY FREQUENCY *OH1 100H2 u j 10KMJ 20KH2 80 KMl - 1 5 Ã¯ -10 200 Hz 13511 IMPEDANCE NOT SPECIFIED BELOW 200 Hz NOtSE MEASUREMENTS (3S51A) DYNAMIC RANGE Nowe-Mth-tone lOdBmio -85 dBm ,80011 9001 It Nowe-to-oround -40 dBm W â€¢ 125 dBm RESOLUTION 1 dB SAMPLE RATE Â¿Mcond DETECTOR TYPE Quasi RMS ACCURACY Meeaage orcuri none Â±1 dB i -20 dBm to -85 dBm). -2 dB |0 dBm w -20 dBm) Nowe-wttvtone ri dB ('20 dBm to -85 dBm) -2 dB (10 d6m to â€¢ as dBm) NCtte-tO-ground :1 dB i-8C dBm lo -125 dBm]. Â±2 dB I -40 dBm to â€¢ 80 dBm} WEIGHTING FLTERS C-manaoa. 3 kHz ftat. 15 kHz tat. program NOISE MEASUREMENTS (35S2A) DYNAMIC RANGE Message a'ciÃ-t nasÂ» -90 dBm lo -5 dBm S*ndSsction FREQUENCY RANGE: 40 Hz K> BO kHz RANGES 40 HZ to 600 Hz 200 Hz to 6 kHz 2 kHz lo 60 kHz. 1004 Hz teed (3551A) or BOO HZ |3552A) Ovwr frequencies avariaole for 3552A HARMONIC OtSTOfmON: - -50dBTHD(100Hzio4kHz) > -40dBTHD(40Hz Â» -00 Hz and 4 kHz to 20 kHz) --55 dB (al hermorao 100 Hz K 4 kHz), â€¢ -60 dB THO ( 1004 Hz fiied) ACCURACY: -1 oount LEVEL RANGE: - *0 dBm lo -60 dBm (40 Hz to 00 kHz) -6 dBm to -60 dBm (1004 Hz fixed) RESOLUTION: 0 1 dB 9 © Copr. 1949-1998 Hewlett-Packard Co. ACCURACY 40 Hz â€¢â€¢Ml MONITOR cu'i-in speaker montÃ³n received ex tranarmned *gn* BALANCED IMPEDANCES: '3511 (3551A|. 15OI1 (3552AJ. 60011 90011 BRIDGING LOSS OlOB RETURN LOSS: â€¢ 30 dB LONGITUDINAL BALANCE: -60 dB at 6 kHz HOLD r 24 TÂ»Â»ampÂ« concur* current >50 kfl impedance, r haÂ» protection INPUT OUTPUT PROTECTION Mocfcs 300 V dc maximum tongrtudneJ vot lag* 200 V rmÂ» BATTERY SUPPLY: 4-6 hours contvuous operafcon on Â«*em*J rechargeable battene* ai25'C Battery dram n aulomaucaiy lurr^ off Â«rnen oHcharged beto* propÂ»' Operating kevel Complete 'Â«charo* m 12 hours ACUNE: 100 120 200 240 V Â±10% 48 Hz to 440 Hz. 14 VA TEMÂ«RATUÂ« RANG* OPERATING 0*C U M'C (32"F lo 130*F) STORAGE 20-C to 65'C ( -4'F to 148*F) RELATIVE HUMnfTY: 0Â»95%(<100*F >40*C) WEIGHT NÂ« 66Hg|l3b) DIMENSIONS: 133 mm M â€¢ 343 mm W * 254 mm D with front-paneÂ» cow (5 25 PRICES IN U.S.A.: 35S1A $1750 35S2A $2000 MANUFACTURING DIVISION: LOVELAND INSTRUMENT DIVISION P O BoÂ» 301 815 Fourteenth Street. S W LoveÃ-and Colorado 00537 A Computer System for Analog Measurements on Voiceband Data Channels Besides making nine data-channel performance tests automatically in less than two minutes, this new Trans mission Parameter Analyzer is capable of a much broader range of measurements. by Stephan G. Cline, Robert H. Perdriau, and Roger F. Rauskolb sentative list of measurements appears in Table I. There are two categories, one consisting of general measurements that may be useful in a broad range of applications, and the other consisting of nine charac teristics of voiceband data channels that are common ly measured in North America. The 5468A Trans- MANKIND TODAY POSSESSES a truly stag gering ability to disseminate information. Much of this ability depends upon networks that transmit in formation in the form of electrical signals to the far corners of the earth and deep into space. Over the past decade, it has become necessary to transmit a grow ing volume of information at higher speeds and with greatly increased accuracy and reliability. This has led to the imposition of increasingly stringent per formance requirements on the transmission networks and the components that comprise them. The 5453A Transmission Parameter Analyzer (TPA), Fig. 1, has been developed to aid in the design, manufacture, installation, and maintenance of today's high-performance voice-grade communication chan nels and components. Transmission parameters are the properties of an electrical path that must be suitably controlled if information is to be successfully transmit ted over the path. The path may be a simple amplifier, for which frequency response, noise, and distortion are the primary parameters of interest, or it may be as complex as a long-haul telephone channel that must be optimized for high-speed data transmission. The 5453A TPA performs digital signal analysis, measuring signal properties using computational techniques rather than analog circuitry. No hardware detectors, demodulators, or signal generators of the type normally found in analog instrumentation are incorporated in the 5453A. New measurements may be implemented, or existing measurements modified, purely through software. Programs can be written in either FORTRAN or BASIC together with a simple calculator-like language. The 5453A also offers sig nificant advantages in terms of speed and accuracy when compared to equivalent analog instrumentation. The system is capable of both stimulus-response (network analyzer) or response-only (spectrum ana lyzer, power meter, counter) measurements. A repre Fig. 1 Model 5453 A Transmission Parameter Analyzer is both a general-purpose stimulus-response test instrument and a special-purpose analyzer for performance tests on voiceband data channels and components. Table I lists some of its capabilities. 10 © Copr. 1949-1998 Hewlett-Packard Co. Stimulus Generation ponder (see box, page 16) is used with the TPA for many of the data-circuit measurements. A simple block diagram of the system appears in Fig. 2. A digital-to-analog converter (DAC), together with suitable filtering and impedance matching, generates the desired stimulus. On the response side of the device being tested, an analog-to-digital con verter (ADC) samples the incoming waveform and prepares it for digital processing. The mass memory provides storage for measurement and analysis pro grams, test stimuli, test results, and other data that may be required for a measurement, such as digital filters. A CRT terminal serves as the sole operator interface. From it, programs can be prepared and executed and tabular results displayed. No manual controls are present or necessary. An optional CRT display provides graphical output. Virtually any desired stimulus may be quickly and easily designed from the CRT terminal. The stimulus may be as simple as a sine wave or as complex as pseudorandom noise. It may be described as a wave form in the time domain or as a spectrum in the fre quency domain. In the latter case, the user has com plete control of both the amplitude and the phase of each spectral component. Spectra and the corresponding waveforms can be designed in minutes, stored in the mass memory, and retrieved either on command from the operator or automatically by the measurement program. Before output, the waveform is represented in the computer memory by a block of 16-bit words, each word repre senting one time sample of the waveform. To convert this to an analog signal, the 5453 A has a 13-bit DAC followed by programmable gain circuitry. A com bination of block scaling and gain setting is used to achieve the desired peak or rms power output. Sig nals can be generated over a range of 0 to â€”40 dBm with distortion products down 60 dB. The number of words per block is variable in powers of 2 from a minimum of 64 to a maximum of 4096. The number is selected along with the scan ning rate to achieve the frequency range and resolu tion required in a given application. Generation of the test signal is accomplished by causing the pro gram to read the block out through one of the com puter's high-speed direct memory access (DMA) channels to the DAC. This process may be contin ued for as long as desired, resulting in a periodic sig nal being applied to the test device. The frequency range is limited by the settling time of the DAC and the DMA rate of the computer. Frequencies up to 10 kHz may be generated by the 5453A. Table I Representative Measurement Capabilities of the 5453A Gain/Loss Phase Envelope Delay Frequency Power Noise Nonlinear Distortion Impedance