What really killed the Apple III
A technology study of the Apple III
The Apple III  was a mysterious Apple offering, one that isn't well known as a system probably due to its short life span. But the 1980 Apple III did represent many firsts for Apple. The III was the first system designed by Apple as a company. It was designed by two key groups, primarily marketing and engineering. It was the first design to cost Apple 'infinite, incalculable amounts' (as quoted by Steve Jobs: 'Playboy' 1985) in research, development and promotional funds. It was also Apple's first market failure. For the long drawn out battle since it's inception in 1979 to the day it left Apple along with Steve Jobs in mid-1985, the Apple III was responsible for bringing out the true character of many people within the company.

The history of the Apple III is rich to say the least, and reads like a disaster story, even though the machine and its design had so much potential. The engineers who worked on the III were highly talented and capable. The support and resourcing required to complete the product was prolific, to a point it seemed almost unlimited. And the market potential for selling the new III to the existing enthusiastic Apple customer base was obviously there. Many people have written about how it all came undone for the III, detailing problems within Apple, and their lack of ability to work as a team being the primary contributor, but also stating that the accelerated design cycle was also to blame for its initial failure. Upper management were accused of enforcing certain design criteria, such as a fanless build and a significantly increased number of integrated circuits that were required to be accommodated inside a fixed case size. All this was topped by an unchecked marketing group, who dictated a day-by-day definition of what the product had to be, and applied maximum pressure to the design teams to have the machine available for sales as soon as possible, due to a perceived threat that Apple II sales would drop off.

The outcome: a potentially great machine was doomed, a large amount of money was spent, a lot of time wasted, and importantly some key lessons were learnt. One was what 'bad press' really meant, the other was 'maybe marketing should stay in marketing.'

The Apple III was dogged by many design faults, such as chips coming out of sockets, real time clocks not working, and excessive heat problems due to over populated boards. Although over a period of twelve months Apple had worked vigilantly to correct these problems, the solutions seemed to come too late. The final release of the machine was in the form of the Apple III Plus (III+), a fully functional and reliable business machine which possesed all the characteristics Apple had hoped would be in their initial release of the III; but were not. The damage that the III's first impression made on the market led to a very poor reputation. This was so significant that the machine as a concept, functional or not, would not sell. In 1984, the Apple III plus, after only two years on the market, was quietly removed from the price list. By mid 1985, the III was no longer part of Apple's business - sales, service or otherwise.

I was lucky enough to come by one of the earlier Apple III machines, which I used to aid the process of getting the design into the FPGA design space, and also took the opportunity to conduct my own design study analysis to get a better understanding of why this machine was considered so bad. These are my findings:

Objective

To see what design elements comprise the infamous Apple III, and to compare them to the rumours and reports to determine which of them would seem to be true.

Claims and Rumours

It is commonly known that the Apple III suffered from:

- Chips coming out of sockets
- Unrealistic high density PCB routing (six tracks between IC pins)
- Real time clock problems
- Memory board problems
- Melting chips and/or floppy disks (thermal issues).

Chips coming out of sockets

This was the first reported issue the Apple III experienced when released in 1981. Accordingly, the chips in their sockets would work loose. In my own experience, I find this very interesting. I'm not sure if I have ever come across this problem happening so often and so soon in a product's life. After sending out a few emails, it appears this was correct, but primarily due to a new "automated assembly procedure" where DIP Chips were inserted automatically by robots. I am awaiting information on where and how the III was assembled, but it appears the process must have been really, really bad (ie, only 2/3 fully seated) to have this effect. I took an Apple III main board from an early revision (820-0043-00)  and conducted a cross section of the IC socket. The socket, understandably, uses the same AMP and RN brands used in the Apple II, and later in the IIe amongst other Apple systems. So the quality and pedigree of the socket itself is not in question. Although the temperature was higher inside the Apple III than any other Apple, I still cannot see how these integrated circuits could possible work loose if they had been seated correctly the first time. To remove themselves using thermal cycling would take years, as the thermal matching between the integrated circuit's pin material, and the material of the IC sockets (both formed tin-plated) are almost identical. Considering this, I strongly suspect the initial assembly process for automatic integrated circuit insertion could be the main contributor to this problem, and not heat, as is commonly speculated.

