While working on the design of my first receiving loop wideband amplifier, I decided I needed some kind of RF generator to do bandwidth tests. So I went to Ebay and did a search and stumbled on a familiar sight – the Wavetek 3000. And it was cheap too. One was only $120. It was described as “working but untested” which I took to mean, that at least one of the front panel lights come on when power is applied. This unit can generate RF from about 500 kHz to 500 MHz, and during its time it was known for its spectral purity and stability. It has a true digital synthesis design, but is designed using a mixture of discrete transistors and basic TTL IC’s. So, I bought it! I figured, what’s the worst case here? I’m a qualified electronics engineer and should be well able to fix anything that’s wrong with it. Optimism never hurts I guess, but of course, when it arrived it was fairly useless — on some settings it did indeed produce a waveform, but highly distorted and nowhere close to the frequency dialed in on the front panel.
To make a long story shorter, I went through it and managed to get it working quite well. I’ll summarize the work I did below, but before getting into the details, here’s a photo of it on my shelf in my lab. It looks good there.
Construction of the unit is a wonderful example of discrete RF design, with most circuitry built by hand using point-to-point techniques in hard tin “cans”. I thought it was handy that Wavetek printed the can interconnect diagram right on the inside of the top cover:
After doing a rough diagnostic on the unit using its schematic, I found that the signal disappeared at the “Output amplifier” can designated as M10W. I pulled out that can, and found inside it a circuit with an obviously cracked and dried-out 10 uF aluminum electrolytic capacitor (C32). Looking further, I found others which were either badly reduced in capacitance or were leaky. In addition, to my horror, the IC’s were socketed, with the point to point wiring done to the socket itself. So, I decided to go through the entire unit and do two things: (1) replace all electrolytic capacitors, even the tantalum ones, and (2) unseat and reseat each socketed IC, to abrade the pins and hopefully reduce the risks of loss of contact due to corrosion.
The cans have pins on the bottom, mostly used for power connections, and are connected at the top by a network of SMC cables. At this point I want to remind everyone that there are 3 types of miniature RF connectors: the very familiar SMA connectors which are commonly used today, the lance-fit SMB connectors, and the smaller and rarer SMC connectors. While working on the unit, I found a frayed SMC cable and had to order a replacement from an Ebay supplier who made one to order. This photo shows the tops of some of the cans, and their SMC cables.
Here’s a look at some of the capacitors I removed and replaced:
Below, a look at a sample can, the M29 module. This shows the point to point and socketed IC construction techniques. At the bottom left is an original 10 microfarad capacitor which I later replaced. Additional wiring is on the other side. In many modules, a sub-can consisting of smaller tin boxes or walled-off sections are used. The Wavetek 3000 is a wonderful lesson in how to push the performance of point to point discrete component design right to its limit.
Below is another example module, M33a, showing some of the sub-can construction. Note the two tiny component areas at top right. These were all hand wired of course. How long do you think it must have taken for the assembler? This kind of construction requires a high skill level of the assembler. These skilled point to point assemblers undoubtedly lost their jobs as this kind of assembly converted to PCB, ASICs, and offshore suppliers. So this is a relic of a different age. At the lower edge (red arrow) you can see a place where I hand wired a replacement 100 uF capacitor, using a surface mount component, plus leads.
Below is my M32A unit. You can see that the MC4044 and the Eprom both have date codes of 1982, but someone much later put new 74S196 IC’s in the sockets because they have data codes of 1990. Many of the linear IC’s have date codes of 1978.
To test the repaired unit, I used my ham radio (FTDX5000) to evaluate the accuracy and spectral purity of the sine wave at various frequencies. The accuracy is excellent — it is within about 150 Hz of the exact frequency on most bands. The amplitude stability is good too, though some noise is still visible when using a scope. Below, I’m checking the noise level with my oscilloscope on the 10 mv scale, comparing a channel looking at ground (red) with one looking at the signal (blue) — the noise seems nearly the same amplitude, so the Wavetek is not contributing a large amount of noise. (Scope: Picoscope MSO 5244D.)
There is some amount of phase noise, causing the frequency to contain a FM signal with about a 100 Hz width. I think this might mean that there is still a problem. But as is, the unit is perfect for sensitivity and bandwidth measurements, and I’m really happy with the result, and I’ll be using it when I do my redesign of the wide bandwidth amplifier for the receiving loop antenna.
My thanks to VK3ZZC, Ralph in Australia, who gave me troubleshooting advice by email. Here is his experience:
In response to your question, nearly every single black tubular electrolytic either was a dead short , or quickly became one once power was applied. The truly wonderful news is that a couple of them are hiding inside tin cans…inside tin cans…all lovingly soldered shut. The part of the implementation that truly caused me to ask , “how could anyone be so daft”, was discovering inside 3 nested tins one of the divider chains, counter ICs in sockets ! At least it was a proper little pcb…but sockets…in a space that was utterly unreachable without destroying the unit. As you know, sockets are always totally reliable. It was this that brought about the death of my Wavetek. It had a stable, noise free output…utterly unrelated to the thumbwheel setting !
I do not understand their fixation on the horribly expensive point to point wiring, in an era where PCB’s were a well established technique. Just look at equivalent contemporary products from HP.
The one other thing you must do, check the +/- 18V regulators. The series pass transistors will fail as a short. Pre-emptively wire in some sacrificial zener diodes to protect the rest of the unit.
I still think that the electronic design of the wavetek 3000 is sublime. The implementation ,however, was a farce.
If you have a Wavetek 3000 I’d be glad to help you out if I can. Leave me an email. Here are the two service manuals I located on the Internet: