Electronics > Repair

Adaura programmable RF attenuator SMA connector replacement + teardown

(1/1)

hanakp:
Short version: I'm looking for flange-mount SMA connector that has protruding dielectric "candlestick" of 2.1 mm diameter and 2.5 mm length.

Long version:
Colleagues asked me to repair somewhat unusual instrument, Adaura programmable RF attenuator AD-USB1AR36G95. It's interesting because it can be directly controlled and powered via Ethernet (with PoE) or USB. Some Adaura's higher models even have internal web server with GUI, LXI-style. But AD-USB1AR36G95 is a basic model and contains pretty much what you'd expect, PIC32MX675F512H with LAN8740A LAN controller and NCP1081 PoE controller driving Pulse PA1134NL transformer. The RF attenuator is a cascade of three pSemi PE43712 ICs.

Anyway, it had broken input SMA connector. Don't ask me how it happened, because nobody knows, and the culprit hasn't stepped forward. Finding a suitable (but problematic) replacement took me quite a while, which is the main reason why I'm posting this. It's 4-hole 12.7x12.7 mm flange type with extended dielectric "candlestick" on its backside. However, this "candlestick" proved to be the main complication, because it has only 2.1 mm diameter. And it tighly fits into a hole in the attenuator's casing. The only similar connector I could find is Cinch Connectivity 142-1701-191, but it's not an exact replacement. The original "candlestick" was only 2.5 mm long, whereas 142-1701-191 has 3.18 mm. Fortunately, the attenuator case is a bit thicker than 2.5 mm, so even the longer "candlestick" fits... if you'd call it that. It looks strongly compressed against the PCB, which is not ideal, but works. I considered shortening it with a scalpel, but teflon dielectric is stiff and I feared I'd accidentally cut the center lead, too. If you happen to know a SMA with the right dimensions, please put it below. I manually checked datasheets of dozens of SMAs on Mouser, Digikey, TE Connectivity and similar sites, but couldn't find it.

Overally, the attenuator is really nicely made. The case looks like stainless steel, but it's completely non-magnetic and too soft for stainless steel. It's not plated copper or brass either, I tried to deep-scratch it. And it's relaively lightweight, so my best guess it's Alloy 770 or similar RF material. The attenuator ICs are separated by a lid from the same material, fitted with RF gaskets. Even the SMA connectors have their RF gasket O-rings, I've never seen that before.

hanakp:
Since I succeeded to repair the first attenuator, colleagues "rewarded" me by bringing another one. This time, the problem was electrical, it worked fine when it was new, but gradually became unreliable. When I got it, it didn't work at all over Ethernet, and when connected to PC via USB, it produced "device descriptor failed" error on Windows and "device descriptor read/64, error -32" on Linux.

I couldn't find anything wrong with usual visual and electrical checks, so I resorted to live comparison with a working unit. I probed the 6-wire serial link between MCU mainboard and RF board which controls PE43712 ICs, see attached image. In a good unit, SCLK ran at 1 MHz, but only at 100 kHz in the bad one. This led me to realize 8 MHz crystal oscillator in the bad unit isn't working. I was aware of it, but couldn't know if it's fault or feature. And I also didn't know that PIC32MX675F512H falls back to some internal oscillator when external one fails. Anyway, it turned out CLKI clock input (pin 39 on MCU) had low resistance to ground, about 12 ohms. It appeared and disappeared as I probed around it -- for example, it ran for about 2 hours after I re-melted crystal's solder joints, then it became shorted again. The crystal is in 4-pin 5x3.2 mm SMT package, with two pins connected to GND. Apparently, some internal structure slightly touched the crystal's case. That surprised me - I've encountered broken (open) or burnt (too high drive level) crystals before, but never a shorted one. For the time being, I replaced it with a small THT crystal I had on hand and the unit has been running for the last 3 days without any problems.

BTW, I tried reading firmware from the MCU, and to my amazement, it wasn't locked in (read-protected). I saved it as HEX files from both units, PM me if you need it. But I suspect it won't help you much, because about 100 bytes differ in the HEX files, even though both units have consecutive serial numbers. The differences are probably calibration data etc. which will be unique for each unit. The MCU uses standard Microchip 5-pin programming header, I marked its orientaton in the image. I also tried reading Winbond W25Q16JWNIQ Flash that's nearby, but it was completely blank on both units.

Navigation

[0] Message Index

There was an error while thanking
Thanking...
Go to full version
Powered by SMFPacks Advanced Attachments Uploader Mod