Hi everybody! I recently bought an HP 8714ES VNA "as-is" off of eBay. It covers 300 kHz to 3 GHz, has an S-parameter test set, and includes 10base-T ethernet support. It's my first VNA. I was hoping it would be fully functional, but sadly, that was not the case.
In hindsight, the eBay listing showed evidence of a problem: the R trace was visible in a couple of the photos, and it was well below -40 dBm. It should be right around 0 dBm when the ports are left open.
Another problem: the packaging job was rather inadequate, so it arrived with a bent right handle and a dented rear-bottom-left corner. It was otherwise in good physical condition, and it fired right up when I gave it power. Unfortunately, it was non-functional as a VNA. The traces were all noisy and showed no response to different inputs.
Here's the bad R trace:
A quick run through the test procedure in the service manual isolated the problem to either the A4 Source Assembly or the A3 Ref/Frac-N assembly.
I opened up the case and checked the output of the Ref/Frac-N assembly. Both the 10 MHz signal and the fractional-n signal looked good, which meant the problem was with the source assembly.
The manual suggested it be replaced. Bah! I was going to fix it!
The source assembly has two outputs: the "source output" which eventually goes to the DUT and the "receiver LO output" (about 27.778 kHz above or below the source output) that goes to the receiver.
I checked both of those outputs from the source assembly on my spectrum analyzer, since my SA goes up to 2.9 GHz but my scope goes to only 300 MHz. I found a strong source output signal on the proper frequency. However, the receiver LO output signal was either not present (for CW frequencies below about 2 GHz) or on the completely wrong frequency (for CW frequencies above 2 GHz). Something appeared to be wrong with the receiver LO signal generator.
I took out the source assembly, removed the RF shields, and looked for obvious problems. Aside from some minor discoloration near a few 2 W resistors in the power supply circuits, and some flux residue from some long-past repair, there were no obvious issues. I would need to dig further.
The 8714ES A4 Source assembly with RF shields removed:
That presented a problem, since the source assembly connects to the rest of the VNA via a 96-pin connector on the backplane, deep in the chassis. HP made a service kit with an extender board so that the assemblies could be operated outside of the chassis, but I didn't have one of those, and the only one I could find was very expensive (thousands of dollars).
The service manual for the 8714ES has a block diagram but no schematics. Fortunately, HP posted the schematics / CLIP for the HP 8714B, which had an almost identical source assembly board. I mean, it's really, really, really similar. In fact, I haven't found any substantial differences between what's on the schematic and what's on the board.
I used the schematic to identify which of the 96 pins were actually used by the source assembly and came up with about 20ish lines, plus maybe a few more for redundant grounding and power. With that knowledge, I soldered jumper wires to the back of the connector on the backplane and plugged them in to the connector on the source assembly, which was by then sitting on top of the chassis. On the front side, I used some longer SMA-to-SMA cables to hook the source assembly to the receiver assembly. I powered up the unit and the source board came to life. I was then able to probe voltages on the board.
My hacked-together extender:
None of the rails were significantly out of range, so I got out my thermal camera to see if that would reveal anything.
A few chips were a little warm, but the only one that seemed to be unreasonably so was a quad-XOR in the receiver LO PLL circuit. Pretty sure that shouldn't be running over 170 degrees F!
Closer inspection revealed that somebody had soldered a jumper wire from ground to the output and one of the inputs of an unused gate on that package, leaving the other input to that gate floating. I'm almost sure that they were one pin off and meant to tie the two inputs to ground, leaving the output floating. The schematic was silent on the issue. However, the receiver LO PLL circuit is nearly identical to the source PLL circuit, so I had a convenient reference. Indeed, in the source version of the circuit, there was a PCB trace grounding the two inputs on that XOR gate and the output was left floating.
I moved the jumper wire over one position, grounding the two inputs on that unused gate, no longer connecting the output, and checked the performance again. Sadly, there was no change in the VNA's behavior -- but the chip was running much cooler (basically room temperature).
As an aside, look how hot those resistors are! The are part of the 5.2 V regulator, which is a Darlington emitter-follower with a zener. I think the resistors serve as current limiters, but I'm not sure. In any case, they were each dissipating about 1 W of power, which was comfortably inside their 2 W power rating. I'm not sure what their temperature rating is, but it's plausible that 450ish degrees is in range for that, too. It's also possible that the emissivity is a little off.
I moved on to checking signals with the scope. Eventually, I got to the 1 MHz reference and 1 MHz feedback in the PLL. In the source version of the circuit, they were both rock-solid at exactly 1 MHz. However, in the receiver LO version of the circuit, the 1 MHz feedback was actually more like 1.05 MHz. The divider upstream in the signal path wasn't the issue; for example, the divide-by-9 right before the 1.05 MHz was showing 9.45 MHz, and so on. In addition, the phase detector was clearly failing to lock.
I could tell I was getting close!
I probed around some more and soon realized that the VCO in that PLL circuit was receiving only about 5 V instead of the proper 15 V. Working backwards from there, I found the culprit: a bad transistor! The low-power BJT in a discrete Darlington pair wasn't turning on fully despite being driven appropriately, which was preventing the high-power BJT (a PZT2222A) from turning on fully, which was causing the VCO to be starved for power. The low-power BJT was a lowly 3904 in a SOT-23 package.
I shorted the base of the 2222A to +15 V, and the VNA jumped to life! Everything suddenly worked great. I checked for excessive current consumption through the base of the 2222A, found it to be reasonable, and concluded that the 3904 was bad.
The culprit: Q19
Although I had a bunch of 3904s in TO-92 packages, I didn't have any SOT-23s around, so I had to order one.
A couple days later, the new transistor arrived. I popped it in, and the VNA was happy as a clam.
Reassembled:
System info:
Measuring a stock BF-888S antenna:
A successful repair!
Home
Help
Search
Login
Register
