Frye reckons that the diodes (which will be schottky barrier) switch in 2ps. Which implies a 3dB
bandwidth per diode of 175GHz.
The nearest you can get to that performance are microwave schottkys, for example this range of GaAs
devices from MACOM which have a bandwidth of
80GHz. They have a junction capacitance of 0.02pF, which implies that the S4/S6 diodes have a
junction capacitance of 0.01pF (ie 10 femto-farads). And scaling from MACOM's drawings, the actual
junction is about 50um square. The S4/S6 diodes are 30um square - so 0.36 times the area (and
capacitance) of an 80GHz device. Scaling that gives a bandwidth of the S4/S6 diodes of 222GHz -
close enough to the 175GHz worked out from Frye's data.
Which is why the darned things are static damageable. Take a human with a body capacitance of 150pF
and an unspecified but almost always significant voltage up to 10kV, and connect that to a junction
with 10fF capacitance. The result ought to surprise no one.
Note from the MACOM datasheet two things. First is a junction breakdown voltage in the range 4.5 -
7V, entirely consistent with the S4/S6 spec that says do not exceed 5V on the input. And secondly,
handling where MACOM say "Static sensitivity: Schottky barrier diodes are ESD sensitive and can be
damaged by static electricity. Proper ESD techniques should be used when handling these Class 0
devices"
Some interesting numbers here though regarding body voltage when properly grounded
nsitive-operations.php . In particular "international ESD losses that now exceed an estimated $90
billion"
All of which is why I take very great care of my unfixable samplers and fast pulse generators. And I
urge others to do so too.
Craig
toggle quoted message
Show quoted text
Per square nanometer. 1 square nanometer is 30um square - which is the size of the S4/S6 diode
junctions.
This is the designer George Frye's explanation of operation, including a picture of the hybrid
I think it can easily be seen that this is not a repairable assembly.
Craig
