Hit the ground running: Detective work revives vintage vector voltmeter
My workbench includes a Hewlett-Packard 8405A VVM (vector voltmeter) of 1960s vintage. The VVM can convert two inputs at frequencies of 1 MHz to 1 GHz to outputs of 20 kHz, complete with modulation and sideband components. It can read signal levels as low as –70 dBm at full-scale and phase-difference readings over a range of 360° with 1.5° accuracy. With an APC (automatic-phase-control) PLL, it locks to the input-signal frequency in 10 msec. You can then measure level and phase angle in the presence of interference as high as –40 dBm and a signal level of –60 dBm.
Measurements of samples of the input signal with 300-psec pulses at a rate of 0.98 to 2 MHz produce the 20-kHz IF output. Such a pulse generator requires large, short, stable current-pulse outputs from the power supply, and even a small inductance can cause instability. The "APC unlocked; check signal/frequency range" front-panel indicator warns that the signal level or the frequency is out of the instrument range or that something is preventing the APC from locking onto the input signal. While I was using the instrument, this indicator came on.
Something in the unit had degenerated. So, I measured every waveform and voltage that could pertain to the PLL. The waveforms in the phase comparator, which resided nowhere near the pulse generator, were unstable, but there was nothing wrong on the pc board. I had changed transistors, bypassed and changed capacitors, and checked every voltage with the voltmeter and scope. I also sought trace breaks on the pc board and found nothing wrong with them. But the PLL still didn't lock!
The 300-psec sample-pulse generator resides on a pc board in a shield cage with sliding clips along the sides to contact the ground plane, and it has four edge-contact pins wired to the chassis. The ground wires were less than an inch long. Clearly, the HP designers knew that this circuit needed a good ground. Assume a ground wire ¾ in. (approximately 2 cm) long for a pc board with the pulse generator. The output is 300 psec—that is, 3×10–10 sec wide. Such an output has energy at 3 GHz. The 2-cm ground wire has approximately 20 nH of inductance, or just about 400Ω reactance at 3 GHz. That ground is not a good one; that's why HP's designers used sliding clips and multiple ground wires.
The ground wires were intact according to the ohmmeter, but, with a circuit like this one, that criterion is not final. You can get the best assurance of adequacy by duplicating the ground connections at the pulse-generator edge connector. So, I connected four more wires, each more than 1 in. long, to a solder lug on the chassis and to the pc-board edge-connector ground pins. Now, the ground arrangement included the sliding clips along the pc-board sides; four wires, each less than 1 in. long, to a ground lug; and four more wires, each more than 1 in. long, to another ground lug—resulting in a circuit with eight wires in parallel!
Even for 3-GHz energy components, this arrangement should provide an adequate ground, and it did. The APC locked immediately. Now that the ground was good enough, the instrument worked correctly. The rivets holding the pc-board sliding clips to the aluminum chassis must have corroded over the years, causing more resistance. Perhaps the same situation was true of the ground-lug fastener that made the edge-contact ground connections.