|bh||Early Embedded Processing:
The Sherwood Micro/CPU-100 FM Tuner
The Sherwood Micro/CPU-100 is an FM stereo tuner dating from the mid-1970s. In its time it was a high-end tuner incorporating then-state-of-the-art technology.
In the mid-70s when the Micro/CPU-100 was introduced, microprocessors were still quite new. About the only consumer items containing a microprocessor were the then-new calculators and the timer/control in the similarly new microwave ovens. Putting a microprocessor in a type of device - such as an FM tuner - which had been around for decades was highly unusual. The Micro/CPU-100 is notable historically as one of the first such 'common' consumer items (albeit expensive in this instance) to contain an embedded microprocessor, one of the early steps to the ubiquity of embedded processors today.
Some of the features of the Micro/CPU-100:
The Micro/CPU-100 was actually OEM'd for Sherwood by Draco Laboratories.
The Micro/CPU-100 units I have worked on needed extensive repairs. Most of the work needed has been the result of damage from leaking NiCd batteries. Having been repaired, the units have been working very well. In the course of these repairs the Micro/CPU-100 has been reverse-engineered to produce the schematic. Contact if interested.
Shortly later some actual PLL/synthesised-tuning designs appeared, such as the Scott 433 and the Heathkit AJ-1510.
These initial entries used SSI/MSI ICs.
The Micro/CPU-100 incorporated a microprocessor to provide additional functionality and features.
|CONTENTS (this page):||TECHNICAL INFORMATION:||RELATED:||EXTERNAL LINKS & REFERENCES:|
The tuning scale, however, has no sliding pointer. Instead, a row of 22 LEDs provides resolution to the appropriate MHz. This linear scale is really superfluous to the more-prominent digital displays: a 4-digit 7-segment LED display of the frequency and a 4-character alphanumeric LED display for station call-signs, making for three displays of the frequency and station.
The linear tuning scale was likely included as a comfort factor for purchasers of the day. In the mid-70's, most people were not used to digital electronic numeric displays for what they were used to seeing and interpreting in analog form, such as radio dials. Many were not fond of the transition from analog watches to digital watches as another example.
The station call-sign display is four 5-by-7 dot-matrix LED modules. The call-signs - or whatever 4-character ID for a station one might wish - are manually programmed by the user. This is accomplished by touching the 'ALPHA' button, then using the tuning knob to dial-in a character, pressing the 'STORE' button to lock in the character, and repeating for the four character positions. The characters are generated by a standard 2513 character-generator ROM (typically used in computer video terminals) so the set of 64 ASCII uppercase, numeric and punctuation characters are available. (Note: a call-sign entry can be cleared by pressing 'ALPHA', followed immediately by 'A').
The tuning knob is a massive chunk of solid aluminum. Inside, on the same shaft, is an additional flywheel mass. Also on the same shaft is an optical chopper wheel, for two optical sensors in quadrature feeding the microprocessor. One rotation of the knob steps through 6 FM channels. As the tuning knob does not have to drive a dial cord or any mechanism, it is very smooth to operate.
The rectangular 'buttons' for the programming, pre-select memory and auto-scan features are actually touch sensors. Each one has an LED behind it which functions as visual feedback of activation and state indicator. The pre-selects also light when the tuned-in frequency matches the pre-select frequency, even if it wasn't selected via the pre-select touch sensor.
The novel touch sensors are based on 555 timer ICs used in monostable mode. Each sensor button has two contacts, the right contact goes to the 555 trigger input, the left contact is ground. These sensors work well enough for the most part, but can fail to trigger if one's skin is very clean and dry. The contacts are pins soldered onto a PCB without much mechanical relief and the solder mountings have been observed to have fractured from being pressed with too much force. Survival of the 555s in a high-static environment may be questionable.
Some conventional switches and adjustments are located behind the flip-down door on the front panel.
Visible in the upper right corner of the bottom view is the optical chopper wheel for the tuning rotor.
The microprocessor system is contained on the board seen mounted on the right side of the tuner. A shield between it and the rest of the tuner helps keep noise from the digital signals out of the analog sections.
Several modifications to the manufacturer's design are present in these photos.
Some of the wiring has been redone and rearranged.
The heatsinks around the power supply are new.
The battery mounted on the rear panel is an alkaline-cell replacement for the original internal NiCd battery.
(See power supply commentary further below.)
|RF Front End|
The electronic tuning involves 6 varactor diodes: 1 for the VCO and 5 for the RF front-end tuning.
Two dual-gate MOSFETS provide gain in the RF front-end, followed by a JFET mixer.
The VCO feeds the mixer and a divide-by-8 counter.
The 3-stage divide-by-8 counter is contained in two of the ICs on this board.
