An external mixer for the HP8555A Plug-in
This article describes an extension of this useful instrument to 24Ghz and above, by the addition of a simple add on mixer.
In its standard form the 8555A plug-in for the HP141T Spectrum Analyser will accept inputs in the range 10MHz to 18GHz via the input N type connector. There is a modification available to allow access to at least 24GHz via this connector, but 24GHz via an N type is not really recommended. Indeed I believe the modification precludes one from using the external mixer capability.
The 8555A will allow the use of an external HP mixer via a front panel BNC socket and this will extend the range of the instrument to 43GHz, just short of the amateur band! To obtain displays at this sort of frequency the 8555A LO is automatically multiplied in the external mixer and mixed with the test signal. In conversation with Brian, GM8BJF it turned out that he was using a "home-brew" external mixer to look at a 24GHz signal quite successfully and its was decided to try this approach to look at 47GHz. Although the amplitude calibration using this method is not really absolute, adjustments for maximum output, sideband selection and signal cleanliness can nevertheless be examined.
With no genuine surplus HP external mixers available, a small unit would have to be made as GM8BJF had done. However without access to a 1N26E diode that Brian had used, experimentation with more readily available diodes was needed.Since the mixer was going to be used at 24GHz and above, it was decided to make the unit in a very short length of WG20 which was to hand.
The figure shows the outcome of using this approach. The SMA socket was mounted on a small 4.5mm thick brass block soldered to the guide broad face, with a ~7.5mm diameter. hole in it to accommodate the mounting of the diode. A short circuiting brass block was soldered in to the end of the guide at the position shown. The probe from the SMA protrudes into the guide via a 2mm hole in the guide wall. The probe itself can be a wire extension of the SMA pin, or better still the SMA pin itself.
A number of diodes where tried including those out of LNBís. All worked well, but in order to obtain a readily available source the commonly available HP HSMS 8101 diodes which are available from Farnell (Stock No. 994-649) were used. Only really applicable for use in a surface mount applications their packaging is not ideal, but since maximum efficiency is not required, they seem quite satisfactory here. Care should be excised in soldering in the diode as static can damage the device. Correct orientation of the diode should also be observed, since a positive bias voltage is applied to it from the 8555A. There are three leads on the diode package, but only two are used so its should be mounted with the lettering as shown ensuring that it is correctly orientated. It is quite possible that other diodes such as the DDC4561 or even a 1N21, could also be used, but these have not been tried.
The external mixer arrangement on the HP 8555A is clever in that the DC bias, IF and LO all share the same cable to the mixer. Thus if a short length of good SHF cable is used, suitable for use at 2-4GHz (e.g SUCCOFLEX) no difficulty should be experienced in obtaining results. Adjustment of the "Ext. mixer bias" pot on the 8555A optimises the mixer /multiplication process for best signal into the analyser.
Since the frequency dial on the 8555A runs out at 43GHz some other means has to be employed to determine the frequency of the wanted signal. Luckily HP also put the LO frequency on the top of the scales, thus one knows what LO frequency the mixer is seeing! The IF centre frequency on the highest ranges of the 8555A is 2.05GHz, thus knowing this, one can, by some arithmetic, work out what the LO frequency should be.
As an example, to display a 47.088GHz signal, 47.088 minus the 2.05GHz IF will require an LO of 45.038GHz. This is obtained by a x12 multiplication of 3.75GHz LO in the mixer. Thus the frequency set pointer should be set as close as possible to the LO frequency of 3.75GHz on the top scale of the instrument. Although I have not tried this, due to lack of a good 3GHz counter, it should be possible to connect a counter to the ext. mixer socket (watch the DC bias!) or the first LO output socket and with the 8555A set to "manual sweep" set the LO to precisely this frequency and then reset to "Int. sweep mode" again with the span set to say 1MHz. However, having said this, it appears that the "Signal Identifier" on the 8555A still operates in this non-standard mode thus one can check by the normal means if the signal being displayed is the correct one by the "usual two divisions to the left" offset.
Note that for the external mixer to function and for the LO to be directed out of the instrument correctly the 8555A must be operated with the band selector set to one of the frequency ranges above 18GHz. Also note that in this mode the "Input Attenuator" of the 8555A is not functional, although the "Log Ref. Level" control is.
If the power of the signal has been measured previously, on a power meter, then a crude calibration of the vertical scale can be made.
The signal being checked into the mixer should be kept to 2mW or less as driving it harder only produces more mixer products and tends to confuse the measurements. Keeping the input to this sort of level also ensures the longevity of the diode!
It is hoped than when Iím QRV on 76GHz that the same method can be used to examine the signal at that frequency.
A similar diode mounting arrangement is being examined with a view to producing a useful waveguide noise source for noise figure measurements above 24GHz. If successful this will form the basis of a future article