Sony SOBAX 2500 Calculator

Manufacturer: Sony
Model: SOBAX 2500
Year: 1970
Form: Desktop
Functions: Basic four, accumulator, 7 user memories, rounding, square root, programmable
Number of Digits: 15
Display Type: NIX
Display:Hitachi CD81
Logic Technology: SSI
Memory Technology: DELAY-MS
Diodes: 137
Transistors: 78
Principal ICs: Sony 500 Series
IC Complement:Sony 501 (138)
Sony 502 (35)
Sony 503 (23)
Sony 504-3 (35)
Sony 504-4 (15)
Sony 504-5 (4)
(250 total)
Tech. Data Level: 2
Tech. Data Source: RE
Tech. Data Pages: 24
Tech. Data:Schematic (pdf)

This calculator is based on a small IC family proprietary to Sony. Unfortunately these ICs have a propensity to fail (see the notes about the ICs), making it difficult to keep these calculators functional. I received two units and was originally able to restore both of them to functionality, but one has since been relegated to being a sacrificial unit for ICs.

Notable technically for the use of a magnetostrictive delay line in longitudinal mode, in contrast to the more usual torsion mode. Also, a channel slice in the delay-line is filled with a fixed pattern at power-up. The bitstream from this channel feeds a phase-comparator in a Phase-Locked-Loop arrangement with the master clock as VCO. The master clock is thus tweaked to keep it in sync with temperature variation of the delay-line latency.

Most of the logic is contained on six printed circuit boards which plug into a backplane. A small amount of logic is on the backplane PCB. The boards slide into the card cage at an angle to reduce the overall height.

The magnetostrictive delay line is in a metal box mounted to the rear-most logic board.

View underneath of the backplane. There are two versions of the backplane with minor differences.

The power supply at the rear is best accessed by removing the aluminum card cage above.

Board B, an example of one of the logic boards.

A repair involving two diodes had been made on this unit, to fix a bad gate input in an IC.

Board F, the main memory board, with the magnetostrictive delay line mounted on the rear.

Inside the magnetostrictive delay line housing.

The actual delay line is the coil of wire. The transducers are two tiny solenoids mounted with white silicon putty a little ways in from each end. The printed circuit board contains a two-transistor read pre-amp.

The delay line wire is 10 loops of ~ 11 cm diameter. With an additional ~ 16 cm around the ends, the total length of the line is:
         11cm*pi*10 + 16cm = 362 cm

The delay is ~ 1600 uS. The pulse velocity then is:
         362 cm / 1600 uS = 2.26 Km/S = 8140 Km/h

The delay line holds:
         16 registers * 16 digits/register * 4 bits/digit = 1024 bits

Thus each bit occupies about:
         3620 mm / 1024 bits = 3.5 mm
of length in the delay line.

Board extender for servicing, built from scrap bits.

- Unit Log -

Serial Number: 620173
Year of Manufacture: 1970 (transistors stamped with 0J, capacitor with 9D)
Other Identification: Acoustic Delay Line marked with "Type: 21.2, Lot No.: 107.2, Serial No.: 06594".
Board F (Master Clock and Acoustic Delay Line) marked with 1-539-22-12.
Front panel engraved with "UBC 29-2000 #7".
Date of Receipt: Sep 1999
Source: UBC SERF.
State upon Receipt: Display active but numeral entry does not work properly.
Current State: Board B need 504 IC and repair of decimal point issue.
Board E needs a 502 IC, a 504 IC and repair of M register issue (May 2014).
At this point, this unit (3) is a sacrifical unit for ICs for unit 620175 (5).

Date: Sep 1999
Procedure: Cleaned extensively, including rebuilding of keyboard with new rubber bumpers.

Date: Oct 1999
Procedure: Original power connector replaced with IEC standard. Power lamp replaced.

Date: Nov 1999
Symptoms: Entering a number and pressing an operation key would result in some numerals (1,2,3,5,8) changing to other numerals. Shifting the decimal point would sometimes reveal the correct numeral.
Analysis: Conjecture and an examination of the incorrect bit patterns suggested the zero-blanking procedure was affecting more than zeroes. Further conjecturing that the zero-blanking procedure involved a pass through the digits to decide where zero-blanking begins, followed by a pass to convert blanked zeroes (0000) to unblanked zeroes (1000), it was determined that the J counter was used to hold the digit position of the transition between unblanked and blanked zeroes and was indeed sometimes incorrect.

