Electrojector and D-Jetronic: Early Electronic Fuel Injection

Summary

The first production electronic fuel injection system was the Bendix Electrojector, offered briefly by Chrysler for 1958. It was dropped due to unreliability, but it inspired Bosch to develop the similar but more successful D-Jetronic system, introduced by Volkswagen for 1968 (which required a patent licensing agreement between Bendix and Bosch) as a way of meeting U.S. emissions standards. In the seventies, both Chevrolet and Cadillac offered a second-generation Bendix electronic injection system, which was very similar to D-Jetronic and used some Bosch parts.

Electrojector vs. D-Jetronic

In any event, the D-Jetronic system was quite similar to the Electrojector in concept and operating principles. Because the Bosch system differed somewhat from application to application and evolved over the years, how closely it resembled the old Bendix system depended somewhat on which D-Jetronic iteration you were comparing, but the systems were sufficiently alike that it’s easier to summarize their salient differences:

  • Fuel pressure: D-Jetronic maintained a fuel pressure of about 28.5 psi (1.97 bar), roughly twice that of the production Electrojector systems, probably for better mixture formation.
  • Pressure sensor: Like the Electrojector, D-Jetronic used speed-density metering, but its pressure sensor used a movable armature attached to a pair of aneroid bellows in a chamber responsive to intake manifold vacuum, rather than a crank-arm potentiometer. The armature had no contacts to wear, and using the aneroids (whose internal air pressure was effectively zero) meant the sensor effectively measured manifold absolute pressure rather than simply manifold vacuum, eliminating the need for separate devices to compensate for barometric pressure.
  • Fuel cutoff: D-Jetronic cut off fuel to the injectors during deceleration and overrun, as the Electrojector did, but the cutoff was controlled electronically, based on the engine speed signal from the triggering selector in the distributor, rather than signaled through changes in manifold vacuum. Some later applications removed the cutoff circuitry, as it tended to produce spikes in hydrocarbon emissions when the fuel supply resumed.
  • Trigger contacts: Although D-Jetronic used mechanical points within the ignition distributor to trigger each pulse and then energize the injectors, it energized the injectors in groups rather than sequentially. This meant that injection wasn’t timed to coincide with valve opening or ignition timing, but Bosch, like Chrysler, had concluded that phased cyclic timing offered little benefit for port injection and made the triggering circuitry more complicated.

Early D-Jetronic systems shared a number of the same design flaws as the Electrojector, including poor cold start performance at very low temperatures (below -13°F/-25°C) and inadequate acceleration enrichment. D-Jetronic initially provided for momentary full load enrichment using a vacuum diaphragm switch, which was similar to the one on the Electrojector and had the same limitations. By the end of 1969, Bosch had instead incorporated this enrichment functionality into the pressure sensor itself, as illustrated below, and added an air temperature sensor to the intake manifold to compensate for increased air density at low temperatures so the mixture wouldn’t become too lean in cold weather.

B&W schematic of the D-Jetronic pressure sensor with full-load enrichment, with the major components labeled "Full-load stop," "Diaphragm," "Part load stop," "Aneroids," "Core," "Secondary winding," "Armature," "Primary winding," "Valve," "Damping," and "Leaf spring" (Robert Bosch GmbH, Nov. 1969 technical report Fig. 6)

Unlike the Electrojector, D-Jetronic used a contactless manifold pressure sensor, where increases in intake manifold pressure would compress the evacuated (aneroid) bellows and move the control cone of the spring-loaded armature into the core (leftward), increasing the impedance in the sensor circuit and thus the pulse duration. The section at the left, added in 1969, served to shift the support for the bellows farther to the left in response to a sudden increase in manifold pressure, further increasing impedance. It had the same purpose as the old pressure switch: providing momentary additional enrichment for smoother transient response. (Illustration copyright Robert Bosch GmbH, Corporate Archives)

At the same time, the throttle-valve switch was redesigned to incorporate an additional set of drag switch contacts, which provided a small amount of additional enrichment when the throttle valve changed position, for better transient response.

B&W schematic of the D-Jetronic throttle-valve switch, with the major components labeled a (idling contact), b (contact plate), c (lever), d (drag switch contacts), and e (contact plate) (Robert Bosch GmbH, Nov. 1969 technical report Fig. 7)

Originally, the D-Jetronic throttle-valve switch (like the one on the Electrojector) was used mostly for idle control. The revised switch in fall 1969 added drag switch contacts (d) that triggered additional enrichment in response to throttle motion, which was intended to avoid “flat spots” in engine response during abrupt changes in throttle position. (Illustration copyright Robert Bosch GmbH, Corporate Archives)

To address the previous cold start problems, D-Jetronic also added an auxiliary injector in the manifold, which Bosch called the cold-start valve, which sprayed finely atomized fuel into the intake manifold if the starter engaged below a certain temperature. The cold-start valve was controlled either by a thermoswitch, which closed the valve upon reaching a certain temperature, or a thermo-time switch, which automatically closed the valve after a certain period to avoid fouling the spark plugs.

Bendix Electronic Fuel Injection, Round 2

For the handful of engineers and managers within Bendix who were still believers in electronic fuel injection, the Bosch deal was at once validating and galling: Electronic injection now had a viable market niche, and was being offered by some prestigious European makes, but it had been a decade since Bendix had made or marketed a fuel injection system under its own name, and the term of the most important Bendix electronic fuel injection patents was already more than half over.

