HP 8566B: Low Band (A) Works Great, High Bands (B, C, D, E) Don't!

[ignilux]

Hi, all-

As the subject states, I'm working on fixing an issue with the 2-22 GHz bands on my HP 8566B spectrum analyzer. When I received the unit four or five months ago the low band (0-2.5 GHz) worked okay, but it had an intermittent YTO unlock problem. Ultimately it came down to the YTO DAC reference voltage reading ~10 mV too low, and an eighth of a turn on that pot fixed it for good. At that point I tested the high bands by connecting a 2.4 GHz wifi antenna to the input, and set a 100 MHz span from 2.4 to 2.5 GHz. There was plenty of signal there, so I thought everything was just fine. Of course, now I realize that despite sitting above the 2 GHz mark, the instrument will actually switch back to the low band unless the stop frequency exceeds 2.5 GHz. So of course, I hadn't tested the high bands at all.

Fast forward to this week, when I built a couple of comb generators using step recovery diodes. They looked fantastic in the low band, with harmonics out the wazoo. When I checked the high bands, though... Nothing! At 1 GHz into the SRD I am able to see a very prominent fundmental and second harmonic at around -20 dBm when viewed in the low band, but switching to the high bands shows just noise. If I turn the sweep speed WAY down (as in 2 - 5 seconds), RBW down, and span down I can just barely see the tones emerging from the noise floor at about -79 dBm. Centering a narrow span on the harmonic frequency, placing a marker on the barely-visible tone and peaking the preselector yields no difference.

At this point I started troubleshooting at a systems level before diving in and twiddling pots. Referring to the RF section block diagram on page 96 of the  service manual, I isolated my efforts to the signal path that runs from attenuator --> switch --> YTX --> ACLU --> second converter. Again, the low band works fine so I've ignored the path on the top of the page. The attenuator has always worked fine, and running through its steps makes no difference in the high bands. I found appropriate voltages at the control inputs of the RF switch, and I confirmed its function by removing it from the unit for bench testing. The YTX is deep inside the unit, and extremely fragile. I've avoided this as much as possible.

Moving on to the ACLU, the diode bias signal is present and changes between the bands. Possibly worth note: when sweeping through one single band, the bias voltage is constant except when the sweep resets. There is a small blip at first, then it stays constant. I am unable to check the 1st LO output, since this is my only instrument capable of measuring signals this high. However, because it is involved in the first conversion of the low band I'm certain it is present. The 321.4 MHz IF output shows more or less what I would expect: nothing throughout most of the sweep, with a pulse when the sweep resets and as it crosses the very small tones from my comb generator. The subsequent low pass filter and PIN switch work okay, too, as I pulled the 321.4 MHz IF output from the second converter and monitored it, which showed the same behavior as earlier in the signal path. PIN switch control voltages check out, too.

Looking down at A6A7,  I monitored the coil voltages from the YTX driver, which showed a linear ramp that followed the sweep rate. The YTX heater control has an LED that shows the heater is in saturation and a separate LED to show that the heater is running. A more detailed schematic is found on page 127. The saturation LED lit for a few seconds at startup before turning off. The heater running LED was burnt out, though. I removed the A6A7 board, and replaced the LED. While I was in there I found that the YTX filter capacitor had its ESR in the hundreds of ohms, despite being a 100 uF electrolytic, so that was replaced, too. Neither of these repairs made a difference in the actual problem, but the heater running LED illuminates now.

It seems to me that the issue lies with the YTX itself or its associated YTF (filter). I don't think its a catastrophic hardware failure, since I am seeing something in the IF. At this point I feel that it's time to start investigating the step gains, linearity correction, IF offset, etc, but I would like to do so in an organized and rational manner. So, are there any massive holes in my reasoning, areas that I've missed, or anything like that? If not, how would you suggest I go about tuning the YTX and its supporting circuitry? Guess and check is NOT an acceptable method, of course, but I don't really know where to begin. I should mention that my highest frequency signal source is an HP 8657A which tops out at 1040 MHz. I've been using my comb generator and/or splitting a 1.024 GHz signal and mixing it with itself to get higher frequencies for testing.

Thanks in advance for reading!

[ignilux]

Update:

To further narrow the scope of my troubleshooting, I decided to follow an approach similar to that taken by the OP in this thread in the HP/Agilent Groups.io... group. I turned on the analyzer, set the start frequency to 2 GHz, and stop frequency to 2.6 GHz to make sure I was in the first harmonic band. I then keyed 'Shift > t' to lock the analyzer to that band. I switched to zero span with a center frequency of 2 GHz, and monitored the IF output on a Tek 385 while feeding a signal into the RF input of the 8566B. As expected, a 2 GHz input produced no IF output. However, when tuned to about 1898 MHz, a faint signal appeared on the scope.

So, it looks like the mixer and filter are working, but the tuning is way off somewhere. Reading through the test and adjustments manual, it became clear that I really had no chance tuning this thing with my haphazard signal generation. Of course, that just means that I get to drink some fresh TEA! I have an HP 8620C with 2-18 GHz plugin en route, which will cover the bands nicely. Wish me luck...