Agilent 8753ES Repair Part 3 – A21 Test Port Coupler

Continuing on from my previous efforts – there is still a problem with this instrument.

This was apparent from looking at the S11 “Open” measurement. It should be a flat line, but it’s not.

I was able to isolate the problem in just minutes by swapping over the input couplers (A.K.A. bridges). I left the good bridge in the port 1 position.

Two bad parts found in this instrument. Both marked with “U/S”, whatever that means. Likely written on there by a technician at the behest of the hapless engineer who damaged it. I think it’s a bit unlikely that this instrument had both a dead CPU board and these parts faulty. More likely that it is a liquidated “bad parts” unit. A common thing to find the the hands of companies who own multiple. I pushed the seller very hard on the price of this unit because I had anticipated this.

Full part number of the bridge of this unit 5087-7007. As with the last repair, there is an EEVBlog thread extensively covering this repair, with help from its creator Joel Dunsmore. This page is a summary of my experiences. There is a lot more detail in that thread.

The bridge is a similar construction to the A24 transfer switch I covered previously. An exotic proprietary hybrid. At first I was rather puzzled as to what could be wrong with it, as there is no silicon in here.

Another contributor to the previously linked post provided this article published in 1991, by, guess who, describing how the bridge works.

The schematic of Joel’s bridge. From his 1991 article. Note that C1 is 10,000pF (10nF) in the 5087-7007 bridge.

The key components of interest for this repair are R1, C1 and R2. But where are these components?

R1 isn’t visible here, but it’s an extremely thin film resistor layered directly on the substrate under the goop holding down the piece of absorber material.

Because R1 is so thin (and it has to be) it’s likely to be burned out by an RF overload on the input, so that is how this thing died.

The original repair proposed was to simply leave the burned out husk of R1 in-place, and try and tack an 0603 resistor between the end of the coax and the top end of C1. I attempted this many times but was unable to. Whatever material those were coated in did not stick to solder. I tried scraping them with a scalpel, adding flux. Nothing worked.

So I gave up on that, scraped the whole lot out and replaced it all with new components like this:

The first repair using 0603 components. Later revised to use an 0201 resistor and 0402 capacitor with negligible improvement.

This immediately returned the bridge to a functional state but with reduced performance. When I pulled the original capacitor off I unfortunately ended up tearing a bit of the stripline off, but it didn’t matter because the new capacitor bridged that gap quite nicely. Note that I used 0603 sized components because that’s what I had to hand. a 267Ω resistor was instead of the original 265. According to Joel it doesn’t matter too much.

The performance degrades because the replacement resistor inevitably is going to be a lot thicker and wider than the original, increasing the parasitic inductance on the main transmissions line, in turn reducing the directivity.

Performance of the repaired bridge?

To test the repair I’m using an Americon APC7 to N adapter with an Amphenol N terminator.

The working un-repaired bridge on port 1. S11/Load. It’s very good as we would expect.
The repaired bridge on port 2. It’s less good, but not too bad.

Looking at S22 of the repaired bridge there are some distinct differences. Firstly at the lower frequencies, we see some near zero / negative return loss, quite a puzzling phenomenon. Joel explained this is a frequency range where the impedance of the load exactly matches the impedance of the bridge, resulting in negative return loss which literally and figuratively means nothing. This would be due to my modifications, which have altered it a little bit, hence the difference to the un-repaired bridge.

The significant difference is at the top end of the scale at 3 GHz, where directivity has been lost to the tune of 10dB. Note that that it is still quite good at 2.4 GHz, just touching 30 dB (these figures can’t be entirely trusted as I don’t have a precision load to test against) but they’ll be somewhere close. Bear in mind that even after my repair the performance at 3 GHz is still significantly better than anything the average hobbyist is likely to possess.

From this we can see this kind of repair is more applicable to 3 GHz instruments than 6 GHz versions as the degradation would be a lot worse at these higher frequencies.