The problem was then traced back to a flip-flop in IC B105 which was supposed to be detecting whether a digit is zero (the DDISZ register). Examination of it's output (pins 9,10) suggested that the K input (pin 7) was stuck high, resulting in the flip-flop alternating instead of latching the data from register D. The J counter thus received improper indication of whether a digit was zero. Certain non-zero digits then had their W bit set in the second pass.

The signal on the K input was correct. An input stuck high would be consistent with an open input.

Solution: IC B105 was removed in anticipation of replacement. Upon reinstallation in a socket the problem was solved. Whether the problem resulted from a poor solder connection or a poor internal connection in the IC which was corrected by heat is unknown. (01 Dec 2000: Repair to Unit 620175 suggests a resolution to this unknown.)

Fully functional.

Date: 26 Nov 2004
Symptoms: Numerals not being displayed. '0's are present at powerup until initialisation completes, key presses result in processing activity, and the CHANGE SIGN and M keys function, but no numerals show up in the display.

Date: 19 Feb 2005
Symptoms: Zeroes appear at startup, but at end of power-on-reset display goes blank instead of showing "0.0...". CHS functions to turn on the negative lamp, and operations seem to be performed, but display remains blank.
Analysis: Swapping board F with board from other unit returns functionality. With original board, MSYNC pulses are present coming out of read amplifier but QF6/QF8 flip-flop does not toggle. TF10 is always low. PCB trace to IC F302-11 cut, QF6/QF8 now toggles. Capture input (F302-11) must be problem, holding line low. Measures as diode drop to ground.
Solution: Heating IC pin F302-11 restores functionality.

Fully functional.

Date: 21 Feb 2005
Symptoms: Some numerals do not display properly after an operation is performed. Pressing "1+" results in ".0000" instead of "1.000". Improper numerals: 1 to blank, 2 to ?, 3 to 5&0, 8 to 6. Problem goes away after some minutes of being powered on. Operations actually produce the correct value, just not displayed properly.
Analysis: Swapping board B with board from other unit returns functionality.
Solution: Not resolved, wait for the problem to become more permanent and hence easier to trace.

Date: 20 May 2006
Symptoms: Problem of 19 Feb 2005 has reappeared.

Date: May 2014
Symptoms: Improper numerals problem of 21 Feb 2005.
Analysis: Swapping with board (5)B restores basic operation.

Date: May 2014
Symptoms: Memory indicators 1, 4, 7 on, M-selection inoperative.
Analysis: Suspect logic around IC E115.

- Unit Log -

Serial Number: 620175
Year of Manufacture: 1970 (ICs stamped with 0F)
Other Identification: Acoustic Delay Line marked with "Type: 21.2, Lot No.: 102-1, Serial No.: 04586".
Board F (Master Clock and Acoustic Delay Line) marked with 1-539-22-11 (this board is slightly different than those marked 1-539-22-12).
Front panel engraved with "UBC 29-2000 #4".
Date of Receipt: 28 Nov 2000
Source: UBC SERF.
State upon Receipt: Display active but showing random flickering numerals.
Current State: Fully functional (May 2014).

Date: 29 Nov 2000
Procedure: Cleaned extensively, including rebuilding of keyboard with new rubber bumpers.

Date: 30 Nov 2000
Procedure: Original power connector replaced with IEC standard. Power lamp replaced.

Date: 30 Nov 2000
Procedure: High voltage (NIXIE supply) adjusted down from 220V to 200V. The higher voltage results in the voltage rating of some capacitors being exceeded.

Date: 30 Nov 2000
Symptoms: When first received the display flickered randomly between numerals but with the same numeral in all digits at any given instant. After cleaning and reassembly this problem seemed to have disappered. However it reoccurs infrequently (once per several hours of operation), coinciding with a clicking of the "READY relay". Powering off and on restores proper operation.
Analysis: The clicking of the relay suggests that the RDY signal is being lost temporarily. This might lead to a loss of sync between the master clock and the Acoustic Delay Line, perhaps as a result of improper data being injected into the sync channel of the delay line.

An attempt to force the problem by shorting the "ready delay" capacitor produces similar but not identical results.

