7

Among other early computers, the Apple II and IBM PC made the consequential decision to provide internal expansion slots.

How much did this add to manufacturing cost? Intuitively, it seems like it should add quite a bit, but I don't really know, and I'd like to get some idea at least to an order of magnitude.

12

Connectors are surprisingly expensive when you build a device:

  • The part itself is usually quite pricey, relative to other components.

  • They require a different assembly step and tooling.

  • Testing can not always be automated.

  • They present complexity in routing, especially in old systems where connectors expose the whole bus.

  • They require additional components, to deliver power but also buffer the signals.

  • When analog signals are also present on connectors, care must be given to make sure that nearby digital signals don’t introduce noise. On the board the sections are usually separated but in a connector, sometimes the two worlds get in close proximity.

Ports are one of the first things to be looked at when redesigning boards.

A funny annecdot: when Atari designed the portable ST, they had musicians in mind as the machine was popular with its MIDI ports; to make it smaller and cheaper, they removed the cartridge port; on the ST, the company that relied the most on it for its products was Steinberg... The leader, at the time, in music software :)

  • 4
    You might want to add that it isn't simply about adding connectors: You also need to re-design the case for "intentional customer access", document stuff and built-in safety measures. – tofro Jan 28 '17 at 8:45
9

If you want to think about cost of manufacturing, you would have to figure in not only cost of the connectors but also motherboard layout, space in the computer (to accommodate width/height of internal expansion cards) and possibly increased wattage of the power supply.

However, I don't think you can measure cost here and frankly I don't think that was the real focus. I would think it was more because of the evolving technology of personal computers at the time and that people were still coming up with new ideas of how to use these tools; if you as a manufacturer decided not to include slots, the consumer could only expand to what was made available port wise. The original Mac tried to take this approach by requiring special tools to open (and then not having any user servicible parts inside) and thus it made it very difficult to expand; you only had serial ports and an external disk drive for the most part as there was no RAM expansion capability. Heck, even adding a hard drive to those machines was a bit of a hack and not very performant.

In comparison, the open architecture might have come from two places:

  1. The S-100 bus model. The Altair 8800 and derivatives didn't have ANY ports and thus the only way you could expand the thing was by installing expansion cards. I assume this was not so much because of a "free and flexible" design but rather they hadn't come up with everything yet (e.g. video card or serial ports). These expansion cards eventually appeared, but not always from the manufacturer (e.g. Cromemco video worked in an Altair machine).

  2. Trying to keep cost down while allowing for flexibility to have your dream machine if you wanted it. The Apple II didn't have a disk at first, but frankly you could have survived with just the cassette ports. It didn't have a serial port, printer port or a hard drive, but you could add those things if you wanted and incur the cost at that point. These machines weren't cheap when they came out - imagine how stratospheric the cost would have been if they had these things from the factory? It isn't like there was much to bulk discount with suppliers when they were still a curiosity to most and thus only selling in the thousands at times (compared to today when Dell probably sells thousands an hour).

On the other hand, then why did the Sinclairs, Commodores and Ataris not have slots? I think we might blame organizations like the FCC as there was a point in the late 70's into early 80's where RF emissions were a major concern and thus these machines needed to be fully shielded thus making it hard to allow expansion slots. I believe the designers of the Atari 400/800 wanted slots but because of their fears of failing FCC tests they had to encase the motherboard in heavy shielding and thus only had a small opening to the outside world in the form of their SIO port. Eventually, companies like Atari had enough ecosystem built around the SIO model that they didn't bother to go back to slots.

For what its worth, Apple originally ignored the FCC bits since I think there was a loophole about it not being a consumer appliance or something, but regardless the rules were refined and better understood over the next few years. As well, if the designer/manufacturer was OK with having a large device, they could build that shielding into their big cases. I would suspect that Sinclair/Atari/Commodore didn't go that route simply because bigger means more expensive and they were trying to keep these things within the budgets of everyday people.