Bendix executives and engineers popped up regularly in the automotive press and in trade publications to assert that electronic fuel injection would soon see widespread industry adoption, particularly following the passage in late 1970 of the Muskie Act (Public Law 91-604), which called for much tougher auto emissions standards. Behind the scenes, however, Bendix remained very reluctant to make any significant investment in fuel injection, and was only willing to fund research and development work to the extent that royalty payments from Bosch covered the cost. John (Jack) Campbell, EFI Systems manager for Bendix, later lamented that he “spent more time selling inside the company than outside the company.”

B&W line drawing of the Chevrolet Cosworth Vega DOHC engine, cut away to illustrate the details of the cam drive, valve gear, and pistons (GM Media Archive X73100-0550 - cropped)

Conceived in 1970 and designed by engineers Cal Wade of Chevrolet and Mike Hall of Cosworth, the fuel-injected twin-cam, 16-valve Cosworth Vega engine was intended to become the basis of a Formula 2 racing engine, although porosity problems with the aluminum block did not allow the engine to develop enough power to be competitive. The racing version was abandoned in 1973, and the street version didn’t arrive until the spring of 1975, two years later than originally planned. (Illustration: General Motors LLC)

Campbell later claimed that by 1971, this reticence kiboshed his negotiations with Ford Motor Company regarding installing Bendix electronic fuel injection in the Lincoln Continental Mark Series. Ford was initially interested, but balked when Bendix Automotive Group senior management demanded that Ford agree to pay up to $4 million upfront to underwrite the development costs. As late as 1974, some industry observers still expected that Bendix electronic injection would show up next on a senior Ford model, but by then, Ford had long since turned its attention elsewhere.

We assume Bendix made similar demands of Chevrolet, which around the same time expressed interest in adopting electronic fuel injection for the Cosworth Vega, a planned high-performance version of the new Chevrolet Vega subcompact with a twin-cam, 16-valve aluminum head designed by England’s Cosworth Engineering. Chevrolet engine design engineer Calvin (Cal) Wade had already approached Rochester, which had made the defunct Ramjet mechanical injection system for Chevrolet and Pontiac, but they weren’t interested in electronic injection. Bendix was more amenable: Chevrolet wasn’t daunted by the cost, since Wade’s boss, general manager John DeLorean, saw the Cosworth project as a halo model and a proof of concept, and while DeLorean hoped to eventually offer fuel injection in the rest of the Vega line, the Cosworth Vega would be a low-production limited edition of perhaps 5,000 units, a modest commitment.

Press photo of engineer Cal Wade, in a gray suit, with a preproduction Cosworth Vega engine and transmission on a test stand (GM Media Archive C2325-0078)

Chevrolet engineer Cal Wade poses with a preproduction example of the Cosworth Vega engine he shepherded through a troubled development. The engine pictured isn’t the final production version, which had two separate throttle bodies instead of the single conjoined version fitted here. (Photo: General Motors LLC)

The first Cosworth Vega test engines were carbureted, but by April 1972, GM president Ed Cole was able to test-drive an injected prototype. DeLorean initially hoped to put the Cosworth Vega in production as a late 1973 model, but development was slow, and it wouldn’t be until early 1974 that Chevrolet actually began the time-consuming EPA certification process for the injected twin-cam engine.

In the meantime, Bendix entered negotiations with Cadillac, whose general manager, Robert D. Lund, invited Campbell to Cadillac headquarters to discuss offering electronic fuel injection system on a new compact model that would eventually become the first K-body Cadillac Seville. Since the new car was notionally intended as a Mercedes fighter, Lund had decided fuel injection was a must, for reasons of prestige as well as performance, and he was amenable to underwriting the upfront development costs. Unlike the Chevrolet deal, however, Cadillac envisioned a much more substantial production commitment: at least 300,000 systems over five years. This apparently unnerved Bendix president W. Michael Blumenthal, who insisted on going over the heads of both Campbell and Lund to negotiate the contract terms directly with Ed Cole. Nonetheless, by late 1973, GM and Bendix had a deal.

At this point, it belatedly became clear that for all the noise Bendix had been making in the press about the future of fuel injection, the company was painfully ill-prepared for actual series production of electronic injection systems. Bendix had no facilities to manufacture the solenoid-controlled injectors, no production experience with complex solid state electronics, and almost no engineers familiar with recent developments in microprocessors.

1976 Cadillac Seville engine compartment, a press photo of the fuel injected Oldsmobile 350 cubic inch V-8 (GM Media Archive C2506-R21-0003)

The 1976–1979 Cadillac Seville used a 350 cu. in. (5,737) cc Oldsmobile V-8, although Oldsmobile never offered the Bendix injection system. (Note the “ELECTRONIC FUEL INJECTION” lettering on the air cleaner.) The sole transmission was the three-speed TH375, a TH400 with a shorter tail shaft. Although the injected V-8 had decent power for the time, the hefty curb weight of the Seville and a standard 2.56:1 axle ratio sapped any performance pretensions. (Photo: General Motors LLC)

Given those issues, one might well ask why Cadillac, at least, didn’t simply go to Bosch. The answer is that the territorial restrictions in the cross-licensing agreement with Bendix appear to have prohibited Bosch from selling electronic fuel injection systems to American automakers, which left Bendix the only real game in town. Cadillac chief engineer Robert J. Templin said later that Cadillac selected Bendix because there were no contemporary alternatives “worth a damn.”

As we mentioned earlier, the agreements between Bendix and Bosch provided for mutual technical assistance. We don’t know to what extent Bendix availed itself of those provisions in the development of the Chevrolet and Cadillac fuel injection systems, but Porter’s case study says that Bendix manager Walter Schauer went to Germany for several months in 1974 to study Bosch manufacturing methods, and Bendix looked to Bosch to supply injectors and certain other components (such as throttle-valve sensors) for the GM injection systems. The latter point became yet another logistical headache: According to Porter, Bosch didn’t have enough existing production capacity to supply the almost half a million injectors a year Bendix needed to meet its Cadillac commitment, and wasn’t willing to add another injector plant without a longer-term commitment, which Bendix management was reluctant to make. The eventual compromise was for Bendix to split its injector orders between Bosch, Diesel Kiki, and Nippon Denso.

