A (re) introduction to 10 GHz via modified satellite LNBs

Dropping in late to the Lincoln Hamfest a few years ago, I found one of the traders selling off a box of about 30 satellite LNBs for £10.00, and just couldn’t say no.

The mmic IF stages were an obvious candidate for re-use, and it was straightforward enough to convert the stages to provide amplifiers with a gain of 40dB or so from HF to a couple of GHz.  With these, built onto the back of a dipole/reflector plate combination, noise free signals from the local 23cm TV repeater could be piped down through long lengths of pretty ropey coax without having to worry about signal loss.

It was a year or so later that another use came to mind. Interest in SETI had re-kindled a general interest in radio astronomy for several of us and the notion of producing an 11 GHz interferometer using two 60cm dishes with their LNB’s began to appeal.  We knew that the RF stages of most LNBs of the time gave about 10dB gain, and that there were always at least 2 stages, so early on we thought about getting in after the second stage on each dish mounted LNB, bringing the feeds out to a combiner (OK, just a bnc ‘t’ piece) and feeding the result into a third LNB, which would only be modified by fitting an sma socket in place of integral antenna probe. The loss in even bog standard quarter inch coax over a few metres would not approach the 20dB gain figure of the RF stages.
 It was soon clear that with many of the LNB types, you could simply drill through both the pcb and the aluminium housing it was fitted to after the second RF amplifier, and be able to open up the casting hole and sweat an sma socket onto the ground side of the pcb. Indeed, it turned out to be very easy to do.

Convertor block diagram

Mid '90s LNB modified as antenna with integral 20dB pre-amp to feed the converter

pcb groundplane side

 

pcb component side

On LNB types where part of the casting did get in the way, a hacksaw and and file soon resolved the problem...

The 1-2 GHz IF from the third LNB was fed via a couple of further mmic stages into a diode detector. After much adjustment (and fun), good interference patterns were obtained from sun transits with this arrangement.

During the following year, the idea of replacing the ceramic resonator based LO for something more stable grew. It had been 20 years since my last experience with 10 GHz (and those free running 723A/B klystrons), but I at least knew that narrowband operation had now made  klystrons and Gunn diodes (and Barrett diodes, if anyone remembers those) all but obsolete.
About half of the surplus LNBs had a discrete diode pair mixer, with mmic IF gain stages and a single gaas fet oscillator. Again, it was easy enough with most of these to drill a hole through both the pcb and casting at the point where the fet oscillator fed the diode mixer, and fit another sma connector so that an external crystal multiplier LO could be applied. It took a while to figure out a suitable multiplier, but in the end, a Philips FR5000 PMR 1W VHF driver module was used to drive a pair of 1N4148 diodes, with a 12 stage(!) pcb interdigital filter following it to give a very clean x6 multiplication up to just over a GHz. Some Heath Robinson mmic amplifiers and pipe cap filters then took the LO up to 10GHz. I can’t remember now what the IF was, but it was a single conversion to something around 150 MHz.

Signal path side of converter

With this converter fed from a second LNB just used as an antenna-with-integral-preamp (one of the interferometer units), a trip down the M11/A130 soon had GB3CMS coming in at good strength on the road side at Ford End.

The real downfall with this converter was the poor stability of the crystal oscillator - but it was a start, and the bug had bitten.

The frequency stability problem eased when it was realised that 10MHz, 12.8MHz and 14.4MHz TCXO’s - all readily available, would multiply up to provide useful LO frequencies. Since 14.4 MHz also multiplies up to exactly 10 368MHz, this later became the basis for a CW/FM transmitter [likewise a 12.8MHz TCXO was/is being used as the basis for a low power beacon in the Cambridge area].

The 10GHz converter described here uses a 0.5ppm 10MHz TCXO in a double conversion configuration. The first IF is 768MHz [9.6GHz LO] and this is brought down to 48MHz using a second LO of 720MHz, all driven from the 10MHz oscillator as below:

A more practical approach to the LO multiplier at 480MHz for anyone who has access to an ex Ionica head would be to use the duplexer to select the 3.2GHz harmonic . These items have a suprisingly wide tuning range, and both rx and tx sides will cover 3.2GHz (9.6 GHz LO) or 3.456GHz (10 368GHz). They can also be operated end to end from the tx to the rx port, giving more/sharper selectivity.