In practical usage, the performance degradation is negligible, both because it barely affects measurement at all, certainly to the 3 GHz the instrument stretches to and because it’s only going to affect S22. Measuring both S11 and S22 simultaneously is a rare (and improbable in the case of a hobbyist like my self) requirement. I can always swap the cables over if I don’t feel I can trust S22.

Practical lessons learned

  1. Once a bridge is damaged in this way, original performance is gone for good. It is not possible for an individual equipped with typical rework tools to re-create the microscopic substrate resistor R1 which is lost during overload.
  2. Don’t touch anything you don’t have to. The insides of these bridges are extremely fragile. On mine I broke the connection to the input while trying to clean out flux with a cotton bud, and I also broke the connection from the coax to the stripline probably doing the same thing. These of course were easy to repair with a dab of solder, but every such modification skews the performance of the part a little bit.
  3. If removing components like the capacitor, don’t rip it out, try extracting it with a soldering iron at each end. I did rip mine out and lost some of the material on the stripline. It wasn’t a big deal in the end but in hindsight I wish I had been more careful.
  4. Mask the striplines. While working on this I put pieces of kapton tape everywhere I didn’t want solder to go. Once you get solder on that gold, it’s not coming off.
  5. Don’t use solder braid. It’ll mop up the solder and the gold stripline which will have mixed with the solder upon application.
  6. Use smaller components where possible. I wasn’t able to get much improvement by doing so but it certainly can’t hurt.
  7. There’s no point in wasting money on exotic components, unless you have equally exotic ways of connecting them. Hypothetically an epoxy bonded flip-chip RF resistor of the correct value may perform better than the repair I’ve performed, but not much better, and for the extra cost and difficulty, probably not worth it.
  8. Don’t re-assemble / re-connect for testing unless the DC resistances test out OK. Check for 0Ω from the input centre pin to the biasing pin on the rear underside, to check that the stipline hasn’t been broken. Check for 267Ω from the input centre pin to C1. Also check for 50Ω across the receiver output connector.
  9. Clean clean clean. The biggest gains in directivity I was able to achieve were not by using smaller components, but instead by cleaning flux and debris remaining from the re-work.
  10. “Don’t F with it”, said by the man himself. Once you’ve got it back to an acceptable level of performance, walk away. There’s no point in going back in again and again like I did, because each time you do, you make it worse.

Other repair options

If a repair like the one described on this page isn’t practical, trying to obtain a replacement 5087-7007 bridge may not be the best idea as there are other options which may be cheaper or more desirable.

5086-7489 Bridge from the 8753D

In the aforementioned thread it was disclosed to us that there is only a single difference between this bridge and the newer 5087-7007 which is that the larger DC blocking capacitor at the rear of the bridge was increased in value from 0.047μF (300 KHz spec) to 1.2μF (30 KHz spec). Replacing this capacitor would of course need all of the care taken to perform the repair I described on this page.

5087-7054 Bridge from the 8753ES-H39

The 8753ES has an oddball 3-port variant (H39) which uses bridges featuring an N connector. The drawback being the instrument has no biasing capability, thus the three biasing connections on the underside are not present, but at least the bridge its self has the same dimensions and screw pattern.

5087-7050 Bridge

Not cheap or common but appears to have N connectors and biasing capability.

5086-7488 “R channel coupler”

Another interesting looking bridge with the same screw pattern and dimensions as the original, instead with SMA connector on the input. Also no biasing as with the 5087-7054. Probably 300 KHz spec.

Raid an 85047A test set for parts

At the time of writing, one of these test sets sells for a lot less than two bridges, and an A24 switch, all of which are contained within, but these’ll of course all be for 300KHz minimum operation.

One thought on “Agilent 8753ES Repair Part 3 – A21 Test Port Coupler

  1. Thanks for sharing this insight. I’ve not been brave enough to open this in the past, but after your post I might have to give it a try.

Leave a Reply

Your email address will not be published. Required fields are marked *