With the Cadillac delivery deadline looming just 18 months after the signature of the contract, Bendix scrambled to establish ECU production in a factory in Newport News alongside existing air pump assembly lines (not the optimum environment for delicate electronics) as Bendix engineers rushed through multiple last-minute redesigns of the ECU circuitry. The Chevrolet system also had to undergo further revisions: One of the prototype Cosworth Vega engines had failed its EPA durability testing with burned exhaust valves, and the production delays meant the engine had to be updated to meet the more stringent 1975 emissions standards, which cut the allowable hydrocarbon and carbon monoxide limits by more than half compared to 1974.

Cutaway illustration of a 1975–1976 Chevrolet Cosworth Vega engine, with the intake passage highlighted in blue, the combustion chamber and exhaust port highlighted in red, the water jacket highlighted in light green, and the catalytic converter and crankcase highlighted in orange (GM Media Archive X73100-1046)

The twin-cam Cosworth Vega engine displaced 122 cu. in. (1,995 cc) and made 110 hp (kW) and 107 lb-ft (145.4 N-m) of torque, much less than originally announced due to tougher regulatory requirements. While the fuel-injected four complied with both federal and California emissions standards for 1975–1976, concerns about emissions durability led to the inclusion of secondary air injection and a two-way catalytic converter. Chevrolet engineers said these imposed little power penalty, attributing the twin-cam four’s disappointing output to the restrictive muffler needed to comply with drive-by noise regulations, which reportedly cost 21 hp (15.7 kW). (Illustration: General Motors LLC)

All this chaos helps explain why the second-generation Bendix electronic injection system was something less than state-of-the-art even for 1975. While electronic injection still seemed novel and futuristic in concept, the Bendix system remained wholly analog and relied on air/fuel maps that were hardwired into the control modulator circuitry. After early service issues, Bendix added some “limp-home” fail-safe modes, but the ECU wasn’t programmable, so there were different iterations of the ECU for each application, adding to production costs. Speed-density metering based on manifold absolute pressure was also looking a little old-hat by 1974–1975. Bosch was already phasing out the D-Jetronic system in favor of the latest L-Jetronic (sequential electronic) and K-Jetronic (continuous mechanical) systems, both of which used mass airflow rather than manifold absolute pressure as their principal control variable. Although adding an airflow meter to the intake path cost some power, Bosch engineers believed that mass airflow metering avoided the acceleration enrichment issues that had been a persistent shortcoming of speed-density metering, provided better idle stability and transient response, was less affected by exhaust gas recirculation (EGR), and was more tolerant of variations in engine wear and tune. Speed-density metering never entirely went away (Toyota, for instance, used both speed-density (EFI-D) and mass airflow (EFI-L) systems on different home-market models throughout the eighties), but mass airflow systems remain vastly more common today, particularly since the introduction in the early eighties (on Bosch LH-Jetronic systems) of hot wire airflow sensors, which are less restrictive than earlier flap-type airflow meters.

Emissions durability was one of the Bendix system’s most significant shortcomings. Its metering was precise enough to let injected engines meet 1975 and 1976 federal and California emissions standards without a catalytic converter so long as the engine was operating at factory spec, but both federal and California emissions rules also required engines to maintain their exhaust emissions compliance for 50,000 miles (81,000 km) with only limited maintenance (the durability test the Cosworth Vega engine had failed in April 1974). The Bendix ECU couldn’t correct for combustion chamber carbon deposits, increased back pressure, vacuum leaks, or other common age- or tuning-related changes in volumetric efficiency, which could throw off the metering enough to eventually put the engine out of compliance. Both Cadillac and Chevrolet therefore opted to use catalytic converters with the injected engines, with the Seville engine also using exhaust gas recirculation. The Cosworth Vega didn’t have EGR, but added a novel air injection system Chevrolet called Pulse Air.

B&W press photo of a 1975 Chevrolet Cosworth Vega engine, showing the twin throttle bodies of the fuel injection system, the HEI distributor, and the hoses across the valve covers for the Pulse Air system (GM Media Archive X73600-0623)

The production-spec Cosworth Vega engine had dual throttle bodies, high-energy ignition, and a secondary air injection system called Pulse Air. The latter used exhaust pulses to draw intake air through one-way check valves (in the pipes extending over the valve covers) into the two-way catalytic converter, which promoted oxidization to reduce hydrocarbon and carbon monoxide emissions. Chevrolet claimed the system was as effective as an air injection pump, but consumed 6 hp (4.5 kW) less than a pump. (Photo: General Motors LLC)

The fuel injection system Bendix provided for Cadillac and Chevrolet was very similar to late D-Jetronic systems, differing mainly in minor details. For instance, since the GM injected engines used high-energy ignition, the distributor used reed switches rather than breaker points to transmit spike signals to the ECU. The GM systems also ran higher fuel rail pressures (nominally 39 psi/2.69 bar) and had dual electric fuel pumps, one on the chassis, and a second, called a boost pump, in the tank itself, to help prevent vapor lock. There were a few other peculiarities as well: The Cadillac system triggered its injectors in alternating groups of four rather than in pairs as on eight-cylinder D-Jetronic systems, while the Chevrolet system, which normally triggered the injectors in alternating pairs, would trigger all four injectors simultaneously at engine speeds over 6,000 rpm.

