Saturday, June 28, 2014

7L12 Repair

While taking some measurement on a 1575MHz SAW filter, the 7L12 screen went blank.  The unit had been in continuous use for several hours (longer than I normally have it on).  The screen reticule illumination changed (brighter) and no trace or text output were visible (span or reference).  Some push button based lamps were lit but the 7 segment red LED indicator for attenuation setting was not.  Even the fixed calibration output signal was dead.  On closer inspection of the schematics, the calibration circuit is a simple 50MHz oscillator requiring +/-15V.

Based on the behavior as well as some reading, my suspicion was that a tantalum supply bypass capacitor had failed in a short mode.  On removing the unit from the frame I measured the resistance between A18(+15) and ground as well as B18(-15V) and ground.  Eventually it was identified that the +15V was shorted to ground.  Fortunately, the problem was not with the YIG oscillator or first mixer (components I had not hope of repairing or replacing).

After a lot of work the offending board was identified as the A1300 board.  This involved basically pulling board connectors until the short is removed.  Sounds easy until you note that: a) the thing is assembled like a Swiss watch, b) the manual does not include a wiring harness diagram.  The A1300 board is referred to as the “Phase Lock Board”, however, there is quite a bit of functionality on it besides a PLL including the YIG driver and marker generator.  With the +15V supply to this board removed, the unit was tested in the frame with a successful display (flat line trace with reference markers) and calibration signal.  The following picture shows the A1300 pulled up from the rest of the assembly (yes, the middle boardset is on rails).
No capacitors on the board showed any visible indication of failure (discoloration, charring) - apparently this is normal.  I started to make a list of candidates to remove, however, quickly realized that part of the board was unpopulated but shown as populated in my manual schematics.  This tells me a couple of things a) my particular unit is not as old as I thought it was, b) my schematics were not entirely correct.  By going through the candidate list and removing the capacitors and measuring +15V to ground at each step I was able to find the offending entity on the second step.  The two removed were bypassing +15V for the PLL supply line.  The offender was a 100uF +20V, of which I seem to be out of.  Based on the location of the capacitors and artwork (under a shield limiting side access with several pads/traces under and around) I could not just use some near value 0805 SMTs.  In the A1300 picture above, the capacitors are located under the shielded area on the lower right under the large cable harness looping over it.

The unit was tested without the bypass capacitors (the thinking being that I did not need the PLL working or working well for basic use and first test).  The unit was tested with the calibration signal at various frequency spans and reference levels with good results. 

As part of this exercise, I decided early on that my probability of success was low so I ordered an alternative that fit within my budget and needs.  This was an RF Instruments SA0314.  No one would confuse it with an Agilent or Rhode and Schwarz with respect to size, performance or cost. Based on initial testing it is exactly what I need and is performing well.  It provides better performance for many measurements than older spectrum analyzers, fits in your pocket, and exports digital data.

Friday, June 20, 2014

Cascaded SAW Filters (315MHz)

As an experiment, a PCB with two cascaded SAW filters with fence option and 3dB pad separating them was built.  The following is a picture of unit #1 populated but without a fence.
Cascaded SAW board.  3dB pad coupling two 315MHz SAWs with shield mounting outline.
The filters used are EPCOS B3711 centered at 315MHz.  These come in a DCC6C package.  The advantage is that there are multiple frequency ranges available in this package allowing the PCB to be re used.  In addition it is large (on a relative basis) providing easier handling and assembly.

The wide band response of a single filter from the datasheet is shown below.

A quick simple setup was used with both the 7L12 and the Si application with single A-B stack to scan the filter response.  Neither device has enough dynamic range to scan the desired response range, however, it is a good sanity check and quick to do.  Those results are shown below.

Measurements show a passband insertion loss of 7dBm.  This aligns nicely with the datasheet insertion loss of 1.7dB for a single filter (2x for the cascade) with a 3dB coupling pad. The lower end peak at ~309MHz is 50dB down for a 2x cascade while the datasheet indicates -25dB for a single device.  All of this tells me the devices are not interacting badly and pulling each other.  In short, this is working quite well.