|
The First Digital Voltmeters and the Birth of Test Automation
It may seem odd dedicating a page of this computer-oriented Web site to digital voltmeters, but digital instrumentation played a key role in the success of the HP 9825 desktop computer. The HP 9825 represented a real breakthrough in instrumentation control. Without digital instrumentation, there would have been nothing to control and the HP 9825 would not have succeeded half as well as it did.
Although Hewlett-Packard, which was founded as an instrumentation company, achieved many firsts in the test and measurement industry, the digital voltmeter (DVM) was not one of these firsts. Not by a long shot. In fact, HP got behind the differential voltmeter, an analog instrument eventually obsoleted by the DVM. HP didn’t enter the DVM market until 1958, but that’s getting ahead of this story.
|
 |
|
|
Cover Design for the 1962 Non-Linear Systems DVM catalog.
|
|
San Diego?
One might easily expect that Stanford Dean Fred Terman’s budding Silicon Valley in the San Francisco Bay area would surely have been the DVM’s original home, but it wasn’t. The DVM’s birthplace and childhood home was southern California, specifically San Diego County.
During World War II, a new MIT engineering graduate named Andrew Kay was working for the Bill Jack Scientific Instrument Company making aerial-reconnaissance equipment. While at Bill Jack, Kay observed relatively untrained technicians and manufacturing people attempting to take precise voltage measurements using a 19th century invention called a Kelvin bridge (more formally, a Kelvin-Varley bridge), co-invented by Lord Kelvin in England. In the right hands, the Kelvin bridge can produce very accurate measurements (with much better resolution than a conventional analog voltmeter) but it requires careful manual adjustment of precision, calibrated potentiometers to null out the needle in an analog meter. It’s a precise, slow, and error-prone operation.
In addition, a careless, clumsy, or sleepy operator can accidentally apply too much current to the meter and destroy it by driving the meter’s needle all the way over to one side or the other. The result is a bent meter needle as it forcibly pegs against its mechanical stops. Enough current can also burn out the sensitive meter coil.
Accidentally destroying meters during World War II was particularly bad because they were expensive and scarce—the Manhattan project was gobbling almost all of them up in its focused efforts to build an atomic bomb before Nazi Germany could do so.
Kay was Bill Jack’s first employee and eventually became its vice president of engineering. By the beginning of the 1950s, the company was flush with Cold War military contracts so it was focused on production and not development. Kay was more interested in engineering and product development. He became bored, left Bill Jack, and started his own engineering company called Non-Linear Systems in Del Mar, California. Kay had decided to develop a voltmeter that could make accurate, high-resolution voltage measurements far more quickly than a Kelvin bridge. He also wanted to make an instrument that required far less skill to operate and was much more rugged. Kay developed the first such instrument in 1952, the same year he founded Non-Linear Systems (NLS).
The First DVM
He named the instrument the digital-readout voltmeter, which he later shortened to digital voltmeter (DVM). The Naval Electronics Laboratory (now called the Naval Ocean Systems Center) bought the first unit early in 1953. The first NLS DVM shares many common attributes with today’s digital voltmeters. The world’s first DVM was a 4-digit instrument with a resolution of 0.01% and a rated accuracy of 0.1% of reading. It had automatic polarity selection, automatic range selection, and a full-scale reading of 999.9V. It sold for $2300. (In 2004, a 3½-digit DMM (digital multimeter)
can cost less than $5, including battery.)
This first NLS DVM employed expensive, high-speed, mechanical relays to select nodes in a resistor divider. This approach emulated and acted as a discrete or digital version of the Kelvin bridge’s precision, calibrated potentiometer. Relays weren’t the least expensive method available to switch among the resistor-divider’s nodes, but they had low impedance, they switched quickly, and they were reliable (lasting millions of cycles). Kay’s design approach to the first DVM ensured that it would work, and work well.
