Thank you for being annoyed. I thought I was the only one who
was bothered by this.
8-) Barry
At 07:26 AM 1/15/03 -0700, Micro Eng wrote:
>Just pondering this on my way to work...after being annoyed by some of the
>new LED based taillights
>
>If you want to vary the brightness of an LED, would it be better to vary the
>voltage, say a PWM circuit driving a voltage-follower transitor, or adjust
>the circuit voltage (ie the resistor using a digital pot) ?
> Thank you for being annoyed. I thought I was the only one who
> was bothered by this.
What about these new lights annoys you? I thought they were pretty cool,
and frankly am suprised they didn't emerge earlier. Do you feel they're
too bright? Just don't like 'em?
At 06:21 PM 1/15/03 -0600, you wrote:
>On Wed, 15 Jan 2003, Barry Gershenfeld wrote:
>
> > Thank you for being annoyed. I thought I was the only one who
> > was bothered by this.
>
>What about these new lights annoys you? I thought they were pretty cool,
>and frankly am suprised they didn't emerge earlier. Do you feel they're
>too bright? Just don't like 'em?
They are too directional, the brightness is too bright right on the
axis and dim off the axis. I saw some really neat LED-look incandescent
tail lamps on a KIA SUV or truck. The effect was as a result of the
design of the acrylic lens, it looks like large number of LEDs arranged
as radiating spokes.
; Do operations 2 thru NSize to compute the least significant bytes.
;
; Operation 2 multiplies the two pairs:
; A<0> by B<1> and A<1> by B<0>
; to compute R<3:2:1>
;
; Operation 3 multiplies the three pairs:
; A<0> by B<2>, A<1> by B<1> and A<2> by B<0>
; to compute R<4:3:2>
;
; The NSize'th operation multiplies the NSize pairs:
; A<0> by B<NSize-1>, A<1> by B<NSize-2> ... A<NSize-1> by B<0>
; to compute R<NSize+1:NSize:NSize-1>
;
; Note that for the I'th operation the sum of the subscript for each pair of
A and B
; bytes multipled together is I-1
LFSR 2,R+1 ; Each operation will increment FSR2
I = 2
while I <= NSize
; point at start of A
LFSR 0,A
; point at first byte of interest in B
LFSR 1,B+NSize-I
; call into the correct location within the MultiplyVector routine
; to multiply I byte pairs, accumulating them in R<FSR2+1:FSR2>
RCALL MultiplyVector + 16 * (NSize-I)
I = I + 1
endw
; Continue computing the output bytes for the high order bytes of the result
;
; Now do operations numbered NSize+1 thru ?
;
; Operation NSize+1 multiplies NSize-1 pairs:
; A<1> by B<NSize-1>, A<2> by B<NSize-2>, ..., A<NSize-1> by B<1>
; to compute R<NSize+2:NSize+1:NSize>
;
; Operation NSize+2 multiples NSize-2 pairs:
; A<2> by B<NSize-1>, A<3> by B<NSize-2>, ..., A<NSize-1> by B<2>
; to compute R<NSize+3:NSize+2:NSize+1>
;
; Operation NSize+3 multiples NSize-3 pairs:
; A<3> by B<NSize-1>, A<4> by B<NSize-2>,..., A<NSize-1> by B<3>
; to compute R<NSize+4:NSize+3:NSize+2>
;
; Operation NSize + NSize - 2 multiplies 2 pairs:
; A<NSize-2> by B<NSize-1>, A<NSize-2> by B<NSize>
; to compute R<NSize*2-1:NSize*2-2:NSize*2-3>
; Note: at this point I = NSize+1
while I <= NSize * 2 - 2
; point at first byte of interest in A
LFSR 0,A+I-NSize
LFSR 1,B+NSize-1
; call into the correct location within the MultiplyVector routine
; to multiply 2*NSize-I byte pairs, accumulating them in R<FSR2+1:FSR2>
RCALL MultiplyVector + 16 * (NSize-(2*NSize-I))
I = I + 1
endw
; Finally the last operation to multiply together A<NSize-1> and B<NSize-1>
and
; finish up computing the result
; Here is the kernel of code that multiplies all the pairs that have the
same significance
; Note that each byte's worth of multiplication takes 6 words of
instructions
MultiplyVector:
I = 0
while I < NSize-1 ; NSize copies of the code
movf POSTINC0,W ; get byte from A
mulwf POSTDEC1,W ; multiply by corresponding byte from B
movf PRODL,W
addwf POSTINC2,F ; accumulate it in correct location
movf PRODH,W
addwfc POSTDEC2,F
skpnc
incf HITEMP,F
I = I + 1
endw
; On the last iteration we leave FSR2 pointing to the next byte of the
product,
; and also bring in the high byte
movf POSTINC0,W ; get byte from A
mulwf POSTDEC1,W ; multiply by corresponding byte from B
movf PRODL,W
addwf POSTINC2,F ; accumulate it correct location
movf PRODH,W
addwf POSTINC2,F ; point to highest byte
movf HITEMP,W
movwf POSTDEC2,F ; point to new middle byte
clrf HITEMP ; get ready for next operation
return
; Timing analysis:
;
; Call, return, initialization and multiply of LSBits takes 12
; Operations I = 2 thru NSize+NSize-2 take 4 + call on MultiplyVector
; Multiply of MSBits takes 6 instructions
; MultiplyVector takes 8N + 5 instructions including call and return.
; MultiplyVector is called once with N = NSize
; and twice each with N=1..NSize-1
;
; The total time used, not counting MultiplyVector is:
; 12 + 4 * (NSize*2 - 3) + 6
; = 12 + 8*NSize - 12 + 6
; = 8*NSize + 6
;
; The total time used by MultiplyVector is:
; 2*8*1 + 5
; + 2*8*2 + 5
; + 2*8*3 + 5
; + ...
; + 2*8*NSize-1 + 5
; + 8*NSize + 5
;
; = 2*8* ( 1+2+3+..+NSize-1) + 8*NSize+5*NSize
; = 2*8*(NSize-1*NSize)/2 + 8 * NSize+5*NSize
; = 8*(NSize-1*NSize) + 8*NSize + 5*NSize
; = 8*(NSize*NSize-NSize+NSize) + 5*NSize
; = 8*NSize*Nsize + 5 * NSize
;
; The grand total time is then:
; = 8*NSize*NSize + 13*NSize + 6
;
; So, for NSize = 32 we get:
; 8*32*32 + 13*32 + 6 = 8192 + 416 + 6 = 8614 instructions.
;
; At 10MIPS that will take 0.8614 msec.
mitvma.mit.edu with SET PICList DIGEST in the body
