I utilize the ATMega48’s hardware counter which is synchronous with the system clock, so it can only measure frequency less than half of its clock speed. I solve this issue by dividing the input frequency by 8 and clocking the chip at 12mhz. This allows me to measure frequencies up to about 48MHz, but can be easily adapted to measure over 700MHz (really?) by dividing the input by 128. Division occurs by a 74HC590 8-bit counter (not a 74HC595 as I accidentally said in the video, which is actually a common shift register), allowing easy selection of input divided by 1, 2, 4, 8, 16, 32, 64, or 128. The following image shows the o-scope showing the original signal (bottom) and the divided-by-8 result (top)

The device outputs graphically to a LCD simply enough. That LCD is from eBay and is only $3.88 shipped! I’m considering buying a big box of them and implementing them in many more of my projects. They’re convenient and sure do look nice!

The signal I test with comes from an oscillator I built several months ago. It’s actually a SA612 style receiver whose oscillator is tapped, amplified, and output through a wire. It’s tunable over almost all of 40m with a varactor diode configuration. It was the start of a transceiver, but I got so much good use out of it as a function generator that I decided to leave it like it is!

THIS IS HOW THE PROGRAM WORKS: I don’t supply a schematic because it’s simple as could be. Divide the input frequency to something relatively slow, <1MHz at least. Configure the 16-bit counter to accept an external pin as the counter source (not a prescaled clock, as I often use in other applications). Then set the timer value to 0, _delay_ms() a certainly amount of time (1/10th second), and read the counter value. Multiply it by 10 to account for the 1/10th second, then multiply it by 8 to account for the divider, and it’s done! It will update 10 times a second, with a resolution down to 10*8 = 80 Hz. It’s well within the range of amateur radio uses! If you’re considering replicating this, read up on how to use hardware counters with ATMEL AVR microcontrollers. That should be enough to get you started! Here’s the code I used…

For the LCD, this code requires LCD library.

#include <stdlib.h>
#include <avr/io.h>
#include <avr/pgmspace.h>
#include <util/delay.h>
#include "lcd.h"
#include "lcd.c"

int main(void)
{
	TCCR1B=0b00000111; // rising edge trigger
    char buffer[8];
	long toshow=0;
	char mhz=0;
	int khz=0;
	int hz=0;
    lcd_init(LCD_DISP_ON);
	for(;;){
    	lcd_clrscr();

    	lcd_gotoxy(0,0);
		itoa(mhz , buffer, 10);
		lcd_puts(buffer);
		lcd_puts(".");

		if (khz<100){lcd_puts("0");}
		itoa(khz , buffer, 10);
		lcd_puts(buffer);

		itoa(hz/100 , buffer, 10);
		lcd_puts(buffer);

		lcd_puts(" MHz");

		TCNT1=0;
		_delay_ms(99);
		_delay_us(312);
		toshow=TCNT1;
		toshow=(long)toshow*16*10; // tenth second period
		mhz=toshow/1000000;
		toshow=toshow-mhz*1000000;
		khz=toshow/1000;
		toshow=toshow-khz*1000;
		hz=toshow;
	}
}

As you can see from the video, it’s good but not good enough. If I could get samples at 2,000 per second I’d probably be OK, but it’s just not quite fast enough with it’s current, ultra-simplistic method of sample recording. I’ll figure out a fancier way to build a spectrum analyzer – it’s obvious the platform is there, it just needs some refinement.

A few stills:

UPDATE! The more I think about it, the more I think this might be just good enough to work! Look at the stagger in those peaks near the top – that’s probably the lines telling which character to display. Data between the peaks indicates the value to be provided, and I should have enough time to accurately measure that… Maybe this is good enough after all? I’ll have to run some more tests tomorrow…

Where’s the code? It kills me to do this, but I need to withhold the chip side code. I’m working on an idiot’s guide to USB connectivity with AVR microcontrollers, and I’d rather post the simplest-case code first, then share complicated stuff like this. I’ll post the python scripts:

# LOGIC.PY - this script will log (or print) raw data from the USB device
from optparse import OptionParser
import time
import usb.core
import usb.util
import os

while True:
        print "\nTrying to communicate with the Gator Keyer ...",
        dev = usb.core.find(find_all=True, idVendor=0x16c0, idProduct=0x5dc)
        if len(dev)==0: print "FAIL"
        dev=dev[0]
        dev.set_configuration()
        print "there it is!"
        break

def readVals():
    x=dev.ctrl_transfer(0xC0, 2, 2, 3, 4).tolist()
    return x

def toBinary(desc):
	bits=[]
	for i in range(7,-1,-1):
		if (2**i>desc):
			bits.append('0')
		else:
			bits.append('1')
			desc=desc-2**i
	return bits

def toStr(lists):
	raw=[]
	for port in lists: raw+=toBinary(port)
	return ''.join(raw)