Unrealistic high density PCB routing (six tracks between IC pins)

As a product designer myself I found this claim very impressive; I have always been inspired by Apple's designs, and regard them as very well made. For a long time I believed the fineline six track routing rumour. One of the rumours was the issue of broken tracks due to such small traces. Even with a conservative annular ring (land) on each integrated circuit pin of 45 thou, with six tracks between the legs of a 2.54mm DIP chip, this would mean tracking of about 3 thou with 3 thou spacing (which is fineline). For 1980, that would be exceptionally impressive, especially for a board that is 39cm x 24cm in size where a linear error of up to 5 thou can occur. As it turns out, after studying the Apple III logic board, its clearly the most densely routed PCB Apple had made to date. Although the layout would be considered a high density route, it certainly is not fineline. After a thorough inspection, I could not locate any more than a mere single track between the legs of an integrated circuit, even then it resembled eight thou in width with approx five thou clearance spacing.

I did, however, locate another manufacturing problem I had yet to hear of, one that is directly associated with the density of the board's routing. As part of the board's design it was covered in risky untented vias. These through hole connections are used to interconnect layers (in this case, to traverse an electrical connection between the top to bottom layers). By today's standards, all vias are tented for a lot of good reasons. But in the 80's it was a common trend to leave them exposed to allow a solder fill to aid reliability of the joint. As a result, when the board is soldered by the wave-soldering equipment during manufacture, not only do each of the through hole pins get soldered, but a small amount of solder is deposited on the exposed vias. This is ok, as long as the via has a clearance large enough to prevent bridging to other exposed joints, such as through hole pins of integrated circuit sockets and/or other vias. The Apple III however, with its high density two layer route, contains exposed vias so close to adjacent pins, that even without wave soldering, it looks as if they could short. Add this to the seven power distributing thermally mismatched Rogers bus bars, and you have a board that would have been a nightmare to produce.

Real time clock problems

One of the more common complaints of the Apple III is the National Semiconductor Real Time Clock, the MM58167. To date, I have not seen a description of the symptoms of what the real time clock would exhibit when it failed in the Apple III. Just that it failed after some time (no pun intended). But there has been a solid trend to point the finger towards National Semiconductor, with claims of low quality and a reputation for shipping known bad devices. If you take a look at how Apple implemented their real time clock in the Apple III, it may reveal a little more. I have designed with the MM58167, as well as the MM58274, so have had the opportunity to develop an understanding of the characteristics of these devices. To this day, I have found both devices to be quite reliable, but also found them to be very implementation-sensitive, to the point where only tight coupling in my designs produced the desired reliability. They have a tendancy to very quickly gain time, and/or lose time when not implemented correctly. On the Apple III the reference oscillator for the RTC, the 32.687KHz crystal oscillator, is located more than 40mm away from the MM58167 itself. The oscillator had also been routed between two busy bus devices, the R6551 UART (B1) and the R6522 Versatile Interface Adapter (VIA) (B4). To place this ultra sensitive high impedance reference device at this location is highly precarious, and could well produce spurious reference oscillations. In my experience, this type of implementation would typically end up swaying more towards being a random number generator than a real time clock.

Memory board problems

The Apple III was designed with a daughter board-based memory card that was mounted above the main board two thirds down. It was mounted using 2.54mm Molex extended pin temporary connectors, and was able to be removed and/or upgraded. This provided several advantages, and also gave Apple the flexibility to sell the machine in several different memory configurations. Yet another common complaint of the AIII was the claim that the memory board was intermittent. The reasons why the memory board failed were again associated with high temperatures inside the Apple III's case. So high as to cause accellerated corrosion of the plated pins. After acquiring a early (81) revision Apple III mainboard, I took the opportunity to look at these connectors in more detail. As it turns out, the interconnectivity between the mainboard and the memory board would appear exceptional. Not only is there a very firm mating force required to mate and release the two parts, but the connectors make contact in three locations on each pin. In my opinion, this type of connection has a very high probability of working correctly the first time, for all time. Admittedly, the tin-plated pins don't fare as well as their gold plated counter parts, however on this thoroughly worked Apple III, there was very little sign of any surface corrosion on the memory board's connectors or the main board's pins.

Melting chips and/or Floppy Disks (thermal issues)

As part of researching the Apple III over many years, and recollections of my own III, I have come across a few interesting claims. Maybe a more unusual one was that the Apple III would "melt" floppy disks in its internal drive. I do recall my own Apple III warming the floppys significantly more than the Apple II. However, it certainly wasn't near melting point. Floppy disks by the way do like being warm, as this prevents moisture inside the jacket on the cleaning liner from gripping onto the mylar suface of the disk, which can occur when damp. Really, to answer this question, maybe taking a look at the Apple III's switch mode power supply will help. I do know the Apple III had a Astec SMPS, with a rated power output of around 200W with a typical efficiency capability of around 80%. To produce this much heat inside the Apple III,  to a point where it could melt the jacket of a floppy disk, sounds more like the machine had the efficiency of a light bulb. Although the aluminium chassis was joked about, it did provide a significant surface area to dissipate stored heat. Common IC's are specified to operate at temperatures of up to 70 degrees celsius, well above the threshold of human pain. So the AIII was operating well inside specification. It would have been other components, such as the electrolytic capacitors, that would have suffered first from the warmer than usual ambient temperature, and even then, that would have taken several years to exhibit any symptoms.