Frequency Modulation Phase Modulation Amplitude Modulation Conversion Loss Characteristics of Four-Wire Data Circuits Measured by the 5453A/5468A Attenuation Distortion Envelope Delay Distortion Level Message Circuit Noise C-Notched Noise Phase Jitter Intermodulation Distortion Frequency Shift Single-Frequency Interference 5478A System Interface Graphic Display (Optional) Digital-toAnalog Converter Channel A Analog-toDigital Converter CRT Terminal Channel B Analog-toDigital Converter Direct Digital-to-Analog 1 Converter Output Anti-Aliasing Filter, Impedance Selection Voiceband Digital-to-Analog Converter Output Direct Channel A Input Anti-Aliasing Filter. Impedance Selection Direct Channel B Input (Optional) 11 © Copr. 1949-1998 Hewlett-Packard Co. Voiceband Analog-to-Oigital Converter Input Fig. 2. Transmission Parameter Analyzer block diagram. A digitalto-analog converter provides stimulus signals for the network being tested. Analog-to-digital converters sample the network's response. Parameters of interest are then computed digitally. Digital Processing tion can be executed and the results displayed im mediately upon entry. In this mode the 5453A is used very much like a general-purpose scientific calculator. Engineering personnel can learn to operate the sys tem in one or two hours. Once the desired keyboard program has been written, it can be executed directly or, more typical ly, it can be stored in the disc memory and retrieved and executed using a CALL statement from a FOR TRAN or BASIC controlling program. Both the con trolling and keyboard programs have access to any data blocks stored in the disc memory. The disc mem ory is also used to transfer data from one program to another. In this manner, measurement data result ing from the execution of a keyboard program is avail able to the controlling program for further computa tion, formatting and output, or decision making. In addition, data may be synthesized by the controlling program and passed to a keyboard program for use in a specific measurement. The disc memory is capable of storing a large number of keyboard programs, any one of which can be executed at will by the con trolling program. The result is that powerful digital signal analysis capabilities are now available in the context of standard engineering-oriented computer languages. The number of possibilities for digital processing of the raw data is large and they cannot all be dis cussed in this article. We will discuss two examples that illustrate processing of steady-state and random signals and then look at an example of using the 5453A to generate or simulate a desired impairment. On the response side of the 5453A, the incoming waveform from the test device is sampled by the ADC and converted to a block of 16-bit words representing successive time values of the input. The incoming waveform may have been generated by the DAC and distorted by the test device, or it may be an external signal. From this point, digital processing in either the time domain or the frequency domain is used to extract the pertinent information. Because sampling is used as the means of gathering the raw data, we must, of course, be aware of the con straints imposed by aliasing, leakage, and quantiz ing noise.1 The balanced, voice-frequency ports of the 5453A are provided with seven-pole elliptical filters that keep aliased products down at least 50 dB. For applications outside of the filter frequency range, the direct DAC output and ADC inputs must be provided with suitable anti-aliasing filters. In put frequencies greater than 100 kHz can be accom modated. Leakage is reduced by using Manning or other appropriate windows on the data or, in some cases, by measuring the amount of leakage and ac counting for it. Dynamic range greater than 70 dB is obtained and the system noise floor over the voice fre quency band is approximately â€”90 dBm. Digital signal analysis, by its very nature, involves operations on or between blocks of data words. These operations include block arithmetic, forward and inverse Fourier and Hubert Transforms, power spectrum, convolution, correlation, integration, and so on.1'2 In most practical situations, an ordered se quence of such operations must be performed on the raw input data. Programs to accomplish this may be written in FORTRAN. However, an alternative soft ware approach, developed for the 5453A, provides a simple keyboard language that may be used to call for any desired sequence of block operations. The name "keyboard" derives from the fact that any block operation may be programmed by pressing at most two keys on the CRT terminal. Each block opera Measuring Insertion Loss and Phase As a first example, suppose it is desired to measure the insertion loss and phase of a two-port network. Insertion loss and phase are defined as follows: Insertion Loss = 20 Log [V0(f)/VN(f)] (1) Insertion Phase = [<f>0(f) - </>N(f)] (2) Fig. 3. spectrum typical insertion loss measurement on a two-port network showing the spectrum (a) of the test (c) generated by the TPA, the test signal itself (b), and the network's insertion loss (c) computed by the TPA. 12 © Copr. 1949-1998 Hewlett-Packard Co. where the "o" subscript refers to conditions at the load with the source directly connected, and the "N" subscript refers to conditions at the load with the network inserted. V(f) and <Â¿>(f) are voltage and phase expressed as a function of frequency. Further assume that the measurement is to be made at the frequencies contained in the spectrum of Fig. 3a. A simple program to accomplish this measure ment might first instruct the operator to bypass the network and connect the source directly to the load. It would then generate the waveform of Fig. 3b, cor responding to the spectrum of Fig. 3a. Next it would sample the DAC output, deriving a 512-word block of data (representing, in this case, a 64-ms time rec ord), and compute the complex spectrum. In practice, several such records would be sampled and averaged to reduce the effect of external noise. The resulting averaged spectrum would be saved, the operator in structed to insert the network, and the process repeated. We now have two complex spectrums, VolO/foolf) and VN(f)/<ftN(f). Performing the calculations indicated in equations 1 and 2 yields the desired re sult. Accuracies of Â±0.1 dB and Â±0.2 degree are ob tainable, and the measurement can be accomplished in only a few seconds. Fig. 3c illustrates the results of an insertion-loss measurement on a simulated communication channel. Noise Measurements The measurement of noise is perhaps one of the most common maintenance activities in telecom munications, and the 5453A offers several capabilities in this area. A conventional measurement of noise might be modeled as shown in Fig. 4. In the case of a telecommunication channel, the input signal spec trum Sxxif) may be zero (input terminated), or it may represent a holding tone intended to bias compandored facilities to their normal operating points for continuous signals. In the latter case, the tone is re moved by including a notch at the appropriate fre quency in the transfer function of the weighting net work. In either event, the weighted spectrum Syy(f) is indicated on an rms-responding power meter. S,x(f) . Weighting Network Rms Responding Power Meter Fig. 4. Model illustrating the measurement of noise. The S,,(f) represent complex voltage spectra as functions of frequency. The transfer function of the weighting network is H(f). From the model we write: Syy(f)=H(f)*[Sxx(f)+Snn(f)] (3) or, for Sxxif) = 0 Syy(f) = H(f)*Snn(f) 4] Since we are interested in power, we multiply each factor in equation 4 by its complex conjugate and write: GÃ-f)= |H(f)|2*GÃ-f) (5) In