Illustration showing photos and drawings of the major components of the Cadillac electronic fuel injection system, superimposed on a drawing of a 1976 Cadillac Coupe De Ville: Throttle position switch; throttle body; fast idle valve (in throttle body); coolant temperature and air temperature sensors; electronic control unit; manifold air pressure [sensor]; fuel pump; fuel filter; injectors (8); fuel pressure regulator; fuel rail; and speed sensor (1976 Cadillac Shop Manual, Fig. 6-110)

The major components of the Cadillac and Chevrolet versions of the Bendix electronic fuel injection system were substantially the same, although of course the Cosworth Vega system had only four injectors rather than eight. Both versions had dual fuel pumps, one in the tank and one on the chassis. (Illustration: General Motors LLC)

Cadillac and Chevrolet Electronic Fuel Injection in Production

Chevrolet and Bendix had expected that the Cosworth Vega would be the first GM car with electronic fuel injection, which it was, but only barely: The injected Cosworth Vega bowed on April 17, 1975, as a late 1975 model, while the new Cadillac Seville arrived just five days later as an early 1976 model.

1976 Chevrolet Cosworth Vega front 3q, a black hatchback coupe with gold pinstripes (Aaron Severson)

The Cosworth Vega was never intended as a big-volume seller, but with its late arrival, disappointing performance, and hefty price tag ($5,916 for 1975, $6,065 for 1976), production didn’t even reach the 5,000 units originally anticipated, totaling 3,507 cars in 16 months. (author photo)

1976 Cadillac Seville front 3q, a press photo of a Brentwood Brown sedan with brown padded roof and whitewall tires (GM Media Archive C2506-0068)

The 1976 Cadillac Seville cost twice as much as a Cosworth Vega — list price started at $12,479, more than an Eldorado convertible — but its sharp styling and more rational exterior dimensions struck a chord with Cadillac customers, and it was a profitable seller, moving 60,127 units for the elongated 1976 model year. (Photo: General Motors LLC)

While the Bendix injection system was standard on both the Cosworth Vega and the Seville, it was also available as an expensive option on most other Cadillac models, presumably to ensure that Cadillac got full use out of the 60,000 systems a year Bob Lund and Ed Cole had committed to buy from Bendix. On other Cadillacs, the option initially listed for about $600, reflecting the system’s high cost: Bendix charged Cadillac more than $325 per system, at least four times the manufacturer cost of a conventional carburetor in that era; the ECU and injectors accounted for roughly 60 percent of the price. Bendix anticipated that the unit cost would come down as production ramped up and the company established its own injector factory, but it doesn’t appear things worked out that way. By 1979, the price tag of the system had risen to $783.

From a performance standpoint, the benefits of the Bendix system were incremental. Cadillac forewent resonance supercharging, opting to use basically the same intake manifolds for carbureted and injected engines, although the ones for injected engines had no manifold heat risers. On the big Cadillac 500 cu. in. (8,194 cc) V-8, fuel injection added 25 horsepower (18.6 kW) and 40 lb-ft (54.2 N-m) of torque, but the 350 cu. in. (5,737 cc) Oldsmobile engine in the Seville gained only 10 hp (7.5 kW) relative to its four-barrel Olds brethren, while the later 425 cu. in. (6,970 cc) Cadillac engine claimed a 15 hp (11.2 kW) edge over its carbureted counterpart. In principle, the injection system improved fuel economy, but even using that term in connection with any mid-seventies Cadillac seems slightly absurd.

1977 Cadillac Seville, front fender "Seville" and "Fuel Injection" badges on a Naples Yellow sedan (Aaron Severson)

Bendix electronic fuel injection was available on all Cadillac models for 1977, but it was standard only on the Seville, helping to justify that model’s higher list price (and mechanical commonalities with other, notionally lesser GM cars). (author photo)

Fuel injection did provide noticeably better driveability than contemporary carbureted engines, at least so long as the system was working properly, but the Bendix system could be very troublesome. A Popular Mechanics owners survey reported numerous problems with the electronic control unit, perhaps reflecting the chaos of early ECU production. Bendix said the reliability goal was a first-year warranty return rate of 1.5 percent, but even before the system went on sale, Bendix engineers publicly admitted it would take some time to reach that target. Owner perception of the system’s reliability may also have been skewed by the Cadillac and Chevrolet service policy of replacing the ECU rather than attempting to repair it, which resulted in some cars having their ECUs replaced more than once.

1977 Cadillac Seville front 3q, a Naples Yellow sedan with yellow padded roof and wire wheel covers (Aaron Severson)

A more vertical grille was the only noticeable change for the 1977 Cadillac Seville (unless you spotted a rare example without the padded roof), but it got four-wheel discs, some minor chassis changes, even more standard equipment, and an $880 price hike. Sales totaled 45,060 — still enough to be very profitable, but short of the original 60,000-unit target. (author photo)

Until near the end of its life, the second-generation Bendix electronic fuel injection system lacked what would prove to be the real “killer app” in fuel delivery technology: “closed-loop” operation, using lambda sensor feedback (see the sidebar below) to keep the mixture as close to stoichiometric as possible. Bendix was certainly aware of the development of closed-loop feedback systems, which Bosch had been pursuing since early in the decade. In fact, Bendix engineers John Camp and Todd Rachel presented a paper on that subject at the SAE Automotive Engineering Congress & Exposition in February 1975, two months before the debut of the Seville. However, Bendix didn’t add oxygen sensor feedback to its electronic fuel injection system until 1979 — two years after the Volvo introduced the Bosch-developed Lambda Sond system — and only on the California-bound 1979 Seville and 1980 Eldorado, which had a new closed-loop control system designed by Pontiac, combined with what GM called its “Phase II” three-way catalytic converter. (The Cosworth Vega was long gone by then, having been dropped in July 1976.)