Conclusion

In my opinion the Apple III was actually a very good design, but simply suffered from one or two minor implementation bugs. In product design, if a single flaw preventing a released or shipped product from being used occurs, then a risk exists that has the potential to destroy new markets and render a product or technology as flawed. First impressions are everything. This is a company's worst nightmare, especially if the product has the potential to become a de facto standard or has existing market demand. It is for this reason that many new technologies introduced to the market are heavily over-engineered. Although this usually means the product is accompanied by a high price, it does give it the greatest chance at market life. Over time as the reputation for the product develops, new cut down versions where the over-engineering is designed out are released, and the price lowers. The Macintosh classic is an excellent example. So is the Platinum IIe. The idea is that there is always at least one reliable model on the market to fall back on, and if there isn't, then the first release must be right first time. It is this important understanding that usually tends to encourage  marketing to not be allowed anywhere near a product until the prototype is complete.

As for the Apple III, over $60 millon 1980 dollars had been spent bringing the machine to fruition, so if something were going to stop it from being the great success Apple had planned, it would not be something trivial. If the Apple III was going to fail, it had to be of such monumental scale as to completely shroud the R&D dollars spent on it; and that failure was most certainly not going to be a few poorly placed components and some badly inserted chips. However, it's hard to look past these facts as the only key contributors.

Fully Seated: The Apple III uses the same proven AMP1 and RN branded sockets as used in other Apple products, as well as both the Apple II and IIe.
Full House: 1980 High Density routing shows a typical area of the Apple III main board with the majority of the PCB being used for routing interconnections.
Under Pinned: A side section shows the contact surfaces of the integrated circuits pin inserted into the blades of its socket.
High Accuracy?: Apple III's Real Time Clock, and it's implementation on the III's main board.
National Semiconductor: The MM58167 Real Time Clock suggested implementation application note.
The Apple III blueprint: The circuit diagram detailing how Apple implemented the National Semiconductor MM58167 Real Time Clock in the Apple III.
The beginning of Memory Modules: The Apple III 256k/512k DRAM memory board in a "plug-on module" format.
keep28@hotmail.com
Memory IO: The extended 2.54mm interconnect pins of the Apple III Main Board, J16 and J17.
Inside Story: Exposed mating characteristics of the Apple III main board to memory board when fully seated.
Precarious Design: Untented exposed vias threaten to short to near by circuits during the production process of wave soldering.
Production nightmare: Tightly grouped vias and pads are soldered simultaneously during wave soldering. Strict manufacturing specifications and tolerances such as solder temperature, quality solder mask printing, flow direction and timing are critical to prevent shorts.
Best Effort: Only one route between the pins of a DIP IC was achieved on the Apple III main board.
Business Connections: The cable securing "Jack Screws" was omitted from the Apple III's peripheral cable IO panel.
Thermal Relief: The "Heat Sink" style of the Apple III chassis, manufactured by an automotive engine producing plant, provides the warm Apple with a passive cooling system.
Team Effort: Apple's first product offering as a company, the Apple III. The demand was there, the design was sound, however Apple's lack of ability to work as a team marred the Apple III to the point of complete market failure.
Industrial Strength: The Apple III's Expansion Card Cage. Four expansion cards can be populated inside the aluminium chassis. Apart from only a few pins, and a slightly different mechanical outline, the AIII expansion cards are almost identical to that of the Apple II's.
Case Open: Inside the Apple III illustrates each of its RFI/EMI sealed aluminium chambers, with the power supply on the left and the main board to the right.
First Cut: Apple's first chips designed for the ill fated Apple III roll off the lines in early 1980. The 341-0032-00 Video Mode Controller was completed in 1979.
Proven Technology: Recycled devices taken from the Apple II system. The 341-0028-01 P6 PROM borrowed from the DISK ][ Controller helped the Apple III design take form quickly.
Card Pool: The Apple II electrically compatible expansion slots are identical to those on the Apple III's, except for pin 35-(IOSYNC) 39-(C02x) and 19-(PH0) - All of which are unused on the Apple II.