words, the weighted noise power spectrum is equal to the unweighted power spectrum multiplied by the squared magnitude of the transfer function of the weighting network. The measurement is implemented with the 5453A TPA by first causing the TPA to gather a record (data block) representing the waveform associated with Snn(f). From this raw data, the complex spectrum Snn(f) is computed, followed by computation of the power spectrum Gnn(f). In practice, this process is repeated several times and the computed Gnn(f) are averaged to obtain a reliable estimate of the noise power spectrum. With an estimate of Gnn(f) available, the desired weighting is applied in accordance with equation 5. The weighting function, |H(f)|2, may be obtained by actual measurement of a physical net work or it may be computed from the ratio of polyno mials that describe the network. In either case, it is most often stored in the mass memory and used as needed. The resulting data block, representing Gyy(f), is then integrated to obtain the total (mean square) weighted noise power. Specific frequencies, such as 60 Hz and its significant harmonics, may be elimina ted by excluding them from the limits of integration. The resultant data block is then passed, via the mass memory, to the controlling program, where it is con verted to the appropriate units (dBm, dBrnC, etc.) for output. Each of the steps is called for by simple keystrokes and, once programmed, may be repeated as desired. It is possible to apply any number of weightings with out repeating the measurement. The power spectrum is available and may be scanned by the controlling program to determine the frequency and level of any interfering tones that may be present. A typical flatweighted noise power spectrum with an interfering tone at 1 kHz and -53 dBm is illustrated in Fig. 5. Finally, if Sxx(f) is non-zero â€” representing, for ex ample, the output of a data set â€” it is possible to check for proper operation as regards both the frequencies and power level transmitted. 13 © Copr. 1949-1998 Hewlett-Packard Co. eÂ¡ (t) = Acosat + Bcos/3t SOdBm 171 Substituting (7) into (6), applying the appropriate identities, and assuming A = B for simplicity, we obtain: + a1 A(cosat+cos/Jt) (cos2at + cos2/3t) + a2A[cos(a-/3) t + cos(a+/3) t ] (8) The output contains a dc component, linear terms at a and /3, second harmonics, and sum and differ ence frequencies. Second-order distortion is the ratio of the power at the sum and difference frequencies to the power at the fundamentals. By selecting a^ and A, it is possible to compute a2 for any desired amount of second-order distortion. The input spectrum can now be entered into a data block, transformed to the time domain, and the cal culations indicated by equation 6 performed. The result is a test signal that can be used directly to evaluate system performance. Fig. 6b illustrates a test signal of this type as it appears on a swept spec trum analyzer while being output by the DAC. This example is also an excellent illustration of the potential speed advantage of digital signal analysis. The spectrum in Fig. 6b was taken with an analog spectrum analyzer over a 2-kHz sweep width with 3-Hz resolution. Approximately one-half hour was required for a single sweep. The same result, over a wider bandwidth, with equivalent resolution and dynamic range can be computed by the 5453A in approximately twelve seconds. 4 kHz Fig. 5. Typical flat-weighted noise power spectrum com puted by the TPA. An interfering tone is apparent. Simulating Impairments Digital signal analysis can also be used to simulate known impairments. Fig. 6a shows the power spec trum of a clean test signal containing energy at 703 and 1172 Hz. Such a signal might be used as a stimu lus when performing measurements of intermodu lation distortion. Suppose that we have devised a system for mea suring the second-order intermodulation distortion of a device using the signal in Fig. 6a as a stimulus, and we wish to test this measurement system. To do this we need a means of creating known and variable degrees of second-order distortion. The design and construction of a physical device to do this is expen sive and time-consuming. An alternative is to use the computational ability of the 5453A to generate known distortions. We begin by assuming a nonlinear device transfer function given by: e 0 ( t ) = a ^ i i f l + a z e f l t ) ( 6 ) Telephone Channel Measurements The testing of telephone circuits used for the trans mission of high-speed data is a difficult problem. While many different types of voice-grade data chan nels are available, we will limit our discussion to twopoint, private-line, four-wire circuits. Such a circuit where a^ is the linear component of the transfer func tion and a2 is related to the degree of second-order nonlinearity. Next, we describe our test stimulus as: Fig. 6. Spectrum of a typical stimulus signal for intermodulation distortion measurements (a), and the spectrum of a typical response (b). In this case the response signal was generated by the 5453A and contains a known amount of intermodulation dis tortion. It could be used to test a distortion-measuring system. Other known impairments are also easily generated. 14 © Copr. 1949-1998 Hewlett-Packard Co. Tâ„¢Â°n Fig. 7. Two-point private fourwire data circuits like this one can be measured in both directions from the test center by the 5453A. Circuits not passing through the test center can also be tested. Transmission System might be laid out as shown in Fig. 7. Several transmis sion media may be used between intermediate points along the circuit, including PCM, FDM cable, and mi crowave systems. The user leases the circuit from a common carrier and it is available to him on a fulltime basis. Most transmission systems were originally de signed to enable people to talk to people. As data traf fic has increased in volume, speed, and importance, a number of circuit characteristics, most of which offer little or no degradation to voice traffic, have be come significant. A list of the parameters presently measured on data circuits by the 5453A appears in Table I. The interested reader unfamiliar with the ter minology will find additional information in refer ences 3 and 4. The parameters encompass such fun damental characteristics as insertion loss, power, fre quency, noise, distortion, and incidental modula tion. A circuit like that of Fig. 7 may be tested from end to end or between any two points at which voice-fre quency access is available. Assume that a 5453A is lo cated in the test center and that we wish to character ize the portion of the circuit from that office to user lo cation B. The transmit and receive sides of the circuit could be connected together at the user location. This would form a loop and the line could be treated as a two-port network by the 5453A. However, this approach does not make it possible to separate the characteristics of the two sides of the circuit. Its major usefulness lies in the ability to characterize a knowngood circuit on a looped basis and to save the result in the mass memory. Subsequent troubles may then be traced to either the circuit or the terminal equip ment by repeating the measurement and detecting changes from the benchmark. This can normally be accomplished without dispatching a trained repair man to the remote location. The 5468A Transponder has been developed to provide for two-way measurements between distant locations. When connected to the circuit at the user location, it can be commanded automatically from the 5453A to generate the test signals required for measurement of the receive line or to process signals generated by the 5453A in a manner that allows