Systron Donner Spectrum Analyzer Type 809-1

Actually, 809-1 is the designation of the RF section on the right. The manual describes it as a model 762-1. Alternatively it can be referred to as ‘Microwave Spectrum Analyzer type AN/USM-394’ which I believe is its US Military designation.

This piece of hardware has been in my possession since 1985, when my boss at the time said to me “If  you can put the boot of your car between the back door and the skip, I’ll give you a hand with the old analyser”. At the time it was probably 10 or 12 years old and was technically MoD property, having been purchased against a one-time top-secret contract. I first came in contact with it in 1979 when its sole purpose was to monitor 2 signals in the region of 9GHz and to verify that one was sweeping across the other. This task was done inside an anechoic chamber with the lights out because the trace intensity was so poor. By 1985 the test method had been refined and the analyzer consigned to storage since it was MoD property. At the time there was an MoD policy which deemed items of test equipment over 10 years old to be classed as ‘end of life’. The thinking behind this policy was based on the premise that they would become uneconomical to service or repair; thus they were disposed of and/or replaced. It was the belief of many at the time that this old analyser was faulty anyway … the fact that it had to be viewed in a darkened room. With the analyzer safely in my car I then paid a visit to the Calibration Lab to retrieve the service manual and showed the technician the paperwork which proved that the analyzer had been scrapped. Back home and with the help of the manual, I found an internal adjustment that turned up the volts on the CRT and the trace has been nice and bright ever since!

The analyzer came in very handy when I modified an Injection-Locked 9.6GHz oscillator for 10.224GHz. The fact that I had done this at home intrigued my boss who asked how I had managed it. When I told him that I had used the old Systron-Donner analyzer he was even more intrigued since he had assumed that I had stripped it for parts. He was even more surprised to learn that all that had been required to get it into working status was to turn up the HT. I think at that point he may have uttered something about the competence of the Cal-Lab technician.

Then about 2 years ago (2013) I noticed that it was ‘deaf’. I was working on a 10GHz transverter for a friend and discovered that the sensitivity of the analyzer had dropped to the extent that it was unusable. It then gathered dust in my radio room for about a year before being ‘shunted’ out into the hall-way to sit alongside other miscellaneous non-working items. Earlier this week (March 2015), I decided to embark upon a systematic approach to identifying the problem.

What sets this analyzer apart from others of its time is the fact that its frequency coverage is so wide. It covers 10MHz thru’ 12.4GHz in 5 ranges without the need to change any ‘plug-Ins’. (The model 809-2 increases this range up to 40GHz in 3 more ranges employing external wave-guide mixers). Perhaps the most popular Spectrum Analyzer of the early 1980s was Hewlett Packard’s one based around the 141T display, requiring a range of RF and IF plug-ins to give a similar range of coverage. The HP equipment with its calibrated display can be used to make relatively accurate power measurements. In contrast, the Systron-Donner analyzer cannot be used in this way since it is not possible to set up a reference level on the display. In essence, only relative measurements can be made. Thus the Systron-Donner spectrum analyzer is more accurately described as a ‘field’ instrument rather than a laboratory instrument. Credit must however be given to Systron-Donner for coming up with an instrument with such a wide frequency coverage in a single unit and keeping it relatively small (smaller than the 141T mainframe) … this was over 40 years ago! … mid 1970s.

But my analyzer was now faulty … What had failed? Essentially everything appeared to be working … it was just VERY deaf. When you pull the 809-1 plug-in out it can appear a bit daunting since the RF-unit is fully enclosed. It is only when you turn it upside down that you read the dismantling instructions. Essentially, the RF-unit splits into 2 parts; the RF-deck and the IF-section. The first thing I checked was the input attenuator since I already knew that one of the positions was u/s. Whilst I did confirm a ‘dead’ element in the step-attenuator, this was not the cause of the chronic deafness.

Turning my attention away from the complicated RF stuff, I investigated the power supplies located on the underside of the display section. Everything was in order, although the 100uF capacitor on the 200V line appeared to have sprouted a ‘wort’ (see below).

