2668-B PC

THE LATHE, aka: LS-76, and the LJ-10 and LJ-12 tape machines
Post Reply
User avatar
Site Admin
Posts: 100
Joined: 03 Apr 2013, 03:13

2668-B PC

Post by Fonotec » 08 Sep 2016, 03:30

The LS-76 Display Processor circuit board uses two, 8 bit, parallel output A/D converters to generate the numerical representation of the feed. The 7-segment led driver chips on the LPI Display card in the Control Panel take the data-selected, 8 bit, parallel signals from the Display Processor board and convert them to display patterns that we recognize as "digital" numbers (from 100 to 600, in 5-LPI increments), or make a blank display (when Feed is off). When illuminated, the three, red-digit leds are showing the carriage pitch, in "LINES/INCH."

One A/D is for generating the bits used to create the display code for the base pitch. It generates these bits by digitizing a copy of the base pitch signal that is going to one of the Feed Servo board's uA748 op amps (trimmed by the "base pitch" selector pot on the Control Panel). The other A/D's outputs are switched on by logic whenever the feed expands. It digitizes the expansion signal, so those bits help generate the display code for when the carriage moves faster than base pitch. (The expansion signal is obtained from the summed output of the base pitch op amp and two other (uA748) op amps on the Feed Servo board - one, for manually-selected expansion current, and another, for automated variable pitch expansion pulses.) By having one A/D converter dedicated to digitizing only the base pitch, the present setting for that can be referenced at any time that Feed is on, by flipping a switch, even when expansion is taking place.)
400_LPI_detail.jpg (35.21 KiB) Viewed 2877 times
I like to keep inventory of NOS discontinued parts for the vintage studio gear so that I can minimize down-time when service is wanted. Since I have only been able to find broken examples of ADC-89A8B from the obsolete market, I decided to depot the module (made by Datel Systems, Inc., who also made the ADC-Econoveter and DAC-98BI modules, used in THE LATHE), so that I could clone some working models for inventory.

Here's how it looks before being depotted:
ADC-89A8B_potted.jpg (5.3 KiB) Viewed 2886 times
Here's the PCB (2668-B PC) that was found inside:
2668-B_back.JPG (50.21 KiB) Viewed 2884 times
The depotted circuit can be viewed in further detail on two examples of 2668-B PC, which I reclaimed from their diallyl phthalate casings, using a Dremel rotary cutter.

http://www.andrewhamiltonmastering.com/ ... B_A&B.html

As you can see from the depotted examples, removal of the plastic casing with the rotary saw leaves the board mostly intact, however, some small sections of track near the edges often get pulled off by the high speed blade, since the plastic casing is in intimate contact with the circuit board, leaving no gap between the inside of the casing and the surfaces of the PCB. Therefore, I opted for PCB cloning, rather than repair, as the simplest way to end up with perfect modules. Wherever a small section of track or a through hole pad is missing, the cloning engineer can correct that easily with the digital imaging software used for generating the drill map and track layout, à la Photoshop. For starters, a "tan line" remains where the trace used to be, and by depotting two examples, I ended up with an unambiguous roadmap for a complete and working board, since the slight damage to each board was random, and, therefore, unique - so, the uncertain areas of the first board were explained by the same areas that were undisturbed on the second board.

After getting Gerber files made from these boards, down in Richardson, Texas (at Micro Technology Services), I hired Sunstone Circuits (in Mulino, Oregon) to manufacture some PCB clones with the solder mask option. I recently packed this first board and tested it in a custom jig (shown here: ADC-89A8B/Test_Jig_ADC-89A8B.jpg), made from a Radio Shack, Copper-plated through hole breadboard and a tomato stake (as legs) - both of which were sawed with a Black & Decker jigsaw.

