Dear John and LF group,
I was interested in G4CNN's experiments with loop antennas - in
response to some of the issues he raises here is another long E-
mail - please delete now if the subject bores you:
Fets have notoriously loose tolerance on their biasing parameters -
for a one-off design, the easiest way to get the bias current you
want is to "select on test" the source bias resistors. A possible
problem with G3LNP's preamp is that there are no source bias
resistors, which means that the drain resistors must be varied to
obtain the correct bias voltage, which also directly varies the gain,
and also means the FETs need to be reasonably matched. But so
long as both have reasonably similar voltages on the drains, they
probably are. Then again, you could find some FETs with the
correct Idss, which is what John has effectively done. On the plus
side, it also runs the FET's at the highest practical drain current,
which usually gives the best noise and dynamic range.
The J310's have better low noise performance, higher operating
current, but not much reduced spread in parameters compared to
the 2N3819 (Idss 24-60mA, against 2 - 20mA). You would want to
use source bias resistors with the J310.
A more significant flaw in the 'LNP preamp as shown in the LF
handbook is that, unless the quiescent drain voltages of the FETs
are matched to within less than 60mV (unlikely), one of the bipolar
output transistors will be cut off all the time, due to the "differential
amplifier" biasing connection of TR3 and TR4. This would
effectively turn it into a single-ended preamp. Again, this could be
got round by using seperate emitter bias resistors & capacitors,
and separating the transformer windings. Or you could lose them
altogether, and redesign the transformer to match the FET outputs
directly to the RX input.
It is certainly possible to tune loops using a transformer to 'step up'
the capacitance of the tuning capacitor. At University of
Hertfordshire, we designed some loops for EMC measurements,
one using a loop of 40-or so uH to tune 100kHz to 1.6MHz in
octave bands, using a 365+365pF variable C by doing just this.
The problem from the LF amateur point of view is that it is difficult
to make transformers with very low losses that would not degrade
the overall Q, especially after you had gone to all the trouble of
using litz wire. This was not an issue with the UofH loops, which
needed a fairly wide bandwidth.
Discrete JFETs have lower noises than FET op-amps - the
relevant paramenter is the eqivalent input noise voltage , "e
suscript n", in nanovolts per root hertz (noise current from most
FETs is too low to make much difference most of the time but
maybe does for the highest Q loops). Eg - OPA111 6nV/root Hz,
2n3819 about 4, J310 1nV/ root Hz. Op - amps have the benefit of
high loop gain, which means less distortion, if the gain-bandwidth
product is large enough. Gain bandwidth product roughly means -
the gain of the closed loop amplifier multiplied by its 3dB cut-off
frequency is a constant. for the OPA111 this is about 2MHz, so for
a unity gain circuit, bandwidth should also be roughly 2MHz. High
voltage gain is not needed; going from 100's of kilohms impedance
at the loop terminals to 50 ohms at the RX results in enormous
power gain even with a voltage gain of 1. However, if the RX input
actually is 50ohms, many op-amps would struggle to drive it.
Similarly, very low noise is not required for High-Q loops. At
136kHz, 3mH inductance, and a Q of a few hundred, the equivalent
parallel resistance of the loop will be several hundred k. The
thermal noise voltage across a 500k resistor is about 90nV/root
Hz, higher than most FETs and op-amps, so amplifier voltage noise
is not that critical. However, this high source impedance might
mean that the current noise is significant, but I haven't worked it
out. I suspect cheaper FET-input op-amps than the OPA111, which
is now quite an old & expensive design, would work just as well.
I have tried a different aproach to an LF receiving loop, which has
some atractive features. Basically, I am using a single turn loop
with 3m sides. This is connected to the shack via a 1:1 balun
transformer and about 20m of RG58 coax. This lot has an
equivalent inductance of some 10's of uH at LF. At the shack end it
is resonated with a series inductor of a few mH , and parallel 500p
+ 500p variable C, connected to the gate of a single-ended FET
preamp. This gives the big advantage of being tuned over a wide
range from inside the shack, using different series inductors. The
tuning circuit is a long way from optimum from the noise point of
view, but the large area of the loop means external noise still
dominates, even under quiet band conditions. You get a good null,
and it is unaffected by rain, being relatively low impedance. The
obvious problem is that it is bigger than the normal loops - mine
uses a 5m wooden pole, with 2.4m bamboo cross-arms. However,
with some optimisation, It could well get quite a bit smaller. It has
been very useful for nulling Rugby on 73k (although it also nulled
G3YXM very effectively last night!).
Hope this is of some interest,
Cheers, Jim Moritz
73 de M0BMU
|
