[GreenKeys] Minimal DSP TU Working!

[epvgk at limpoc.com]

On Mon, Mar 14, 2022 at 07:02:51PM -0700, Harold Hallikainen via GreenKeys wrote:
> I've wanted to learn a bit about DSP, so a TU seemed like the perfect
> project. It is now driving my model 15 with ITTY!
> 
> https://drive.google.com/file/d/1VSPhojioKn0XA0mof5SyRESE7UNYKEyh/view?usp=sharing
> 
> This is running on a PIC32MZ. I plan on checking how much time it is
> spending doing stuff and perhaps it would run on the PIC32MX.

Nice!  A while back I was interested in doing a software-only TU board and was able
to get a minimal RTTY demod working on an 8 bit AVR (atmega328p iirc). It was quite
basic using an integer FFT implementation, but if that was able to work, your 
substantially more advanced one should be pretty portable as well!

eric

> 
> Here's my code as of right now:
> 
> int main ( void ){
>     biquad *MarkFilter; // Audio BPF for mark
>     biquad *SpaceFilter; // Audio BPF for space
>     biquad *MarkDataFilter; // Mark LPF after filtered data squaring
>     biquad *SpaceDataFilter;
>     uint16_t AdcSample;
>     double DdsFreq=100.0;
>     /* Initialize all modules */
>     SYS_Initialize ( NULL );
>     TMR2_Start();
>     OCMP1_Enable();
>     AudioPwmSet(0.0);
>     ADCHS_ChannelConversionStart(4);
>     MarkFilter = BiQuad_new(BPF, 0.0, 2125.0, 8000.0, 50.0);    // Q-50
>     SpaceFilter = BiQuad_new(BPF, 0.0, 2295.0, 8000.0, 50.0);
>     // Create biquad BPF for mark and space, 0 dB gain, 2.125 or 2.296
> kHz, 8 kHz sample freq,, 50 Jz BW
>     MarkDataFilter=BiQuad_new(LPF, 0.0, 136.0, 8000.0, 0.707 );
>     SpaceDataFilter=BiQuad_new(LPF, 0.0, 136.0, 8000.0, 0.707 );
> 
>     while ( true )
>     {
>         if(Timer2TimeoutCounter==0){        // We have timed out 10 times,
> so it has been 125 us
>             if(ADCHS_ChannelResultIsReady(4)){
>                 AdcSample=ADCHS_ChannelResultGet(4);
>                 AdcSamplef=(double)(AdcSample-2048)/2048.0;
>                 ADCHS_ChannelConversionStart(4);
>             }
>             // Tone Sweep
>             DdsFreq*=1.00001;
>             if(DdsFreq>2295.0) DdsFreq=2125.0;
>             DdsFreqSet(DdsFreq);
>             // End tone sweep
>             //sample=DdsNextSample();
>             MarkSample=BiQuad(AdcSamplef,MarkFilter);
>             SpaceSample=BiQuad(AdcSamplef,SpaceFilter);
>             DiscrimOut=BiQuad(fabs(MarkSample),
> MarkDataFilter)-BiQuad(fabs(SpaceSample), SpaceDataFilter);
>             if(DiscrimOut>=0){      // Mark
>                 LATDbits.LATD4=1;
>                 LATBbits.LATB1=0;   // light green LED
>                 LATBbits.LATB5=1;   // red LED off
>             }else{
>                 LATDbits.LATD4=0;
>                 LATBbits.LATB1=1;       // green LED off
>                 LATBbits.LATB5=0;       // red LED on
>             }
>             AudioPwmSet(AdcSamplef);
>             Timer2TimeoutCounter=10;  // come back in 125 us. PWM
> frequency is 80 kHz, so change every 10 cycles
>         }
>         /* Maintain state machines of all polled MPLAB Harmony modules. */
>         SYS_Tasks ( );
>     }
> 
> 
> Timer 2 is generating 80 kHz PWM. Every 10 times around, the ADC is read
> and started again. The ADC is converted to a double precision float called
> AdcSamplef. This runs through a bandpass filter for mark and another for
> space. The absolute values of the outputs of the filters are computed
> (same as full wave rectification). The results go into a low pass filter
> for mark and another for space. If the absolute value of the mark LPF is
> greater than that of the space LPF, the SSR is turned on causing loop
> current (mark). Otherwise, the loop current is interrupted (space).
> 
> There is also some Direct Digital Synthesis code in there for test. When a
> frequency is set, DdsRadiansPerSample is set telling the DDS code how many
> radians to add to the last value, then do a sine lookup. The result of the
> DDS can be sent to the audio output (for transmit audio).
> 
> The audio output is 80 kHz PWM generated by timer 2 and an output compare
> module. The duty cycle is updated at 8 kHz. The PWM goes to a single pole
> RC LPF with a cutoff of 4 kHz. This is followed by a 2 pole active LPF
> with the same cutoff. For testing right now, the AdcSamplef is being fed
> to the audio output. I can hear it on speakers and see it on the scope.
> 
> Next steps:
> 
> 1. I plan to do a long transmission from punched tape and record the audio
> on an SDR (probably KFS). I will then vary filter parameters and see what
> gives me the lowest error rate with noise and QSB.
> 
> 2. I'll get the transmit side running. That will include an opto detecting
> loop current which will then FSK the DDS. I'll put an LPF between the data
> and the DDS to minimize keyclicks.
> 
> Fun stuff!
> 
> Harold
> https://w6iwi.org
> 
> 
> 
> 
> 
> -- 
> FCC Rules Updated Daily at http://www.hallikainen.com
> Not sent from an iPhone.
> ______________________________________________________________
> GreenKeys mailing list
> Home: http://mailman.qth.net/mailman/listinfo/greenkeys
> Help: http://mailman.qth.net/mmfaq.htm
> Post: mailto:GreenKeys at mailman.qth.net
> 
> >>> Jordan Spencer Cunningham's GreenKeys Search Tool: https://teletype.net/gksearch
> >>> 2002-to-present greenkeys archive: http://mailman.qth.net/pipermail/greenkeys/
> >>> 1998-to-2001 greenkeys archive: http://mailman.qth.net/archive/greenkeys/greenkeys.html
> >>> Randy Guttery's 2001-to-2009 GreenKeys Search Tool: http://comcents.com/tty/greenkeyssearch.html
> 
> This list hosted by: http://www.qsl.net
> Please help support this email list: http://www.qsl.net/donate.html
> Message delivered to epvgk at limpoc.com