After that, Cadillac abandoned the Bendix system in favor of a more modern digital engine management system, developed by the GM corporate Engineering Staff, which combined mixture control with digital management of spark timing, exhaust gas recirculation, idle speed, and other emissions-related functions. This system was offered in both carbureted and injected versions. The injected version used a new in-house throttle-body injection (TBI) system, still with speed-density metering, but now with only one injector per cylinder bank. This was a compromise, but it was significantly cheaper than port injection, and the system’s digital controls were far more sophisticated than the analog ECU of the Bendix port injection system. (Bendix subsequently developed its own digitally managed single-point injection system, which debuted on the 1983 Renault Alliance.)

20 Comments

Add a Comment
  1. Great article

  2. I’ve heard that at least at first, Bosch sold Volkswagen the D-Jetronic system at cost because they wanted to get field experience with it. Do you know if that’s true?

    1. I haven’t read anything regarding how much Bosch charged Volkswagen for the early D-Jetronic systems, and I don’t claim to have any idea how their contracts were structured (which I suspect both Bosch and Volkswagen would be reluctant to divulge). However, I’m skeptical because that assertion is hard to reconcile with the scale of Volkswagen’s production commitment. Volkswagen saw themselves as up against the wall in a regulatory sense, so they didn’t dabble. Their initial orders were for 10,000 systems per month, with what I have to assume were pretty ironclad delivery deadlines! This wasn’t a matter of buying a few early ECGI systems to offer optionally in high-end models for familiarization purposes; Volkswagen was making D-Jetronic standard equipment on all U.S. Type 3 models as means of regulatory compliance (without which those models would have to be withdrawn from VW’s biggest export market). Also, given the magnitude of the investments Bosch had to make in development and manufacturing capacity, plus the Bendix royalty situation, even determining what “at cost” might mean under the circumstances would seem to be a complicated and subjective question.

      The only specific figure I have for the cost of the system was what Volkswagen charged for D-Jetronic as an option on the German 1600 TL, which was initially 600 marks. This was at a time when the exchange rate of the Deutschmark to the dollar was still fixed at 4:1, so that’s the equivalent of $150 USD, which does admittedly seem very cheap even for the time.

    2. I was thinking about this, and it occurred to me that this might be conflating the production D-Jetronic systems with the 40 preproduction prototype systems Bosch built in the autumn of 1965. Bosch retained some of those for testing and provided some to Volkswagen for evaluation and familiarization. I can readily believe that Bosch may have sold those systems to Volkswagen for a nominal cost, which is a different matter than the regular production orders that began two years later.

  3. Ridiculously comprehensive, as usual!

  4. Anecdotally, I worked on T3s in the mid-70s for about five years at an independent shop and a couple of dealers. D-Jetronic was extremely reliable except for leaky injector lines, the ~3 inch fabric covered rubber lines that are part of the injector. We’d simply cut off the integral swage and put on a new line using a small hose clamp.

    The most reliable component was the analog computer. As I recall there wasn’t a single electrolytic capacitor used. Rather, much more reliable tantalum capacitors were used which were much more suitable for automotive use, more resistant to temp swings and vibration. I probably worked on a couple thousand T3s and don’t remember replacing a single computer. I did replace a few fuel pumps and manifold pressure sensors.

    That was a long time ago so I might not remember exactly; I do remember customers praising their T3’s gas mileage compared to their carbureted T1s.

    1. According to Hermann Scholl’s remarks, the most persistent issue Bosch faced with the ECU was its wiring harness and connectors rather than its individual components, although the large number of the latter was a headache from a manufacturing quality control standpoint, and drove up the price. Other comments I’ve heard suggest that the aneroid bellows were the weak point of the MAP sensor — over time, the rubber of the bellows would crack, and the resulting air leaks would cause the sensor to malfunction, since it was set up with the assumption that the pressure inside the aneroids was effectively zero.

      1. I thought the T3 MPS sensor was a bit more sophisticated than a modern-day MAP sensor in that modern FI systems have downstream sensors – 02 sensors in particular – that provide additional info to the ECU. Maybe I’m wrong.

        As far as reliability, I wrenched T3’s in the humid Midwest. Perhaps the bellows were prone to crack in drier environments. And of course that wrenching was done 50 years ago. Ten years after that it’s likely a higher proportion of those diaphrams developed leaks.

        1. There’s a lot of detailed technical and troubleshooting information on D-Jetronic in Paul B. Anders’ 914 pages at https://members.rennlist.com/pbanders/ — Anders is an electrical engineer (which I am not!) who has also spent a lot of time wrenching on 914 systems (which I have not). He mentions that the MPS aneroids would eventually crack with a lot of full-load operation (see https://members.rennlist.com/pbanders/manifold_pressure_sensor.htm). It does seem reasonable to surmise that this might be exacerbated by aggressive driving styles and hotter/drier conditions, although Anders sees it as a design flaw, which also seems fair.

          The D-Jetronic MPS was unquestionably more sophisticated than the MPS in the Bendix Electrojector system, which wasn’t contactless, responded only to differential (not absolute) manifold pressure, and had a tendency to move to the extremes of its travel (something the Bendix patents note) rather than providing a really progressive response. However, the D-Jetronic sensor was still analog, it was mechanically complicated compared to a modern capacitive or varistor MAP sensor, it didn’t have its own temperature sensors (requiring separate correction for intake temp), and it didn’t have any kind of self-diagnosis capability. So, it was a step forward, but it definitely wasn’t the last word even for speed-density metering systems.

