--- /dev/null	Thu Jan 01 00:00:00 1970 +0000
+++ b/venus/doc/Star-points-in-core	Mon Nov 22 09:54:48 2021 +0000
@@ -0,0 +1,66 @@
+iWOW's TR-800 module, whose reverse-engineered PCB layout will serve as a
+reference for the Calypso core of FC Venus, is nothing but a mass-produced
+version of TI's Leonardo reference design.  We don't have any original PCB
+layout files for Leonardo, but we do have schematics, in 3 different versions:
+
+ftp://ftp.freecalypso.org/pub/GSM/Calypso/Leonardo_plus_RD122.pdf
+ftp://ftp.freecalypso.org/pub/GSM/Calypso/Leonardo_plus_quadband_schem.pdf
+ftp://ftp.freecalypso.org/pub/GSM/Calypso/Leonardo_rev05.pdf
+
+TR-800 is based on quadband Leonardo+ (that's what makes it interesting), but
+the hobbled 2-band version of Leonardo is the only one for which we have TI's
+original DSN (OrCAD) schematic file, in addition to the corresponding PDF:
+
+ftp://ftp.freecalypso.org/pub/GSM/Calypso/Leonardo_rev05.dsn
+
+TI's group in Aalborg (Denmark) that designed and built their development boards
+used OrCAD for schematics and PADS for PCB layout.  We have their PADS PCB
+design file for a different board from about the same time frame (E-Sample),
+but not for any Leonardo variant.
+
+If you look at the baseband page of Leonardo schematics (this part is the same
+in all 3 versions), you will see 3 symbols denoting star routing: HST201 and
+HST202 (labeled "STAR_CONFIG") split V-DBB net into VDD-PLL and VDD-CORE
+branches, and HST203 (labeled "GND_POINT") creates a special ground for the
+32.768 kHz oscillator circuit, connected to the rest of board ground at just
+one point.
+
+This star routing in TI's reference version of Calypso/Leonardo core exists not
+only in TI's schematics, but also in the physical PCB layout which we have
+recovered from TR-800 by reverse engineering.  If you look at the Gerber files
+which we received from the PCB RE company (published on our FTP site in 2021),
+specifically the OKPCBM~1.G1 gerber file corresponding to the inner layer just
+below the surface (L2 or L5 depending on which way you count), you can see the
+physical features that were once produced, presumably in PADS, to effect the
+star routing depicted in the schematics.
+
+The physical PCB star routing feature corresponding to HST201 & HST202 resides
+just underneath C214 footprint.  There is a microvia (blind via going only one
+layer in) inside C214 pad 1 (the left pad in the Gerber view), and on the inner
+layer where this microvia goes, one can clearly see the feature that splits the
+V-DBB net coming from C214.1 into VDD-PLL and VDD-CORE branches.  The lower (in
+the Gerber view) branch that goes to two microvias back to the surface is
+VDD-PLL, whereas the other branch that goes to an inner buried via (spanning L2
+through L5) is VDD-CORE.
+
+The other star routing feature in the TR-800 PCB corresponding to TI's HST203
+can be seen in the far lower right corner of the same OKPCBM~1.G1 gerber file:
+look for an L shape that connects a short thin trace (going to a microvia to the
+surface) to the surrounding GND copper pour.  The short thin trace connected to
+the GND copper pour by the L shape is what we named GND_32khz in
+src/core/dbb_block.v, and the L shape on the inner layer is what eventually
+became of HST203.
+
+The present ueda netlist source code for FC Venus is written with the
+expectation that the same pseudo-component EDA trick for star routing will be
+used on our board, using intnoconn feature of pcb-rnd:
+
+http://repo.hu/cgi-bin/pool.cgi?project=pcb-rnd&cmd=show&node=intnoconn
+
+There is one difference in this regard between TI's original version and ours.
+In TI's original version all 3 pseudo-component objects HST201, HST202 and
+HST203 were 2-way stars, with HST201 and HST202 being somehow combined in PADS
+layout to form the desired 3-way star.  In our version 3-way star HST200
+explicitly replaces TI's HST201+HST202 pair.  The actual pcb-rnd subcircuits
+for the 3-way star HST200 and the 2-way star HST203 will need to be created by
+the PCB layout engineer.