Protecting a modern spectrum analyser from voltages in valve equipment

The cost of a spectrum analyser can run into tens of thousands of pounds and modern types often have quite a modest limit on the voltage at the input socket. My analyser is the DSA815 and it has a limit specified at 50 volts DC and a miniscule 2.25 volts of CW RF and is therefore really unsuitable for use with equipment using valves where the typical HT voltage can be anything from 200 to 350 volts. Of course a transmitter can have even greater voltages present and it's prudent, even if one intends to restrict measurements to points carrying low voltages, to prevent inadvertent destruction of the analyser front end. It's also worth considering other avenues through which high voltages can get into the input such as connecting to a mains operated radio chassis. Of course it would be foolhardy to experiment with an AC/DC equipment without employing a proper isolation transformer. I say "proper" as there are not only isolation transformers around, but also autotransformers which at first sight appear to be similar to isolation transformers but do quite definitely not provide any isolation. One form of autotransformer might be a device for converting UK mains voltage to US mains voltage. If you use one of these then under most circumstances you will also need an isolation transformer. Even if the set being worked on has a proper mains transformer it may have only a 2-wire mains connection. Because of this the chassis will not be earthed and can easily assume a high voltage with respect to the analyser chassis. This can arise from decoupling capacitors connecting a mains input to the chassis and can often be felt as a rubbery sensation through ones fingers. Also present on an old equipment could be an electrical fault such as a bad earthing point. In summary therefore, connecting a spectrum analyser to a mains powered equipment is fraught with risks if proper earthing is inadequate. See this article The prudent solution is to always employ a protection device and to this end one of the advantages of building test aids for isolation and voltage protection for a spectrum analyser is the fact you can readily check its performance. One can of course purchase protection devices, but these are often extremely pricey and, as most are very simple in concept, it is quite easy to make your own. Many amateurs and experimenters are not too concerned with the more esoteric aspects of measurements but mainly interested in relative readings such as determining the shape of a curve. Simple DIY test gear can be built which has really flat characteristics over typical ranges met in receiver or transmitter adjustment, but if you need to work in the realm of precise specifications then commercial test aids are there for purchase. I'll describe two items which I made very quickly after working out the mechanical details. The first is a DC blocker and the second an RF probe (in fact two of these). I made both in miniature diecast boxes. The purpose of what follows is to give experimenters a rough idea of what's required and not a article on their detailed construction.

DC Blocker This is a very simple device which uses two capacitors, a pair of zener diodes and a miniature choke.

The aim was twofold. First to block a DC voltage getting through and second to limit a spike getting into the spectrum analyser input. I used parts that were readily available so I built it into an old attenuator casing with BNC connectors already fitted. I chose two capacitors so that I could add a clamping circuit that would not interfere with either the circuit under test or the analyser input. I discovered that a pair of 0.22uF 400 volt capacitors worked well in terms of flatness of response. Two 18 volt zener diodes wired back-to-back limit the voltage nicely and to minimise the damping effect of these diodes I added a small choke from a scrap VCR board. The choke is not very critical and can be selected using the spectrum analyser to see its effect. I was concerned about the charging effects of connecting the DC blocker to a high voltage hence the addition of the zener diodes. Connecting these zeners back-to-back clamps both positive and negative spikes to ground and limits the level to around 18 volts which is very much greater than most monitored AC signals. In the design of the case are four bolts connecting the front and back end plates (removed for the photos).

High Impedance RF Probe See the improved Mk2 version The probe's chief purpose is to limit any DC voltage present at the measurement point to something much less than the spectrum analyser's input rating. It's actual amplification or attenuation is to a large extent unimportant as long as the spectrum analyser can cope with the output signal level. I must stress that this type of device is not for the purist, being designed for an amateur experimenter. It will be used mainly for measurements up to 50MHz.

The probe is made in a small diecast box and, as you can see, most of the space is taken by the 6 volt power supply.

The aim of the design, besides protecting the spectrum analyser from voltages met in valve equipment, was to minimise the effect of the measurement device on the circuit being measured. It's not a good idea to place even a few picofarads across a tuned circuit to observe twiddling effects as, when this is removed, the circuit will no longer be tuned as one desired. An ideal active device on which to base design is a FET (field effect transistor) as this will have a very high input impedance (meaning it will draw close to zero power from the circuit under measurement). A surface-mounted FET is best because its geometry will make it even less interactive with the measured circuit. In fact I found I could add a series resistor of 1.1 Megohms without affecting the overall design objective of the probe. This resistance effectively kills off any capacitive effect of the probe, but because, RF'wise, the resulting input circuit is a potentiometer it results in the voltage output of the probe being less than that being measured. Measurements show that the output of the probe with a 100mV input was something like a quarter to a half of this. The spectrum analyser showed that the output of its tracking generator at a level of 0dBm was about -20dBm to -30dBm at the probe output, but this was pretty much flat from 100KHz to 50MHz. Proof of the effect is that connecting the tracking generator directly to the 100pF input capacitor results in a probe output of just marginally less than 0dBM.