The board takes two 7493 (4 bit binary counter) DIP ICs, one 7474 (D-type flip-flop), and one 7400 (NAND gate), which I put in the four, screw machine sockets. I'm glad I socketed the chips, since the two, 4 bit, binary counters that I tried first would pass the logic checker test but would not work correctly in this circuit. The gold legs were of no benefit. When I tried other 7493s, the circuit turned on and made the right patterns on the oscilloscope for all parallel bit outputs.

(The analog test signal I used was from the Neumann console oscillator. Until I changed the ground connection for that oscillator's return to 2668-B PC's Power Ground pin (rather than its Analog Ground), whenever the sine wave entering the Analog input was either muted by turning the oscillator off, with the switch, or with the oscillator's gain pot turned, fully CCW, the LSB pin would emit a heavily-attenuated copy of the Krohn-Hite oscillator's square wave (4 kHz), used for this test as the A/D's clock signal, which was (still) going into the SC pin. I subsequently made a clock board, which has only three legs, so I called it 'Cricket', as in a 'cricket table', which has only three legs, and because it basically keeps chirping when it's turned on...ADC-89A8B/TTL_clock.jpg)

Out of curiosity, to see if I could reconstruct the digitized signals, I then made a test jig for the 8 bit, parallel input D/A converter, DAC-98BI, which the 6 bit, parallel output A/D converter, ADC-Econoverter (see "Depotting," in this site's Exotics section: viewtopic.php?f=15&t=124), signals to on the Variable Pitch and Variable Depth boards. In those circuits, since ADC-Econoverter only sends 6 parallel bits to DAC-98BI, by way of six, 5 bit shift registers (7496), the D/A's two least significant bits are grounded. (Green circle in drawing, below, indicates grounded input bit pins.)

{By digitizing the control audio (on the Variable Pitch and Variable Depth cards) and sending the bits through shift registers and immediately reconstructing them (as analog) in unipolar mode, the lathe automation ends up with scaled pulses of control bursts for the feed motor current that are synchronized by a turntable velocity compensation clock for the feed, on board the Variable Pitch card. Switching the turntable speed selector switch to a given turntable angular velocity's step changes not only the reference sub-frequency modulus logic for the TT Servo Drive board's 432 kHz crystal oscillator, but also the reference frequency tuning network resistors for the 555 feed timer on the Variable Pitch card. Since the automation is designed to store peak audio control pulses 25 times per revolution (and act on those stored values 5 times per revolution), the 555 feed timer is clocked at 13.888... Hz for 33+1/3 rpm, since 13.888... x 1.8 seconds (the time required for one revolution at 33+1/3 rpm) = 25. For 45 rpm, the 555 feed timer is automatically re-clocked to ~ 19.23 Hz, using the same 4.7 uF tantalum capacitor, across the threshold, discharge, and trigger pins, but with different resistors selected in parallel. A trimpot in the network allows fine-tuning of the time constant. Once it's set for 33+1/3 rpm, ideally, it's already trimmed for the other selectable platter speed compensations, since they, all, share the same trimpot (setting). This can be fined-tuned, and must be, for ideal operation (See: viewtopic.php?f=3&t=219). The timer's output is connected to the Clear inputs (pin 16) of the six, 5 bit shift registers that are in between the A/D and the D/A. A one-shot (9602) synchronizes the Start Convert pins of the A/D converters on both automation boards.}
2_LSB's_Earthed.jpg (87.13 KiB) Viewed 2837 times

When the newly-cloned 8 bit ADC and an original 8 bit DAC were interconnected between the two test jigs (directly), I was able to confirm on an oscilloscope that the (40, 200, and 1,000 cps sine wave) test tones which I digitized, emerged, as analog, again, in unipolar mode, from the DAC's analog output.


Picture, below, shows the first clone PCB (2668-B PC) which I packed with mostly new, and some NOS, components - shown without any DIP chips in the screw machine sockets or gold pins in the pin holes. Identical to the originals, except that the clones can be serviced, since they aren't potted.


