Peak detector

[Leo22]

I did this code for a sine peak detector , but it's not so precise . Does some have any idea how can I improve it to get more precison?And another question , I'm just put a diode for the negative cicle of the sine wave , is only that going to protect the Pic16f877 againt negative inputs. And the peak it's lower than 5V. Thanks

void main()
{


Long int cont=0;
float value,value1,value2, peak=0,peak1=0;
set_tris_d(0);
set_tris_c(0);
set_tris_b(0b01000000);
output_d(0);
output_c(0);


LCD_init();
setup_adc_ports(RA0_RA1_RA3_ANALOG);
setup_adc(ADC_CLOCK_DIV_32);
setup_timer_1(T1_INTERNAL | T1_DIV_BY_1);



set_adc_channel (0);
delay_us(10);


while (cont<250>peak)
peak=value;
value2=read_adc();
if (value1>peak)
peak=value1;
if (value2>peak)
peak=value2;

cont++;
}


peak= (5*(peak+1)/1023);
}

[kender]

[SherpaDoug]

What are you going to use this peak for? If you know the signal is a really good sine then kender's idea will give a very good answer. But if the further the signal differs from a true sign the more error.

If you are measuring AC voltage or power you probably want RMS not peak. You can calculate RMS from kender's integral.

If you want the peak value so you can adjust gain to prevent clipping then you want the real peak whether it is a sine or not.

To protect the A/D from the negative half cycle a resistor and a Schottky diode work well. A common signal diode (1N914) will work up to audio frequencies. Don't use 1N4000 series diodes above about 400 Hz.
_________________
The search for better is endless. Instead simply find very good and get the job done.

[Leo22]

I have to mesuare a gain between to sine waves, but to mesuare the RMS value I need a external circuit, I want to put the sine wave directly to the PIC.

[languer]

I think you need to define your requirements a little better and then look at the underlying math.

1. Are you detecting fixed frequency waveforms?
   
2. Are the waveforms same frequency, different amplitude?

The easiest way, for a fixed frequency system is a true peak detector (i.e. series diode, shunt cap). Anything more complex than this (or using the MCU for the sampling/peak detection/smoothing) will be very waveform dependent (unless you can achieve tru-RMS performance).

[SherpaDoug]

[Leo22]

It's 1khz , can I get the rms value with Pic? How can I do that? Thanks for your attention

[SherpaDoug]

It is the Root of the Mean of the Square. So you take each datapoint, square it, average it over some period of time like one cycle, and take the square root.
_________________
The search for better is endless. Instead simply find very good and get the job done.

[Rocket]

Have a look at 'EX_RMSDB.C' in the Examples directory.

SHALOM!
_________________
J.I.L.

[languer]

One thing you may want to keep in mind is that for either peak detection or RMS to actually work you must sample the signal using the Nyquist criteria. On the analog domain this is not really a problem, but on the digital domain you must sample the signal at least twice as fast as the maximum frequency. For 1kHz you are probably ok, but nonetheless you should verify it.