Brent Brown <brent.brown <at> clear.net.nz> writes:

> Very cool circuit. Tidy, low parts count, low cost, good performance, I like
it. > I guess it could work on 3.3V ok, make R1 approx 470R?

Actually scale R1 so the minimum clamp current you need is available. I.e. if
you need Iclamp = 10mA then set R1 for Iclamp/20 through the mirror or so.
Simulation with Vcc=3V and the circuit shown, with R1=1K shows that the clamp
still works fine (with slightly higher drop but also higher current). If you
analyze the circuit you will see that Q4 and Q3 (and the other side likewise)
form a standard current mirror. When the clamp input is at the rail the current
it that in the mirror. When the clamp goes about 100mV above the rail the 
entire current through R1 times beta of Q3 is available as clamp current, as Q4 
is shunted by Q3's emitter being pulled above the rail. Thus the maximum 
current in Q3's emitter is (f.ex. Vcc-3V, beta=50, Ir3=1.7mA -> 85 mA !). 
Clearly you can increase R1 to reduce quiescent about four times and still be
able to clamp 10 mA safely. E.g. with R1=5k the mirror current is ~350uA and
the clamp will go to +/-50mA and not exceed +/- 250mV from the rails.

Note: as you increase R1 you want to add decoupling to the bias circuit. For
this, for R1 below 1k add one 10nF cap in parallel with each: R1, Q1 and Q3.

Other:

When the power is off the clamp capability is reduced to Ibmax of the output
transistors. Also when the power is off the clamp will try to generate reverse
voltage on the rails ! A reverse-protecting TVS or diode must be installed
across the rails to avoid this.

Crosstalk between 'channels' is -60dB (with 3 x 10nF fitted as above).


Turn-on with 3 x 10nF will cause a small current spike with 10nF across R1.
Adjust (lower the value) of the cap across R1 to avoid it.

Leakage with Vdd+0.1V <= Vin <= Vss-0.1V is <100nA, subject to crosstalk as 
above.

This clamp gets better as temperature increases (and Vbe decreases). This 
is the opposite of how other clamps work ...

The circuit is given as a concept, known to work. Use it at your own risk, and
please run tests as needed to assure it works in your system.

thanks,
Peter P.


-- 
http://www.piclist.com PIC/SX FAQ & list archive
View/change your membership options at
http://mailman.mit.edu/mailman/listinfo/piclist