I drilled a large hole in the end of the box to minimise capacitance. Although a nice material to work with, the diecasting won't solder so I used stand-offs to hold the components in place and a solder tag screwed to the case for ground connections. You can just make out the black case of the FET which has four legs. The wire loop is the connection to gate 2. The drain of the FET is soldered to the stand-off carrying the orange wire which connects to the 6 volt power supply. The pale blue part just visible below the input capacitor is the 10Mohm resistor.

The above picture shows the FET amplifier constructed on four stand off posts. The black capacitor is a DC blocker connected to FET gate 1. The FET is a dual gate BF998 in a 4-pin package for which a brief spec follows:-. Drain-Source Voltage: 12.0V max. Drain Current: 30mA max. Total Power Dissipation: 200mW max. Forward Transfer Admittance: 24mS typ. Input Capacitance @ Gate 1: 2.1pF typ.. Reverse Transfer Capacitance (f = 1MHz): 25pF typ. Noise Figure (f = 800MHz): 1.0dB typ. Operating Junction Temperature: 150C max. The basic circuit is a unity voltage gain "cathode follower".The source (cathode) connection has a 47ohm resistor to ground and coupled via a 10nF capacitor and a short length of coax to the rear BNC socket. Gate 1 is biased, via a 10Mohm resistor, by 6.8kohm and 4.7kohm resistors which essentially determine the drain current. Gate 2 (the thin wire link near the orange wire) and the drain are connected to Vcc (6 volts). The chief practical consideration was how to power the attenuator. The smallest practical diecast box enabled me to fit four alkaline AAA size cells in holders, a small toggle switch and an LED as well as the FET and its circuitry. The LED hopefully will let me readily see if the probe is on or off. To reduce drain I'm using an 8.2kohm resistor to feed the LED.

Physically building the RF circuit was solved by using stand-off posts. By drawing the various parts on paper I was able to work out drilling details. I assume that anyone wishing to build an attenuator will be able to find suitable components. My only purchases were an FET (plus a spare if something went wrong), a diecast box and a pair of double battery holders.

Using a spectrum analyser with radio equipment designed for WW2 needs some thought. Firstly you need to consider the voltages, which in valve equipment can be around 250 to 350 volts and, because old capacitors can leak it's not good enough to assume you're measuring earthy levels, not to mention finger trouble such as accidentally touching an HT rail with a test probe. Secondly you need to consider the typical RF voltage you'll be looking at. Many oscilloscopes won't display very low signals and 10mV might be their limit, so trying to look at these will be tricky. A decent spectrum analyser should be OK for low level signal analysis but if you have a tracking generator you need to think about its RF level. For example a 0dBm output will be a milliwatt across 50 ohms. This works out at 225mV RMS or nearly quarter of a volt. A communications receiver will have a sensitivity down to a microvolt or less, so pushing quarter of a volt into its aerial socket will be a trifle high. My spectrum analyser tracking generator can be set to a minimum of -20dBm which still represents 22mV. Still a bit on the high side so I add an external 20dB attenuator which results in 2.2mV which is more in keeping with the receiver's sensitivity. Similarly, when testing an amplifier you should consider its maximum output as it's easy to overdrive an amplifier and produce all sorts of unwanted by-products. Below is a table showing the relationship between common power levels and voltages. Note the red numbers which could be easily met in an RF amplifier. The figures in red are above the recommended limit for the DSA815.

Setting in dBm  Power in 50 ohms  Volts RMS  Add 20dB of power  Add 30dB of power  Add 40dB of power  0  1mW  225mV  2.2V  6.6V  22V  -10  0.1mW  66mV  0.7V  2.2V  6.6V  -20  10uW  22mV  225mV  0.7V  2.2V  -30  1uW  6.6mV  66mV  225mV  0.7V  -40  0.1uW  2.2mV  22mV  66mV  225mV  -50  0.01uW  0.7mV  6.6mV  22mV  66mV  -60  0.001uW  220uV  2.2mV  6.6mV  22mV  -70  0.0001uW  70uV  0.7mV  2.2mV  6.6mV  -80  0.00001uW  22uV  220uV  0.7mV  2.2mV  -90  0.0000001uW  7uV  70uV  220uV  0.7mV  -100  0.00000001uW  2uV  22uV  66uV  220uV

Next, I'll describe the second version of the RF probe, and also go into some theory. This uses the same active device as the first probe, a BF998 FET, but several components are different. I improved the LF response by increasing coupling capacitors and made a few minor changes. Note that the probe is not a voltage amplifier; it's designed to monitor tiny RF signals in the presence of valve receiver HT rails. and to present approximately a 50 ohm impedance to my spectrum analyser. A major design feature is the extremely high impedance of the input circuit.

