I've been continuing my work to develop the "ultimate" beam steering
160 meter antenna system.  The current 4 element broadside array of 2
element end-fire arrays works well, but does not seem to work as well
as the models predict.  Part of the problem is accurately measuring
the gain and phase shift of each of the 4 antenna elements.  If the
gain and phase is known, then the differences can be compensated in
software.  If a 4-channel receiver has known gain and phase errors,
then the antennas can be more accurately measured.

Theoretically, calibration with an external signal could provide data
to compensate for the combined antenna element and receiver accuracy.
So far, my calibration attempts have been only mildly successful.

Along the way, I decided that development would be easier if I put all
4 softrock receivers on the same circuit board, and eliminate the
redundant circuitry.  Also, if I used high accuracy components, maybe
the 4 receivers would track each other better and simplify my
calibration study.

To this end, I've laid out a 4-receiver board based on Tony's Softrock
v6.2 design.  This board has 4 receivers fed from a common local
oscillator and divider chain.  I added pads for an external VFO.  I
also added a power LED.  I used all surface mount components to keep
the size down, and perhaps to achieve better receiver matching.  I
used 3 pole bandpass filters in the front-end because odd-pole filters
seem to be easier to design for constant phase shift than 2 pole
filters.  However, that finding could be my lack of knowledge about
filter design.

Also, there are pads to add bypass capacitors and RF chokes (actually,
ferrite beads) so that one can feed DC through the antenna feedlines
(these parts are not on the BOM).  You'll note that the Softrock
design keeps the antenna "ground" isolated from the receiver ground
via the transformer.  I use this feature in my existing Softrock
receivers to control a relay in each end-fire array.

I used ExpressPCB's $51 3 board "mini-board service" to get my one and
only set of boards.  The layout has only one error, which is the fact
that I ran the +5 Vdc trace under the surface mount voltage regulator.
The data sheet for the LM78L05 didn't mention the fact that the ground
pin connected to the heat sink tab on the outside of the regulator
case, which causes a short circuit when the part is soldered onto the
board (moral: get samples before committing the design).  I'd make a
couple of other minor changes as well if and when I do a board turn.

I forgot to order 6.3v 10uF tantalum capacitors for C8 and the 4 C18s,
so I used some 20v 1uF tantalum capacitors that I had on hand.  The
20v 10uF capacitor used for C6 won't fit in the C18 space.  The
smaller value reduces low frequency response at baseband, but does not
adversely affect overall operation.

I used a TO-92 package LM78L05 instead of the SMD package to avoid the
short, and all else seems to work.  I have not yet measured any
performance parameters, but since its the same circuit as the proven
V6.2, I would expect comparable performance.

On the schematic, I got lazy and only drew one copy of the quadrature
detector circuit.  When I am no longer lazy (or someone objects
strenuously), I'll add the connection symbols that show power and the
quadrature clocks going to the other copies of the detector circuit.

The schematic shows values for only a 160 meter bandpass filter.  If
you really want a different band, go find the Softrock v6.1 schematics
and use those values for the filter circuit.  Also, my filter values
reflect a slightly different design choice than Tony may have used.  I
chose a Butterworth filter design because the Butterworth filter has a
nice compromise between constant group delay and band-stop rejection.
Theoretically, the constant group delay will make multi-channel
tracking simpler.

Victor, K1LT
06 August 2008