During one of the Orwell truck-stop saturday morning breakfasts that a dozen or so amateurs regularly attend, it was mentioned that someone in the UK (it was actually John, MW1FGQ) had come up with an idea that would result in a satellite LNB's LO frequency drift being totally compensated for, and that this could be achieved without modification to the LNB, perhaps providing an easy, low cost way of monitoring the 3cm amateur band.

Over the following few months a good deal of head-scratching was needed in order to work out how this might be achieved - after all, similar things had been done prior to the development of digital synthesisers. The Racal RA17 receiver's use of theWadley loop being a fairly famous example. Here, the 1st LO VFO does not have to be too stable because it is used to mix not once, but twice, allowing (say) hf drift in a first mixer to be counteracted by an equal amount of lf drift in a second mixer, resulting in zero net shift. The overall LO effectiveness relies on band-pass selectivity between the mixers, so the VFO must not drift too far, but the requirement was easily met, even in the days of valve VFOs.

To avoid the need for LNB modification, LNB LO drift has somehow to be lived with, yet not show any effect on the final demodulated signal.  The method to be described achieves this.

      BTW, I've since discovered that Kerry Banke N6IZU  has already done the same, and there is an article in the Microwave Update 2008 Proceedings on his findings.

It's nothing more than intermod...

True (though lets call it 'parametric mixing') - It needn't be as appalling as it sounds, either:

If a stable carrier is leaked into the LNB input at say 10.318 GHz (ie, 50MHz lower than 10.368GHz),  a 568MHz down-converted signal will appear at the IF output (assuming a current UK satellite LNB with LO operating at 9.75GHz). Increase the 10.318GHz input signal level and eventually the final IF amplifier stage will start to clip.  Now also add a low level input signal at 10.368GHz. This will be down-converted to 618MHz. Since the LNB final IF amplifier is clipping, the strong 568MHz component will mix with the weak 618MHz signal to produce a component at 50MHz, and a receiver tuned at this frequency will hear a signal  when connected to the LNB IF output port.

A number of points may be noted:

a) If the LNB LO frequency increases 10KHz, the IF ouput component at 50MHz will remain at its original frequency, because both the 568 and 618MHz components that generated it have increased by the same 10KHz, resulting in a differance frequency that does not change. The LNB LO drift is a common mode effect, in this example.

b) Since changes in LO frequency do not affect the difference frequency, LO phase noise will also not appear as components on this difference signal.

c) The arrangement is double conversion, with a first IF of 618MHz. Image rejection at this IF, using the LNBs internal image filter alone, will be about 45dB (based on measurements I have made on six current/recent LNB brands). Image rejection at the second IF will be zero, if operated as described, but external selectivity (at 618MHz, and followed by an IF amplifier which does the clipping) is capable of giving adequate image suppression at 2nd IFs of 10MHz and above.

d) Sensivity should be good (NF better than 1.5dB - assuming at least 10dB image rejection is being achieved).

e) Frequency stability will be as good as that of the external 10.318GHz source.

f) As described so far, the strong signal handling performance will be quite poor (ok, appalling), but this can be improved by splitting out the 568MHz and 618MHz components externally (by adding selectivity to each of two paths) and putting a lot of gain in the 568MHz path and feeding the two signals together again in a common IF amplifier stage/mixer - more later. In any case, how often do you have another signal on the band, let alone a strong one...

But have you tried it?

Yes - and although results with just an LNB alone (ie, no external IF amplifier) were not particularly exciting, it doesn't take much externally to bring everything to life. Fortunately, my initial attempts used a dc supply inject/IF dc blocking assembly that had been used to feed an SDR-14 harmonic sampling receiver, so had also included a 618MHz band-pass filter followed by a single stage mmic IF amplifier (the SDR-14 is fairly deaf when operated at the 9th harmonic of the sampling frequency, and of-course requires selectivity to avoid reception at the other sampling harmonic integers). This arrangement gave remarkeably good results.

One reason for un-exciting results with an LNB alone is that the IF stages in these assemblies are run with intentional high-pass filtering, and the last stage output loading looks pretty much like a short circuit at well below 800MHz, so the 50MHz IF component is very much attenuated.

The output load of the external mmic amplifier had plenty of reactance at 50MHz, so there was no shortage of level. Even at 2nd IFs of 2 MHz or so, there was enough output (but almost zero image rejection, of-course).


Most basic form:

lnb The power required from the 10.318GHz signal source was about -40dBm to
produce a 10dB lift in noise floor from the IC-706 that was being used to monitor the 50MHz output. This is enough to produce 4 -5 dB front end NF, remembering that there is no image rejection with this arrangement.

Increasing the 10.318MHz level to -30dBm did not increase the conversion gain, and further increase in level resulted in a decrease in gain.

A useful feature of satellite LNBs is that their high conversion gain allows this sort of tune up/assessment on band noise only (although this was constantly confirmed by having a 10.368GHz beacon running in the background).
There is a short video clip of this set-up here.