          It’s definitely true that the MPS was substantially more critical for D-Jetronic than the MAP sensors in later digital systems. D-Jetronic (and early L-Jetronic systems) had hardwired air-fuel maps, and they didn’t have memory in the way modern engine computers do. With digital controls, you can potentially store several sets of air-fuel maps as lookup tables and use those along with other sensor input data (coolant temperature, throttle position, lambda information from the O2 sensor) as a supplement or fail-safe in the event of sensor malfunction; the computer can also self-adjust for variations in volumetric efficiency due to age or state of tune. D-Jetronic couldn’t do any of that, and so it was heavily dependent on the proper function of the MPS, and on the engine’s volumetric efficiency remaining pretty close to the expected values.

  5. Thank you for this (well researched, as always) article. It answers a question that has puzzled me since the 1970’s- why didn’t Detroit go immediately to EFI in the 1970’s? The malaise era Detroit Carburetored engines were a such disaster and fuel injection was such an obvious answer. The answer as to why not, it seems, is Bendix. As the (then) owner of a mechanically fuel injected 1974 Alfa Romeo, I knew that fuel injection was the answer to fuel economy, drivability, and emission concerns, and I also knew that mechanical fuel injection was already obsolete. EFI was obviously the answer but “no one” adopted it.

    I had always thought that it was merely the big 3’s stinginess that kept them wed to carbs. VW had shown that EFI could be used economically and ‘reliably’ on inexpensive cars. However I didn’t know of Bendix’ stranglehold on the U.S. market. So while I was correct that it was old cigar-smoking pennywise, pound-foolish bean-counting fools, I misplaced them – they were at Bendix.

    1. I also came away feeling like Bendix really flubbed the whole thing: They pushed this technology to production before it was ready, alienated at least two major U.S. automakers in the process, and were then apparently prepared to just sit on their jealously guarded patents until the clock ran out. Even when they had a new customer waving money at them, they remained so reticent about making any big investment that it kept their costs and prices prohibitively high.

      That said, it isn’t the whole story. As Michael Porter notes, U.S. manufacturers had significant sunk costs in carburetor development and manufacture, and so from a capital investment standpoint, it was cheaper for them to keep modifying carburetor designs for emissions compliance than to switch to even single-point injection. Volkswagen was an unusual case because they decided their backs were to the wall AND they were prepared to eat more of the cost than most U.S. manufacturers would have deemed acceptable in order to maintain their price point. This is something that became a growing issue for Volkswagen during the waning days of the air-cooled era, as Bernhard Rieger discusses in his 2013 book The People’s Car: VW recognized that they needed to make improvements, but they also had to hold the line on price, so their profit margins began to shrink precipitously.

      The situation in Japan was perhaps more broadly representative. By the mid-seventies, Japanese automakers were also facing significant challenges in emissions compliance (their domestic standards were tougher than ours by 1978), so they were very interested in electronic fuel injection, but they weren’t prepared to swallow the significant price premium (especially since local manufacture was still royalty-encumbered). As a result, JDM adoption followed a pattern similar to what eventually happened here: Senior grades got fuel injection first, and it gradually spread downward, not becoming universal until much later. Some lower-grade models still had feedback carburetors well into the nineties. Their domestic trend was a couple of years ahead of ours, for a number of reasons, but the trajectory was roughly the same.

      The other oddity here is that Bosch was extremely successful for years with K-Jetronic, which arrived about five years after D-Jetronic and came very close to killing the Bosch electronic systems. K-Jetronic was a clever piece of work, much less complex than the older Bosch mechanical systems or the SPICA system in your Alfa and yet precise enough for emissions compliance, particularly with feedback control. Many European manufacturers traded D-Jetronic for K-Jetronic, even Mercedes and Porsche, and stuck with it for a surprisingly long time. Mercedes stuck with the electromechanical KE-Jetronic system through the early years of the R129 SL-Class, on which cost was presumably not a big issue. (I think the reason they finally switched to LH-Jetronic was that Bosch was phasing out the electromechanical systems, although OBD-II requirements would have forced the issue eventually anyway.)

    2. The other consideration is that many of the factors that have made modern electronic fuel injection so useful depended on the availability of relatively cheap, reliable digital processors. The earlier analog systems, even with closed-loop feedback control, didn’t have a memory, were only programmable by making physical changes to the control circuitry, and had little to no self-adjustment capability. Sixties digital electronics were not yet up to the task (the early Autonetics VERDAN computer used by the USN had a mean time between failure of only 15 minutes!), and while the technology evolved very quickly in the late seventies, I can’t really seeing it being successfully implemented in series automobile production much earlier than it actually was (which started around 1979–1980).

  6. Thank you, Aaron, for this discussion of the Bendix Electrojector system and its influences. My interest stems from back when I owned a Chrysler C300 many years ago, and as a member of the Chrysler 300 Club, read about this system. I always wondered about it, especially since its influence has become so widespread in the automotive world. I wondered, ‘What went wrong with the Electrojector?’

    As you describe it, the first implementations by Bendix were crude. But, the carburetor is a crude device, too, and the Rochester mechanical system was not so far removed from a carburetor, using venturi vacuum signals and metering rods. It was crude, too. But both the Bendix and the Rochester systems offered the ability to improve on air/fuel distribution, and there’s not much reason why the Bendix couldn’t have been successful in performing this task, if not successful in the market, like the Rochester was, more or less.

    As a general observation, I think that Detroit’s developmental weaknesses are highlighted by your description. This is especially due to the dismal performance of so many electronic aspects, such as the first HEI modules. Not much learning took place there, as I had an early Ford Taurus that was overall a good car, except for the TFI ignition module, which consistently failed.

