This 50 MHz pulse generator was built as solid as a rock. It’s almost all metal. We take the unit completely apart and look at it from a system and board level.
On a visit to the MIT swapfest in 2024, I came across an HP 8112A 50 MHz pulse generator. For $20, I brought it home to take apart. This instrument, assembled in the Federal Republic of Germany circa 1987 before the fall of the Berlin Wall, is a classic example of a solidly built piece of test equipment. Here’s what’s inside.

Taking apart the 8112A was both a challenge and an adventure because the instrument is constructed almost entirely of metal. The only removable plastic parts I found were the operating buttons and switches, an arm connecting the power button to the AC mains power switch, and a strip that hides some screws. I needed a cup to hold all the screws. Why? The inside structure is entirely made from metal; no wonder the 8112A is so heavy. The transformer alone weighs 1750 g. All that weight at the back of the unit surprised me when carrying the 8112A out of the swapfest and placing it on a shelf.

As the label in Figure 1 says, it’s for parts or repair. I did try applying power to the unit though didn’t start. The good news is it didn’t produce any smoke. After opening the case, I could see why the instrument didn’t start. As you’ll see, I’m not the first person to remove the cover.
After trying a few ways to get the case opened, I found that strip of plastic behind the front bezel. Popping it off with a screwdriver revealed several screws (Figure 2) holding the front panel bezel to the top cover. Now I could get a look inside.

Removing the front panel revealed three boards. Figure 3 shows the boards: the main board (bottom), control board (middle), and processor board (top). In the video, you’ll get a tour of the boards. Read on to get additional insights.

After removing the top cover, I found markings on its underside indicating the locations of adjustments for the instrument’s power circuits on the main board. Figure 4 shows the main board (right) and the corresponding designations inside the cover. Clearly, this test instrument was designed for serviceability. The operation/service manual provides board layouts, schematics, parts grids, parts lists, calibration instructions, and service instructions. Keysight does not provide the service manual for this obsolete instrument but it’s still available online. As I explain in the video and below, the microprocessor board is different from the layout in the manual. I’ll explain why in the processor board section below.
Main board

The main board contains the power supply and pulse-shaping circuits. Figure 5 shows the power supply, with its transformer, diode bridge, linear voltage regulators, and power transistors reside in the rear, shown on the right. The two ICs with the small heatsinks handle the shape and slope. They’re connected by a wire under the board to three sockets that should contain the period generator, delay generator, and width generator. Someone removed them. Clearly, I’m not the first person to open the case. The missing parts appear to be identical according to the schematic. Figure 6 shows the block diagram of the three missing timing ICs.

The visible wires connect the pulse-shaping circuits — the final signal processing stage — to the front-panel connectors. The top wire in Figure 6 is for the trigger input while the bottom wire is for the trigger output. The middle wire is a ±20 V control input. Other circuits include:
- Trigger level circuit and output amplifier
- Level checker
- Output amplifier
- Offset control
- Address decoder
- Reference circuit
- Current sources

Looking at the power supply, you’ll see that Figure 7 shows the power transformer, the AC mains power switch, and one of three diode bridges, which is just above the switch. The AC mains input connector is at the top right corner of the photo.

Figure 8 moves down the board. The transformer (out of view) is above the large capacitors. You can see the heat sink rotated to expose one of two rows of voltage regulators and power transistors. Two voltage regulators appear on the right. The far right regulator is for +15 V. Moving left is a +23 V regulator. The negative voltage regulators are on the opposite side of the black plate. The regulators all have HP part numbers with National Semiconductor (now Texas Instruments) logos. Moving left again, you can see two power transistors. Each form output stage of a Darlington pair. The transistor on the right (next to the regulator) is an NPN 2N5191 (Motorola marking). To its left is a PNP 2N5194, also Motorola (later Freescale, now NXP).
Control board

The control board (Figure 9) resides in the middle of the board stack shown in Figure 3. The control board generates the output pulses with data from the processor board. It then sends the pulses to the main board for signal conditioning before they go to the outside world. The board contains the following circuits:
- Timing circuits
- Address decoders
- Timing range decoder
- ADCs and DAC reference circuit
- Byte offset and offset DAC
- Parameter control
The heart of the control board consists of six 10-bit D/A converters (DACs). Figure 9 shows the HP part number with an Analog Devices logo. The schematic from the service manual identifies them as AD7522. Of the six DACs, two drive current sources and one each drives the amplitude control, period control, pulse-width control, and delay control.
As with the other two boards, the control board contains some 74LS series logic from Texas Instruments and Signetics (acquired by Philips in 1975 and now NXP). The Signetics parts are 74LS273 octal D-type flip flops.
The control board contains 22 trim potentiometers. They let factory and service technicians adjust op-amp circuits that connect to the DACs, producing current sources and control the output amplitude. I wonder how long it took to calibrate the HP 8112A’s analog circuits. The service manual describes the adjustment procedures.
Processor board

Figure 10 shows the processor board and it’s Motorola MC6802 8-bit microprocessor (marked as HP part number 1820-2099). The other 40-pin IC (upper right) is a GPIB controller. Given that HP invented HPIB (GPIB/IEEE 488) You’d expect that the controller IC would be an HP part, and it is. The schematic refers to it as HPIB644 although the part itself is marked 1820-2219.
Here’s where things get interesting. I found a difference between the board and the service manual. Note the five ROM ICs on the board’s top edge. The four to the right are TI TMS2532, a 32k EPROM. The one on the left is a Hitachi HN462532, also a 32k EPROM, bringing the total ROM capacity to 160 kB. The service manual shows the schematic and layout of a later board, taking advantage of greater ROM capacity. The schematic shows the five devices replaced by a single 256 kB EPROM highlighted in Figure 11. The RAM IC in the upper left corner remains an HM6116, a 2048-word × 8-bit high-speed (in its day) static CMOS RAM.

The rest of the processor board contains the address and data buses, decoders, and 74LS series logic gates. It also contains a power section and battery for the RAM IC. The board also uses several LM324 op amps configured as comparators and resistor ladder circuits for generating a reset signal to the main processor. Transistors control +5 V power to the RAM. There’s also an Intersil display driver on the processor board that drives the front-panel display. LM324 quad op amps are still available, even in through-hole packages.
They don’t make them like they used to
Taking an HP8112A pulse generator apart showed how HP built a rugged, serviceable test instrument that can produce test pulses for many years. The rugged, solid metal design made for a heavy package. Today, lighter materials would surely take over. Power consumption is certainly less than it was in the 1980s. You’d never find a linear power supply today except in cases where any noise from switching supplies is unacceptable. A switching power supply would result in a much smaller, lighter power transformer weighing far less than 1750 g. The smaller, more efficient power supply would also produce less heat, which translates into smaller, lighter heat sinks.
As for the circuits, all of the logic, addressing, and decoding can today take place in a single ASIC or FPGA. There’s no need for a GPIB controller as today’s instruments use USB and Ethernet connections, both of which are serial buses. Today, a pulse generator also needs to produce pulse patterns for testing these serial buses.
While producing pulses at 50 MHz was fast in the 1980s, that’s painfully slow today. Indeed, Keysight’s pulse generators now attain speeds up to 3 GHz. Today’s pulse generators can produce overshoot and undershoot, things not possible in the 8112A.
They don’t make them like they used to. Today’s instruments are faster, lighter, use less power, and do more than the HP8112A. Are today’s instruments as rugged as those from 40 years ago? Ask that question in 40 years.
The service manual doesn’t tell you how to put the unit back together.