the transmit-line characteristics to be calculated (see box, page 16). With the equipment in place, the operator requests any or all of the transmission parameter measure ments listed in Table I. The 5453A will make all nine measurements in both directions on the circuit in ap proximately two minutes. Fig. 8 illustrates the data output. Other capabilities of the program include storage, retrieval, and purging of test results in the mass me mory. It is also possible to compare data to a bench mark or to specifications the circuit must meet. Data taken on two segments of a circuit may be combined to yield the overall characteristics. Operator interac tion with the program is purely conversational, al lowing him to accomplish complex tasks rapidly with a minimum amount of training. Other Applications The 5453A is not limited to testing installed com munication channels. The same approach could be applied equally well, for example, to end-to-end checkout of a complete communications system on the production floor. The speed of digital tech niques makes it feasible to do 100% testing and have complete records even for high-capacity systems. Additional 5453A applications are to be found in the design and testing of all types of communica tions equipment, such as data sets, facsimile trans ceivers, equalizers, multiplex-channel modems, tele phone sets, and loop extenders. Correction In our April 1975 issue, page 10. it is stated that Model 5308A time-interval measurements are guaranteed accurate within one nanosecond That sentence should have read. Measure ments accurate guaranteed accurate within five nanoseconds and are typically accurate within Fig. 8. 5453/4 TPA printout of data-circuit test results. Nine tests are made in less than two minutes. one nanosecond.' The editors apologize for losing a crucial line of type. 15 © Copr. 1949-1998 Hewlett-Packard Co. Portable Transponder Allows Two- Way Data Channel Measurements When used with the 5453A Transmission Parameter Analyzer (TPA), the HP 5468A Transponder, Fig. 1 , provides the capabi lity to characterize four-wire voice-grade facilities automatically in both directions of transmission. Control of the transponder is by means of coded command tones generated by the TPA. The transponder provides the test signals needed to characterize the receive line and conditions test signals received from the TPA so the system can compute the transmit line characteris tics. The automatic feature can be overriden, allowing manual measurements of received level and noise without tying up the TPA. Output Fig. 2. Transversal filters are used in the transponder to generate low-distortion test signals. length) pseudorandom binary sequence. The final output signal is then obtained by multiplication using the circuit shown in Fig. 3. For measurements on the transmit line command tones are sent from the TPA to program the transponder into its signalconditioning modes of operation. Attenuation and envelope delay distortion are measured by causing the transponder to provide an equal-level loopback. The characteristics of the transponder and the receive line are then subtracted from the measurement of the entire loop. Measurement of noise with tone on the transmit line is accom plished by first passing the received signal through a 20-dB notch en With the tone reduced in amplitude by 20 dB, the en tire spectrum (noise plus tone) is given 20 dB of gain before be ing looped back on the receive line. Thus noise on the receive line has a negligible effect on the measurement. The TPA re moves the weighting effect of the receive line. The measurement of intermodulation distortion is achieved by notching out the 703-Hz tone prior to loopback. Therefore, while of distortion products may be formed as a result of transmission over the receive line, there are no Â¡ntermodulation products. Once again, the previously measured frequency re sponse of the receive line is used by the TPA to compute distortion. Fig. 1. 5468 A Transponder works with 5453 A Transmission Parameter Analyzer to characterize transmit and receive lines of data circuits. Three test signals are provided by the transponder to charac terize the receive line. A pure 1015.625-Hz holding tone is used to measure frequency shift, phase jitter, noise with tone, and 1-kHz loss. Intermodulation distortion requires at least two tones, and a third tone is added at V* power to better simulate an actual data signal. Attenuation and envelope delay distortion are measured using a broadband signal containing 16 tone pairs. The transponder also provides a 600-ohm termination on the receive line for no-tone noise measurements. The 1 01 5.625-Hz tone must have low incidental phase modu lation (<0.1Â°), stable amplitude (<0.05 dB drift), and an accu rate frequency (Â±0.025 Hz). The frequency accuracy re quirement implied that a crystal was necessary, while the low phase modulation requirement ruled out a phase-lock loop. The approach taken in the transponder is to use a transversal filter to convert a stable digital clock into a sine wave with the de sired Sec The circuit is illustrated in Fig. 2. Sec ond-order distortion of lesÂ£ than 70 dB is typical of such a filter. Three transversal filters are used to generate the intermodulation distortion test signal. The three frequencies are 703, 1172, and 1218 Hz. The third test signal contains 16 pairs of sidebands spaced Â±78 Hz about suppressed carriers spaced 250 Hz apart. Amp litudes and relative phase differences must be stable and uni form from one transponder to another. The 78-Hz modulation signal is generated using the transversal filter approach. The carriers are generated using a 63-clock-period (4 ms total Subtracts 78 Hz from Multi-Tone Signal Output Fig. 3. Tesf signa/ for envelope delay distortion and attenuaton measurements is generated by multiplying a 78-Hz sig nal by a pseudorandom binary sequence (PRBS). 16 © Copr. 1949-1998 Hewlett-Packard Co. Acknowledgments Hewlett-Packard Instruments for Checking Voice-Grade Telephone Lines The authors wish to thank Peter Roth of HewlettPackard and David Favin of the Bell Telephone Laboratories who jointly originated the idea of apply ing digital techniques to the testing of data circuits and who have contributed consistently throughout the development. Ron Potter made significant contri butions with regard to several of the more difficult measurements. Special thanks are also due to Earle Ellis for the application program, Al Low for some ex cellent product design, and Dennis Kwan for the in troduction to production. Pete Appel contributed to helping complete this 5468A design as well as help ing with the system programming. Finally, thanks are due to Dave Snyder for his help in getting the first prototype running and to Melba Lindgren for her sup port of the programming effort.E The August 1 974 issue of the Hewlett-Packard Journal contained a chart comparing the capabilities of six Hewlett-Packard instruments designed to measure vahous parameters of voice-grade telephone channels used tor data transmission All of these Journal. have now been described in the Hewlett-Packard Journal. The instruments, and the issues in which they appear are: CCITT Standards 3552A Transmission Test Set (May 1975) 3770A Amplitude Delay Distortion Analyzer (November 1974) 3581C described Voltmeter (Related to 3580A Spectrum Analyzer, described in September 1973) North American Standards 3551A Transmission Test Set (May 1975) 4940A Transmission Impairment Measunng Set (August 1974) 5453A Transmission Parameter Analyzer with 5468A Transponder (May 1975) 3581C Selective Voltmeter (see above) References 1. P.R. Roth, "Digital Fourier Analysis", Hewlett-Packard Journal, June 1970. 