In some ways, this was simply an attempt to use the on-board ceramic
puck
to add further selectivity to the LO of an external high
stability multiplier chain, though it was also hoped that the internal
oscillator could be locked to a sub-multiple of the LO frequency. The
page is written in chronlogical order of measurement.
Initial
set-up
An easy way to determine lock is to feed the LNB output
into a
spectrum analyser, and have a reasonably strong in-band signal running
nearby.
With the injection source switched off, the down converted signal is
centred on the analyser display. Switching on the injection
source (I used a signal generator) and setting to +10dBm, swish the
frequency around the nominal LO frequency. When out-of-lock, you see
that not only does the main LNB IF signal not follow, but you can see
additional IF components due to the mix of injection and signal
frequencies and intermod products of these with the original IF. As the
injection frequency approaches that of the internal LO, all the IF
components converge to one point, and the internal LO should lock onto
the injection frequency. At this point, as the injection frequency is
moved around, the internal LO will follow, until at some point it drops
out of lock.
Grundig AUN14S
This LNB is housed in a flat backed casting, so access to the
back of the pcb is very easy, just requiring a hole to be drilled
through the rear face of the casting, as the first picture show.
| The
pictures show the location of the drilled hole (red/white
concentric circle near largest ceramic puck) and where it connects into
the oscillator circuit |
With this LNB, a mmic (Eudyna FMM5201MLT4E1) is used in the oscillator
stage. It uses a single coupling track to the puck, as shown. A
convenient place to inject an external locking signal is at the
terminating resistor for this track.
The nominal LO frequency is 9.75 GHz.
| LOCKING
RANGE FOR COMBINATIONS OF INJECTION DRIVE LEVEL AND FREQUENCY (AUN14S) |
||||
| Injection level |
F.inj = LO |
F.inj = LO/2 |
F.inj = LO/3 |
F.inj = LO/4 |
| +10
dBm |
4.26
MHz |
0.18
MHz |
no
lock |
no
lock |
| 0
dBm |
1.09
MHz |
no
lock |
no
lock |
no
lock |
| -10
dBm |
0.36
MHz |
no
lock |
no
lock |
no
lock |
| -20
dBm |
0.1
MHz |
no
lock |
no
lock |
no
lock |
Here is a video (1.8 Mb) showing the LNB LO following an external oscillator.
Cambridge Industries AE88
The injection input can be brought out from the back of the pcb and through the main casting with this LNB also, as can be seen from the following pictures:
With this configuration, the
following results were obtained (nom 9.75 GHz):
|
LNB LO too low in
frequency
response at near lock
LNB LO too high
in frequency |
| LOCKING
RANGE FOR COMBINATIONS OF INJECTION DRIVE LEVEL AND FREQUENCY (AE88) |
||||
| Injection level |
F.inj = LO |
F.inj
= LO/2 |
F.inj
= LO/3 |
F.inj
= LO/4 |
| +10
dBm |
200
MHz + |
4.85
MHz |
0.57
MHz |
0.63
MHz |
| 0
dBm |
23.1
MHz |
0.354
MHz |
0.03
MHz |
no
lock |
| -10
dBm |
2.42
MHz |
0.04
MHz |
no
lock |
no
lock |
| -20
dBm |
1.02
MHz |
no
lock |
no
lock |
no
lock |
EASS LNB
LO frequency: 9.75 GHz. Injection applied to 50R
coupling line load resistor, as in previous examples
| LOCKING
RANGE FOR COMBINATIONS OF INJECTION DRIVE LEVEL AND FREQUENCY (EASS) |
||||
| Injection level |
F.inj
= LO |
F.inj
= LO/2 |
F.inj
= LO/3 |
F.inj
= LO/4 |
| +
10 dBm |
200
MHz + |
0.58
MHz |
0.44
MHz |
0.52
MHz |
| 0
dBm |
7.65
MHz |
no
lock |
no
lock |
no
lock |
| -10
dBm |
1.556
MHz |
no
lock |
no
lock |
no
lock |
| -20
dBm |
0.336
MHz |
no
lock |
no
lock |
no
lock |
Simple
x23 multiplier
As
well as providing a modest gain, the LNBs internal puck oscillator
ought to add selectivity too. A simple single x23 frequency multiplier
was built, as below, to see if any selectivity was evident.
The
HSMS-8102 schottky pair (often used as an LNB mixer, and marked '2R')
is only rated at 75mW dissapation, but since no matching was used, the
drive was initially set at 100mW (well, 100mW into 50 ohms, that
is). Output from the pipe-cap filter was -27dBm, so 37dB of gain was
required to bring this up to 10mW. Two modified LNBs were used for this
purpose.
To
provide better frequency stability, the HP8640 used in the previous
test was replaced with a crystal multiplier chain from an old M296 UHF
receiver. This runs x8 from a 53MHz 3rd overtone crystal. Output was
about 40mW, resulting in several mW from the x23 set-up above. This was
used for the set-up to provide the following pictures, which
shows the various LNB IF spectrums (LH three pictures) and x23
multiplier output spectrum.
The
injection locked LNB in this test was the Cambridge Industries AE88
detailed earlier.
|
IF output - internal LO
only IF output -
LO locked to ext multiplier
IF output - LO locked, no
signal
External multiplier spectrum |
For
comparison, the LH picture shows the IF output spectrum for a strong
3cm signal when only the internal un-locked LNB was used. Second left
shows the same IF otput, but this time with the LNB LO injection
locked. At first glance, the added spurious responses look
dissapointing, but removing the 3cm signal still results in IF output
port components (third LH pic). These are mainly the 424 MHz multiplier
drive leak-through, and its second harmonic, caused by poor screening
of the test set-up - see below:
Many of the IF spurious products are due to intermod between
the LO leak-through and the 3cm signal, and not the sidebands that are
a part of the x23 multiplier spectrum (RH pic of the four spectrum
pictures), so the LNBs internal LO oscillator has provided selectivity
- just a pity that the 424MHz leakage hides the effect.
Another factor to be considered is that the 616MHz IF is some
20dB down the slope of the LNBs IF high pass filter, with this
particular LNB. Some of the newer LNBs (ie, Thomson 13553) do not slope
the IF until about 300MHz, so would have been more suitable
(the IF frequency can't be varied that much, because it is determined
by the injection locked oscillators nominal puck controlled frequency
(+- a few hundred MHz), and the standard
'lo' oscillator for the UK market is 9750 Mhz).
Conclusions
Injection locking is worthy of consideration. The phase noise
of the resulting oscillator is much improved close in, and the far out
noise was good anyway - but no figures here - absolute value
measurements are for another day...