Scott Dattalo wrote:
>
> Paul Mathews wrote:
>
> ... The correct way to detect a 2kHz square wave. However, Paul
> don't forget the quadrature waveform! It's possible that the incoming
> signal and the square wave you're beating it against  (i.e. multiplying)
> are 90 degrees out of phase. In which case the dot product is zero.
>
> I've been looking at this fairly closely in the last couple of weeks
> and I've come up with a VERY SIMPLE algorithm that will perform the
> filtering Paul describes. It also takes care of the quadrature waveform
> too. I originally was going to post the theory too, but it's too long
> for the current PIC List temperament. Those of you who enjoy the
> theoretical enlightenment then drop me a line.
>
> If you have seen the theory then.....
> Finally, all of the above can be combined to create the following
> (UNTESTED) code. It takes 15 cycles/sample regardless of the path
> that the code takes (recall the recent "Isochronous Code" thread).
>
> TONE_FREQ       EQU     2000  ;Hz
> SAMPLING_TIME   EQU     15    ;uS
> RESOLUTION      EQU     2<<13 ;8192
> DPH             EQU     SAMPLING_TIME*TONE_FREQ*RESOLUTION/1000000
> Qbit_c          EQU     13-8  ;Bit 5
> Qbit_s          EQU     12-8  ;Bit 4
>
> start:
>         MOVLW   NUM_OF_SAMPLES
>         MOVWF   samples
> samp_loop:
>         MOVLW   DPH
>         ADDWF   ph_lo,F       ;Low byte of phase
>         SKPNC
>          INCF   ph_hi,F
>
>         BTFSS   INPUT_PORT,INPUT_BIT
>          goto   low
>
>     ;Input is high
>         BTFSC   ref_wave,Qbit_c ;
>          INCF   DFT,F           ;c=1,s=x,in=1
>         nop
>         goto    next
>
> low:
>     ;Input is low
>         BTFSS   ref_wave,Qbit_c ;
>          goto   $-2
>         BTFSS   ref_wave,Qbit_s
>          INCF   DFT_s,F         ;c=1,s=0,in=0
>         nop
> next
>     ;Put a delay loop here if your crystal frequency>4Mhz
>
>         DECFSZ  samples,F
>           goto  samp_loop
>
>     ;At this point the variable "DFT" has the sum of both
>     ;quadratures. We can decide/experiment with thresholds
>     ;to see  how well they correlate to actual signal.
>     ;
>     ;  Do  some more system stuff...
>
>         goto     start
>
> Scott
>
> PS The theory behind this is almost identical to the
> theory behind the DTMF decoding.


I'm glad that I started things off....however, note that the algorithm I
sketched out (admittedly sketchy and non-optimal) has starting edge
detection, so I don't understand your point about the quadrature wave.
Upon calculating a dot product of zero, the process restarts.

I've implemented more sophisticated versions of this algorithm in many
ways: beginning in 1978 using 9601 monostable multivibrators, in 1980
using custom ASICs, in 1982 using microcontrollers, in 1992 using PGA,
etc., so I'm quite confident that it works.
--

Paul Mathews, consulting engineer
AEngineering Co.
optoeng@whidbey.com
non-contact sensing and optoelectronics specialists