Hi All,
Today's rainy-day project is the tear-down and repair of a 30dB, DC-1.5Ghz,150W, Weinschel attenuator. Its been sitting on my shelf for a year waiting to be fixed.
These attenuators use cascading pads internally so one end can only handle 150W and the other can only handle 20W thus care needs to be taken to make sure you're putting all the power into the big end.
I've had this unit for some time and in a moment of brain fade, I had attempted to pass 100W into the low power end of the unit. The amp I was testing quickly shut down due to high SWR and I was left wondering what was wrong until I realised I had ruined the attenuator. A quick test of the unit showed it was now measuring 200+ ohms instead of 50.
Somewhere in the past year I had purchased a RES-NET 30dB 250W, DC – 2GHz, chip attenuator for a project and I decided to use that as the basis of repair for this unit since the internal resistor pads are either $$$$+ or unobtanium.
Disassembling the attenuator is fairly easy with only 10 screws to be undone and gently prising the two halves apart with a fine screw driver.
Internally we find two ceramic substrates with precision, laser trimmed resistors on their surface. The substrates sit in a groove milled into the heat sinks with their edges forming both a ground connection and a heat transfer point for the resistors.
Connection between the substrates and the N connectors is via gold plated, spring loaded contacts and connection between the substrates themselves is via a tiny gold plated back-to-back S shaped spring arrangement.
See the attached pictures for detail.
To fit the RES-NET module into the assembly I had to mill out a recess and drill and tap for M3 bolts. I initially chose to fit the module to the lower half but botched one of the holes, so I then proceeded to machine the top half being more careful with the drilling.
The RES-NET module was then torqued down at 0.5N with a liberal coating of thermal grease to ensure good heat conduction into the heat sink.
I used Belden 83242 M17, PTFE insulated, silver coated braid and core, coax to make the connection between the N connectors and the module. I connected the shields at both ends by bolted down copper shim.
I also had to mill a recess in the opposite side of the heat sink to give clearance to both the module and the bolt heads holding it down.
Reassembling and testing the attenuator on my HP E6380A shows that its within +/- 2 dB of of 30dB attenuation from 40kHz to 1GHz, additionally I tested the return loss and that too is more than acceptable for use. If I want to get picky when working at a particular frequency, I can characterise it before use.
Additional testing using an Agilent 4432B as source signal and an HP 8566B spectrum analyser shows that beyond 1GHz the attenuation begins to get funky. This is because the coax begins to look like resonant lengths and we see that affect exhibited in the final test from 1.7GHz to 2GHz.
Care will have to be taken if I intend to use this attenuator beyond about 1.5gHz.
Home
Help
Search
Login
Register