2. E.O. Brigham, "The Fast Fourier Transform", PrenticeHall Inc., 1974. 3. "Transmission Systems for Communications", Bell Telephone Laboratories, Inc., Fourth Edition, 1970. 4. E.G. Smith, "Glossary of Communications)", Tele phony Publishing Corp., 1971. Robert H. Perdriau Bob Perdriau is product market ing engineer for digital signal analyzers at HP's Santa Clara ^â€¢*L I Division. Before assuming that f | I post in 1973 he had served as a 'Mi} I design engineer and as an appli cations engineer for three HP divisions. Born in Bosiun. Mas sachusetts, he graduated from the University of Massachusetts in 1963 with a BSEE degree. Except for three years in the U.S. Army, he's been with HP ever since. Bob is married, has two children, and lives in Los Altos, California, where for the last two years he's been busy building a major addition to his for With that project about finished, he's looking for ward to having more time for his other interests, which include fishing and hunting, woodworking, and bicycling. HP Model 5453A Transmission Parameter Analyzer Contact the factory or your local Hewlett-Packard sales office for specifications. PRICES IN U.S.A.: 5453A Transmission Parameter Analyzer, $62.800. 5468A Transponder. $2500. MANUFACTURING DIVISION: SANTA CLARA DIVISION 5301 Stevens Creek Boulevard Santa Clara. California 95050 U.S.A. Stephan G. Cline Now a laser interferometer ap plications engineer, Steve Cline until recently was involved in Fourier analyzer hardware and software design. With HP since 1968, he wrote much of the softC% â€¢ J ware for the 5450A Fourier AnaII JT l| lyzer. helped design the 5470A 9 M >*^ Fast Fourier Processor, and served â€¢w â€¢ L. ^H as project leader for the 5471 A ff 41 AvjB FFT Arithmetic Unit and the 5468A Transponder. Born in Camp McCoy, Wisconsin, he received his BSEE degree from Michigan State University in 1967 and his MSEE degree from Stanford University in 1968. Especially interested in meeting people from different cultures, Steve enjoys travelling and serves as treasurer of American Field Service, a foreign-student exchange program. His main rec reational activity is golf, but he enjoys biking, too. He and his wife live in Los Gatos, California. Roger F. Rauskolb Roger Rauskolb, a native of Maehrisch-Ostrava, Czecho slovakia, received his Dipl. Ing. degree from the Technische Hochschule of Darmstadt, Ger many, in 1961. His career at HP has been a varied one that began in 1962 and includes service as a microwave project engineer, spectrometer project manager, digital signal analysis (and 5453A) group leader, and now, member of the HP Laboratories technical -Â»Â»^. staff. In 1965 he received his MSEE degree from Stanford University. A resident of Palo Alto, California. Roger is married and has two daughters. He's an audiophile, a photographer, a swimmer, and a skier. His interest in building a better world goes back many years and currently expresses itself in his membership in Project Survival, a group concerned with energy problems and education for long-term survival on earth. 17 © Copr. 1949-1998 Hewlett-Packard Co. A Precision Spectrum Analyzer for the 10-Hz-to-1 3-MHz Range Adaptable to automatic systems or bench use, a new spec trum analyzer has measurement resolution of 0.01 dB, passbands as narrow as 3 Hz, and a dynamic range of 70 dB. by Jerry W. Daniels and Robert L. Atchley THE SPECTRUM ANALYZER and the network analyzer are assuming greater and greater im portance as means of evaluating the performance of electronic circuits and devices. The network ana lyzer gives complete information about the ampli tude and phase performance of linear networks while the spectrum analyzer evaluates the amplitude performance of both linear and nonlinear networks. The spectrum analyzer is a single-channel instru ment that selects and measures the amplitudes of the individual frequencies that make up a complex sig nal. It is thus able to detect and measure the distor tion and intermodulation products of nonlinear net works. The network analyzer is a dual-channel instru ment that compares the amplitudes and phases of two signals, usually the input and output of a net work or device. It is normally not suitable for mea surements involving nonlinear networks because it is designed on the assumption that only one fre quency at a time will be at its input. The method of heterodyning signals within the network analyzer could cause spurious responses if the input signal were distorted or otherwise contained more than one frequency. modulation responses are below the measurement range of the instrument. The new spectrum analyzer has accuracy and pre cision normally not associated with spectrum ana lyzers. It has an amplitude resolution of 0.01 dB, an accuracy of Â±0.05 to Â±1.15 dB, depending on the sig nal level and frequency, a dynamic range of 70 dB and a measurement range of 150 dB. The analyzer's passband is selectable in steps from 10 kHz down to 3 Hz, the stability of both the analyzer and the synthe sizer used as the local oscillator being such that the 3-Hz bandwidth is practical even at 13 MHz. The fre quency of a signal component can be determined well within 1 Hz. Typical measurement results are shown in Fig. 2. Some of the measurements for which this high-pre cision instrument is especially useful are harmonic and intermodulation distortion in amplifiers, powerline sidebands and harmonic levels in oscillators, RF and LO feedthrough in mixers, and frequency transla tion errors in communications repeaters. It can also serve as a frequency-response test set for high-preci sion measurements, such as amplitude errors in the up-converted channels of multiplex communica tions systems. A New Spectrum Analyzer Remote Control Some two years ago, a network analyzer for measur ing the magnitude and phase characteristics of sig nals in linear networks over a frequency range of 50 Hz to 13 MHz was introduced (Model 3570A).1 The accuracy and ready adaptability of this instrument to systems use have now been incorporated in a new spectrum analyzer, Model 3571A, for measurements of complex signals over this same frequency range (and down to 10 Hz). This new instrument (Fig. 1) performs waveform signal analysis with full assur ance that all internally-generated image and inter- Besides being operable from the front panel, the new spectrum analyzer is also programmable through the HP Interface Bus.2 Every front-panel switch position (except for the ON/OFF switch) is as signed an ASCII code so it can be selected by a sys tem controller. The analyzer can thus be incorpo rated into calculator-controlled automatic measure ment systems (Fig. 3) that can manipulate the data so it can be presented in more meaningful form. For ex ample, it can function as a high-precision distortion analyzer by providing a mathematically exact sumâ€¢6 © Copr. 1949-1998 Hewlett-Packard Co. Fig. 1. Model 3571A Tracking Spectrum Analyzer (lower unit) works over a 10-Hz-to- 13-MHz range using one of the HP syn thesizers (upper unit) as a local oscillator. The combination of analyzer and synthesizer is known as the Model 3044A Spectrum Analyzer. spectrum analyzer but it differs in the characteristics of its selective filters, which have a rounded re sponse curve that minimizes ringing during a fre quency sweep, rather than the wave analyzer's squared-off response curve. Because the front-panel of the new Model 3571A Tracking Spectrum Analyzer resembles neither the traditional spectrum analyzer nor a wave analyzer, a look at the controls can be informative. First of all, there is no tuning control on the instrument itself. It was designed to work with the offset frequency sig nal from either the HP