Becoming Analog Again
11 Nov. 2009
It's fall of 1980 and I'm working on the Osborne-1 design at my office in Berkeley. Pat McGuire is working alongside me, designing the electronics board for the floppy disc drive. Pat had been one of the engineers designing the ill-fated dynamic RAM board for Osborne Associates, a project I helped kill with a negative report on the construction technique. I needed another engineer to design the disc electronics and Adam recommended him.
Pat is a competent engineer with a “P.E.” credential after his name – someone who has passed the “Professional Engineer” exam given by the state. You actually can't call yourself an engineer in California unless you've either passed this exam or hold a bachelor's degree in engineering from an accredited college. Pat is shorter than me with wavy red hair, penetrating dark eyes, and a very clearly-articulated speaking style. He keeps careful, neat notes of all that he works on.
One of the foundations of the Osborne-1 project was that we could buy bare floppy disc drives with no electronics on them at a very good price. We have found three suppliers who will do this, and we need to design an electronics board that will mount on all three and work. Designing this board is Pat's task. He researches the drive electronics boards of all three companies and works out a design that will accommodate all the drives.
Floppy disc drives (FDDs for short) are pretty simple, consisting of a motor driving a hub, or spindle, a stepper motor that moves a read/write magnetic head, and a mechanical system for clamping the head down to the diskette – the mylar disc with the magnetic coating inside the envelope (these are 5 1/4” floppies, not the more recent 3 1/2” ones with a hard shell). The same mechanism clamps the diskette to the hub when the head is lowered. There are also a couple of photosensors that tell when the index hole passes a certain point in the rotation of the diskette.
The electronics board has to handle several functions – regulate current to the motor to keep the speed constant, send current pulses to the stepper motor to move the head, read the photosensor and both drive and read the read/write head. These components deal in analog signals, and the external interface driven from the cable is purely digital. The electronics board therefore performs conversion between the two types of signal.
Like all magnetic record/playback systems, the FDD “head” is a single coil wound around a ferrite (magnetic ceramic) core that is carefully formed to slide over the diskette surface with minimal wear. Unlike hard disk drives, the head does not float on a cushion of air but drags along surface that is both lubricated with stuff like graphite and is turning fairly slowly. I am familiar with the electronic characteristics of this kind of system from the magnetic tape designs I did at Ampex.
You write data by feeding signals to the head that are basically digital – the magnetic field produced drives the magnetic particles on the surface of the diskette into “saturation”, where they stay magnetized in one direction or the other. During readout the coil generates a small voltage when the boundaries between saturated areas passes under the ferrite core. This voltage is only a few millivolts – thousandths of a volt – and the electronics is operating at 5 volts, so an amplification factor in the hundreds is necessary. Fortunately, there are chips that perform this amplification well – the analog 733 type video amplifiers. Pat designs these in, providing the same test points present on all FDD electronics boards, allowing a technician to connect an oscilloscope and observe the strength and shape of the signals.
At one point Pat looks up from his work and says to me “You know, I'm really glad to have the opportunity to take cost into account when I'm designing.” I am surprised – it turns out that Pat's prior experience had been in the aerospace field where the sums of money charged were so high that the mere cost of the electronics mattered little – in fact, the contracts were often “cost-plus” where the higher the costs, the more the “plus” part amounts to. I've been sweating the costs from the very beginning of the project, and I somehow assumed Pat had been doing the same. However, his design work is very good and I've been reviewing it regularly, so there are no surprises here.
Some time in this process Adam brings two candidates for general manager to see me, separately. The first is well-mannered and seems intelligent, though I learn little about him. The second is striking – a short, stocky guy with grey hair named Tom Davidson who speaks loudly and profanely with a broad “dis 'n dat” Brooklyn street accent. His one-page resume has him graduating from the Wharton School of Economics at the University of Pennsylvania with an MBA, and shows several management jobs culminating with a turn-around of the Cermetek corporation, a manufacturer of electronic components.
Of these resume entries, only the last one is factual, as we are later to discover. Adam apparently falls in love with him, the first candidate never gets a call back (as he tells me decades later), and Tom becomes the General Manager – the most critical position in the company, as Adam does not pretend to know how to manage the operations of a startup manufacturing company. Tom is the son of a New York City policeman, was trained as a machinist, served in the Korean war where he was captured and spent several years as a P.O.W., and generally seems the polar opposite of Adam, the British PhD who speaks in measured cadences and projects an aura of calm all-knowingness.
