Testing my new seek logic for the Virtual 2315 Cartridge Facility - part 8

FINISHED MODIFYING TERMINATOR BOARD WITH NEW RESISTOR VALUES

The terminator board that fits into the Virtual 2315 Cartridge Facility (V2315CF) main unit is the same one used successfully with the unit in its original role as George Wiley's RK-05 Emulator connecting to a DEC computer to emulate a disk drive that is a derivative of the IBM drive. The terminator implemented the same resistance values as the M903 terminator board built by DEC. 

The impedance of the terminator connections is just over 122 ohms, which may match the impedance of the cabling in the DEC computers. The RK-05 emulator kit offers an option A1 RK-05 to RK8-E that has two IDC 40 pin ribbon cables to make the connections to the controller card, since the RK-05 drive needs 36 signal lines for the interface between disk and controller, deploying ground lines between each signal line on the ribbon cables. 

Ribbon cable characteristic impedance for single ended signals (not differential pairs) is 80 to 100 ohms with interspersed ground wires. Using the DEC terminator board yields a big mismatch in impedance, which will lead to reflections of about 20% of the signal level. If the rise time of signals is faster than 6 nanoseconds the signal will ring and appear to have spurious transitions. 

I determined resistor values to drop the terminator impedance to about 100 ohms, which would minimize the issues even with the current sharp rise times. The Lattice FPGA does not support the means to slow the signal transitions to 7ns or slower instead typically rising at 1 to 1.5 ns.

I removed all the original resistors - a mix of 1206 size and 0603 size for the pullup to 5V and the pulldown to ground respectively - and cleaned up the board.

The original terminator board has a few unpopulated resistor locations - the crude circles show you were to concentrate to see them. 

I installed my 179 ohm pullup and 249 ohm pulldown resistors onto all the positions where I have signals connected to the ribbon cables. The new terminator board was tested and then plugged into the V2315CF to do more testing. 

DETAILED VIEW OF EFFECT OF SIGNAL GLITCHES ON SEEK BEHAVIOR

A glitch on the +Access Ready signal causes the 1130 disk controller logic to malfunction, thus not forming correct step requests to the drive. The disk drive responds to brief glitches in -Access Go that are induced by the +Access Ready glitch, so it does move to the requested cylinder. The V2315CF fails to recognize the malformed sequences thus it does not reach the same cylinder number as either the drive or the intent of the XIO Seek commands. 

The dance of signals we should see is the controller drop -Access Go with -10 Mil Step low, holding it until the drive drops +Access Ready about 5 milliseconds later. When +Access Ready drops, the signal -10 Mil Step will rise from low to high since only the first seek should be a 10 mil step. 

Also at the time when +Access Ready drops, if the count register (Bits 7 to 14 plus the -10 Mil Step which is bit 15) has all bits on, the -Full Word Count signal is dropped which terminates the seek in the 1130 disk controller logic.

The disk drive, meanwhile, moves the arm for another 9-10 milliseconds then raises +Access Ready to indicate the seek step is complete. If the count register is not all ones yet, another cycle of -Access Go dropping with its +Access Ready responses will occur, iterating until -Full Word Count drops. 

Above, we will examine a successful seek of 1 cylinder. We can see the signals conforming exactly to the protocol. After -Access Go is dropped with -10 Mil Step low, we see +Access Ready drop 5 milliseconds later. At that time, the controller changes -10 Mil Step to high, which turns on -Full Word Count, stopping the controller from trying any more steps. At 15 ms from the start of the step, +Access Ready returns to high indicating the move is complete. 

When I did a move of 3 cylinders, however, the glitches arrived and the process breaks down. The arm moved three cylinders but the V2315CF only saw a single cylinder step. Lets look in detail below, zooming in to the time when the glitch first arrives:

At the 5 ms point, when +Access Ready should drop for 10 ms, we see a very short blip downward, circled in red. This turns off the -Access Go and makes -10 Mil Step go high as pointed by the green arrow.

We see another -Access Go, but very brief, indicated by the blue arrow. Since -10 Mil Step is now high, this asks for a 20 mil step. The short drop of +Access Ready just to the left of the blue arrow causes the disk controller to add 2 to the count register, making it reach the all-ones state and turning on -Full Word Count pointed to by the pink arrow. The seek is over so no further step requests. 

The drive did respond to the additional -Access Go, in spite of the prior seek not being finished. It is only a 20 nanosecond pulse, borderline for the Solid Logic Technology (SLT) edge detectors to see but obviously long enough. Somehow the drive managed to do the second step correctly, as the arm did end at the proper spot. 

Since the V2315CF logic is generating the correct protocol dance - keeping +Access Ready low for 10 ms starting 5ms after the falling edge of -Access Go. It doesn't watch for the next -Access Go falling edge until after the 5ms and it misses the defective short mistimed pulse that the disk drive responded to. 

The drive even responded to a 5 cylinder seek with a correct movement. The signal pathology is even uglier, as seen below:

We see that a single step is observed as far as the V2315CF, based on the 15 ms dance expected. Zooming in on the glitch area, we can see how it managed to move 10 mil, 20 mil and 20 mil to achieve the 5 cylinder goal. 

The glitchy +Access Ready is see to the left of the green arrow, resulting in the -10 Mil Step signal going high where the green arrow points. The first step we began is 1 track, but the subsequent moves will be 2 tracks each. A glitch from the controller shown with the first red arrow drives the arm another 2 tracks. Later we see the -Access Go drop for a longer interval, driving the last 2 track move. This adds 2 to the count register, completing the count of 5 and turning on -Full Word Count to complete the XIO Seek.