The first NLS DVM also used a new type of digital display based on stacks of edge-lit, engraved Lucite plates. Each stack (representing one digit) consisted of 11 plates arranged so that they recede from the viewer. Ten of the stacked plates have a numeral deeply engraved on it (digits 0 through 9). The eleventh plate has a decimal-point engraving. A small grain-of-wheat incandescent lamp located along the edge of each plate illuminates the associated plate from the edge. If the lamp is lit, its light travels down the plate, which acts as a light pipe. Eventually, the light strikes the plate’s engraved character. The deep groove of the engraving interrupts the light as it travels down the Lucite plate and scatters it towards the front of the instrument where an operator sees the engraved numeral light up.
Lucite in Disguise
Each numeral display stack thus has 11 grain-of-wheat bulbs, so the 4-digit NLS DVM had 44 incandescent bulbs, plus two more for the plus and minus signs. Consequently, bulb burnout was a serious failure mechanism and the NLS DVM was designed to make bulb replacement easy. (A 1962 NLS instrument catalog touts the advantage of the 20-second bulb-replacement time for the company’ DVMs.)
NLS described this new type of display as “in-line, but not in-plane.” It contrasted with other early digital displays commonly used on frequency meters, such as those HP was building at the time. The frequency counters used a columnar, thermometer, or pin-ball display. Like the stacked numerical displays, the thermometer displays also used ten incandescent bulbs. However, these bulbs were arranged in a vertical column on the face of the instrument.
Each bulb in the display was placed behind a small circle of plastic with a numeral painted on it. Only one bulb in the column lights at a time, which illuminated one of the numerals in the column. To read such a display, the operator scanned each column from left to right and noted the illuminated numbers. The operator’s eye must move up and down each column as the operator scans the reading. It’s not really practical, but it’s all that was available at the time.
The NLS readout also stacked numbers, but in a direction leading away from the operator, not up and down. As various numerals in a stack are illuminated, they appear to approach or recede from the operator because the Lucite stack has some depth to it. NLS DVM’s employed this display until late in the 1960s when the company switched to Nixie display tubes like its competitors.
Non-Linear Systems’ first DVM competitor, Electro Instruments (EI), was founded in 1954 by Jonathan Edwards, who had been Kay’s first employee at NLS. Edwards teamed with an NLS sales representative named Walter East to found EI. Kay, Edwards, and East had a lot of history together. They all graduated from MIT, they all had worked at Bill Jack Scientific Instrument, and Kay had previously hired Edwards and East for NLS. Although the San Diego area isn’t Silicon Valley, the departure of Edwards and East from NLS to found their own company demonstrates that the high-tech entrepreneurial spirit was clearly just as strong south of Los Angeles as it was south of San Francisco.
Electro Instruments (EI) produced a DVM based on a less costly telephone stepping relay or switch, thus exploiting Kay’s selection of higher-priced mechanical relays at NLS. These stepping switches were relatively cheap because they were being manufactured in the tens of millions for the Bell System’s newly automated central telephone offices. However, stepping switches are slower than relays. They’re also noisier (who cares about the racket in a telephone switching room?) and they wear out more quickly. Use of stepping switches allowed EI to make a less expensive DVM at the expense of measurement speed, operating noise, and product reliability. These same sorts of tradeoffs are made every day in the 21st century (50 years later) by today’s design engineering teams.
EI’s DVM put marketing pressure on NLS, which countered by producing its own DVMs models based on stepper switches. However, to reduce the operating noise and improve reliability, NLS packed its DVM stepper switches in sealed cans containing oil. The oil lubricated the switches (which extended their life from about 100 million cycles to more than 400 million cycles) and somewhat dampened the noise. Eventually, NLS also produced a top-end line of DVMs based on really expensive, mercury-wetted relays that could operate for more than a billion cycles each.
