9. Tek Type 109 Pulse Generator
µ Ω ± °
The Tek 109 pulse generator continually oscillates at 640Hz nominal, using a bounce-free mercury (Hg) relay to switch coaxial charge lines that produce output pulses up to +50V or -50V (±300V with an external supply) having a nominal rise time of 250ps.
I found a working Tek 109 pulse generator for €50 on eBay France and bought it, along with a couple of spare Hg relays and two GR874 charging coax lines. I had already bought GR-874 to BNC adaptors.
When it arrived the 109 no longer worked, so I set about determining the cause.
109 OPERATION
The 109 produces a fixed 640Hz (nominal) output from two charge lines switched by a 360Hz oscillator. Mine can be adjusted down to about 505Hz using pot R614 intended solely for calibration to 640Hz.
Examining the 109's manual [I39] and schematic (above), the 640Hz oscillator is formed by a pair of cross-coupled PNP transistors that directly drive an electro-magnetic transformer near the relay, effectively forming a feedback astable multivibrator whose frequency is limited by the saturation flux of the transformer and the voltage applied to the circuit. The reed relay is biased by the flux of a simple horseshoe magnet; the transformer flux adds to this bias causing the reed to switch its pole between changeover contacts. The voltage is derived from a 10V zener-based regulator with a transistor drive output passing '+8V' from a poorly regulated +17Vdc supply (ripple is quoted at 1.5V pk-pk).
Note - although it's identified as signal +8V, it can be set between 7V and 10V according to the frequency adjust pot R614. For this reason I've put '+8V' in inverted commas.
One of the first things I check on old equipment is electrolytic capacitors as they can often dry out. I checked the ESR of the two 2000µF low voltage supply reservoir capacitors C610 (good: 2960µF, 60mΩ) and C611 (bad: 7.8µF, >20Ω). Given their age, I replaced both with 105°C 25V 2200µF 20mΩ electrolytics.
Bottom left - inside RHS. C610 & C611 are the two large cans at the bottom:
It still didn't work.
TRACING THE FAULT
I connected both GR-874 cables in series to create the longest delay:
The oscillator frequency is determined by the voltage on signal '8V'. Conveniently, inside the 109 this signal is the bare wire on the nearest end viewed from the front, of the ceramic component carrier housed in the left side. On the top joint is a WE/OE wire [T16] that leads through a rubber grommet between drive transistors Q725 & Q735, into the instrument right side, where it is connected to Q627 emitter and Q725 collector.
Something to note about these ceramic carriers is they are sensitive to heat and require a special silver content solder. There is actually a note to this effect on the metalwork inside, and on page 5-2 of the manual. Tek used to supply a small reel of it in some of their instruments for future repairs. By chance I had already bought 3m of a similar '96SC105' 95.5% tin + 3.5% silver 0.46mm dia solder for this exact purpose (eBay UK, £3).
This 109 had been repaired at some time in the past and some carrier joints were poor, maybe they used ordinary solder? I reworked them with the silver solder just in case.
I probed '+8V' on the carrier and found even though there was no output, the oscillator was working:
Above right - I began to suspect the magnet might have shifted in transit so I loosened and moved the red horseshoe magnet. A pulse came out, but it didn't have the expected fast risetime.
Suspecting the worst, I reluctantly decided to replace the Hg switch.
REPLACING THE HG SWITCH
The magnetic switch assembly has a biasing red horseshoe magnet on the left, and a white electro-magnet on the right.
The Hg switch is located within the vertical metal tube behind the front panel below the assembly, its pole contact mating with the GR-874 output connector, where it is held in place by a grub screw:
Two small carbon film resistors first have to
be unsoldered from the small pcb to which the Hg switch is also soldered.
The grub screw is loosened using a 0.05" Allen key:
I fitted the new Hg switch and re-assembled it but it still didn't work.
Frustrated, I unscrewed the electromagnet to examine it more closely.
Below left, the 109 frequency is adjusted by variable resistor R614, visible to the right.
Left - A 4mm spanner fits around a T10 bit
Right - Some people have suggested the horseshoe magnet might lose its strength over time, so I measured it.
When I refitted the Solenoid, the output appeared. It was simply a loose screw.
Right, I noted the magnet's best working position for future reference.
Below, The 7104 confirmed the expected waveforms.
The slow curve at the top of the rising edge is very likely due to the cheap 50Ω coax I used between the 109 output and the 7104 7A29 input:
µ Ω ± °
µ Ω ± °
µ Ω ± °
µ Ω ± °
µ Ω ± °
µ Ω ± °
µ Ω ± °
µ Ω ± °
µ Ω ± °
µ Ω ± °
µ Ω ± °
µ Ω ± °
µ Ω ± °