47GHz Narrowband - the way forward?
©Lehane Kellett - G8KMH
To date most activity(!) that there has been is on wideband, with varying IF’s of 10.7MHz, 30MHz and 100MHz. However, even with the best manageable antennas and with better mixers than generally in use it is likely that long paths are going to be difficult but perhaps not impossible - the Dartmoor to Prescelly path comes to mind as a ‘nice’ goal but still someway short of the world record.
Longer term the aim has to be to move to narrowband but most likely to nbfm/mcw rather than SSB. The reason for this is two fold. A number of stations have, or are contemplating, direct multipliers to 47.088GHz with power ranges from 0 to +13dBm. Secondly, phase locking a Gunn, especially one of the +18-20dBm units in use, would be very attractive and these lend themselves to lower deviation FM.
It therefore looks like the future will consist of about 4 station types: The wb systems in use today, phase locked versions with nbfm modulation and a separate receive mixer, the multiplier approach with a separate receive mixer and those based upon the DB6NT transverter design with c.-10dBm output. In case this seems far fetched, three of these types were at Martlesham with only a phase locked Gunn missing - out of 9 sets on display, probably a record in itself.
As someone who has one of the high power Gunn’s, the phase locked approach would appear to give the best of all worlds. Using almost all of the existing components together with a small LNA/filter near the current (poor) mixer, a low power 47G source and a PLL assembly will give a potent nbfm transmitter. The LO only needs to generate maybe -6 to 3dBm. Possibilities include a varactor, driving a mixer diode ‘silly’ and, probably easiest now, using a HEMT doubler - as demonstrated by Toshi JE1AAH(Microwave Update '95).
A totally separate receive multiplier/mixer driven by the same LO chain is used on receive. This has some advantages compared to the classic G3BNL system by not using the ‘noisy’ Gunn on receive and using the best mixer here. A good IF would normally considered to be 1296MHz but this results in a messy frequency for the PLL reference crystal, fine if you are a commercial manufacturer but expensive to get an OCXO/TCXO made. However 1280/256 gives 5MHz and there are plenty of 5 and 10MHz(1280/128) OCXO/TCXO’s available at rallies, or you could make your own OCXO using a CPU clock oscillator sat on top of a bunch of 1206 resistors (insulate!). 128 or 256 is convenient division ratio from modern 1.2/1.5GHz prescalers(CA3179, 501/SA701, SPxxxx, U893BSE, etc). A simple PLL(4568) and loop filter(LF356) drives the gunn supply(723) with modulation applied in the normal way.
Part of a JNT004 1296 system could be used for the receive downconverter with a 94.666MHz crystal or you could build one using dead bug techniques (ERA3/Toko filter/TUF1/BFR9xx) and a DDK001(B). A feed is also taken from the reference mixer to allow a number of tests to be performed.
Whilst I happen to have a multiplier to 23GHz which has two outputs, something similar could be fabricated based upon the DB6NT 12/24G modules, either splitting the LO at 12GHz or at 23GHz in waveguide. The starting point for the LO chain is the normal DDK001 followed by a x4 multiplier and a filter. A DDK009(on PTFE) could also probably be used with one of the ERA1 modamps as the last stage.
Any of the other normal methods of getting to 23G could also be used ( WDG009/DB6NT/JE1AAH).
So there you have it, a system capable of working many line of site paths built pretty much upon normal 24G parts. The most difficult and expensive bit is probably the 47G converter/mixer which uses an HPSCH53xx (Si) or HPSCH9xxx (GaAs)part silver epoxied to the board.
Who is going to be first to 150km in the UK?
Work on the driver chain is going well, if not quite in line with the diagram shown below. I have used a 2.3544G LO chain (the DDK004) as I already had one built with an Xtal which comes out on 47.088GHz (+/- a few KHz). That is driving a Toshi(JE1AAH) doubler which has a integral side coupled microstrip filter - nominally on 4.48G which is close enough for now. That then drives the ex-commercial 'cube' which gives me +6dBm on each of two outputs. The next stage is to build a HEMT doubler on 5mil Duroid - this is being CADded at the moment (I had one but I lost it on one of my many computer upgrades!). I may take a temporary sidetrack after that and do a quick and dirty since I have a 40G mixer in WG22 and try a bit of operation (after re-xtalling the LO) - the UK record is calling..... Even if that doesn't work on TX but is OK on RX I only need to build another DDK004 (For the TX freq) and I can use the doubler, 'cube' and 47G source for full tx/rx. Or use a DB6NT 47G board.
No doubt this will end up with one of my legendary lash ups, as my 24G system was for a long while, with boxes laid out on the ground or top of the car. I hear Steve, G4KNZ, groaning already :-)
Merry Christmas.
However, the 47G doubler is now been through the CAD and I have the film ready to do the board. I also have obtained some more Duroid for 47G, which I'll have to spray up with photo-resist. I'd like to put a filter in the output of the doubler, just to make sure, so I've been looking at the two post design which works well as an LO filter on 24G and I'll scale that and give it a try using the 47G gunn to align it as it is already on 47.088GHz. Since I don't have milling facilities the mounting for the board will be on brass angle.
Also see for more on 47GHz narrowband
<PL calc here for 150km (+17dBm, +27dBi ant, 8dB NF, 15KHz b/w, nbfm detector)>
Revision 1.3