The next company to enter the DVM arena was Cubic, a company founded by Walter Zable in 1951 to produce calorimeters that measured the output power of magnetron tubes (used in radar). Zable started Cubic in a San Diego storefront. Cubic produced its first DVM in 1957 because Zable believed any self-respecting instrument company needed to be in the DVM business. However, the DVM business proved unprofitable for Cubic and it exited the business in 1960. The company is still active in other businesses and Zable still runs the company, but Cubic’s Web site mentions nothing of its foray into the DVM market more than 40 years ago.
Next came DVM vendor Cohu, which introduced a stepper-switch DVM in 1958. Cohu’s engineers decided to overcome the stepper switch’s slow speed by simply overdriving the electromechanical switch at 60 steps/second, three times the component’s 20-steps/second rating. (The phone company’s always very conservative when rating components, isn’t it?) This was a disastrous error on the part of Cohu’s design engineers and the badly abused steppers in Cohu’s DVMs quickly self destructed. Cohu backed the switching rate down to 20 steps/second and restored its DVM’s reliability but it forfeited the competitive advantage of a higher speed, lower cost instrument. Cohu, now located in a northern suburb of San Diego (Poway), is still engaged in a variety of electronics businesses but it no longer produces DVMs.
HP Finally Gets With the DVM Program
In 1958, Hewlett-Packard became at least the fifth company to enter the DVM arena with its HP 405AR DVM. While the DVMs discussed previously from a variety of vendors all employed some variant of the automated 19th-century resistor-bridge approach, HP design engineers Ted Anderson and Noel Pace took an HP-like, all-electronic, 20th-century approach to DVM design. The HP 405AR DVM measured voltage using a precise voltage ramp (an integrator), a precise frequency time base, and a digital counter.
HP’s Frequency and Time (F&T) Division was already making frequency counters with precise time bases, an instrument much favored by the US nuclear scientists whose tests were blowing up portions of the New Mexico landscape and vaporizing an unexpectedly large assortment of small islands associated with the Bikini Atoll in the central Pacific Ocean’s Marshall Islands. The electronic guts of a frequency counter make an excellent technological foundation for an integrating DVM and voltmeters probably outsell frequency counters by a factor of 1000x or more. So the addition of a DVM to its growing instrument catalog was a reasonable evolutionary move for HP to make in its quest to grow the company’s revenues while controlling development costs.
Today, we’d call the HP 405AR’s ramp-circuit design a single-slope-integration analog-to-digital (A-to-D) converter. Back then, it was just a ramp circuit. Here’s a summary of the ramp-circuit’s operation: The DVM’s internal ramp voltage starts at zero volts and starts to rise as the counter starts to count pulses from the time base. The DVM compares its internal ramp voltage to the input signal (or an attenuated version of the input signal) and it stops counting the time-base pulses when the ramp voltage just exceeds the input voltage being measured. At that point, the counter contains a digital representation that is proportional to the measured voltage.
For example, if the DVM’s internal ramp rises at 1V/second and the measurement takes 0.576 seconds, then the counter might contain the value 576 (assuming the time base generates 1000 pulses per second). A count of 576 might represent 0.576V if the measured voltage is directly compared to the final ramp voltage. That same count might also represent 5.76V if the measured voltage is first attenuated by a factor of 10 by a resistor divider on the DVM’s input before being compared to the final ramp voltage.
To complete its more modern, all-electronic nature, the HP 405AR DVM used a “modern” Nixie-tube display. Nixie tubes are cold-cathode, neon displays that also happen to be “in-line, but not in-plane” display devices like the Lucite displays that NLS used. Significantly, the HP 405AR DVM also had a digital output port for driving a data logger. Digital outputs were also available on NLS DVMs by the early 1960s and NLS offered a wide range of peripheral devices that could be attached to its instruments. These peripherals included paper-tape and card punches, typewriters, and high-speed printers.
Digital Voltmeters with digital output ports, like this HP 3440, set the stage for the development of instrumentation computers and controllers such as the HP 9825. Both David Cochran (left) and Chuck Near (right) worked on the HP 3440 and both would go on to help develop HP’s first desktop calculator, the HP 9100. Photo courtesy of Hewlett-Packard.