Model 3320A/B Frequency Synthesizer or the Model 3330A/B Automatic Fre quency Synthesizer. Tuning the synthesizer tunes the analyzer, the frequency of the synthesizer corre sponding to the center frequency of the analyzer's mation of the individually-measured distortion pro ducts. An automatic system not only speeds measure ments, removes the tedium from repetitive measure ments, and facilitates a high degree of data manipula tion, but it also provides a means for enhancing mea surement accuracy by using calibration routines to store the results of reference measurements and then using these to correct actual measurements. Information Display To obtain the high resolution that the accuracy of this instrument makes possible, it has a digital read out rather than the CRT display commonly asso ciated with spectrum analyzers (however it has an analog output for a CRT display). Superficially it re sembles the traditional wave analyzer more than a dBV 20 3BV -40 -40-- -60- - 60-- -â€¢O- - 5th order 7th order 80-- l Â ° Â ° J / J t M U . ^ . Â · + M ^ Ã- 1 1 U * ' 100- â€¢ ' * v * ^ - J j ^ 1 â€” r â€” I 1 1 H 97 98 99 100 101 102 103 104 105 106kHz H â€” I â€” h - 1 kHz 11 MHz - 1 kHz 19 © Copr. 1949-1998 Hewlett-Packard Co. Fig. 2. Typical spectra recorded by a 3045A calculator-controlled spectrum analyzer system based on the Model 3571A Tracking Spectrum Analyzer. The recording at far left was made during a twotone intermodulation test of an amplifier using input frequencies of 101 and 102 kHz. The oddorder intermodu/ation products are clearly shown. The recording at near left shows the output of a double-balanced mixer fed by a high-level 11 -MHz carrier through the LO port and a mixture of 1-kHz and low-level 180-Hz tones through the RF port. The up-con verted 180-Hz signal is clearly resolved about the 1 1-MHz carrier. frequency components from stronger frequencies close by. Because of the stability of the instrument, two signals only 15 Hz apart but with an amplitude difference of 55 dB can be resolved. Line-related side bands more than 70 dB down can be resolved. Measurement results may be displayed in a variety of measurement units. With the DISPLAY REFERENCE switch in the dBV position, the display reads dB with respect to 1 volt no matter what the input impedance may be. With the switch in the dBm position, the instrument displays the measured power in the selec ted input impedance (either 50ÃÃ, 75Q, or an external 600ÃÃ). With the switch in the RELATIVE position the instrument displays a dB reading relative to a pre viously established reference. For example, a reading in either the dBV or dBm position can be stored as a reference by pressing the ENTER OFFSET button, es tablishing this signal level as the 0.00 dB level. Then with the switch moved to the RELATIVE position, all further readings are displayed as so many dB above or below this reference. This is handy for reading the level of harmonics with respect to the fundamental. This arrangement also allows the user to calibrate the instrument with respect to some other impe dance level. With an external termination of the de sired impedance attached, the user supplies a known 0-dBm signal and presses the ENTER OFFSET button. Subsequent measurements made with the switch in the RELATIVE position will then be direct reading in terms of this impedance level. The analog equivalent of the stored reference is subtracted from the analog output voltage allowing expanded-scale visualization of a portion of the spec trum on a CRT display or X-Y recording. Fig. 3. The Model 3571 A Tracking Spectrum Analyzer also functions under programmable calculator control through the HP Interface Bus, giving an automatic measurement and datareducing system that shortens test time and decreases the possibility of measurement error. A packaged calculatorbased system that includes the analyzer is known as the Model 3045/4 Automatic Spectrum Analyzer. passband (it is for this reason that it is known as a Tracking Spectrum Analyzer). The controls that are on the analyzer have to do with bandwidth, signal level, units of measurement in the display, and input impedance. The input impe dance is selectable to allow use of the analyzer in a variety of measurement situations. The 50ÃÃ and 75Ã1 input impedances match a wide range of high-fre quency devices and are especially useful for measure ments in communications systems. The IMfl input impedance allows the user to supply his own termin ation for other impedance levels, and it also allows a conventional oscilloscope probe to be used for highimpedance circuit probing. Full-scale input amplitude ranges are from 3.16V rms to 1 mV rms in eight 10-dB steps with a full 70-dB dynamic range on each step. An OVERLOAD indicator flashes if the signal exceeds the input range. As mentioned earlier, the passband is selectable from 3 Hz to 10 kHz in a 1-3-10 sequence. The wider passbands permit relatively fast sweeps over a wide band for a quick, overall look at a spectrum. The nar row passbands make it possible to isolate low-level Block Diagram Overview As other spectrum/wave analyzers do, the Model 3571A heterodynes the input signal to an interme diate frequency for narrowband filtering. A block dia gram is shown in Fig. 4. The input signal is mixed with the synthesizer offset frequency to derive a 20-MHz intermediate frequency (the synthesizer off set frequency is precisely 20 MHz higher than the synthesizer main output). The up-conversion to 20 MHz places the image frequencies in a range of 40 to 53 MHz, which are easily attenuated more than 80 dB by a low-pass filter at the input. The 20-MHz output of the mixer is filtered to re move unwanted mixer products and then down-con verted to 100 kHz for the filtering that establishes the instrument's passband. An IF of 100 kHz was chosen to permit the use of narrow-band crystal filters. The filtered IF is then processed by an amplifier whose output is logarithmically proportional to the input. This amplifier, a hybrid 1C similar to that used 20 © Copr. 1949-1998 Hewlett-Packard Co. 20-33 MHz From Synthesizer Input Amplifier 20 MHz 100 kHz LÂ°9 Amplifier Analog Output 10 Hz-13 MHz v 19.9 MHz Overload Detector A-to-D K D-to-A Converter I Converter 1 MHz From Synthesizer Fig. Spectrum Analyzer. block diagram of the Model 357 T A Tracking Spectrum Analyzer. in other HP instruments,3 converts the signal vol tage level to the equivalent dB level by compressing signals in proportion to their amplitude. The detector is a peak-to-peak type. The detector output, a dc voltage proportional to the log of the sig nal amplitude, is smoothed in a low-pass filter and then provided at a rear-panel connector as a Y-axis output for use by a CRT display or by an X-Y plotter (an X-axis output is available from the associated synthesizer). A front-panel switch can slow the filter response by a factor of 20 to smooth noisy signals. The detector output is also converted to an equiva lent digital number by an analog-to-digital converter and sent to the digital processor. The number is proc essed in accordance with the format established by the settings of the front-panel switches. Offset, dB, dBV, dBm 50ÃÃ, dBm 75ÃÃ, dBm 600Ã1, and input range all affect the number that is finally displayed. The difference between the displayed number and the raw digitized number is converted to an analog voltage and applied as an offset to the rear-panel ana log output voltage. The analog output is thus consis tent with the digital readout in terms of measure ment units (0.1 V = 1 dB). Analog Circuit Details Now to examine some of the considerations in volved in the design of this instrument. The dynamic range of a spectrum analyzer is