    Bosch and Volkswagen seem to have tried a little harder, and stuck with the problem a little more. I would guess that Volkswagen customers in the late 60s and early 70s might have had a stronger affinity for the marque, as well. I recall that it was common to replace the Bosch system with carburetors back in those days.

    The Japanese seemed to do a better job at this, which was reflected in the marketplace. (The Acura/Sterling comparison is an interesting aside to this: taking a very reliable engine and making a disaster of it also demonstrates how car makers can be tripped up by electronics.)

    Even in my old Volvo/ Bosch S60, there are glitches in the computer operation that should have been worked out, but were passed on to the consumer, even though my car does work fine as long as I observe the limitations.

    I also found your mention of the aftermarket Conelec system interesting. I recall seeing Conelec fuel pumps available on the aftermarket decades ago, and wondered why this odd name popped up to compete against traditional brand products. Now, thanks to your thorough research, I know!

    Great article, thank you

    1. The conceptual challenge presented by moving from carburetion to electronic fuel injection is akin to the difference between learning to walk and trying to create a mathematical model to make a robot walk. This was particularly troublesome with starting (and especially cold start), which I sort of skimmed over, but was a big development hassle: You know you need a richer mixture for starting, especially when cold, but exactly how much for how long? And what do you do if the extra enrichment is fouling the plugs? It was a matter of trying to precisely define factors that previously got by with workable approximation, without the benefit of sensory feedback to allow self-correction or the ability to remember what did and didn’t work last time.

      Where the Electrojector really fell down was in the difference between theoretical models of electrical circuits and the more complicated physical realities. In theory, the injector valve operation was quite simple (energize the solenoid to open the valve, hold it for the requisite duration, and then cut the current and let the spring push the valve closed), but that didn’t take into account things like eddy currents generating enough of a residual magnetic field to keep the solenoid partially energized for too long. Even Bosch, which had extensive experience with mechanical injection valves for diesel engines, struggled with that at first, although they got it sorted more quickly than Bendix had.

      I don’t know that I’d call the Rochester Ramjet system crude; it’s not as intricate as the Bosch jerk-pump mechanical systems, but it’s a clever approach to mass airflow metering. It seems to have been designed as something that could eventually be mass pass produced relatively cheaply, although it would likely always have been more costly than a carburetor, and low volume kept it expensive.

      There’s no question that automotive electronics have had a rough learning curve. The environmental conditions are difficult (heat-cycling, vibration, dirt), and since the adoption of the technology has often been driven primarily by regulatory compliance, certain performance metrics have to be prioritized over all others. For obvious reasons, this has tended to be a bigger problem during periods where the regulatory requirements (usually emissions-related) are a moving target. A lot of what’s made such a difference in recent decades is integrated digital electronics with greater fault tolerance — both of faults with the electronics themselves and of mechanical variance — and the ability to learn and self-adjust. To return to my original analogy, the reason toddlers are able to learn to walk is not that they become mathematical geniuses, but that they’re able to adjust based on feedback and experience; they may still wobble and sway, but as they learn to compensate, they fall down less and less.

      1. Your pointing out the difference between learning to walk vs programming a robot to walk is well-taken, especially without feedback systems. The cold start issue, I also see now and take into acount.

        I would have done better to ponder the intricacies of injectors regarding magnetic properties that evidently tripped up early development efforts.

        When I used the word “crude” regarding the Rochester injection, I perhaps could have chosen my words better. I was thinking in terms of carburetion — which in one sense are simple analog mixing devices with provision for acceleration and cold start (and maybe a few others, depending on the implementation) — but in another sense do have complexity in those simple concepts which, when refined, can make for a pretty capable system. The Rochester uses the same analog systems and signal concepts, but rather than, say booster venturis and emulsion tubes, uses a straightforward constant injection into the port.

        Comparing this with later port injection systems with lots of signal inputs and a fairly complex computer program (which also incorporates spark control, an additional advantage), my comparison was in the vein of so many electronically controlled devices of today compared to manually of mechanically controlled analog devices from the 50s and before.

        You are right about the price of the Rochester system. As I recall from the 60s, many took it off for carburetors, while my neighbor would buy the systems for small sums and run them on his Chevys with performance and economy simultaneously. He claimed that it was only a matter of following the service manual correctly to achieve this — no surprise there!

        re: electronics, HEI only replaced points in its early ~1975 iteration. I recall it being of questionable reliability, such that one might want to keep an extra module in the glove box. However, by 1977, my vehicle with HEI was trouble free. I don’t see the technology changing sufficiently in those two years to justify a “learning curve” if adequate testing was done.

        Certainly not over 10 years later, with my Taurus’s thin film module, which was a failure looking for a place to happen, which it did on a number of unfortunate occasions.

        1. In a 1957 article I looked at, Roger Huntington reported the surly comments of some unnamed Detroit engineers to the effect that all early injection systems were essentially continuous because getting the injector valves in intermittent systems to close completely and on schedule was wishful thinking — not so much on and off as “flow” vs. “dribble.” That was essentially Chrysler’s experience with the Electrojector injectors, so they had a point, particularly given the short duration of timed injection pulses. (Pulse width for the Electrojector in normal driving was supposed to be between 1.0 and 4.5 ms, so you can see how a solenoid injector valve that took 8 or 9 ms to close fully would be a problem!) Also, Chrysler found, as Bosch later determined as well, that it really didn’t matter whether the valves were closed at the time of the pulse, which is why D-Jetronic and the seventies Bendix system triggered their injectors in pairs or groups rather than one at a time. For direct injection, it was useful to time the injection pulse to arrive after the intake valve had closed, so as not to spray fuel back into the port, but with port injection, it turned out not to make a meaningful difference. So, in that respect, timed intermittent injection did not have a particular advantage over continuous injection in sophistication or performance, which is also part of why K-Jetronic remained competitive for as long as it did.

