[hpsdr] ALEX - Call for Comments - II

[Lyle Johnson]

> 2.) Control...
> This is a receiver front end preselector.  I would prefer
> to not have any in-band oscillators, clocks, or noisy high speed
> buses appear on the card.  For that reason I would propose that
> this be a "dumb" slave card, controlled by a three wire SPI bus,
> which would be filtered upon entrance.  I2C would also work.
> Phil H wanted I2C, I personally prefer SPI, Lyle seems to like SPI.
> 
> It could be controlled by any other (one) CPLD on the Atlas bus.
> 
> Space is also an issue. The universal CPLD bus interface takes a
> noticeable portion of a 100mm x 120mm card.  I am mostly worried
> about noise. (Read Henry's comments again.)

What if we use the CPLD but don't connect it to the entire bus.  Connect 
it to the pins defined for I2C, SPI, JTAG (in case we want to reprogram 
the CPLD).  Then it is just an SPI or I2C device that happens to have a 
few extra pins.  Can still filter the control lines...

>  ...
> I humbly apologize in advance for proposing to put a card in
> a Software Defined Radio, which has no software on it.  :-)

Hey, don't feel bad.  I did the Pinocchio card!  And Phil C did Atlas 
before he did Ozy :-)

> 4.) Physical design...
> If we go with SPI or I2C control, I intend to try to make
> this a two sided PCB design. The inductors are thru-hole
> shielded cans, about 1/2 inch square, three per filter
> section.  Ground plane, inductors and control on the top side.  
> RF interconnect, and tuning chip caps on the bottom side.  
> Option to put a shield (0.031 in. one sided PCB) over the entire
> bottom, if necessary to control noise pick-up.  

My greatest concern for a two-layer board is isolating the control lines 
for the Hittite switches.  Since the same pair of control signals has to 
go to both ends of the filter, there is a real opportunity for coupling 
of signals to the control lines.  It is easier to isolate/decouple this 
on a 4-layer board, but I'm sure it can be done on a 2-layer board.  But 
it might require more decoupling components or a more complex decoupling 
topology, offsetting the cost savings.

73,

Lyle KK7P


 1173220532.0