Of course, if all of these peripherals could be connected to record a DVM’s measurements, so could a computer. By the early 1960s, the stage had been set by the development of the computerized test-automation market. All that was needed was a reasonably priced computer that could be dedicated to making measurements. Eventually, engineers at HP would recognize that opportunity, but not quite yet.
The HP 405AR and one of its designers, Noel Pace, moved out to Colorado in 1960 when HP opens the Loveland Division. However, the rack-mounted HP 405AR DVM was not as commercially successful as it might have been. It was a 3-digit DVM with automatic polarity selection and automatic range selection but HP’s first DVM was introduced five years after NLS introduced 4- and 5-digit DVMs with the same features. By 1959, just one year after HP introduced its 3-digit DVM, NLS would demonstrate a 6-digit DVM with 1000 times the HP 405AR’s resolution. In the 21st century, even $5 DVMs have 3½-digit resolution.
HP Tries Again
 |
|
|
David Cochran joined HP in 1956 as a technician and then became an engineer at HP after earning his EE degree at Stanford. He worked on the HP 3440 DVM project with Chuck Near and would then work with Near again on the HP 9100 programmable desktop calculator. Photo courtesy of Hewlett-Packard.
|
|
|
|
|
HP’s next foray into DVMs started in 1961. A freshly minted electrical engineer from Michigan State University named Chuck Near joined HP in California and started working on a DVM design project that would produce HP’s first commercially successful bench DVM: the HP 3440A. He worked on the project with an engineer named Dave Cochran who had previously worked on the HP 405AR. Cochran had been an electronics technician in the Navy and then entered Stanford to earn his EE degree. (He’d grown up in Palo Alto.) However, the GI Bill money ran out and, with a wife and kids to support, Cochran joined HP in 1956 as a technician. He completed his EE degree at Stanford while working for HP and then became an HP engineer. After the HP 3440 project, Cochran will again work with Near on HP’s first programmable scientific desktop calculator, the HP 9100.
.
Near’s boss starts the new engineer out right by suggesting to him that he mine the
|
|
|
Chuck Near joined HP in 1961 and worked on the HP 3440A DVM project. He will be associated in some way with nearly every desktop calculator and computer that HP introduces from 1968 to 1980.
Photo courtesy of Hewlett-Packard.
|
|
|
|
|
company’s existing cache of technology whenever possible. In other words, Near is told not to reinvent the wheel. Design engineers do have a tendency to believe they can always design something better than anything that’s been done before and they often waste time doing exactly that, producing something that’s better but not sufficiently better to merit the extra design costs. It’s an attitude called “NIH” or “not invented here.”
Near discovered that HP’s F&T Division had developed a circuit board with a transistor-based decade counter and an integrated Nixie-tube display. Four of these boards would form the HP 3440A’s counter.
Unlike the HP 405AR, the HP 3440A DVM employs a dual-slope A-to-D converter that gives it better accuracy. A dual-slope A-to-D converter eliminates many sources of measurement error caused by imperfect electronic components in the DVM’s ramp circuit. (Note: All electronic components are imperfect. The measure of their imperfection is a characteristic called “tolerance.” Designers must create circuits that tolerate these imperfections.) As a consequence of the improved ramp-circuit design, the HP 3440A has 4-digit resolution and it became an extremely successful product for HP. In fact, it became one of HP’s top-selling products during the early 1960s.
However, as with most engineering projects, development of the HP 3440 didn’t always go smoothly. For example, all HP equipment had to pass environmental tests including a test that checked a product’s ability to run at high ambient temperature. The HP 3440 was certified to operate at 40 degrees Centigrade. However, the HP 3440 prototype failed its first thermal test. Its Nixie-tube displays got hot and the voltmeter overheated when the ambient temperature went up.