          A key point that’s easy to misunderstand is that fuel injection for passenger car applications has always been computerized, whether it’s mechanical or electronic. A mechanical system is analog, of course, but so is D-Jetronic, and the fact that its computations were based on voltage and resistance doesn’t actually make them more computerized. This becomes especially apparent if you look at the workings of the early Bosch mechanical systems, which had a complex assortment of different mechanical sensors. They did the same things as the analog electronic sensors in the Electrojector, just not quite in the same ways. The advantage of the electronic system was ultimately in growth potential: It could be more readily adapted to incorporate digital memory and additional input parameters, although none of that was present in the early systems.

          The seventies were a time of very rapid development in electronic components, so there were sometimes significant differences in the span of two or three years. I haven’t studied HEI systems in any detail (there was an in-depth three-part article on Curbside Classic a couple years ago that covers the subject well), but early systems were dubiously reliable in ways that were worked out in relatively short order.

          The challenge of “adequate testing” during periods like the seventies and early eighties is that automakers were on a tight development schedule for regulatory compliance, which was a moving target. For instance, between 1975 and 1977, federal standards for oxides of nitrogen were lowered from 3.1 g/mi to 2.0 g/mi, while California NOx standards went from 2.0 to 1.5 g/mi and HC fell from 0.9 to 0.41 g/mi. Powertrain engineers were rushing to certify powertrains that met one set of standards while also scrambling to revise the hardware to meet the next set of standards, which meant there were a lot of essentially interim features that worked well enough for shorter-term needs, but weren’t going to be continued for long. This is the conundrum of regulation-driven change: It resulted in many headaches, but without it, a lot of the ultimately beneficial improvements in engine technology would never have become universal; cheaper cars would probably still be carbureted and have breaker-point ignition. I don’t think automakers necessarily handled these challenges well, but it would be a mistake to overlook how much of a challenge it frequently was.

  7. Excellent and fascinating article. I enjoyed the article so much I sent you $. Haha. The thing I find amazing is Bendix or Chrysler didn’t consider just running two larger injectors (instead of the 8 port injectors) at the throttle body. This would have slowed down the injector pulse width dramatically. Likely to a speed the fairly primitive control unit could have handled. As you seem to suggest they got pretty close then maddeningly just basically gave up

    1. Bendix did definitely consider throttle-body injection — in fact, the first illustration in their most important electronic fuel injection patent is of a throttle-body injection iteration of their concept. Throttle-body injection was discussed and proposed throughout the fifties, but nobody put it into production because the cost-to-benefit ratio just wasn’t worth it. Once you put mixture formation back into the throttle body, you lose most of the actual performance benefits of fuel injection: The manifold isn’t dry, it still needs to be heated (except for pure racing applications), and mixture distribution isn’t much better than with a carburetor. Single-point injection (i.e., one injector per bank) provides better fuel vaporization and more precise metering, which is better for emissions than carburetion, but that wasn’t very high on the agenda in the late fifties, and it was not enough to justify what still would have been a $200+ price premium. It would also have had much poorer performance than contemporary mechanical injection systems.

      The problem with the Electrojector was not that the control unit couldn’t handle the speed required for the pulse width range involved (although there were early problems getting enough amplification), it was that Bendix and Chrysler had very great difficulty getting the solenoid injector valves to close fast enough and completely enough to make the calculated pulse width values meaningful. Chrysler eventually got the closing time down to 0.8 ms, which was an order of magnitude improvement over the valves they initially got from Bendix, but the production tolerances were still so poor that they had to basically just hand-select matched sets of injectors. Having fewer injectors would not have helped that, but it would have meant that there wasn’t even a nominal performance benefit to the injection system.

      The Electrojector had some design limitations, but its principal failing was that getting the relatively straightforward concept to work reliably ended up being a lot more difficult than anticipated. Even Bosch, which had lots of experience with diesel injection valves, struggled with it, although the pressure of their sizable production commitments gave them strong incentive to work it out.

  8. My father bought my mother a 1 year old ’76 Seville in ’77. They loved the looks and didn’t want to hear my brother’s and my criticism of it’s ox-cart suspension and Nova roots.
    In any case, expecially for the era, the advantages of the EFI were immediately apparent. The engine started immediately with no need to “set” a choke, it never stumbled or hesitated, and considering the weight of the car, it got pretty good gas mileage and was reasonably fast. I recall burying the 85MPH speedometer so the pointer completely disappeared inside the dash and then set the cruise control.
    It did need to be towed a couple (several?) times due to not starting. It seems to me that I heard it was the HEI, not EFI that kept failing. I can tell you this though, my parents were not impressed when one of those tows was when I was using the Seville, when home from college on break, and it wouldn’t start in front of a liquor store about 1 AM on a Saturday morning….THAT I clearly remember.

    1. And I completely forgot to compliment on a wonderfully written and very informative story. I had wondered for a long time why AMC and Chrysler ended (or aborted in AMC’s case) their Electrojet option, and then, why it never reappeared.
      BTW, Bendix had a manufacturing facility in Towson, MD, which I believe made radar and radio systems. By coincidence, I work next to the old plant. In any case, I heard stories about 20 years ago, that they made the original EFI systems at that plant – I’d say that was nothing more than a story.
      Fantastic article!

Leave a Reply

Your email address will not be published. Required fields are marked *

Comments may be moderated. Submitting a comment signifies your acceptance of our Comment Policy — please read it first! You must be at least 18 to comment. PLEASE DON'T SUBMIT COPYRIGHTED CONTENT YOU AREN'T AUTHORIZED TO USE!