Friday, April 27, 2018

Seemingly possessed Diablo disk drive now working


As readers may remember, we have been battling this drive which reads just fine but has problems writing correctly. If we take a cartridge written on other drives, this one will read it perfectly. If this drive writes anything, it is very likely to have errors reading back on this or any other drive.

The problems occur with both top and bottom heads. The oscilloscope showed that the previous contents of the sector were not fully erased. We swapped PCBs with other drives but the problem remained.

Today we decided to swap out the heads on the drive with known good heads from another drive, just to eliminate the heads as a cause of these symptoms. Even with all PCBs and the heads from a good drive swapped into this unit, the problems continued. 

There was not much left that could be wrong - a cable from a PCB to the disk head connectors, the backplane carrying signals between PCBs, and the electromechanical drive and arm mechanisms. That is, if this is not a case of possession by malevolent spirits. 

We ruled out the backplane after checking that all the signals required for writing were reaching the PCB that connects to the heads. We ruled out the cable and connectors to the heads by using a Time Domain Reflectometer and also measuring the resistance of each line. The TDR would show us shorts, open wires or other anomalies that cause reflected signals which may impact the quality of recording. 

At this point, Marc began to closely examine the mechanism that loads the heads down onto the rotating disk surface. This is a pair of brass prongs that pivot to force the two heads together, pinching them against the disk surface as it rotates. The back of the disk head has a raised metal ball that sits under the end of the brass prong, through which the prong exerts the diskward pressure. 

Marc could see the ball because the prong did not extend over the ball. The prong was about 3/32" shy of the ball. Looking at our other drives, we saw that the prong sits right over the ball on correctly operating drives.
Prong does not cover small metal ball on disk head
We looked at the bracket holding the prongs and saw they were bent out a bit from the machined metal structure that holds the heads. Using pliers, Marc bent the bracket to compensate, putting the prongs back over the metal balls on the disk heads. 

Adjusting bracket holding prongs

The failure mode was failure of the prongs to press the heads down against the disk surface as fully as on a correctly working drive. Just as seriously, the point where pressure was applied is now offset compared to the aerodynamic center of the head. 

This would cause the head to fly at an angle, not parallel to the disk surface, and have reduced pressure to force them down. The spacing from head magnetic poles to the disk surface was larger and uneven.

Prong properly positioned over metal ball
With the prongs adjusted, we fired up the drive and tested again. First I wrote a known image on the cartridge using my FPGA based disk tool, then did a read of the cartridge contents back to the tool. Zero checksum errors and a perfect match to the image being written!

We swapped back all the PCBs to the ones from this drive, moved the drive to the Alto and booted up to ensure it was working well. Problem solved! All that remains is to swap the heads back between our drive and this one, run an alignment, and do a final set of tests. 

We believe this problem is related to the bad disk crash we experienced early on in the repair of this drive. It was shipped to us by the owner, who followed our recommendation to secure the disk arm from movement by using a cable tie. It is likely that the cable tie was tightened much too much, bending the bracket outward and shifting the prongs out of position.

Very slight gap of bracket against metal of disk head holder
Our early crash, after swapping out the original bad heads for a replacement set, was associated with bent prongs after the loud noise and scraping sound. We have to assume that the bending during shipping caused both the early bad crash and the continuing write errors. 

The misalignment was very small and subtle. Worse, it is not a position measurement that is called out for checking in the maintenance manual, so we didn't spot it in all these months. At least it is corrected now and the drive can go home to support its Alto system. 

No comments:

Post a Comment