Its really NO mystery why a narrow "roofing" filter didn't help. It just
means that the IMD isn't limited by the first IF and second mixer
whether close in or with wide spread signals. It means that the IMD is
determined by the RF and first mixer stages. It does require that the
second mixer not intermod with a signal stronger than that was at the
first mixer, by at least the first IF gain. That's entirely practical.
Though it means the second mixer may well have to handle +20dBm signal
and need a +27dBm (1/2 watt!) local oscillator in order to have 20 dB
gain in the first IF and to handle 0 dBm at the first mixer. Makes that
first IF stage and mixer and its local oscillator look like a PA stage
with lots of idling current and heat sinks. It can require careful
construction of the crystal filter to keep it from being a source of
intermod.
When (as in the case of many radios from Japan) the IMD is far worse
inside the passband of the "roofing" filter it shows that the intermod
performance of the stages between that filter and the real selectivity
are not built to handle the signal levels possible there going through
the "roofing" filter and outside the main selectivity and so not cut
down by AGC.
Its vitally important that the mixer ports be properly terminated over a
wide band also. Any time a mixer product (fundamental or oscillator
harmonic or any mixing product of any of those) is reflected back to the
mixer it makes more signal for the mixer to handle and so lowers the
threshold of intermod. A crystal filter is a rotten termination for a
mixer, probably matched only in the passband, very reflective at all
other frequencies. Sometimes the reflected signal also increases mixer
conversion loss. It is possible to tailor the reflection to reduce mixer
conversion loss, but in the process it cuts the IMD performance by 20 or
30 dB and makes the mixer noise figure significantly greater than the
insertion loss.
When the IF bandwidth is fairly wide, its possible to use a termination
filter that passes the IF unchanged, while loading frequencies outside
the IF bandpass. When the IF is going right into a crystal filter, there
most often is a wide band gain stage with lots of feedback to improve
its bandwidth and impedance performance. Like the RF stage in the
Corsair II. Or like the RF stage in the Omni V and VI with a gaggle of
JFETs in parallel grounded gate. There was a TI JFET made in Europe that
was liked for that purpose, took only one but its been out of production
a long time. The last attractive yet effective terminator is a well
built broad band attenuator, say 6 to 10 dB right after the mixer. Also
a 3 or more dB pad on the local oscillator or signal input to the mixer
is part of that necessary wide band termination.
This need for proper mixer termination has been known for several
decades, and the Yaesu 767 VHF modules do it and perform quite well. The
later Yaesu 736 terminates the mixers with a "roofing" filter and it has
rotten intermod inside or outside the "roofing" filter passband and
terrible sensitivity. A design failure.
There is a very good book on mixers by Maas, and Ulrich Rohde has three
editions of his book on receivers. Between them they (Maas and Rohde)
explain good and bad modern receiver design.
As for first IF filters when the radio up converts to a low VHF first
IF, that "roofing" filter selectivity isn't very good, surely to pass FM
and I think often to allow the first conversion oscillator to move in 10
KHz steps while the DDS oscillator only has to cover a narrow range
while doing the fine steps. That way the general coverage PLL for the
first mixer can have fairly decent phase noise performance. Its really
hard to have good phase noise and to cover DC to 30 MHz (or to light) on
1 Hz steps. Rohde's books spend a lot of time about that problem too.
Then those receivers that up convert to a VHF first IF often have
nothing but a 30 MHz low pass filter before the RF and first mixer
stages. That's all it takes for noise figure performance, but it exposes
those stages to BC and SW BC and other band operations strong signals
that are handily rejected in the ham band only Tentecs by a bandpass
filter before the RF stage. Which should make an Omni V or VI (or
Corsair II) a far better performer in a multi band FD or contest
station. Though there are bandpass filters on the market to improve
those radios lacking that fundamental part when in the multi band FD or
contest station.
It is difficult if not impossible to make a CW bandwidth filter at 30 to
70 MHz but I may try anyway. Looks like in a ladder filter it may take
20 or 30 stages (of matched crystals) to get adequate rejection because
the shunt C of the crystal is about the same as the computed C to ground
between crystals. That makes the rejection only 6 dB per crystal.
73, Jerry, K0CQ
--
Entire content copyright Dr. Gerald N. Johnson, electrical engineer.
Reproduction by permission only.
|