External filtering and gain:

This arrangement is much more predictable, because the 618MHz band pass filter reduces the IF bandwidth from 1GHz down to 5MHz or so. It requires less LO drive, and the IF output is greater due to the mmic resistive loading. One slight problem is that the BPF restricts the IF output range to 20MHz or so and below, similarly the image rejection is poor since there the BPF response requires a compromise to be made between image response and down-converted LO throughput level. However, enough image rejection can be obtained to bring the ssb noise figure down to 1 - 2dB.


dc filt mmic

The image rejection at an IF of 18MHz was 25dB with the filter shown above. Looking at this from the other direction, 10dB image rejection was obtained at an IF of 7MHz. This also implies that for an 18MHz IF, the down-converted LO signal at 600MHz is being attenuated by just over 10dB, showing the compromise of needing an extra 10dB of 10.350GHz level at the probe - this isn't as bad as it may seem, since only -60dBm of signal was required to obtain full sensitivity. More signal gave more conversion gain (optimum for this was -40dBm), but did not improve sensitivity (see video below).

There is a short video clip of this set-up here

The low level of LO signal required (60dB less than that in a conventional mixer) obviously implies a much simpler/leaner LO multiplier chain.

Do all LNBs work equally well?

No - it's all down to how sharp the IF amp high pass filter response is (after all, the lowest IF frequency in satellite TV use would be 950MHz, and we are operating at 618MHz). The Thomson 13553 is particularly good since unusually, the roll-off doesn't start to about 200MHz!. Others that work, but are 10 - 20dB less sensitive are:

13553 ae88
Thomson 13553
Cambridge AE88
 Lidl IP-401 MTI AP8-XT2EBL Thomson Quad Wistron Quad

Non of the Grundig units seem to work that well.

But what about all that intermod - and then there's the radiation from the LO probe 

I can see that you are not completely convinced.

A little more complexity post-LNB can assist here, if you really think it's necessary:

lnb full monty

The down-converted signal and LO components (shown here with values set for a 50MHz IF) are split using a pair of band pass filters so that the LO component can be amplified much more without causing much damage to the received signals, which are left un-amplified. The two components are then re-combined in a mixer (which could be yet another mmic amplifier in practice) which further down-converts to the final IF frequency.

With this arrangement, the 10.318GHz level at the probe could be reduced down to -70 or -80dBm, which would then meet the usual ETSI spurious radiated signal spec. Also, the two BPF's remove the compromise previosly mentioned regarding a single BPF solution.

This arrangement is yet to be tried.

A further 20dB reduction in the amount of LO injection in the signal path could be achieved by using a dual or quad LNB, and taking the LO band pass filter direct from the vertical polarisation output (and mounting the probe in that plane too, of-course). This would require a 16v supply though in addition to 12v one in order to achieve the switching (not necessary with an Quatro LNB).

Is that 'simplified LO chain' just talk?

Let's see.

In the scrap box was a dual pipe cap filter for 10GHz. Tuning this up to10.350MHz and looking at +-450MHz showed 50dB selectivity. Why 450MHz? Well, this is what you would require if using a 10MHz reference as the starting point of a multplier, ie:

10MHz  x  5  =  50MHz  x  3  =  150MHz  x  3  =  450MHz  x  23  =  10350MHz

So the last stage is a x23 multiplier driven at 450MHz. The 50dB selectivity is good enough for a receive converter.

An old LNB was found that had a dual schottky diode mixer. The particular one found was an HSMS-8101 in an SOT-23 package, and the easiest way to use it was to solder it directly across the coax feed to the pipe cap filter. As the pic shows, the feed is semi-rigid, so it was easy enough to expose the outer shield to gain access to the centre core.


There is no matching, either on the filter side or the input side. The feed was connected to an HP8640 signal generator, which was set to 20dBm output at 450MHz. Tuning the two cavities to 10.350GHz and measuring the output power revealed -30dBm was present - far more than was needed!

Here is the proof of the pudding.

The next step was to replace the HP8640 with something more stable and a lot smaller, and this was the result:

450MHz mult  450MHz LOa

Unfortunately, the 10MHz reference is not very good and wonders around over a range of 800Hz or so (at 10GHz). This was changed for a Rakon unit, and the board housed in as a die-cast box:



A further board was built up using an NDK 3119A 10MHz reference osc.
These 2ppm TCXOs seem to be stable enough (as long as the module is not subjected to draught).

ndk osc


450MHz 15dBm source:

Single housing

single housing
The three modules fit quite well  into an Eddystone
box (120 x 95 x 30), which Maplin stock under their
part number N90BQ. This arrangement is useful where a dish is not going to be used.

On this build, the input and output coax on the x23
multiplier was RG316, and was a little easier to cut to
fit the multiplier diode.

The 450MHz output filter has also been replaced with
a pair of Toko dual helical filters. Likewise, a Toko
helical filter replaces the IF semi-rigid coax filter.

Remote LNB

An elderly G88, with semi-rigid probe attached, was mounted next to the normal twin Tx/Rx dish assembly on the roof of the bungalow. The probe and IF connections were connected via 4.5m lengths of RG 223 cable to the converter, which was placed in the false roof.

When compared with the dish fed LNB, the local beacon was only some 12 dB down in S/N when listened to on the G88 based unit - I don't know whether to be pleased or worried!

It will be interesting to listen during lifts andor contests to see how the system performs.

A kit was eventually put together with boards manufactured by Top Tech pcbs Ltd. A run of 25 units (with the Cambridge and District Amateur Radio club in mind) was put together mid year 2009. Details of the build are here

There are some pictures of the build/alignment day here and here.