This picture shows the probe partly assembled. I'm using stand-off insulators on which to mount the various parts. The top stand-off is for the source pin of the BF998 and the lower one for connection to gate 1. The stand-off can be slid in and out of their nylon insulators to provide exactly the right amount of space in which to fit the transistor. These, and the input pin which is a chassis mounted feedthrough component are superglued to the case to prevent movement if the probe were to be dropped. The other three stand-offs provide mounting points for connections that are not especially sensitive to capacitance. The top one will connect to a short length of 50 ohm coax going to the rear BNC socket. The lower one is for Vcc which is the positive battery connection and the one on the lower right is for gate 1 bias connections. The battery ground connection is under the upper stand-off and a brass 6BA screw (top right) is used for a common circuit ground connection. Another method of construction could use a small piece of Veroboard. Surface-mount resistors and capacitors might help with improving linearity over frequency.

Now with the transistor mounted. The lead to gate 2 is temporarily connected to Vcc and in this configuration the battery drain was 31.5mA so I biased g2 to reduce this to 10mA but the performance was poor so I connected g2 through a 100Kohm resistor to Vcc and decoupled it to ground. During testing I tried setting gate 2 bias at a lower voltage to decrease drain current. This worked but for some odd reason the output dropped by around 10dB after something like a second. It was as if the input circuit was charging up before coming to rest. This effect vanished once gate 2 was set at the drain voltage. I later changed the 100nF input capacitor from a rating of 50 volts to one of 10nF and 400 volts. I may change this again to a surface-mount 100nF rated at 500 volts but I've mislaid them....

For those that haven't handled surface mounted parts, here's the BF998 transistor next to a 6BA nut.

Below, the probe ready for testing with four AAA cells installed. The bias resistors (ie 22Kohm etc) are yet to be changed to improve overall performance.

The finished probe. I used my spectrum analyser with its tracking generator together with a signal generator to test it. My first test gave very odd results before I discovered I'd used a BNC cable sold for audio applications.

The probe delivers an output of -37dBm +/- 2dBm from a 100mV input (=-7dBm) from around 1MHz to something around 100MHz. Respectable output is obtained down to a few tens of KHz and up to a few hundred MHz. Feeding in a signal generator set to 10mV@50ohms (=-27dBm) gives -57dBm +/-2dBm across the same frequency range which is consistent with the above. One of the problems met with this sort of test equipment is non-linearity. As the input voltage is raised the amplifier may saturate and produce harmonics. (see this page) Eventually, if the input is raised too much the output will be a square wave so it's important to know the limitations. This can be done by increasing the RF input to the point where a given number of dBs input fails to produce that same number at its output. As long as the overall gain is reasonably flat across frequency bands of interest the gain (or in practical terms, loss) is not too important because the probe will be used for comparative measurements only. Typically it'll be used for checking the response of an IF strip or a filter or for monitoring an RF voltage. The gain of the probe is coming out at -30dB. This represents a significant loss at the expense of protecting a piece of modern test equipment but it does benefit by introducing only a tiny effect on the circuit being monitored. As long as the spectrum analyser can handle the sort of RF input levels met in practice the 30dB loss isn't important as long as this is constant across the bandwidth being monitored. A typical error in measurements can be made by monitoring a voltage as trimmers are twiddled. You'd think that the higher the voltage the more accurate the tuning, but this isn't necessarily the case. A good example is that of a Microwave Modules 2-meter Linear amplifier. There are a few trimmers on the circuit board and it's possible to view a power monitor in the 2m aerial lead and twiddle the trimmers for maximum output. I distinctly remember the MM proprietor telling me that a spectrum analyser needs to be used and sure enough... judicious twiddling gave me 100 watts output, but some of this represented 288MHz and 432MHz power. One of the trimmers is for adjusting a harmonic trap. When this is set correctly the RF output dropped from 100watts to 80watts (see this page).

This is now much simplified. Pictures below show the band up to 30MHz with the probe off then ON with no input. Fed with a 10mV or -27dBm sinewave signal of 2MHz the output measured -57dBm, a loss of 30dB with the majority (20dB) due to the 1M input resistor. Note that the BF998 has both gates protected by back-to-back zener diodes.

in progress