| Manufacturer:  | Casio |
| Model:  | AL-1000 |
| Year:  | 1968 ? |
| Form:  | Desktop |
| Functions:  | Basic four, square root, 4 memories, 30 step linear program capability |
| Number of Digits:  | 14 |
| Display Type:  | NIX |
| Display: | NEC LD866 | | Logic Technology:  | DIS |
| Memory Technology:  | CORE |
| Diodes:  | 1530 |
| Transistors:  | 421 |
| Tech. Data Level:  | 3 |
| Tech. Data Source:  | RE |
| Tech. Data Pages:  | 26 |
| Tech. Data: | list |
|
The unit shown in the photo is my reconstituted unit.
What I received was actually just the card cage and a chopped off keyboard.
The case and power supply were missing.
A replacement power supply was constructed and some plexiglass and aluminum used to hold it all together.
I was not inclined to try to recreate a full case, rather this unit serves as a display of the internals
of a discrete-component calculator from the 1960's.
This calculator model appears to be very sophisticated for its age,
with it's programmability and 4 memories.
However, it suffers from a very poorly designed user interface which severely limits it's
capabilities.
| Anomalies: |
- There is no overflow indication.
There is no indication of incorrect results of too-large multiplications and divisions.
- If a large number is added or subtracted with a number containing
many fractional digits, the result may be incorrect as the decimal point alignment
procedure may shift the large number too far up and the most significant digits lost (overflow).
The fractional number is not automatically shifted down to avoid the overflow.
- The M2 register is used internally by the divide procedure.
A number stored in M2 by the user will be lost (corrupted) if a divide is performed.
- Negative numbers are not handled properly in all cases for multiply and divide.
The resultant sign and/or value may be incorrect.
- To enter a fractional number (0<n<1), "0" must be pressed before pressing the decimal point.
- Taking the square root of a number <= 0.001 will produce an incorrect result (the resultant decimal point will be incorrect).
- Data in program storage and the 4 memory registers are retained when the unit is powered off.
Operand data are not.
|
Gotta love NIXIEs ..
Top view showing the boards.
The little blue boxes in a row on the board near the rear are pulse transformers for the core memory address lines.
Also observable is the 3D layering of the NIXIE numerals.
The core memory board removed for observation.
Closer view of the little core memory planar array.
4 bit-arrays, each 8 by 14.
The red-black twisted pairs are for the sense wires, the red-white twisted pairs for the inhibit wires,
and the singular black leads are for the matrix of address wires.
Note how each address wire goes through two bit-planes, then reverses and comes back through the other two bit-planes.
A view of one of the boards to give an idea of the component density.
This is board 4: in essence that's the sequencing logic for the function execution state machine in view.
Signal flow is primarily right-to-left on this board:
- Starting on the right are a couple of flip-flops (note transistor and capacitor pairs)
for the BET and GAM flags and some miscellaneous logic.
- Moving left, the four columns of blue diodes are AND gates.
These gates serve to decide the next state of the state machine.
- A column of load resistors for the AND gates.
- A column of diodes composing OR gates, fed by the AND gates.
- Columns of transistors and resistors making each OR gate into a NOR gate.
- The left-most column of blue diodes is a 1-of-32-to-5-bit-binary encoder to encode the selected state into a binary value.
A closer view of components on board 4.
Also visible is a typical source of problems in calculators of this era:
observe the only-partially-filled solder fillets around some of the feed-thru leads where they connect to the top foil.
A failure of one of these joints in my unit was one of the problems with it.
The new power supply.
The transformer was scavenged from an HP2912 instrumentation reed-scanner of the same vintage as the AL-1000
(it had a secondary for a NIXIE anode supply).
Modern integrated regulators were used rather than bothering to make up regulators from discrete components.
- Unit Log -
| Serial Number:  | (unknown) (boards and cage stamped with 7530) |
| Year of Manufacture:  | 1968 ? (some components stamped with 8F and such) |
| Date of Receipt:  | 21 Sep 2003 |
| Source:  | SPARC contact. |
| State upon Receipt:  | Missing case and power supply.
Wire harnesses from logic chassis to keyboard cut.
Digit 12 (3rd from left) Nixie tube broken. |
| Date:  | Sep 2003 |
| Procedure:  | Cleaned. |
| Date:  | 2 Mar 2004 |
| Procedure:  | Resistor moved slightly to avoid possibility of short around nG3 and G2 on board 6. |
| Date:  | 2 Mar 2004 |
| Procedure:  | Powered up with temporary power supply. Display shows all zeroes. |
| Date:  | 3 Mar 2004 |
| Procedure:  | Decimal point not displayed. PCB trace cracked near edge of board 1 where board
had been cracked. Resoldered with jumper. |
| Date:  | 10 Dec 2004 |
| Procedure:  | Reverse engineering completed including state graph and descriptions (level 3). |
| Date:  | 17 Feb 2005 |
| Procedure:  | Simulation complete. |
| Date:  | 18 Mar 2006 |
| Procedure:  | Construction of replacement power supply and reconnecting of keyboard complete.
With the unit powered up, pressing All-Clear leaves the display filled with '4's.
Pressing 1 results in a 5 in the display, 2->6, 3->7, 4->4, etc.
After 5-10 minutes of warm-up, the display can be cleared to 0 and functions
operate correctly except square root goes into a loop with the two LSDs spinning.
The program mode switch has some intermittent contacts. |
| Date:  | 20 Mar 2006 |
| Procedure:  | Broken NIXIE (3rd from left) replaced with Rodan GR112.
Close match but not identical to original type. |
| Date:  | 21 Mar 2006 |
| Symptoms:  | Square root results in a perpetual loop with the two LSDs spinning.
At times it will switch to working properly. |
| Analysis:  | State machine is looping through states 4 and 5.
State 4 should head to state 1, not 5.
Signal nNST4 is 0 during state 4: incorrect.
Traced to bad solder joint on board 4 on top side of a feed-thru stub
(not enough solder, partial crack in solder)
in connection from ST11 (pin 4A29) to diode 4B9. |
| Solution:  | Feed-thru stub resoldered. |
| Date:  | 26 Mar 2006 |
| Symptoms:  | Program mode switch is intermittent. |
| Solution:  | Contact cleaner and light oil applied. |
| Date:  | 31 Mar 2006 |
| Symptoms:  | In program execution mode only one instruction is executed for each press of the
execute (add) key, rather than executing the entire program. |
| Analysis:  | T state machine is not looping into the fetch state after completion
of the first instruction.
Traced to 2ST flip-flop not being set.
Adding a small capacitance (~100pF) to the base of the Q transistor of 2ST
results in proper execution.
Substituting transistors does not help. |
| Solution:  | 3 trigger diodes replaced, now works properly 80% of the time.
100pF capacitor to ground added at diode-cap junction of SE input. |
| Date:  | 22 Oct 2007 |
| Symptoms:  | On power-up or clear, display shows "4444444444.4444".
Entering numerals "123456789" display as "567456767".
Add after clear displays 0 but hangs, add after numeral results in "404040..." and hangs. |
| Analysis:  | 4 bit appears to be stuck on at some point in primary data loop.
Not the display decoder because 0 can be displayed.
Circuit analysis suggests "89" are translated to "67" due to sum correction
circuitry when 4 and 8 are on.
When display has "0"s in it, signal A4 does drop, but W4 does not.
Further isolated to inverter 7N6 (S4), transistor has open junctions. |
| Solution:  | Transistor 7N6 (C371) replaced with a C372. Unit functional. |