G4KNZ's 47GHz Wideband Equipment

G4KNZ with 47GHz equipment....Martlesham Nov 1997

For those interested, here is a more detailed description of the 47GHz wideband FM equipment in use by G4KNZ, using MA-86147 49GHz oscillators. This is a summary of part of a talk given by G4KNZ at the Martlesham Microwave Round Table in November 1994, with some updates.

The original application of the equipment was a 4GHz Earth Station low noise parametric amplifier assembly. Two parametric amplifiers with Peltier cooling were contained in the assembly, the pump frequency being at 49GHz.

Acknowledgements: Assistance/contribution by G8KMH, additional helpful suggestions by G8AGN, and loan of test accessories by G4DDK. Measurements made using a HP 8565E 50GHz spectrum analyser.

Waveguides and Components

See the comments in the section entitled: Have you considered using WG22 on this band?

The surplus Gunn assembly and other components here are all in WG23, but WG22 is also used in places, connecting directly to the WG23 with apparently little ill effect.

Gunn Assembly

Probably the Gunn diode is rated at 100mW min. output, covering 40 to 50GHz, with a typical output power at +55C of 110mW, which goes UP as the temperature goes down. So it will pay to make sure the assembly is well heatsunk!

Two slightly different versions of the assembly were encountered: one with a '79 date code, and one with a newer '82 date code.

The Gunn is originally set [at 49GHz] close to it's maximum frequency - it will tune to 50GHz or so. It will also tune a lot lower, but earlier versions misbehaved around 47GHz. A later, better behaved sample, tuned down to 44GHz. At around 47GHz, the earlier Gunns jump frequency suddenly - tuning down, the jump can be from 47.5 to 46.5, for example.

Loading the cavity with PTFE, fitting it round the post - by using a cylinder of PTFE was tried. This lowered the frequency by 1 to 2GHz, and moved the problem frequency lower. The Gunn would still misbehave, but maybe not till 46GHz - which is not a problem for WB operation with a low (eg 10.7MHz) IF.

The Gunn tuning is fairly course and it is quite tricky to tune in a signal using this - so some fine tuning, eg voltage tuning, is desirable.

Power supply requirements

The Gunn supply current is 1.6A max, with a starting peak of 2.4A - so a conventional 10 or 24GHz Gunn PSU with a 1A regulator won't do! The maximum diode voltage [NOT nominal] is 4.5V, and minimum is 2.5V. At, say 4V, the assembly will be dissipating something of the order of 6W.

Measurements from one (79nn) Gunn assembly of voltage vs current were:

Voltage (V) Current (A)
3.0 1.58
3.5 1.48
4.0 1.40
4.4 1.32

After 5 mins, the assembly gets luke warm - OK to test without heatsinking for a few minutes, but better heatsinking is needed for continuous operation.

G4KNZ's PSU is based on the RSGB Microwave Handbook vol 3, page 18.34 design, replacing the 7805 regulator with a variable regulator, driving a series pass transistor (well heatsunk). A tone generator and mic amp is provided on the PCB. After the regulator is a 3A fuse, and an overvoltage protection 5.1V zener (to blow the fuse).

Voltage pulling

Some frequency tuning is available by varying the Gunn voltage. This pulling is not constant but depends upon the loading of the cavity with PTFE - this variation can be as much as a factor of two! The newer date coded model is different again, with a much lower voltage pulling.

Older (79nn) assembly:

For a voltage operating range of 4.2 to 5.2V, the frequency variation is 47103-46960, ie 143MHz tuning range. Thus, for WBFM operation, a modulation voltage of the order of 1mV peak to peak is required. Depending upon the amount of PTFE fitted to the post, the tuning range for an input voltage of 4.2 to 5.2V varied from 60MHz (with no PTFE) to 180MHz (largest amount of PTFE).

Newer (82nn) assembly:

For the newer assembly, only 5MHz tuning range for a 0.4V input voltage variation (ie 12.5MHz/V pulling). Thus a much higher modulation signal, by a factor of 10 (ie around 10mV pp) is needed.


At 49GHz, two samples measured seem to have an unmeasurable insertion loss (less than 0.2dB), while the isolation was measured at 21dB.

At 47GHz there is a problem with the isolators, which may explain why the newer Gunn assembly still shows a small frequency jumping problem around this frequency. The isolator seems to have some resonance close to 47GHz. The insertion loss rises, to 1 to 2dB, and the isolation drops - one was measured as having only 2dB isolation at 47.1GHz, the other as having only 6dB. With performance this bad, G4KNZ has removed the isolator.


A detector is provided with the paramp assembly, used to detect the Gunn power level -15dB from the cross coupler, ie at about 3mW. This mount has different characteristics compared to a typical detector mount. Two of these were measured below (marked 97 and 87).

A HP WG22 detector was also measured. The flange holes on the detector match up with WG23 holes, so it can be used with the step without a transition.

Warning - do not to put 100mW from the Gunn straight into the detector!

RF Power HP WG22 Detector no 97 Detector no 87
-10dBm 28mV 8mV 6mV
0dBm 170mV 80mV 85mV
+8dBm 630mV 1700mV 440mV

At low RF input, the levelling detectors produce hardly any output, but in the region 0 to +8dBm, the output rises sharply. Given that they are intended to operate at +3dBm, this is perhaps the ideal type of diode for levelling! These diodes do not work well at detecting weak signals, and were found to have very poor noise figures of some tens of dB.