Liberal use of surplus satellite LNBs was [deliberately] made in this converter, though no attempt was made to use the integral image filter as part of the 9.6GHz LO multiplier, but this did work well on the transmitter strip to multiply 3456MHz up to 10 368MHz, giving 40dB plus rejection of all other harmonics (without any modification to the etched filter). Instead, a couple of (15mm) pipe cap filters were used. These together with the single image pipe cap filter can be seen in the pic of the convertor above:

Yet to be tried is modifying the LNB oscillator so that the stage acts as both buffer and selectivity block using the integral ceramic puck. Has anyone already tried this approach?

The single image filter seems to have been quite adequate with such a high 1st IF, and the through loss not an issue when fed via  a remote  LNB antenna-with-integral-pre-amp arrangement. Incidentally, RG223 seems to have a loss of 4 to 5 dB/m at 10GHz, so up to 2 or 3 metres of interconnection lead between antenna and converter is OK. For longer lengths, such as chimney to shack use, it just needs another LNB configured as a 20dB amplifier added as a line amp.
Two LNBs are used in the LO path (laziness rather than necessity) configured as 20dB amplifiers. The first takes the 2.4GHz feed at about 0dBm, ensuring that this amplifier clips. The fourth harmonic is selected by the pipe cap filter and fed to the second LNB for amplification. About 10dBm is then available, so another pipe cap filter was put in for good measure prior to feeding the third LNB, taken from a mini-dish, and fully used as the 10 368 to 768MHz converter stage. These early Cambridge mini-dish LNBs seem to work very well, and the converter is fed from another one of these:

Yet to be tried is modifying the LNB oscillator so that the stage acts as both buffer and selectivity block using the integral ceramic puck. Has anyone already tried this approach?

The single image filter seems to have been quite adequate with such a high 1st IF, and the through loss not an issue when fed via  a remote  LNB antenna-with-integral-pre-amp arrangement. Incidentally, RG223 seems to have a loss of 4 to 5 dB/m at 10GHz, so up to 2 or 3 metres of interconnection lead between antenna and converter is OK. For longer lengths, such as chimney to shack use, it just needs another LNB configured as a 20dB amplifier added as a line amp.
Two LNBs are used in the LO path (laziness rather than necessity) configured as 20dB amplifiers. The first takes the 2.4GHz feed at about 0dBm, ensuring that this amplifier clips. The fourth harmonic is selected by the pipe cap filter and fed to the second LNB for amplification. About 10dBm is then available, so another pipe cap filter was put in for good measure prior to feeding the third LNB, taken from a mini-dish, and fully used as the 10 368 to 768MHz converter stage. These early Cambridge mini-dish LNBs seem to work very well, and the converter is fed from another one of these:

Laziness dictated the use of a seperate mini-dish for transmit. With 300mW of CW, there has been no need to power down the receive LNB when on transmit.
Since there is already a very noticeable increase in background noise as the dish is dropped down to the horizon, no attempt has been made to optimise the LNB probe feed at 10 368MHz.
GB3CCX at 150km is pretty much always copyable with this set up, and thermal noise from nearby trees and buildings can be noticed as the antenna is rotated.

So all in all, the LNBs have proved to be a good buy. Every type of unit that was modified to act as antenna/pre-amp units had a noise figure that were good enough to detect ground noise easily.  Some of the mid 90’s (analog) Cambridge units had three cascaded stages following the antenna, but most have only two. A good rule of thumb is to expect 10dB of gain per stage. Using a hot air gun to remove devices has proved reliable. Careless handling after that point has destroyed a few devices through static damage, but that is all The only comment I would make is that  current limiting on the gaas fet stages never seems to be provided, so that shorting the gate to ground always causes device destruction.

This piece was written in early 2005 as a filler for 'Scatterpoint', and the converter has since been added to to form a complete receiver for portable/monitoring use