My HP 8447F RF Amplifier is Repaired! (Patreon)

I thought this repair was going to be trivial and a bit boring, so I didn't film it. However it predictably degenerated into the usual bunch of unexpected surprises, so here is the story in pictures. 

Remember that cute HP fellow from one of my previous posts? It's an HP 8447F, 100 kHz to 1.3 GHz RF preamp and power amplifier. I got it as an "extra" in an auction lot that I purchased for another HP instrument. But, wanted child or not, I still cleaned it and tested it as I would do for any wayward vintage instrument I get. The preamp worked, but I had to affix the ignominious sticker of non-workingness to the bottom power amp channel, which gave me no output.

It's too bad because the amplifier module that makes it tick looks downright gorgeous. It's a beautiful ceramic hybrid, looks like a piece of electronic jewelry.

Being no RF guru, I had asked Sharhiar over at the Signal Path for a hand on how to repair it. But I got no answer. So I decided the simplest and most cost effective "repair" was to replace the module with  a modern low cost RF amp. There are Chinese boards aplenty on ePay for less than $10, but they have no specs - you all know what that means. Instead, I opted for a $60 eval board for the new TI Silicon-Germanium wideband amplifier, with specs: 40MHz-4000 MHz, 18 dB gain, 20 dBm out, 19.5 dB output at -1dB compression. Close enough to the original HP amp without spending a fortune (the HP module specs are 0.1 to 1300 MHz, 20 dB gain, 20 dBm out, 16 dBm out at -1dB compression). I reasoned that if I used it only up to 1,300 MHz, I might be able to get instrument grade performance out of it. 

And it fits right in, it's the little red board at the bottom.

Except it's powered by 5V only (miracle of modern day RF electronics), instead of the 20V of the original. So I added a miniature DC-DC converter. Another miracle of modern electronics, 95% efficient and good for 500mA for a mighty $7. 

I still made sure all the mods I made were reversible, so I could get back to the original config if I ever repaired the original RF amp.

Confident in my oh-so-simple repair, I reassembled the whole instrument, connected it to the VNA, turned it on and... NOTHING! Signal 40-50 dB down instead of being 20 dB up!

How could I have messed up such a simple repair? I checked the 5V, or if I had reversed a connection. Nope, all looked good. What else could go wrong? The indestructible, semi-rigid RF cables to the front panel? And if that was true, could my HP amplifier still be good after all? 

I was suddenly overtaken by electro-guilt. I quickly wired up the HP amp up to VNA, gave it the required 20V. And voila, 20 dB of picture perfect HP gain, flat as a pancake to 1300MHz (and most of the way up to 2 GHz while we were at it). It was working all along!

So I took the instrument apart, extracted the two semi-rigid SMA to BNC coaxes. The first one (input) looked OK. But the output one did not even register: below 50 dB of loss, blank screen on the VNA.

What can possibly go wrong with a sturdy, rigid, passive assembly that is entirely protected inside the instrument? A quick look in the bino at the BNC connector revealed the problem - there was no center contact! Nothing, gone. Just high quality ceramic. Not a connecting connector anymore.

I opened up the good coax and the bad coax, and you can see how the center contact is missing from the one of the left. But fortunately, the center coax conductor is still there - the contact must just have de-soldered or broken off and fallen away. And that's likely why the  instrument was junked.

Luckily, I had gotten a whole bunch of semi-rigid RF cables from Patreon John Lawson at our last equipment exchange (see older post below), which I just stashed away as spare parts. And now that came in very handy. I did not have one with the right length nor a BNC that would fit in the HP chassis, but I just stole the center contact from said BNC and re-soldered it onto my HP coax. As good as new:

The repaired coax assembly gets tested on the VNA. Now that looks a lot better!

I first put the coaxes back in the instrument with the TI amp still in place, to see if it was better than the vintage HP. But it was not:

Only 18 dB of gain (as advertised), big 10 dB dip around 1.3 GHz, not very clean. Does not go down to 100 kHz either (as advertised too). Now, mind you, I'm comparing apple to oranges. This is a $60 eval part on a PCB. And it goes to 4GHz and power consumption is quite low. Impressive, really! Maybe if it were mounted on a well designed ceramic substrate with high quality connectors in a metal enclosure like the HP, it would look perfect. I could also have gotten a $500 Mini-Circuit or NAGRA amp module (which are exactly that: modern MMICs mounted on properly designed ceramics with nice connectors and metal enclosures), and that would have outperformed the HP in noise, gain, power, bandwidth and efficiency. 

Anyhow, the original HP amp went back in, and the instrument looks again as Bill and Dave intended:

And it performs as intended too:

There is some extra ripple that was not present in my previous measurement outside of the instrument. It must be from the semi-rigid cables. Likely a limitation of the BNC connectors. But still very awesome. There is a version of this instrument with N connectors that is probably even better.

Anyhow, I can now remove the sticker of infamy. One more HP instrument saved!

As always, thanks for the support, and for those in Northern America, have a good Thanksgiving!

Marc