[hpsdr] HORTON

[Eric Blossom]

I've been doing some thinking about HORTON.

It seems to me that the biggest problem with getting the QSD to run
fast is finding switches/muxes that are fast enough.

In my searching, the fastest CMOS FET mux I've come across to date
is the TI SN74CB3Q3253.  
http://focus.ti.com/docs/prod/folders/print/sn74cb3q3253.html

The problem with these isn't the "High-bandwidth data path (upto 500 MHz)",
but rather it's the relatively slow speed at which you can switch
them.  According the specs, this part maxes out at 20 MHz, but perhaps
it could be pushed a bit past that.

The ADG901 has a similar problem.  It's flat out past 1 GHz, but
the typical switching time is 4.7ns (worst case 7.75ns: 129 MHz).

I'm interested in getting a QSD running at at least 100 MHz, but why
not try for 145 MHz while we're at it.

To run at 100 MHz (passive radar application), the switches need to
work at 400 MHz.  You're feeding it quadrature LO at the center freq,
but the difference between I & Q is 90 degrees.

Sythesizing the control of the switches shouldn't be a problem.  E.g.,
use a DDS to generate the I & Q LO, filter, run that into one of the
very nice ADI clock distribution parts (AD9511 and friends).  These
will give us differential LVPECL out.  A bit of LVPECL fanout
buffering and some AND/NAND ECL will generate the switch enable
signals, no problem (modulo level translation).  We should be able to
run this at 700 - 800 MHz, no problem.

So, then we return to the original question: What's the fastest analog
switch we can find? 

Are there faster FETs available in technologies other than CMOS?
E.g., GaAs or SiGe?  

I don't think we need the signal path to be flat out past 150 MHz,
but we do need them to switch quickly ;)


Regarding the converter, the AD7762 (parallel) or AD7763 (serial) look
like a better fit than the AD7760.  The '62 and '63 sample at 625kS/s
(or lower, it's a delta-sigma part), with better dynamic range than
the 7760.


I'd like to end up with a system that contained 4 HORTONs, a board to
generate the quadrature LO and sample clock for all 4 HORTONs, and
some way to get it all back into the PC.

Comments, suggestions, etc?

Eric