From: bill.sloman@ieee.org   
      
   On 18/02/2026 4:34 am, Phil Hobbs wrote:   
   > On 2026-02-17 06:40, piglet wrote:   
   >> On 16/02/2026 22:08, Christopher Howard wrote:   
   >>>>>>   
   >>>>>> Circuit A is the basic squaring concept, relies on modern open-drain   
   >>>>>> output comparators being pretty good switches to ground. Assumes you   
   >>>>>> already have a source of sawtooth or triangle waves with defined zero   
   >>>>>> and peak values in the system.   
   >>>>>>   
   >>>>   
   >>>   
   >>> I had a few follow questions, if you please, regarding the PWM squarer   
   >>> (circuit A):   
   >>>   
   >>> - Does this circuit work in quadrant one only? If so, would any slight   
   >>>    modification of the circuit allow for handling negative input   
   >>> voltages   
   >>>    also?   
   >>>   
   >>> - Could you please tell me a little bit more about the low pass filter   
   >>>    portion of the circuit? Does that particular configuration have a   
   >>>    name?   
   >>>   
   >>   
   >> The circuit I showed gives output of zero for negative inputs (the   
   >> input would always be below even the lowest point of the ramp so the   
   >> comparator output would always be low).   
   >>   
   >> To handle negative inputs precede it with a precision full wave   
   >> rectifier aka absolute value circuit.   
   >>   
   >> Good place to look for examples:   
   >>    
   >   
   > It's not applicable here on account of the huge offset voltage and   
   > one-quadrant operation, but FWIW you can make all sorts of integer and   
   > fractional powers using glorified current mirrors.   
   >   
   > Example:   
   >  Screen shot   
   >    
   >    
   >   
   > The shunt resistor effectively nails one or more of the emitters still,   
   > which produces the change in power law.   
   >   
   > I'm using a square-root version of this one in a demo for a circuit   
   > whose bandwidth wants to be quadratic in the supply voltage.  It's fine   
   > inside a feedback loop, and may eventually get put in an ASIC.   
   > (Unfortunately they don't make the CA3096 anymore.)   
      
   But Renesas still makes the HFA3096, which is faster, even if the base   
   series resistance is a bit high, as you pointed out when I last drew   
   their parts to your attention.   
      
   https://www.renesas.com/en/products/hfa3096?srsltid=AfmBOopj2-kt   
   GNo1-1Vks_ne_gw3vh8ftXdGh1voCdQ5u-c6I3y0GeU   
      
   --   
   Bill Sloman, Sydney   
      
   --- SoupGate-Win32 v1.05   
    * Origin: you cannot sedate... all the things you hate (1:229/2)