User avatar
Site Admin
Posts: 100
Joined: 03 Apr 2013, 03:13

Re: 2668-B PC

Post by Fonotec » 13 Oct 2016, 18:59

Today, I finally finished packing two more DATEL ADC-89A8B clones, after having to wait many weeks for some 0.1% tolerance, 1/4 Watt, film resistors of the right values for the 8 bit, parallel output, A/D converter's circuit board (2668-B PC). I just tested them in the homemade test jig and saw that all 8, bit-output pins are working on both of the new clones. (Note: As the reply, below, explains, the bit-pins can be firing and not be valid unless the comparator is also properly operating and has not lost its output signal.)

These two boards will now join the first two (tested and confirmed) clones in the 100-year old Mosler wall safe that's in the studio repair shop. I hope I never have to use these, of course, but it's nice to know that there is hope, against the day...
Clones_3&4.JPG (139.14 KiB) Viewed 2771 times

User avatar
Site Admin
Posts: 100
Joined: 03 Apr 2013, 03:13

Re: 2668-B PC ("solder bridge jumper"!)

Post by Fonotec » 16 Jun 2022, 00:08

I normally only cut with Base Pitch settings above 250 LPI, so, for the past six years, I didn't realize that my cloned converters (which I've not yet had to use, since my original, potted modules are still working) would not convert signals over +5V, meaning, I would only get the first 7 out of the 8, total, bits of precision that are available from the ADC-89A8B (the last two characters identifying its '8[-bit] B[inary]' capability). This is because, although I could see all, 8 bits sending out a pulse train when I tested them in the test jig, the comparator (LM310CH) would turn off when the input signal was just over +5V, rendering the output from the still-firing bit-pins, invalid... I had no idea this was the case until someone asked if one of the clones could be purchased, so I did more testing and discovered this malfunction that only affected (unipolar) input signals over +5VDC.

Depotting my last, working original module with the Dremel rotary cutter revealed that all of the componenet interconnections on the clones were correct, however, when the original, depotted board was depopulated of its components by desoldering them, the contractor inevitably also removed a small, but crucial, 'solder bridge jumper' that spans a deliberate gap that's in the circuit trace... The bridging of the gap allows a transistor to turn on and provide additional current that's needed for making the 8 parallel output pins operate when input signals over +5VDC are sent to the converter.
Traces_Gb.jpg (336.33 KiB) Viewed 378 times

Here's an old picture of the solder side of one of the original depotted modules before I sent it to Texas, showing the necessary 'solder bridge':
PC-2668-8_rear.jpg (211.84 KiB) Viewed 378 times
Adding a tinned jumper made of 24 AWG solid-core silver wire across this gap turned on the extra current capability of the circuit, and now my clones can convert signals up to just over +10 VDC, which is needed when the Base Pitch is set to lower than 200 LPI.

Here's an article I found online about this practice: https://resources.altium.com/p/solder-b ... pcb-design

It explains that leaving a gap for a 'solder bridge jumper' in the circuit trace allows the same PCB to be used for different versions of the circuit. I suspect that the same PCB for ADC-89A8B is also used for the "Two-Digit BCD" version called, ADC-89A8D (mentioned in the datasheet: ADC-89A8B/ADC-89.pdf ). It outputs the signal in a form of 'binary coded decimal' code, while the standard version of the converter outputs the signal as a standard binary code. By adding a solder bridge jumper across one of the two, additional, deliberately-made gaps in the traces (or both, so, all three?), it would complete the slightly different circuit path needed for that alternate model.

(By the way, while it might be true that potting a circuit hides the design from manufacturing competitors, it's known that digital signal converters operate best when the temperature is very stable. The Lavry Gold DAC has an 'oven' for elevating the temperature of the components because it tends to make them all be at the same temperature. Similarly, although without any logical feedback from a thermal servo, potting an A/D or D/A converter and filling it with thermal grease (as DATEL did with their modules) would help keep the components at the same temperature, regardless of external conditions. )

Post Reply