COMMENT ⊗   VALID 00005 PAGES
C REC  PAGE   DESCRIPTION
C00001 00001
C00002 00002	FIXING AMPEX DR CARDS
C00004 00003		PARTS SCREENING
C00008 00004	THE +12 NEEDS TO BE SET TO 10 VOLTS TO STARVE THE 711 COMPARITOR FOR
C00011 00005		SYSTEM LEVEL ADJUSTMENTS
C00013 ENDMK
C⊗;
FIXING AMPEX DR CARDS

	WHAT GOES WRONG

BY ACTUAL COUNT, HALF OF THE CARDS REMOVED FROM OUR SYSTEM BECAUSE OF
BIT DROPPING/PICKING HAVE NOTHING WRONG WITH THEM.  THE PROBABLE CAUSE
OF THE ERROR IS IN THE CABLING, AND CHANGING THE CARD JUST CAUSED THE
CONNECTION TO HEAL.  PLASTIC EDGING MUST BE ADDED TO ALL THE SUPPES
CORE BOX EDGES TO LESSEN THE FORCES ON THE CABLES.

THE OTHER HALF OF THE TIME, SOMETHING AGES ON A DR CARD AND THE CARD
NO LONGER MATCHES THE SYSTEM MARGINS ON THE VARIOUS POWER SUPPLIES AND
TIMING SETTINGS, CAUSING AN INFREQUENT ERROR.  A HARD ERROR IS ALMOST
ALWAYS DUE TO A FRIED INHIBIT SWITCH MODULE ON THE DR CARD.

	HOW TO FIX IT

HALF THE TIME, THERE IS NOTHING TO FIX AFTER YOU HAVE JOSTLED THE CABLES.

IF AN INHIBIT SWITCH HAS SERIOUSLY DIED (NO FUNCTION AT ALL) THE 7440
WHICH DRIVES IT MUST BE REPLACED WHETHER OR NOT IT IS WORKING AT THE TIME.
ALMOST ANY NUMBER OF COMPONENTS CAN BE DESTROYED WHEN AN INHIBIT MODULE
GIVES UP, SINCE THE MAIN WAY OF DYING CONNECTS 26 VOLTS AT UNLIMITED
AMPS INTO SOME 12V AND TTL CIRCUITRY.
	PARTS SCREENING

ESSENTIALLY ALL INCOMING AMPEX INHIBIT SWITCH MODULES ARE GOOD.  IT IS
NOT ADVISABLE TO BOTHER TO TEST THEM INDIVIDUALLY, AND THEY HAVE LITTLE
ANALOG "PERSONALITY" TO TEST ANYWAY.

HOWEVER, THE COMPARITORS USED AS SENSE AMPS ARE TOTALLY DIFFERENT.
BUYING ONLY MILITARY GRADE, ONLY ABOUT 1/3 WILL BE ACCEPTABLE.  THEY
MUST BE TESTED BY PLACING THEM INTO THE ACTUAL CIRCUIT ON A DR CARD
WHICH HAS HAD A SOCKET INSTALLED ON IT, AND RUNNING THE CARD UNDER
PESSIMISTIC ACTUAL OPERATING CONDITIONS IN THE EXERCISER.

SENSE AMPS FAIL BY GOING TOTALLY BANANNAS FOR NO REASON AT ALL,
BY RESPONDING FEEBLY TO PERFECTLY FINE SIGNALS, OR BY HAVING SUCH
A TREMENDOUS OFFSET VOLTAGE THAT THEY RESPOND ALL THE TIME TO
NOTHING IN PARTICULAR OR HALFWAY TO GOOD PULSES.  MOST COMMONLY, BOTH
HALVES ARE THE SAME.

	DR TESTING IN THE EXERCISER

THERE ARE FOUR SECTIONS IN A DR CARD.  THE EXERCISER SELECTS A SECTION
(ONE AFTER ANOTHER) AND SIMULATES A SENSE AND INHIBIT CYCLE OF A CORE
MEMORY IN ABOUT A MICROSECOND. THE TWO OPERATIONS ARE SIMULTANEOUS AND
INTERFERE SOMEWHAT, GIVING AN OPPORTUNITY TO INFER ACTUAL SYSTEM NOISE
REJECTION ABILITY OF THE DR CARD BEING TESTED.

THE INHIBIT SWITCHES ARE EXERCISED IN ROTATION ELECTRONICALLY, IE
BY A 1 OF 4 DECODER. THE SENSE AMP SECTION BEING TESTED IS SELECTED
MANUALLY: THE INPUTS TO THE 3 UNTESTED SECTIONS BEING SHORTED TOGETHER
AND FLOATING DURING THIS TEST.