limited by noise at the low end and intermodulation distortion at the high end. The design of the input circuits is directed towards maximizing the difference between these two extremes. Low noise is achieved by use of a J-FET buffer am plifier which also gives high input impedance. The use of the complementary-symmetry configuration obtains high linearity and very good frequency re sponse. By making the "straight-through" input range 10 mV rms, the noise level allows the desired 70-dB range to be obtained with the widest band width (10 kHz). The output of the input amplifier is monitored by the overload detector, a peak detector driven by an amplifier that has greater than 60-MHz bandwidth to enable response to out-of-band signals. As long as the front-panel overload indicator is not illuminated, the user is assured that the input signal is within the linear range of the amplifiers, which means that internally generated distortion and intermodulation products are more than 80 dB below the full-scale in put range. Only Two Conversions The frequency conversion to 20 MHz occurs in an active double-balanced mixer. After bandpass filter ing, the 20-MHz signal is presented to the second mixer for conversion down to 100 kHz. The second mixer's local oscillator frequency (19.9 MHz) is phase-locked to a 1-MHz signal from the synthesizer and is thus in precise relationship to the first local os cillator frequency. Since the 100-kHz IF is derived by mixing a 20MHz signal with 19.9 MHz, a 19.8-MHz signal in the first IF channel would also be converted to 100 kHz. Normally, this situation would be avoided by having another IF conversion between the 20-MHz and 100-kHz IP's. Fortunately, we were able to avoid the extra cost of a third IF channel, not to mention the ad- © Copr. 1949-1998 Hewlett-Packard Co. Attenuating the Classical Attenuator Problem The classical attenuator problem is encountered anytime a device under test is placed between a single-ended source and a single-ended detector. It may manifest itself in several ways: â€¢ Apparent detector inaccuracies at low signal levels; â€¢ Reduction of dynamic range at low frequencies; â€¢ Spurious responses caused by common-mode signals. The basics of the problem are outlined in the drawing. This represents a signal source driving an attenuator that is moni tored by a detector. Coaxial cables are used, and the test is being conducted at frequencies in the audio range (at high fre quencies, the coaxial cables behave more like baluns and the problem is not so acute). To simplify the discussion, the attenuator is set for infinite at tenuation. It is easily seen that return currents through the cable shield to the signal source can develop a voltage, ea, across the finite resistance of th tages across Rc2 and Z If the detector input resistance, RD, is high, the voltage drop across Rc2 is seen by the detector, so a residual signal is mea sured even with infinite attenuation. When (RL2 + RD) Â» Rc2, (R s (RC From the figure it is easily seen that if Z were zero, which means that the source and detector would be referenced to the same ground, the full voltage of ea would be measured by the detector. Consider a 50fi system using two 4-foot lengths of RG/58U cable with an infinite attenuator between. What would be the real attenuation? Rci = Rc2 = 20 mil and Rs + RL1 = 100Ã1. IfZwereequal to zero, then: e^ (20 x 1Q-J) (20 x 1Cr3) Â»-80 dB 100(40 x 10-3) Increasing Z to 111 yields: (20 x 1Q~3) (20 x 10~3) =>-108dB 100( 1. 0004) thus, a small increase in Z results in a significant reduction in eD . A similar analysis shows that common-mode voltages caused by ground loops are also reduced by increasing Z. Increasing Z, however, would not allow the barrel of the frontpanel BNC connector, which is connected to signal ground, to be tied directly to earth ground. The classical attenuator error was reduced in the Model 3571 A without fully floating the input connector by making Z a saturable-core inductor wound with #17 wire. This has practically zero impedance at dc but a finite impedance at frequencies where the classical attenuator prob lem exists. On the other hand a large powerline signal through Z, such as might occur with a grounding error, would saturate the core, reducing the impedance of Z to less than one ohm. This is why the input to the Model 3571 A is not described as "floating", but as "isolated at low frequencies". The potential reduction in measurement errors achieved by this arrangement is shown by the graph below. This was made by the Model 3571 A measuring the output of a 120-dB attenua tor fed by a one-volt signal supplied through a 4-foot cable (bandwidth: 3 Hz; range: 1 mV; smoothing: on). Z) + Rcl (RC2 + Z) and, The object is to reduce the detector signal, eD, to zero, or at least to insignificant proportions. This would occur if either RC1 or Rc2 were zero, but this would be difficult to achieve. Increasing (Rs + RL)) and/or Z would also reduce eD but Rs and RL1 are fixed by the measuring system, which leaves Z as the only variable available for manipulation. dBV Detector Source -70 -p 80 With Signal Ground Connected to Chassis Ground 90 100 -110 With Z Between Signal and Chassis Grounds 120- 130 10 Hz 100 Hz 1k ditional problems with intermodulation and noise that another mixer would introduce, by use of a 20-MHz filter that attenuates 19.8 MHz more than 80 dB. The 20-MHz filter consists of two cascaded twopole crystal filters, one of which is shown in Fig. 5. Stagger-tuning the crystals gives a bandwidth of 30 kHz. However, at 19.8 MHz the currents through the â€¢ 10 k 100 k 1 MHz shunt capacitances of the two crystals are exactly equal and of opposite phase, cancelling at point A and giving a transmission zero at 19.8 MHz. Signals at 19.8 MHz are attenuated more than 50 dB in each stage, more than adequate to meet demands. IF Stability The 100-kHz IF is where all of the bandwidth selec22 © Copr. 1949-1998 Hewlett-Packard Co. ly programmable with all control executed through the PS register. During local control, the front-panel switches are parallel loaded into the PS register but under remote control the front-panel controls are locked out and the contents of the PS register may be changed only by data from the HP-IB. Pushing the ENTER OFFSET switch during local con trol sets the data flag. The controller checks to see if it is in local control, and finding that it is, takes the pre sent dBV reading, which has been stored in memory, and places it into memory as the reference for a rela tive dB display. Data coming from the HP-IB is parallel loaded into the 8-bit input (I) register. The controller uses a quali fier to sense this condition and upon receiving this information it shifts the data serially through an opti cal isolator to the Q register for decoding (both the HP-IB input and HP-IB output circuitry are isolated from measurement ground). After the data has been deciphered, it may be used to program an arbitrary offset into memory, or to initiate some immediate command. It could also be recognized as an unused command and be ignored. The controller also calculates an offset voltage to be subtracted from the log amp output so the analog output will correspond to the display. During the time that the controller is not in a measurement rou tine or data entry, the controller uses the digital-toanalog portions of the successive-approximation digi tizer to construct an analog voltage. For a given dis play reference and input range, the number is con stant so a follow-and-hold circuit can retain the ana log voltage while the controller is occupied with the other routines. This analog signal is then added to the normal (straight through) analog output to obtain