I'm fixing up a magnetic step attenuator salvaged from an 8555A SA plug-in, to use in my TG project. I have some of these nice old attenuators, and the special rotary wafer switches that encode the binary sequences for the 10/20/40 dB step coils, and trigger the drive pulse. After some grief trying to figure out how the coil driver is supposed to work, and modifying the switch, I've got it going nicely. I built a pulse driver to run the coils according to the operational info in the 8555A manual - the coils take 12V, with the polarity selected by the switch, and are pulsed for about 150 mSec.
In the 8555A, the attenuator was only used over a 0-50 dB range, but I want the full 0-70 capability for the TG. This required some simple mods to the switch. First, the CCW stop on the shaft was eliminated, giving eight positions instead of six. 0 dB is at full CW, which is what I wanted. When I took it apart the first time, it looked like the switch was fully set for all the coil states, but when I got the whole thing operational for experimenting, it still only had the 0-50 dB range - the last two added spots were dead. On closer inspection, I found that one tab on one of the 10 dB wafer rotors was not included, but only on one coil side - the other side's rotor has all four tabs, and properly connects, for some reason. After some consideration, I decided the easiest fix was to do some simple switch surgery, to add a couple of contacts on the wafer, to pick up another tab when it went by, so the last two steps would fully encode. These contacts are redundant with some of the others, except in the last two spots, where they do their thing. It worked beautifully.
I'm working on optimizing the timing now, so am wondering about how well the original 8555A implementation worked - user interface-wise. Since the coil drive relies on open-loop triggering and pulsing, it's possible to overrun the state control by turning the switch knob too quickly, which doesn't hurt anything, but may result in the attenuator's actual state not matching the switch position. For example, if it's rotated quickly through a few steps, it can misfire. Changing it one more step either way will trip the cycle and restore it to the right condition.
Does anyone who has or has used the 8555A (or anything else with a similar setup) remember if this was an issue or minor nuisance? I had some many years ago, but don't recall being able to spin the knob fast enough on the cramped front panel, to misfire. Also, it only had five steps that could be rapidly switched anyway, since the last one had a manual release button to enable the 0 dB setting. On the TG, the knob has lots of room, so it's easy to whip through the steps fast, and I have to remember to take it easy, and be more deliberate in changing the setting. It seems to work fine up to about three or four steps per second. Of course, beyond that, there's less and less time to put out valid coil pulse duration and reset everything.
For the coil driver, I made a small board that attaches right to the coil pins on the attenuator. The board space is kind of tight, and it needs to run from the raw +33V supply, and include various protection circuitry. I decided to go old-school, and made the pulse driver with a relay and capacitor, where the cap discharges through the coil resistance for timing. The trigger from the switch fires the relay coil initially through a small cap, then the relay hooks up the main cap, and applies the juice to the attenuator coils through the encoding switch. At the end of the cycle, the relay flips, disconnecting the load, and connecting the timing cap back to the recharging source. The way the 8555A did it, and the way the switch is set up, is that all three coils are fired simultaneously, forcing it to agree with the setting. The coils are about 70 ohms each, so firing them all takes about half an amp at 12V. I didn't want to add extra power supplies or circuits for a lower voltage, so I took advantage of the non-continuous and pulsed nature of the load, and used the resistive drop of a string of PTC thermistors from the raw supply, with a big cap in the middle to give some isolation against the edges of the pulsed load. The PTCs provide intrinsic overload protection, in case something gets stuck, or in case someone like me gets carried away and spins the knob too often and too fast - the PTCs are clumped together, and will gradually all heat up and trip out. It ain't efficient, but it is simple - there's not a transistor or IC in there, just a few diodes and Zeners. The only active component is the relay.
Anyway, if anyone is contemplating re-using any of these fine old attenuators, this info may help. I have had to learn a lot about them lately, and have renewed appreciation for how they work.
Ed
