The case of the consequent pole

-March 20, 2013

I was employed by a company that produced disk and drum memory systems, back in the era when disk drives were typically the size of washing machines and people carried disk packs around containing their projects. This company's first drive would fit on a desktop, but required two people to safely carry it. There were two 14-in. disks—one "fixed" (permanently installed) and one in a removable cartridge, driven by a line-powered induction motor with capacitor start and dynamic braking. For track following, the head position relative to the data tracks was derived from prerecorded "sector servo" tracks in the gaps between data sectors. Before stopping the spindle, the drive automatically retracted the heads, which were lifted off the disk surfaces by ramps on the long cantilever springs as the heads were withdrawn. The head-positioning servo was a linear "voice coil" motor with a long magnet and a short coil. Inside the center pole of the linear motor (LIMO) was the coil of a velocity transducer. A rod magnet attached to the carriage extended back into this coil, providing velocity feedback.

Since most of the sales of this product were in the UK, most of the drives were manufactured by our subsidiary in Surrey, England. Being the servo guru, I was sent to Surrey to resolve servo problems. While I was there, a rather memorable problem appeared in a few newly assembled units. The servo worked fine over most of the position range, but "unloading" (retracting the heads) and "loading" were anything but smooth operations.

To eliminate the head-retract circuitry from the test, I left the servo in track-following mode and tried forcibly pushing the carriage in (toward the LIMO) far enough to unload the heads. The position loop opened as servo playback was lost, leaving only the velocity loop. A little ways past the unloading ramp (heads off the disk), the damping force decreased sharply. Before I could react to the loss of resistance, the carriage abruptly shot to the "full retract" position, bounced hard off the stop (oblivious to the parking detent), and zinged forward again with the same vigor. It stopped very nicely just past the spot where it had gotten away from me. (Glad I'd kept my fingers out of harm's way.) I don't remember the force constant, but the peak current during a long seek was well over 5A at about 25V.

Dusting off my wits, I opened the LIMO current path via a service jumper and tried this again while watching a pair of LEDs on the servo amplifier, which indicated drive direction. Normally, with the jumper removed and when moving the heads on the disk, the LEDs would toggle rapidly as track groups were crossed, but with very rapid movement or no position signal, the velocity feedback would cause the LEDs to always indicate a control effort counter to the motion. With the heads unloaded from the disk, there was no servo position signal and any dc offset would be evident.

On this unit, the LEDs indicated “wrong drive direction” when I moved the carriage either way with the heads unloaded, but “correct drive direction” with rapid movement on the disk. This could result only from a reversal of the velocity signal, which a scope check quickly confirmed. The system had positive velocity feedback, in the "unloaded" region only. The coil physically couldn't be installed improperly, so assembly error was unlikely.

"Coil shorted to ground?" I pondered. No. One end of the coil was grounded already, so that would just reduce the signal level (albeit not uniformly) by effectively eliminating part of the winding. I checked it anyway. Not shorted. Either there was something awfully strange about the winding of the velocity transducer coil, or the rod magnet's poles were reversed—but only near the mounting end of the rod!

It occurred to me that there may be coils with two windings for a magnet with both poles inside the coil. "Wrong coil?" I didn't think so; the reversal should then be near the midpoint of travel. The magnet was a cylindrical Alnico rod about 1/8 in. in diameter by about 7 in. long to accommodate the extra movement required for retracting the heads. The Brits called it the "tacho rod," and the term stuck. Replacing it with one from another lot corrected the problem.

Now I had another problem: explaining what was wrong with the magnet. If we were going to reject magnets, we'd have to explain what was wrong with them. We also needed a way to test them before assembly into a disk drive. My hypothesis that the magnet was reversed near one end was dismissed as absurd.

I asked whether there was a small compass available. There wasn't. In desperation, I asked where I might find some iron filings. We were going to do that high-school science experiment on these magnets and see what the field pattern really looked like. The crew decided to humor me and cooperate. We'd put this to rest one way or another. I expected to be directed to a vice somewhere and invited to collect filings, but such an untidy workspace was not to be found. My request was supplied quite promptly, though, by one of the technicians.

There were quite a few witnesses to the experiment, perhaps amused by my low-tech methodology, but I think most of them were anticipating my embarrassment. I set the rod on a wooden desktop and laid a piece of paper over it. Then I sprinkled the filings over the paper, and voila! There was a short loop from the mounting end to a point about an inch from it, and a long loop from that point to the tip. Jaws dropped, eyes popped, and incredulity gave way to astonishment as I pointed out the anomaly. There were no cracks or other visible defects on the magnets, and none of us could explain a rod magnet having three poles except by south being at both ends and north being in between them, or vice versa. A handful of magnets were subjected to this test so all could see the difference between a good one and a bad one.

The supplier was called, and we were told this is called a "consequent pole" and can result if the magnet is struck by another magnet during shipping. These magnets were pretty well packaged and we couldn't see any way that could have happened, but clearly the supplier was responsible for the remedy, and I never found out how that was accomplished. Suffice it to say, it was.

When the heads were on the disk, the carriage to which the tacho rod was attached was extended, and the consequent pole was outside the tacho coil. Only the flux from the tip (S) induced voltage. This was identical to operation with a good magnet, even though the external flux terminated at the consequent pole. As the heads were being retracted, the consequent (N) pole entered the coil. This flux loop then passed through the coil in both directions (radially outwards and radially inwards), linking a constant length of the uniformly wound coil as the magnet's inward motion continued. Now, it induced no net voltage. The flux from the consequent (N) pole to the mounting-end (S) pole (always outside the coil and should have been N), however, did induce voltage, which was inverted because north was inside the coil. Testing was simple: Just observe the induced voltage while moving the magnet inside a tacho coil.

Dick Neubert has a long and diverse history in electronics and (mostly real-time) programming, ranging from high-performance disk head servo systems to computerized automation systems for sawmills.

Loading comments...

Write a Comment

To comment please Log In