the required offset. Because only the offset is obtained by the D-to-A converter, small pertubations in the signal amplitude are transferred to the analog output signal. The resolution then is that of the log amplifier, rather than the digitizer. tion and most of the gain occurs. The filters are all 5-pole synchronously-tuned types with a -3dB to â€” 60dB shape factor of 10. The response curve is ap proximately Gaussian. The three widest passbands are derived from high-Q LC tanks while the rest are crystal derived. A single set of five crystals is used with loading resistors to broaden the bandwidth when required. Frequency drift in narrowband filters can cause problems. When using the 3-Hz bandwidth, a drift of 1 Hz, although only 0.001% at 100 kHz, would cause significant measurement errors. This problem was minimized by incorporating a 100-kHz oscillator in the 19.9-MHz phase-locked local oscillator circuit (Fig. 4). The 100-kHz crystal of this oscillator is the same type used for the IF filters; in fact, all six crys tals are supplied as a matched set. Therefore, any drift in the IF center frequency is matched by a com pensating drift in the local oscillator. Digital System Details The digital machine in the 3571A is a 16-bit binary serial processor using 8K of ROM. It has four major functions: the measurement routine, the data entry, the data output, and the calmlation of the analog output. The measurement routine is the data gathering process. Here the controller commands the analog-todigital converter to digitize the output of the log am plifier. The output of the digitizer is then manipu lated according to the program in the program storage (PS) register. The controller does this by interrogating the "PS" register and checking to see what amplitude range is in use. It then subtracts or adds a number to correct the reading to dBV (IV = OdBV). The controller now must check to see if the answer is to be displayed in dBV, dBm, or dB relative to a stored number. If it is to be displayed in dBV, it con tinues with the binary-to-BCD conversion for dis play. If the displayed answer is to be in dB relative to some reference, this reference, which was stored in memory by the data routine, is subtracted from the dBV answer. If the program calls for a dBm display, the controller again corrects the dBV answer by the appropriate factor for the selected terminating impe dance (50Ã1, 75Ã1, or 600Q). In any case the binary answer must be converted to BCD for display purposes. This is done by hardware. The binary data is shifted into a binary-to-decimal converter with the most significant bits first. At the end of the shifting a BCD answer is stored in the regis ter and is latched into the displays. It may also be output to the HP interface bus (HP-IB). The data entry routine is primarily used to bring in data from the HP-IB for remote control of the 3571 A. The 3571 A, except for the power switch, is complete Fig. 5. 20-MHz IF filter precedes the second mixer. The adjustable capacitors balance the crystal distributed capa citance to achieve a zero at 19.8 MHz 23 © Copr. 1949-1998 Hewlett-Packard Co. Acknowledgments References Group Leader Paul Thomas was responsible for the basic block diagram and he contributed to the design of the log amplifier and the method of phase-locking to an external synthesizer. Howard Hilton was re sponsible for the input amplifier, overload detector, image filter, and 100-kHz reference oscillator. The 100-kHzIF was designed by Tom Rodine. Product de sign was by Jim Saar. Virgil Leenerts was responsible for the front-panel design and provided the lowfrequency isolation scheme (see box, page 22). Â£ 1 . G.E. Nelson, P.L. Thomas, and R.L. Atchley, "Faster GainPhase Measurements with New Automatic 50 Hz-to-13 MHz Network Analyzers," Hewlett-Packard Journal, October 1972. 2. D.W. Ricci and P.S. Stone, "Putting Together Instru mentation Systems at Minimum Cost," Hewlett-Packard Journal, January 1975. 3. R. Jeremiasin, "Logarithmic Amplifier Accepts 100-dB Signal Range," Hewlett-Packard Journal, March 1974. ABRIDGED SPECIFICATIONS FREQUENCY RESPONSE: - 0 25 OB from 10 Hi lo 13 MHi referei HP Model 3571A Tracking Spectrum Analyzer Frequency RANGE- 10 Hi to 13MHz SELECTIVITY: 3 Hz W 10 K Â» Shape li Amplitude ME SUREMENT RANGE 150 dB DV AMtC RANGE: 70 dB SE SmviTY Nominally 126 dBV ai 1 NH; on all ranges Mai 45 d6V on 60 dBV range wi!h 3 Hz BW from 0 1 to 10 MHz 33 30 A, B AB OLUTE ACCURACY: calibrated to 33308 3330A or external LINEARITY 30 06 o maÂ» input level at 250 kHz RESPONSE TIME: 0 4 ms ( 10 *Hi BW) lo 1 25s (3 Hz BW). with SMOOTHING on. becomÂ«s some 20 â€¢ longer up to max ol 2 5 s DISTORTION RESPONSES. 80 dB below lull scale SPURIOUS RESPONSES: 70 dB I3330A.B) of 60 dB I3320A B) Below full scale POWER LINE RELATED RESPONSES' 70 dB on - 10 dBV throogh 40 08V â€¢anges 60 dB on 50 dBV range and 50 dB on 60 dBV range IF REJECTION: 80 dB OUT-OF-BANO REJECTION: 70 dB DISPLAY RANGE - 1 99 99 dB READING RATE: 4 readings per second ANALOG OUTPUT: 10 OBVDC Â± 13 5V range SYNTHESIZER INPUTS irear panel) 20-33 MHz tracking signal and 1-MHz . 60 d Input Characteristics 13MHz. 50 orims 30 dB return loss lo 13 MHz ATTENUATOR ACCURACY: â€¢ 0 07 dB per slop total accumulation 0 15 dB General OPERATING TEMPERATURE â€¢ 20 C lo â€¢ 30 C can work trom 0Â°C lo -55 with degraded accuracy POWER: 100 120 220 240V 10*. -S". 46-66 Hz. 230 VA max DIMENSIONS: 425mmW - 133mmH â€¢ W3mmD(l6B5 â€¢ 522 â€¢ 21 27,ncne; WEIGHT 16 7 Kg 137 IDS) PRICES IN U.S.A.: Model 3044A opt 100 (standard Model 3571A| $6250 Model 3044A opt 200 I3571A and 3320A Synthesize-) $6495 Model 3045A opt 100I3571A 33306 Automatic Synthesizer 9620A Calculat with pertinent ROMs interface and cabkng) $22 400 MANUFACTURING DIVISION: LOVELANO INSTRUMENT DIVISION PO Box 301 B15 Fourteenth Slreet S W Loveland Colorado 80537 Robert L. Atchley Bob Atchley joined HewlettPackard in 1968, going right to work on the Model 3570A Network Analyzer where he eventually assumed responsibility for the digital processor. He did the same for the Model 3571 A Spectrum Analyzer. Married, and with two children, Bob once took his family to Bangkok, Thailand, (1967) to work on a study project for Colo rado State University, where he had earned BSEE and MSEE degrees. Bob also dabbles in photography. Jerry W. Daniels Jerry Daniels worked on aero space projects while earning BSEE and MSEE degrees at the University of California at Berkeley in a work-study program. He joined Hewlett-Packard in Janu ary 1969 where he worked on the mixer and D-to-A converter for the Models 3320A/B and 3330A/B Frequency Synthesizers. He then contributed to the Model 3570A Network Analyzer before moving on to the 357 1 A. Married, but with no youngsters, he enjoys skiing and backpacking. Bulk Rate U.S. Postage Paid Hewlett-Packard Company Hewlett-Packard Company, 1501 Page Mill Road, Palo Alto, California 94304 Volume 26 â€¢ Nu Technical Information from the Laboratories of Hewlett-Packard Company Hewlett-Packard S.A . CH-1217 Meyrin 2 Geneva. Switzerland Yokogawa-Hewlett-Packard Ltd.. Shibuya-Ku Tokyo 151 Japan C A 8LACK8URN Administrative Services. Typography â€¢ Anne S. LoPresti European Production Manager â€¢ Michel Fogiia 20910. dele'te your nameVorri bur MM ^ ÃÃ J.c f,,c=ol. jl.tlj?,__ . is lable lit peels off) /*"* I I A K \f~^ off) f^l A I "1 I 1 R F-Ã‡ Ã‡ â€¢ To change your address or delete oryour Ã±Ã¡mÃ©lrorrT'ou'rtTfJItfiQIÃ, v_y I! /\ I N vj L V_X I f\ \â€” J I â€” ' 11 d. O O . Send changes to Hewlett-Packard Journal. 1501 Page Mill Road. Palo Alto California 94304 USA Allow 60 days © Copr. 1949-1998 Hewlett-Packard Co.
* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project