THE +5 V POWER MUST BE PRECISELY SET. THE -6 IS NOT CRITICAL, BUT SHOULD
BE CONSTANT AT LEAST DURING TESTING.  THE +12 IS VARIED DURING TESTING
AND MUST BE ACCURATELY READABLE.  IT MUST BE SET TO 10 V DURING INHIBIT
MODULE TESTING (TO PROVIDE MINIMUM DRIVE TO THE SWITCH TRANSISTOR TO TEST
IT FOR BETA UNDER LOAD) AND TO 12V TO TRY TO GET THE INHIBIT MODULE AS
HOT AS POSSIBLE TO TEST FOR HIGH-TEMPERATURE BETA PROBLEMS. HAVING A
LIGHT DIRECTLY AGAINST THE CARD HELPS TO BAKE OUT CERTAIN PROBLEMS.
THE +12 NEEDS TO BE SET TO 10 VOLTS TO STARVE THE 711 COMPARITOR FOR
GAIN, BUT THE COMPARITOR RESPONSE MUST ALSO BE NOTED AT THE NORMAL
+12 TO TEST FOR INSTABILITY, WHICH IS A SERIOUS PROBLEM WITH THESE
CARDS DUE TO POOR ON-CARD POWER SUPPLY BYPASSING.

TO TEST THE COMPARITORS FOR FUNCTION, A SIMULATED CORE SIGNAL IS
INTRODUCED INTO EACH HALF IN TURN.  THIS MAY BE PRODUCED BY DIFFERENTIATING
A PULSE OF VARIABLE WIDTH FROM A PULSE GENERATOR WHICH HAS BEEN TRIGGERED
BY THE CLOCK WHICH SELECTS ANOTHER DR QUADRANT TO TEST. THE POSITIVE
PULSE TESTS ONE SIDE OF THE COMPARITOR, AND THE NEGATIVE PULSE TESTS THE OTHER.

THE AMPLITUDE OF THE PULSE PAIR IS CAREFULLY CONTROLLED SO AS TO MARGINALLY
EXCITE THE SENSE AMPS, THUS LETTING YOU OBSERVE THEIR GAIN. IF YOU SUPPLY
TOO LARGE A SIGNAL THEY WILL ONLY RESPOND OVERWHELMINGLY AND TELL YOU NOTHING.
CHOOSING THIS AMPLITUDE IS A MATTER OF TRICKY JUDGEMENT. MAINTAING IT
CONSTANT IS A PAIN AND ABSOLUTELY NECESSARY.

THE COMPARITOR THRESHOLD (DERRIVED FROM THE +12 POWER) MUST BE SET BY HAND
TO SOME REASONABLE VOLTAGE, SUCH THAT WHEN A PULSE LIKE A CORE SWITCHING
IS SHOWN TO THE 711, IT RESPONDS LIKE A SENSE AMP SHOULD. A TYPICAL VOLTAGE
IS 9 VOLTS.

IT HAS BEEN FOUND ADEQUATE TO OPERATE THE INHIBIT POWER AT ONLY 12 VOLTS
(CONNECTED TO THE +12 POWER). THIS CIRCUIT NORMALLY FLOATS, SO THE RETURN
NEEDS TO BE CONNECTED TO GND.
	SYSTEM LEVEL ADJUSTMENTS

WHAT WE HAVE DONE IS TO FIX DR CARDS SO WELL THAT SHMOOING THE SYSTEM
IS UNNECESSARY, WHICH IS JUST AS WELL SINCE WE ARENT GOING TO DO IT.

JUST SET THE INDIVIDUAL CORE BOXES TO EXACTLY THE SAME MAJOR POWER SUPPLY
VOLTAGES GIVEN AS TYPICAL IN THE DOCUMENTATION (35 VOLTS, 26 VOLTS ETC)
AND FIX CARDS TO BE IDENTICAL WITH GOOD ONES REMOVED FROM THE MEMORIES.

IF THERE IS EVER ANY DOUBT OF HOW TO SET UP A DR CARD, MAKE IT THE
SAME AS ALL THE HUNDREDS OF THEM WE HAVE ALREADY. FIX THEM BY COMPARISON.
CHOOSE A TYPICAL CARD. REMOVE THE 711 CLOSEST TO PIN 1B AND INSTALL
A SOCKET FOR TESTING 711S. REMOVE ONE OF THE INHIBIT MODULES, AND INSTALL
A SOCKET FOR TESTING MODULES.

BEFORE CHANGING ANYTHING, EXAMINE ENOUGH CARDS WHICH ARE WORKING TO
KNOW WHAT IS NORMAL VARIATION OF COMPONENTS. THEY CANNOT BE MADE ALL
EXACTLY ALIKE IN A REASONABLE PERIOD OF TIME.