Not having a 100uF capacitor to hand, I fitted a high-spec 68uF one instead, with no obvious detrimental consequences. I did find a completely u/s 470uF tantalum (I think) capacitor on the Sweep and Vertical Amplifier board. One lead had completely corroded away! I replaced it with a modern aluminium type. This capacitor could well have failed a long time ago. It is the timing capacitor responsible for the slowest horizontal sweep and not a setting that I have ever used. It was partially hidden by the small timing board in the photograph below right. Having ascertained that all the power supplies were ‘good’, it was now time to investigate the RF-Unit.

Ask anyone who has experience with RF Spectrum Analyzers and they will tell you that the worst-case scenario failure is a failed First-Mixer. In Hewlett-Packard units such as the the HP 8555A (mentioned earlier) the diode is bonded to a ceramic substrate and locked away in a sealed box. Systron-Donner employed a completely different style of mixer assembly. In their case, the diode itself is a replaceable component accessible from outside the unit. Thus it is easily replaced if blown. The diode is accessible through a hole in the unit side-wall … see photograph below.

But it wasn’t the mixer-diode that was faulty. Not only was the analyzer deaf, but there appeared to be a frequency discrepancy. Thus I concluded that one of the two Yig-Oscillators might be off frequency … and if off frequency, they might also be delivering a lower than normal output. That was the notion anyway … hmmm?

The service manual contains a very comprehensive fault diagnosis series of flow-charts covering no less than six A3-size fold-out pages. However there is a problem … whoever wrote them used the 809-2 with its external mixer drive socket in mind. So when I got to the bit where it tells me to measure the level of first LO drive, I don’t have the appropriate front-panel connector. My only option was to measure the level somewhere inside the RF-deck. This introduced another issue, in that the RF-unit needs to be inside the main assembly since I don’t have the necessary extender cable … not to mention that running the RF-unit ‘split’ would be a near impossibility anyway. However, as Baldric would (almost) say, ‘I had a cunning plan’. Removing the input step-attenuator was relatively easy. This left two good size holes in the front panel since the RF-Input connector is integral with the attenuator. It was then simply a case of attaching one end of an SMA lead to the output of the second LO Yig or in the case of the first LO Yig, the associated isolator and feeding the other end out through the hole where the N-type connector once fitted. Off course the whole assembly needs to be disassembled and reassembled each time.

The third and fourth IFs are 60MHz and 6.5MHz respectively. The non-availability of any form of extender meant that chasing a fault on one of the boards in the IF-section was going to be a nightmare, so I still held out hope that the fault was still on the RF-deck. The first and second IFs are on the RF-deck. The frequency of the first IF is dependant on what band is selected, so that ruled that one out, leaving only the second IF which is 260MHz.

I had a brand-new Watkins-Johnson gain-block which I connected up to my Network Analyzer and verified the gain to be around 13dB. The supply voltage was the same (+15V), so wiring it into the RF-deck was not going to be a problem. However the W-J amp used SMA connectors so I had to be creative in connecting up the output! BINGO! It worked … Or more precisely, the Spectrum Analyzer now worked. I had a nice ‘fat’ trace on the display for +10dBm at 400MHz with 30dB of input attenuation. The photograph on the left shows the ‘belt and braces’ lash-up, but it worked. I could have left it like it was and lived with it, but I suspected that the failed amplifier might be an Avantek MMIC and I was sure I had one tucked away somewhere. I put it on my VNWA (Vector Network Analyzer) and confirmed that it was now an attenuator (about 30dB). It was an easy matter getting the PCB out of the enclosure. I was right, it was indeed an Avantek GPD-462. See the above photograph. Note the corrosion around the base of the can. I found my ‘spare’, a GPD-405 which is essentially a beefed up 462, capable of delivering higher power. I cleaned up the tiny PCB and fitted the GPD-405 only to discover that the can was slightly ‘taller’ … so it was necessary to drill a hole right through the enclosure to allow it to fit. See the photograph below.