>It is traditional in step down converters to "feed forward" the
>effect of the input voltage to reduce the on time as input
>voltage rises.

/In the context of switching regulators, feedforward is a technique
/used (when it is needed and appropriate) for a very specific and
/narrow purpose: to improve a regulator's line transient response.
/The essence of this technique is to provide a means whereby sudden
/voltage changes at the input are quickly, but only approximately,
/cancelled out by the feedforward mechanism, while allowing the
/feedBACK mechanism (notice the emphasis) to take care of the
/longer-term input variations and maintain the output voltage against
/changes in load.

PRESCRIPT: I'm sure that you are aware of the details of the circuit's
operation and its fine points so I think that the difference is mainly one
of terminology between us. However, the conversation looks interesting so
here we go ...

Indeed. Dynamic response is indeed the major target of fedforward. This is
not to say that it isn't, cannot be, or even shouldn't be used to respond to
more static conditions. As Saint Albert is quoted as saying (I'm always
suspicious of the vast number of quotes attributed to Einstein)

     "Everything should be made as simple as possible,
        but not simpler."   --Albert Einstein

As I imagine you will agree, there is no holy writ about what techniques
should or must be used in electronics. As long as the pertinent laws of
physics are understood and obeyed all will be well. Many designs incorporate
"loops within loops" for various reasons such as a current feedback and
voltage feedback loop within the same design.

/Appropriate use of feedforward techniques can make the design of a
/feedBACK (emphasis, again) system easier by partially relieving it of
/the burden of having to make large adjustments very rapidly in
/response to sudden process changes.

Agree.
The point I was making was that, in the present application, the hysteresis
mechanism is applying POSITIVE feedback to the control loop when NEGATIVE
feedback is what the main control loop is trying to provide. ie here a
rising input voltage tends to produce a rising output voltage and the
regulator will produce a shorter duty cycle to compensate. However, the
hysteresis resistor from the hot side of the buck coil helps to increase the
duty cycle with increasing input voltage at a time when we want it to be
decreasing. I was concerned with the relatively static case but In a dynamic
situation this feed forward would also be of the opposite sense to that
desired.

/Other than for this particular purpose, I have never seen feedforward
/used in a switching regulator.

I have. Mayhaps it is less usual in modern designs as ICs become more able
to integrate all desired functions.

/And if I had, I would have considered it a Band-Aid applied by the
/designer to cover up some shortcoming in his feedback system.

As above.
If you can simplify a design and/or improve its performance overall and to
do so you use a mechanism such as static feed forward then I sugesst that
Einstein's adage firmly applies.

/ As Richard pointed out, these appliques tend to be a "select on test"
/ sort of thing.

I think the select on test nature in this case is due to the fact that , as
noted above,  this resistor applies the opposite sense of "feed forward" to
that desired. This is entirely acceptable in such a simple and elegant
design if it meets the user's needs in all cases - such a circuit makes some
compromises to achieve its simplicity. One way of preventing this
undesirable increasing of the reference voltage would be to provide a
voltage which toggles to a fixed high value when the switch is on and to
ground when it is off. This would add complexity and cost - the zener
clamped resistor is a reasonable compromise and your zener-less
simplification is OK if transient response and restricted input range are
OK - as they may be in many cases. It will be interesting to examine
transient response of the various designs proposed so far. I will happily
admit that I do not (yet) know the formal transient performance of "my"
design - what I do know is that it is entirely adequate in the application I
have primmarily put it to so far. I also suspect that it will prove adequate
in many other cases but when I build the low power version I have been
threatening I will try some step load and input variations and report the
results.  I'll also compare the simulated and actual results and comment.


regards


                Russell McMahon

--
http://www.piclist.com hint: PICList Posts must start with ONE topic:
[PIC]:,[SX]:,[AVR]: ->uP ONLY! [EE]:,[OT]: ->Other [BUY]:,[AD]: ->Ads