Solution: See repair of 10 Feb 2001

Date: 01 Dec 2000
Symptoms: Thousands separators not being dimmed. The 2500 produces a thousands separator by turning on the appropriately positioned decimal point indicators, but dimmed relative to the proper decimal point. The dimming was not working: the thousands separators were in the correct position but of the same intensity as the proper decimal point.
Analysis: The dimming technique (reducing the duty cycle) had been identifed during the reverse engineering of Unit 620173. The symptoms suggested the duty cycle reduction was not occurring. The problem was traced to the Q counter capture-clock gate (IC B206, output pin 6). The signals on the inputs (pins 7,8,9) were correct but the output (pin 6) was not responding correctly. Specifically, it was not responding correctly to the signal on pin 8. Isolation of all inputs to this gate and resistance testing with positive test lead on +4V (pin 2) and negative lead on the inputs resulted in:
  • pin 7: "diode drop" continuity to pin 2
  • pin 8: open to pin 2
  • pin 9: "diode drop" continuity to pin 2
The open input would result in the input being treated as TRUE. The operation that would result from this logic state was consistent with the observed operation.
Solution: Heating of pin 8 for 10 to 20 seconds with a soldering iron resulted in the input achieving "diode drop" continuity to pin 2. This solved the problem. (02 Dec 2000: Temporarily as it turned out.)

This is essentially the same problem as occurred wth the K input to IC B105 on Unit 620173.

Date: 02 Dec 2000
Symptoms: Thousands separators were again not being dimmed.
Analysis: Apparently the IC had reverted to it's old ways. Reheating of pin 8 failed to bring back continuity.
Solution: A new input to the gate was created using two diodes to feed into one of the good inputs. Specifically: pin 7 (a good input) and pin 8 (the bad input) were rolled out of their holes in the PCB and the cathodes of two diodes inserted into these holes. The anodes were both connected to pin 7.

Logically, this is equivalent to feeding one input of the existing NAND gate with a new 2-input AND gate.

There was some question as to whether this technique would be successful as it adds an additional diode drop which must be overcome for switching to occur in a low logic supply (4V) environment. However, the same technique is used elsewhere in the machine (by design) to expand gates. It was successful.

Fully functional.

Date: 10 Feb 2001
Symptoms: Instability problem continued from 30 Nov 2000.
Analysis: 4V supply occasionally not present at startup.
Solution: Variable resistor for 4V supply tweaked to clean contact between wiper and resistance. Unit now seems to be stable: no problems over a 4 hour run.

Fully functional.

Date: 08 Jan 2003
Symptoms: After being off for several days or more, unit does not power up immediately. Will come on after some seconds or if power switch is flicked on/off.
Analysis: -10V supply not present at power-up, taking 4V supply down with it.
Solution: Slight crack in solder connections observed around pins of 2SD28 power transistors, resoldered. Stable over 14 hour run.

Fully functional.

Date: 22 Feb 2005
Symptoms: After power-on-reset, display flashes randomly.
Analysis: Speculation is that after being off for a long period of time the delay line needs to be magnetised via the capacitor discharge pulse that is sent through it at power-off.
Solution: Turning the unit off and on several times restores functionality.

Fully functional.

Date: May 2014
Symptoms: Display does not clear, several digits flickering between numerals while other digits are blanked.
Analysis: Swapping with board (3)A brings up a properly cleared display.

Date: May 2014
Symptoms: No numeral entry, although display flickers with keypress, indicating the keypress is being seen by the logic.
Analysis: Swapping with board (3)E restores basic functionality. Problem traced to N register clock logic, TE4 indicates stuck high. IC (5)E303 bad. All pins appear open to V+ & GND. IC (5)E303 replaced with 502 from (3)E303. OK.

Date: May 2014
Symptoms: Back to digits problem on board A.
Analysis: Scope obervation indicates K input at (5)A107.7 of digit counter is behaving as if it is stuck high, foreshortening the numeral cycle. IC (5)A107 replaced with 504 from (3)E110. OK.

Fully functional.

Date: Mar 2020
Symptoms: Function keys are not recognised. After entering a number and pressing a function key, digits for the 2nd number simply continue the first number. The press of the function key is seen by the logic in that the display flickers for processing.
Analysis: Swapping of boards B,C,D from (3) does not help. Problem likely on board E.

  Sony SOBAX 2500
Calculators | Integrated Circuits | Displays | Simulations