This cross-coupler (Moreno cross type) from the paramp assembly measures -15dB, and is also fairly directional at 47GHz! It is ideal for injecting the LO into the mixer (100mW Gunn gives 3mW LO injection), with one exception - the integral termination on one port!

The load was removed and a flange fitted, then only the antenna would need to be swapped from the TX to the RX port. If two antennas were used, no changeover is needed at all.

A replacement flange was made from a WG22 square flange and small piece of WG22, since the overall size and fixing holes are the same as WG23 flanges. The flange/filed-out guide was pressed on to the coupler in a vice.

With the coupler modified, G8KMH and G4KNZ have opted initially for full duplex operation, with two horns mounted next to each other, vertically polarized.

Signal sources for testing/calibration

The Gunn modules were tested using a spectrum analyzer with accurate frequency calibration. In the field, it is important to be able to check the frequency.

In the absence of an accurate wavemeter, an alternative approach is to listen to a calibrator signal. Two options for calibrator signals sprung to mind:

24GHz Gunn 2nd harmonic

Three 24GHz Gunns were measured to see how much 2nd harmonic is present. These were (a) GDO33, (b) GDHM32, and (c) a 50mW M/A Gunn.

The GDO33 and the 50mW M/A unit were tuned down to 23.55GHz using the fitted screw. The GDHM32 has the original frequency setting screw removed, and a large polythene screw used instead to tune down to 23.55GHz.

On 23.55GHz, the output levels of the GDO33 and GDHM32 were around 4 to 5mW. The nominal 50mW source was measured at about 40mW output.

In all three cases, an output of around -10dBm was measured at 47.1GHz.

The Gunn (in WG20) was connected directly to a WG22-coax transition - the 47.1GHz o/p depends upon this mismatch between waveguides - it varied from at best -9dBm, to more like -15dBm, as the guides were aligned/misaligned. The level at 47.1GHz could probably be improved by more careful matching. By fitting a filter to attenuate the fundamental, a low power WB TX is made!

If reasonably accurate frequency measurement is available at 23.5GHz, the 47GHz signal can be calibrated.

Harmonic generator

The original paramp contains a SRD mounted in a milled aluminium block, which forms a very reduced height waveguide (about 0.5mm high), and a 4GHz 3 circulator assembly which feeds the 4GHz low noise signal into and out of the srd assembly. The SRD assembly was adapted to make a harmonic generator.

G4KNZ removed the circulators assembly, and the choke, which is probably designed to prevent 49GHz from escaping from the SRD. An SMA with small semirigid was connected to the SRD by a fine piece of wire a few mm long, and still attached from the original assembly. Thus there is no choke or matching.

A number of input frequencies and levels were tried, from 2242.3MHz (17dBm input, gave -44dBm output at 47088MHz), to 11772MHz (+14dBm input, gave -52dBm output at 47088MHz). A comb of frequencies is generated in the 40 to 50GHz region, since there is no output filter at 47GHz.

By adjusting the stub in the opposite channel to the output, and also the tuning screw which enters the broad face of the guide opposite the diode, one particular harmonic could be peaked - in this case the one on 47088GHz. Other harmonics are then 10 to 20dB lower. Providing a DC return for the diode seemed to make little difference, on this unoptimised mount.

For example, for 2354.4 MHz in at +17dBm, at 47088 the output is -35dBm, then n-1 level is -59dBm, and n+1 level -47dBm. This frequency was selected for the calibrator used by G4KNZ.

Conversion loss is quite high, a little over 50dB [cf 27dB for my 24G TX], but (a) the output is untuned, energy at all harmonics escaping, and (b) the input has no tuning nor now a proper 47GHz choke.

At 11.772GHz input, the level was low (-52dBm), until the input SMA was loosened off. Suddenly the level rose to as high as -20dBm! This was with the pin of the SMA male just about to be inserted into the female, but not actually contacting. It would seem that by chance, input matching had been provided, which greatly increased the efficiency.

G8KMH instead of removing the choke assembly, kept this intact, and fitted a semirigid cable (0.141") in through the entry to this assembly. Initial measurements show little difference between this and the arrangement tried by G4KNZ.

In summary, a simple SRD assembly can be used as a comb generator for frequency calibration. At a level of around -40 to -50dBm, it is strong enough to be easily heard on the WB gear across the kitchen table! With some work, higher output, eg -20dBm or more might be achieved.


G4KNZ and G8KMH used WG22 horns for initial tests, separate horns for transmit and receive. With a 55 x 70 mm aperture, 150 mm long (tapered part), the theoretical gain is calculated at 27.5dBi, and the beamwidth at 7 degrees.

G8KMH has reduced the WG22 guide size of one of the horns, using the K&S 264 brass section, so the step from WG23 is smaller, and there is no possibility of overmoding. The performance difference has not as yet been measured.

G4KNZ now uses an 18 inch diameter dish, with a WG22 feed. This feed enables the dish to be first fed with a 24GHz signal and aligned on this band, before swapping to 47GHz. The feed remains undisturbed, and with care not to move the dish, the alignment on 47GHz is then spot on.