Path: utzoo!utgpu!jarvis.csri.toronto.edu!cs.utexas.edu!samsung!think!mintaka!ogicse!orstcs!guille.ECE.ORST.EDU!daver
From: daver@guille.ECE.ORST.EDU (Dave Rabinowitz)
Newsgroups: comp.sys.handhelds
Subject: Re: SPEED vs FAST question (HP28S)
Message-ID: <16144@orstcs.CS.ORST.EDU>
Date: 22 Feb 90 20:14:17 GMT
References: <4282@rouge.usl.edu>
Sender: usenet@orstcs.CS.ORST.EDU
Reply-To: daver@guille.ECE.ORST.EDU.ECE.ORST.EDU (Dave Rabinowitz)
Organization: Oregon State University, E&CE, Corvallis
Lines: 44

In article <4282@rouge.usl.edu> cs303130@pb.usl.edu writes:
>Also, who determined that FAST causes the calc to use more power than
>in normal mode?  How was that determined, or did the info come from
>HP? What exactly does FAST do (I mean, does it directly affect OS
>software or hardware operation) 

The HP28S uses a 32,768 Hz crystal, like the ones used in most digital watches,
as its reference clock.  The CPU has to run at a much higher speed, so it's
clock is generated by a voltage-controlled oscillator (VCO) whose output is 
divided down by some number and then frequency-locked (patent pending) to the 
reference crystal.  For example, to get 8 MHz from the VCO you divide its 
output by 256 and adjust the VCO frequency until the divided frequency matches 
the crystal frequency.  The frequency divider is programmable, and the FAST 
program simply changes the divisor.  For example, if you divide by 128 instead 
of 256 the clock frequency will become 4 MHz, etc..

The speed of CMOS circuitry varies significantly with voltage.  The voltage
obtainable from batteries goes down as the batteries age.  Since the 28S has to
operate reliably with relatively old batteries whose voltage has dropped, the
operating speed was set to guarantee that the calculator will operate correctly
until the battery voltage has dropped well below the low-battery indication.
In fact, the calculator has two separate low-battery detection circuits: the
first, which triggers at about 4V (new batteries put out a little over 6V), 
tells the software to turn on the low-battery annunciator in the display, while
the second, which trips about .5V lower than the first, forces the calculator
to shut down, preventing incorrect operation from destroying memory.  These
levels were determined to be safe for a calculator operating at the nominal
speed, and are not safe for calculators operating at a higher speed, so it is
possible for a FAST calculator to lose memory or otherwise get incorrect
results even without the low-battery annunciator being on.  The FAST program
should be used only with NEW batteries, and even then there are no guarantees.

As for current drain, the power dissipated by CMOS circuitry varies linearly
with clock speed, so doubling the CPU speed would normally double the power.
However, some of the power in the 28S is used in analog circuitry (mostly
switched-capacitor) which is clocked from the 32 KHz crystal, independent of
the CPU speed, so only some of the power will be doubled and the total power
will increase by less than 2X (as an aside, the power varies as V^2, so the
calculator dissipates more power with new batteries regardless of speed).  The
CPU speed only impacts CPU power, and the CPU runs only when the calculator is
actually calculating something or when a key is being pressed, so when the
calculator is simply displaying results and waiting for something else to do
the CPU is not running and the power drain (very low) is independent of CPU
speed.