A thermionic diode noise generator for radio alignment
by ralph klimek  VK3ZZC   August 2020
the last thing the world needs now is yet more noise......oh wait!

This project began about 20 years ago.  I was examining some dumpsters at work, looking for non compliant waste. This is the line I use when challenged, dumpster divers are all , off course, dangerous individuals.  I proclaim "this waste is non-compliant",  waving my hardboard clip board around. That usually does it, and I can recover interesting electronic equipment from landfill.  Nobody want to be "non-compliant" ! But I digress.  I found this little thermionic diode amongst the ash tray contents and banana peels.  It was a rare CV1271 thermionic diode.  This B7G miniature valve contains just one diode, a tungsten filament filament and cathode, surrounded by a cylinderical anode and nothing else.  It is pretty useless even as a diode. 

This little diode went to the bottom of my most neglected and spider infested junk box and suddenly reappeared during a spring clean.  It was begging to be reused. They are extremely rare these days, they always were scarce as they were made only for speciallized test equipment.  The CV1271 was made for the British MOD for the alingment of radar recievers, there would have been a global market for about 10 of these.  Nevertheless,  I have seen these appear in Hamfest valve hoards, and I shall be looking harder next time because now I want a spare.

There are perfectly satisfactory solid state noise generators using low voltage zener diodes, so why this?

This noise power spectrum of the zener diode is not sensibly flat due to its inherent high distributed capacitance. There is also no sensible relationship between the zener current and the noise power produced. I use zener noise sources   , see this project,   and they work well if you just want some noise and you do not care how much noise .

The thermionic noise diode has an extremely simple relationship  between emitted noise power and anode current.  It is actually amazing why this is so.  It is possible to produce a KNOWN amount of noise power from just the fundamental physics that is based on things that can be easily measured.  I this case it is the mean anode current  and litterally nothing else !

How is this possible ?

Its all to do with statistical physics. Electrons are emitted from a heated filament. The are emitted with random energies.  They are emitted in random order.  These form random bunches. The random bunches are accelerated to the anode by the high eletric field.   The random bunches deliver most of their kinetic energy and just plain old heat. Some of that energy becomes electromagnetic energy that can be coupled out of the valve structure. How much ? The random bunches over time cannot be individually measured but can be seen as random flutations in the anode current
We can measure the mean current with just an DC ammeter.  The random fluctations of the electron bunches appear as radio noise.  Given that the statistical distribution is fundamental and not a strong function of anode voltage
the intensity of the generated noise is dependant only apon  the mean anode current and just about nothing else.

The anode current is required to be "saturated" . This means that changes in anode voltage should have no further impact on the mean anode current.  I have provided a negative fixed voltage source of -200V. The mean anode current is controlled by variation of the filament heating current. Filament emission is a strong function of cathode temperature. In the case of the pure tungsten filament of the CV1271 it goes from zero at dull yellow heat to over 25mA at bright white heat.  I have used a multiturn  25 ohm wirewound pot to control the filament current. I recommend the use of a multi turn control  as the anode current is a very strong function of filament temperature.

Do not exceed the filament current rating. They are not like conventional valve filaments, this filament resembles a thin light bulb filament and will not tolerate over current.

There remains only the technical challenge of extracting the RF component of the anode current. So here the real world gets in the way of a good simple physical model.  Real RF chokes do not have infinite and constant impedance, stray capacitance is hard to account for.  If I can achieve a practical bandwidth of 500Mhz then this project succeeds.
To minimise the stray cicuit capacitance, I have grounded the anode and supplied the cathode filament through a bifilar choke. There I have capacitively coupled the cathode directly to a precision 50 ohm resistor made from surface mount 200 ohm chip resistors directly bonded to a brass chassis and thence directly coupled to the output BNC socketon the front panel. The construction method has been designed to minimise the connection  between noise source and output socket.  This is hard to do with the old vacuum technology and means the usefull upper limte of my construction is about 400Mhz,   good enough for my purpose.

Keep the filament current low, when the instrument is idle.  The CV1271 has only a rated 1000 hours at max current !

Some interesting things about my design here. Grounded anode,  neccesitating a negative cathode supply. The valve is upside down !  There is no law against that ( at least in Australia ) .  The regulator is a fixed shunt high voltage zener that is amplified by a high voltage transistor.  

view showing power supply and -200Volt regulatorclose up of filament current control pot and slow motion drive
front panel. Meter shows anode current FSD 20mA
recycled cabinet, anyone want two used quality holes ?

on powering the thing up we had a good stable -200V supply for the diode.  We had a nice yellow glow from the filament. No anode current ! Did I have a dud tube ?  Its a tungsten filament, at yellow heat. No anode current for you ! We need to be a bright yellow/white heat. Now we see 10mA.  At white heat and at rated filament current we have our 25mA .

As for noise, using my TS700  2 Meter amatuer radio as a sensor, at 10 mA we hear a distinct excess, compoared with 50ohm dummy load. We get 3db more noise at 20mA.  This means my radio is not that great. Thats OK, it would have been state of the art  30 years ago.


Yet to come for this article,  a qwick discussion of the mathematics of this thing and precise radio noise figure measurement using the thermionic source.