Agilent 8753ES Repair Part 2 – A24 Transfer Switch

Following on from a previous post about this instrument, I discover a second fault.


It became apparent quite quickly when looking at S21 and S12 of a piece of wire with a ferrite bead on it, a symmetrical structure. The resulting traces should be identical, but they are not.

S21 is correct, but what’s going on with S12???

Unlike in my previous post were a significant effort was mounted to track down the problem, this time it’s not going to be so difficult. First thing to check was the output level of port 2 while the instrument is measuring S22. Could this crazy looking zig-zag response be a result of uneven output?


I plugged port 2 into my spectrum analyser, slowed the sweep on the 8753ES down to about two seconds (so I can watch it from end to end) only to find that it’s a perfectly flat 0 dBm across the entire sweep. The issue must be with port 1.

I then set the instrument to measure S11, with the spectrum analyser attached to port 1 with the same two second sweep, once again measuring the output on my spectrum analyser, and I find that it’s attenuated by 20-30 dB. This is bad.

Looking at the block diagram of the instrument in the service manual, we can see that a likely culprit is the A24 transfer switch, whose exact part number is 5087-7067.

This part is clearly marked “STATIC SENSITIVE” thus I would not be surprised to find that it is at fault. I pulled it out of the instrument, hooked it up to a test supply & spectrum analyser and observed the same attenuation across it as was observed on port 1’s output. Fault located.

The A24 transfer switch as found. I’d bet that “U/S” was written on there by the twit that damaged it, or perhaps someone working for them. Unserviceable? Not the first time I’ll be finding this marking in the instrument as we’ll see later on.

I had a quick look on eBay to see what the availability of this part is, and it’s bad news. It’s not quite as expensive as the A9 CPU board I was previously facing the prospect of having to buy, but it’s still very pricey. Worrying me further is that there are many dead units for sale, with port 1 marked as “BAD”, just the same as mine. One seller is describing them as “For repair”?

The warning in the yellow box spells out the limits of the A24 transfer switch which apparently the previous owner of this instrument didn’t take the time to read.

Realistically, could an ordinary person like myself repair one of these? Let’s open it up and have a look:

The components we’re looking at here are all control and power supply, none of which’ll be at fault. The faulty bit is inside the gold block – a proprietary hybrid. These usually aren’t even repairable by highly experienced, well equipped technicians as the dies within are not obtainable.

Rather than roll the dice on a very expensive second hand part off eBay, I’d rather see if I can make a replacement for this part. I’ll be detailing that effort in a separate post.


I started to put together a design based on what I could possibly assume about the existing switch, but feeling a bit unsure I decided to put the question to the EEVBlog forums and in an almost improbable turn of events, my question attracts a response from industry grandee Joel Dunsmore, the designer of the A24 transfer switch and quite a few other parts in this instrument. In a handful of posts – specifications, layout, design challenges are all revealed. He also made a suggestion about how I could return mine to life.

So I’m going to do just that.

MINIATURE GOLD PLATED WORLD: Key components of interest inside the hybrid. Here we can see how the 100 dB+ of isolation is achieved. Firstly, using two switches are used back to back, then, additionally the PIN diodes are biased to shunt the transmission line on the inactive path. There are also many bypass capacitors both inside this module and on the PCB for isolation at lower frequencies.

At the input there is a diode string to prevent the switch some protection against ESD and power overloads. If these diodes are blown, they’ll attenuate the input as I’m seeing on mine. The theory is that if I simply pull this component out, the switch will be usable again, albeit with no input protection.

My switch, with the input protection on port 1 removed

This technique successfully fixed the problem. I can no longer measure any loss through the switch, or to the termination when blocking. The big caveat of this solution is that port 1 is now extremely ESD sensitive. Beats being completely unusable.

Another problem???

Looking at the S11 “Open” measurement, something is still not right. The trace should be flat, but we’ve got some strange looking hump shape.

You guessed it. There’s another dead part in this instrument. Read the next repair story here.

Leave a Reply

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