The final structure reveals itself. These are the remains (except for that grotesque little piece of hardline hanging on one leg!)
Given what this poor mixer has seen, I thought a quick 'n dirty bodge of my own could hardly be worse. This is the "crossroads" point in the life of the kit. When things are so bad it all represents a time-wasting loss, you get to think you have nothing more to lose, and are willing to meddle with a high risk irrepairable damage.
I don't happen to have a 4-tab diode ring (crossover type) component, so I made up one out of a pair of HSMS-8202 Schottly dual diodes SOT23, stacked one on top the other, and the top one flipped on its back with the tabs bent down.
My "repair"
Kind of tricky to get the lower diode junction connected, finally done with about 2mm of wire to carry the solder. You can just see the end sneaking out from under the diodes and to make it onto the top of the fed strip. Similar for the top junction connection to the hardline outer - a tiny piece of wire.
Harder to see, but the 68 ohm termination resistor (actually required by the original circuit) is now made up of two 0805 surface mount resistors; a 100 ohm and a 220 ohm mounted right at the hardline entry point either side of the hardline.
Next page shows the termination better.
This fact! - that there is most of a termination right at the input, that confuses me about the feed method. I feel sure the real clever fellas out there who do hard sums with Smith Charts in their heads will let me know how this balun works. Recall - we had a starting point here
The 24dB of missing gain is now found! The trace is nearly level right across the screen to 1GHz, but with wiggles and squiggles! It no longer dips and dives crazily through 30dB of ups and downs. It sags badly below 50MHz. (Pictures to come..)
But we are not through yet. It is time to lose that piece of cable between the tracking generator and the input. 2 metres long of old RG58 with knots in it that I had laying about and various BNC adaptors to make it to N-type connect will not do. Its a struggle, because all my good link cables have been grabbed for "desperate site stuff". Eventually, a 1 metre piece with 50 ohms solid inner and fine N-type ends is found. Do not use anything with right angle end connectors for this. This yield is immediate. Clearly some of the wobbles were real measures of the response of piece of faulty link cable. When fixed, I was surprised how good a piece of thin RG223 could be. Progress!
You can see some wobbly displays here
CARRY ON TO THE NEXT PAGE
Page 1 Special Mixer Mess (description of the problem)
Page 2 - Spectrum Analyser Mixer Dissection - Taking things apart
Page 3 - Spectrum Analyser Mixer Dissection - A closer look
Page 4 - Spectrum Analyser Mixer Dissection - Oh Yes - it is still full of surprises!
Page 5 - Spectrum Analyser Mixer Dissection - How it ended up
How We Finally Got it Right (and all the secrets too!)
The Result - Performance Traces
Some (not so good) Traces