The first two stages of the divide-by-8 are ECL D flip-flops.
These bring the count frequency into the range of TTL, used for the third stage flip-flop.
The fixed divide-by-8 counter is followed by a programmable 12-bit binary counter on the CPU board for the PLL frequency selection.
The third IC is a Fairchild 11C44 phase-detector/charge pump to control the VCO.
|IF Amplifiers and 'Digital' FM Demodulator|
The IF amplifiers are primarily a series of LM703 ICs and 10.7MHz ceramic filters, along with some JFETs and a CA3089 IC. No tuning coils or L/C trimmers are present as the ceramic filters are used for the IF filters. An on-board relay selects the narrow-band or wide-band IF amplifier paths. In wide-band operation there is only 1 ceramic filter in the signal path. In narrow-band operation, 4 ceramic filters are present in the signal path.
The 'digital' FM demodulator is a common TTL 74121 monostable triggerred by the 10.7MHz IF carrier to produce a pulse-width-modulated representation of the audio. The monostable pulse width is set to slightly longer than the period of the 10.7MHz IF (~ 108nS vs. 93nS), so the 74121 is actually triggerred at half the rate of the IF frequency. Both the true and inverted outputs from the monostable are filtered and fed to a CA3140 MOSFET-input op-amp in differential mode to become the composite audio signal.
The only adjustment for the entire IF amplifier and demodulator stages is the demodulator monostable pulse width.
Two versions of the audio board are shown here, notice the different positioning of the ICs. The versions differ in some of the circuit configuration for degree of stereo separation and muting. They both use an HA1156 PLL stereo decoder IC.
De-emphasis is selectable between 75, 50 and 25 uS. In addition to the normal multiplex filters, a 19 KHz active notch filter is present to improve removal of the 19 KHz stereo pilot.
There appears to be a third version of the audio board, going from a photo seen on the
Whether it is earlier or later than the versions shown here, I do not know.
The microprocessor manages all the functions related to frequency selection: interpreting tuning knob and touch-sensor activity, managing the pre-selects memory, managing the call-sign memory, auto-scanning, updating the displays, and calculating and setting the division factor for the phase-locked-loop. The only actual control the microprocessor executes over the tuner proper is setting the PLL division factor and muting the audio, which it does while changing the frequency.
The microprocessor is an RCA COSMAC 1802. The 1802 was RCA's entry into the microprocessor market of the 1970s, notable for a somewhat unusual register architecture and low-power CMOS technology.
Firmware for the 1802 is contained in two ROMs, a 256-byte fuse-programmed 74S471 and a 512-byte mask-programmed CDP1831. RAM for the system is comprised of two 5101 ICs (256*4) for 256 bytes of storage. 240 bytes of this are used for the call-sign memory. At 5 bytes per entry, this provides for 48 call-sign entries. Another 5 bytes store the current channel and the preselect channels. The 5101s are low-power CMOS devices and are powered by a separate AC power supply not controlled by the power switch, as well as a battery backup, for preservation of all this state information.
The connector on the wires seen in the photos is a non-original modification.
The wires are for the PLL signals with the RF board and were originally soldered in with no connector.
The connector was added and the wiring replaced so the CPU board could be removed without involving soldering.
|Reading the ROMs|
The CDP1831 ROM is soldered in, so it was read in-situ. This was accomplished by removing the socketed 1802 and wiring in a simulator to exercise the bus and address lines and read the data lines. Shown in the photos is the CPU board wired up for the dumping process. The bus simulator and ROM reader is a SWTPC 6800, seen here on it's side behind the Micro/CPU-100. The 6800 is running an assembly program written for the task. The ROM contents are read through the 6800 and uploaded to a modern computer acting as the console to the 6800. The SWTPC 6800 - one of the first microcomputers - coincidentally is a contemporary of the Micro/CPU-100.
The ROM contents have been disassembled and flowcharted.
|Power Supply and Battery Issues|
The battery is present for memory retention when AC power is lost but leaking or off-gassing from the NiCd battery can destroy the printed circuit board(s) and other parts. In both of the two units I worked on, the battery had destroyed significant portions of the circuitry. In one unit, the entire power supply circuit board had been wrecked, along with components, wiring around the power supply, and nearby areas of the chassis. NiCds are not a good choice for the long-term low-current requirements of CMOS memory retention, so I replaced the battery with two Alkaline cells mounted in a holder on the exterior of the rear panel and modified the power supply circuity to remove the NiCd charging function.
Shown in the photo is a rebuilt power supply board, two new heatsinks replacing the original single smaller heatsink.
The aluminum bracket at the bottom of the photo is another modification, a replacement for the original flimsy posts supporting the display board.