Don Schulz, the normally even-tempered manager in charge of the HP 3440 project, lost his temper at the design engineers for their design oversight. However, the fix was simple. By removing metal at the front and back of the inner aluminum mounting deck, the engineers created a circulation path for air inside of the instrument. Air heated by the Nixie tubes would rise and then travel along the underside of the top panel to the back of the instrument where it was cooler. The air would then cool and drop below the aluminum deck through the opening between the aluminum inner deck and the back panel. It would then flow underneath the deck towards the front of the instrument and would be heated again. This simple design modification created a convection cooling loop and dropped the internal temperature of the instrument by 15 degrees. Problem solved. More important, this same convection-cooling design would be used again in a few years for the HP 9100 calculator.
The HP 3440A soon shipped out to the Loveland Division in Colorado, which would start building HP 3440A DVMs from its date of introduction in 1963. Production continued for several years. By 1968, more than 10,000 HP 3440A digital voltmeters would ship and a long line of HP 34xx series DVMs would follow. Despite a sweetheart offer from Loveland Division’s R&D manager Marco Negrete that included paid leave and tuition to earn his MSEE, Near didn’t follow his creation to HP Loveland in Colorado. He preferred to stay in California, where he earned that higher degree anyway. However, Negrete will get Near only a few years later as part of HP’s first desktop calculator team.
NLS continued as a successful DVM innovator and manufacturer for many more years. It developed its own variation of the Kelvin bridge called the iso-ohmic bridge that simplified the design of the resistive voltage divider and reduced its manufacturing cost. In 1962, psychologist Abraham Maslow spent a summer observing Andrew Kay’s then-unorthodox, team-centric management style and then develops and writes about his theory of “enlightened management” based on his NLS experience. Coincidentally, Hewlett and Packard’s management style, dubbed the “HP Way,” will also be referred to as “enlightened” through the years.
Kay will form a division of NLS called Kaypro in 1982 to manufacture a very successful family of portable personal computers that run the CP/M operating system. By 1983, the entire company will be renamed Kaypro and will go public. That year, Kaypro was rated the fifth largest personal computer manufacturer of the world. NLS became Kaypro’s subsidiary when the company’s computer sales (more than $70 million) completely overshadow DVM sales (only about $3.5 million).
IBM’s introduction of its model 5150 PC (the original IBM PC) will start to erode Kaypro’s success in computers (an effect IBM had on many computer companies in the 1980s including HP). Kaypro quickly developed PC clones but by 1990, Kaypro will file for bankruptcy protection. NLS has since been purchased by and operates as a division of Linear Measurements, Inc. NLS is still in the instrumentation business, producing digital panel meters (small DVMs).
Andrew Kay is still in San Diego County (Solana Beach) making and selling PC-based computers at Kay Computers, which Andrew and his brother Steven started in 1992. For his many contributions to the electronics industry, Andrew Kay was inducted into the Computer Museum Hall of Fame in 1998.
Chuck Near finally retired from HP and now lives near San Diego.
Source materials for this page on DVMs include:
George Rotsky, “Gauging the impact of DVMs,” published in Elecronic Engineering Times in 1997 as part of the publication’s 25th anniversary celebration. The article is now located at www.eet.com/anniversary/designclassics/gauging.html.
Non-Linear Systems Digital Voltmeters, second edition (instrumentation catalog), Del Mar, California, 1962.
Non-Linear Systems Digital Voltmeters, catalog supplement, Del Mar, California, 1963.
Kay Computers Web site, www.kaycomputers.com.
Tom Jennings, “Cubic Corporation V-45 Digital Voltmeter,” World Power Systems Web site, www.wps.com.
HP 3440A Operating and Service Manual, Hewlett-Packard Company, Loveland, Colorado, 1961.
Personal telephone interviews with Chuck Near, 2004.
Kenneth Jessen, “How it all began, Hewlett-Packard’s Loveland Facility,” J. V. Publications, Loveland, Colorado, 1999.
|
|
