Anritsu MH680A1 tracking generator
Posted: Tue Jun 16, 2026 5:04 pm
I volunteered to do some work for a community radio station that site shares at a fairly well populated Arqiva site in Surrey and one of the things I'd need to do was take some measurements that would require a spectrum analyser with a tracking generator. I have two Anritsu MS2601A analysers, they're very repairable compared to anything HP made, one is ex-Ofcom, too, for added authenticity (Ofcom originally commissioned the transmitter I'll be looking at, possibly when they were still called the Radio Authority).
I'd also need some notch and bandstop filters, as I need to show that the new transmitter is playing well with all the Arqiva kit up there, so I made a notch filter, tested it and found a bandstop that I'd made earlier. It was when I was testing this that I saw the trace vanish into the noise on my analyser. I power cycled everything and the same thing happened, sometimes the signal returned, sometimes it didn't. It wasn't consistent. Sometimes the amplitude of the tracking generator output dropped by 30dB but it kept working, sometimes it stopped completely.
I wasn't sure what could cause this but after a power supply check I started by investigating the PLL in the tracking generator that locks its own 2.5214GHz oscillator to the 2.5GHz fixed LO on the analyser. This oscillator is a varicap tuned cavity with >28V of tuning voltage swing and it locks to a mix of 21.4MHz from a VCXO (for manual fine tuning the tracking) and the incoming 2.5GHz from the analyser. The 21.4MHz out of the mixer goes to a surprisingly archaic MC4044 phase comparator, then op-amp loop filter and non-inverting buffer. Periodically this was failing to lock and swinging the tuning wide, hitting the buffer's buffers, a maximum voltage of 14.4V (it's bipolar, so this is about half the tuning range).
After a distraction caused by misreading the block diagram, and a fruitless investigation into the ALC, I found I'd got it working again but this was a trick and I suspected it was a trick. I left it an hour or so, returned and it was broken again. That suggested that it was another of my least favourite thermal issues, though hopefully not a 1N4148 again.
I decided to break the loop of the PLL and give it a fixed tuning voltage to see if the oscillator or the 74LS390 divider chain on the VCO side misbehaved. It did, the oscillator amplitude was varying wildly and sometimes the output vanished entirely. When it was completely cold it generally worked, but only for a few seconds.
There's no schematic for the VCO, it's a blob of surface mounted components built in 3D with a silver coupling loop around it, suspended by a metal tube inside the cavity. Power is fed in through feedthrough capacitors. Coupling can be changed by rotating the tube.
This 3D assemblage of components had what was once expanded foam, presumably to dampen any microphonic tendencies, unfortunately it had done The Thing and turned solid, crumbly and caustic. This mess had eaten through a solder joint, so I fixed that and reassembled it.
I wasn't seeing any oscillation after reassembly, but a bit of tuning and monitoring the VCO voltage quickly brought it back - but there was a problem - no output. I nearly panicked at this point, wondering if I'd broken something else. The oscillator was working, I could see the PLL locking and the voltage moving as I'd expect it to as I adjusted the coarse tuning screw.
The problem was simpler, I'd locked it to a harmonic of 21.4MHz rather than the fundamental, there's a very ample limiting amplifier that made this really easy. Further coarse adjustment brought it all back as it should be and I'm rather relieved.
Now wondering if I should get a more modern spectrum analyser.
I'd also need some notch and bandstop filters, as I need to show that the new transmitter is playing well with all the Arqiva kit up there, so I made a notch filter, tested it and found a bandstop that I'd made earlier. It was when I was testing this that I saw the trace vanish into the noise on my analyser. I power cycled everything and the same thing happened, sometimes the signal returned, sometimes it didn't. It wasn't consistent. Sometimes the amplitude of the tracking generator output dropped by 30dB but it kept working, sometimes it stopped completely.
I wasn't sure what could cause this but after a power supply check I started by investigating the PLL in the tracking generator that locks its own 2.5214GHz oscillator to the 2.5GHz fixed LO on the analyser. This oscillator is a varicap tuned cavity with >28V of tuning voltage swing and it locks to a mix of 21.4MHz from a VCXO (for manual fine tuning the tracking) and the incoming 2.5GHz from the analyser. The 21.4MHz out of the mixer goes to a surprisingly archaic MC4044 phase comparator, then op-amp loop filter and non-inverting buffer. Periodically this was failing to lock and swinging the tuning wide, hitting the buffer's buffers, a maximum voltage of 14.4V (it's bipolar, so this is about half the tuning range).
After a distraction caused by misreading the block diagram, and a fruitless investigation into the ALC, I found I'd got it working again but this was a trick and I suspected it was a trick. I left it an hour or so, returned and it was broken again. That suggested that it was another of my least favourite thermal issues, though hopefully not a 1N4148 again.
I decided to break the loop of the PLL and give it a fixed tuning voltage to see if the oscillator or the 74LS390 divider chain on the VCO side misbehaved. It did, the oscillator amplitude was varying wildly and sometimes the output vanished entirely. When it was completely cold it generally worked, but only for a few seconds.
There's no schematic for the VCO, it's a blob of surface mounted components built in 3D with a silver coupling loop around it, suspended by a metal tube inside the cavity. Power is fed in through feedthrough capacitors. Coupling can be changed by rotating the tube.
This 3D assemblage of components had what was once expanded foam, presumably to dampen any microphonic tendencies, unfortunately it had done The Thing and turned solid, crumbly and caustic. This mess had eaten through a solder joint, so I fixed that and reassembled it.
I wasn't seeing any oscillation after reassembly, but a bit of tuning and monitoring the VCO voltage quickly brought it back - but there was a problem - no output. I nearly panicked at this point, wondering if I'd broken something else. The oscillator was working, I could see the PLL locking and the voltage moving as I'd expect it to as I adjusted the coarse tuning screw.
The problem was simpler, I'd locked it to a harmonic of 21.4MHz rather than the fundamental, there's a very ample limiting amplifier that made this really easy. Further coarse adjustment brought it all back as it should be and I'm rather relieved.
Now wondering if I should get a more modern spectrum analyser.