  • 1
    Upvoted for making a lot of good points! But while it is difficult to estimate the cost, I don't think it has to be an ineffable mystery. There is a fact of the matter regarding whether it was more like $100 or $1000, and it should be possible to find out or estimate, at least to that order of magnitude. – rwallace Jan 27 '17 at 21:00
  • 2
    The FCC loophole was to not provide the TV adapter thingy, and people had to buy it separately; don't remember why, but it made a difference in the device's classification – Thomas Jan 28 '17 at 17:57
  • 2
    @Thomas: Before personal computers came along, the main devices that used high frequency signals did so to interface with RF receivers (generally wirelessly, but sometimes via wires), and rules were set up with such devices in mind. According to Joe Decuir (a major designer of the Atari 2600) TI came up with an optical-fiber-connected television interface with the intention of saying that the main part of their computer wasn't designed to have any electrical connection with a TV, but the FCC didn't accept that notion. – supercat Jan 30 '17 at 16:31
  • 2
    @Thomas: Under the old rules, if a device that was designed to communicate with an RF receiver (even via wires), all of its emissions were subject to scrutiny, but since the FCC didn't expect other devices to use such frequencies there was no need to regulate them. New rules were written in the late 1970s which subjected almost all devices to scrutiny, but recognized that radiation limits which could be easily met with something like a VCR were harder to meet with digital logic, and thus allowed a higher level of emissions than had been previously allowed. – supercat Jan 30 '17 at 16:36
2

Like others have said, expansion ports are surprisingly expensive. When you look at the two examples you provided (Apple II and IBM), they weren't exactly the cheapest options for computers back in the day.

Even today, those 50 pin edge connectors can be $3+ each even in large quantities. I imagine (but cannot confirm) that Steve Jobs worked his magic back in the 70's and 80's and managed to get them for a bargain. But I'm only speculating.

However, the edge connector is hardly the entire cost of expansion ports. You have "glue" logic for device selection (such as 74LS138 multiplexers), buffers, shielding, etc. Plus, it is NOT a trivial task to design expansion buses in a machine. Especially for anything over a couple MHz.

The Apple II (and IBM, really) wouldn't have been a very exciting machines if it weren't for the expansion buses. But all of these ports came at a price. Which is why the more "budget friendly" computers such as the VIC-20, Commodore 64, Tandy CoCo, etc. used external edge connectors that had fewer pins. And, they typically only had one or two of them if any at all.

So it's impossible to say that expansion slots "cost 20% more" or "40% more". There are too many variables involved.

With the exception of the Apple II line, you only really ever saw them in droves on computers marketed for business (IBM, Amiga 2000/3000/4000, etc.). In fact, I would argue that the only reason you saw them on PC compatibles (like Compaq) is because they were trying to mimic the IBM completely. If IBM had created the 5150 with no expansion slots and used all serial/parallel connectors, the clone business may have been a completely different animal.

So, I think it's safe to say that expansion ports are expensive.

2

One problem adding substantial manufacturing cost and complexity when adding any socket taking edge connectors in modern hardware, especially consumer grade, is that it basically forces you to do a mixed SMT/THT design of your circuit board, and provide mechanical stability. While you could use SMT-only sockets nowadays, these are far more prone to get sheared or broken right off the board if it is manhandled.

Drilling thousands of small holes into a PCB is a cost factor (though you need some for vias anyway) - PCB laminates are among the worst materials regarding tool wear. Putting tall THT components on is either manual work or needs additional machinery. Plastic parts in such connectors need to be of either a very heat proof material due to the way PCBs with SMT parts on all sides are soldered. They're baked like a pizza, at similar temperatures to a non-premium pizza. Add in the trouble with a huge amount of metal nails being a very unwelcome thermal mass. Alternatively, a press fit connector could be used after soldering, but that is relatively recent technology.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.