In the meantime, Adam has found a mechanical designer named Housh Ghorbani – a recent Iranian emigre working from a drafting board in his living room. Housh lays out the shape of the case needed to contain all the Osborne-1's components and makes the mechanical drawings necessary for a model-maker to build a single example. In consultation with me he works out a kind of gull-winged platform on which the CRT display and the FDDs will be mounted, with the circuit board suspended from brackets below. The platform will be screwed into the case and will flex under shock to protect the CRT from breakage. This mechanical design will move from one home to another, through three construction techniques, but it survives through the entire life span of the computer.
Things are, of course, pretty closely positioned in this design. The 5-inch CRT monitor sits in the center flanked by the FDDs, and has to be open to air circulation for cooling. Inside the monitor is the high-voltage generating transformer, known as the “flyback”, which creates sharp pulses at 7.5 kilovolts - that's 7,500 volts. The sharpness of the pulses means that echoes of them will appear almost everywhere else in the circuitry, due to physical phenomena that cannot be avoided. The FDDS on either side are susceptible to these pulses, especially to the pulsing magnetic field generated by a flyback with no pretense of shielding. To make matters worse, the ribbon cable connecting the FDDs to the main circuit board drape right over the monitor.
After the product is introduced we notice that its performance is erratic – characters randomly appear on the display screen where none should be. This is embarrassing when we show at the National Computer Conference in Chicago in 1981, though no visitors to the booth point it out. Tom is puzzled - “How kin dat be?” he keeps asking, arguing that there's no direct, line-of-sight path from the monitor to the FDDs – we had put a U-shaped piece of aluminum over the drives as a cover.
I try to tell him that you wouldn't need a line of sight path to get interference – that you could stretch a clothesline between two rooms running it through a tiny crack in the doorway, then wobble one side and the other side would wobble even though the crack through which it passed would not permit waves of such amplitude to pass. My electromagnetic analogies fall on deaf ears. Tom's solution at the NCC is to dismantle some of the demo units and poke around inside them aimlessly.
Back on the workbench I attack the problem. With the case off, I connect the probes of an oscilloscope to the test points on the FDD electronics board – two of them are needed because this is a ”differential signal” that manifests as the difference between the two voltages. Then I take another probe and connect it to the “trigger” input of the oscilloscope. When enough voltage is sensed by this probe the scope will fire off a single sweep of its electron beam, resulting in a visible trace on the screen.
I take the trigger probe and simply suspend it near the monitor. The 7500 volt pulses running at 15,750 pulses per second couple easily to the probe through thin air resulting in nice, strong trigger signals. The scope triggers reliably, and I see a definite pulse show up on the FDD read heads caused by the monitor flyback. What to do to get rid of it?
I take some aluminum foil and cover the rear edge of the FDD cover. There is not much response in reducing the unwanted pulse. I think a bit – aluminum is known to react with air to develop a film of aluminum oxide – the same stuff sapphire is made of and an electrical insulator – on its surface. Maybe if I pressed hard I could break this oxide film, which is only a few atoms thick.
I press on the foil – and see the noise pulse decrease. I've got it – we'll need to change the covers from simple, open-ended U-shapes to shapes with a flap bent down at the rear. This flap will have to be part of the aluminum forming the rest of the cover – made from one piece of metal. I make a sketch and take it to Tom to be passed to the metal fabrication shop.
However, Tom is a very good negotiator, and when the shop points out the cost, he negotiates his way around it. Since he was a metalworker in his youth, Tom apparently feels no need to involve me in the discussions, even though I had told him that the continuity of metal was essential. Besides, I would only reduce his room to maneuver. I am presented with a large number of covers made by adding a separate U-shaped strap across the rear of the original U-shaped covers. Worst of all, these straps are secured by a mere rivet on each side! They have an inviting gap along the rear corner which should have been a continuous metal bend according to my design. I protest but some of these seem to find their way into production units before we got the design right.
Yes, Tom is a great negotiator – aggressive, shrewd, and somehow able to get supplier companies to give low prices without receiving anything in writing – he boasts that he doesn't use purchase orders. He calls it “beatin' up suppliers” and it earns him Adam's great admiration. However, it won't help him in one case. The idea Adam had that he could buy FDDs and simply connect our electronics to them without any adjustment turns out to be very, very wrong. They will need to have the full course of adjustments made to them – adjustments that reconcile the particular circuit board to the particular mechanism.
Workbenches must be set up for the drive-adjustment technicians, workbenches that claim floor space originally intended for white-collar personnel (Tom had laid out a very small space for manufacturing). The planning spread sheets will have to show additional costs – you can't negotiate that away.