### PROGRAM START ##################
live=False
#live=True
start=time.time()
if live==True:
	while True:
		a,b,c,d=readVals()
		if not a==123: continue #bad data
		elapsed=time.time()-start
		print "%.010f,%s"%(elapsed,toStr([b,c,d]))
else:
	times=0
	data=''
	f=open("out.txt",'a')
	while True:
		a,b,c,d=readVals()
		if not a==123: continue #bad data
		elapsed=time.time()-start
		data+="%.010f,%s\n"%(elapsed,toStr([b,c,d]))
		times+=1
		if times%1000==999:
			print "%d readings / %.02f = %.02f /sec"%(times,elapsed,times/elapsed)
			f.write(data)
			data=""

#logicGraph.py - this will show the data in a pretty way
import matplotlib.pyplot as plt
import numpy

c={
0:"",
1:"",
2:"blk sol",
3:"yel str",
4:"yel sol",
5:"pur sol",
6:"pur str",
7:"",
8:"",
9:"",
10:"blu sol",
11:"blu str",
12:"orn sol",
13:"orn str",
14:"pnk sol",
15:"pnk str",
16:"",
17:"",
18:"",
19:"",
20:"",
21:"",
22:"",
23:"",
24:"",
}

print "loading"
f=open("out.txt")
raw=f.readlines()
f.close()

print "crunching"
times=numpy.array([])
data=numpy.array([])
for line in raw:
	if len(line)<10: continue
	line=line.replace("\n",'').split(',')
	times=numpy.append(times,float(line[0]))
	bits = []
	for bit in line[1]:
		if bit=="1":bits.append(1)
		else:bits.append(0)
	data=numpy.append(data,bits)

columns=24
rows=len(data)/columns
data=numpy.reshape(data,[rows,columns])
print "DONE processing",len(data),"lines\n\n"
print "plotting..."
plt.figure()
plt.grid()
for i in range(len(c.keys())):
	if c[i]=="": continue
	plt.plot(times,data[:,i]+i*1.1,'-',label=c[i])
plt.legend()
plt.show()

UPDATE: This story was featured on this post of HackADay.com!

This is the finished product ready to send data to a PC:

MY PROBLEM: I want to send data from a simple microcontroller to a PC. While USART and a serial port is the common solution like I’ve done before, it’s not convenient because it requires a level converter (like a MAX232, about $4), crystal (specific values based on bit and error rate, if you’re lucky you might have a right value in your junk box), and an archaic PC which actually has a serial port. A usb serial port adapter sounds clever, but many aren’t supported on Linux, Windows Vista, or Windows 7. Also, many small chips (most of the ATTiny series) don’t have built in serial capabilities, so it has to be bit-banged in software! Yuk! The second choice would be USB. This requires a crystal too, zener diodes, and bit-banging the USB protocol with something like V-USB since most of the AVR series don’t have built in USB (do they even make breadbordable DIP chips with USB?). Even so, it requires drivers, custom software, cross-platform frustrations, etc. I know PIC has some 18f series chips with USB, but I don’t feel like switching architectures just to send a few bytes of data to a PC. FDTI has a FT232R chip which is a USB serial port adapter, but it’s expensive (about $5) and doesn’t come in dip, so no breadboarding! Sure there are adapter boards, but that just adds the cost. I’m not excited about a $5 solution for a $1 microcontroller. I even did a bit of trolling on AVR Freaks to see if anyone could help me out – just more of the same!

MY SOLUTION: Send data through the sound card! USB sound cards are $1.30 (shipped) on eBay! It couldn’t be simpler. Send pulses, measure distance between pulses. Short pulses are a zero, longer ones are a 1, and very long pulses are number separators. A Python solution with PyAudio allows 1 script which will work on Mac, Linux, Windows, etc, and because it calibrates itself, this will work on any chip at any clock rate. Data is initiated with calibration pulses so timing is not critical – the PC figures out how fast the data is coming in. Check it out! (scroll way down for a bidirectional communication solution)

Here is a sound card I used for bidirectional communication:

Output graph (python and excel) of temperature when I put a soldering iron near the sensor:

~ UNIDIRECTIONAL SOLUTION ~

The following code is designed to have a chip send data to your PC automatically. This can be run on any micro-controller (PIC or AVR I guess, the concept is the same) at any clock rate. Just make sure the sound card is recording fast enough to differentiate pulses. (keep scrolling down for a bidirectional solution)

A NOTE ABOUT MY CODE: This is just the code I used for my demonstration. It might be more logical for you to write your own since the concept is so simple. I’m a dental student, not a programmer, so I’m sure it’s not coded very elegantly. I didn’t work hard to make this code easy to read or easy to share. With that being said, help yourself!

/*The following code is written in AVR-GCC for an ATTiny44a.
It reads ADC values on 3 pins and reports it each second along
 with a number which increments each time data is sent.
It's designed as a starting point, allowing anyone to
customize it from here!*/

#include <avr/io.h>
#include <avr/delay.h>
#include <avr/interrupt.h>

// bytes we want to send to the PC
volatile int data1=0;
volatile int data2=0;
volatile int data3=0;
volatile int data4=0;

void solid(){  // dont touch
	_delay_ms(1);
	pulse(1);pulse(1);pulse(1);pulse(3);pulse(3);
	pulse(3);pulse(5);pulse(5);// CALIBRATION PULSES
}
void pulse(char size){ // dont touch
	PORTA|=_BV(PA3);
	_delay_us(100);
	PORTA&=~_BV(PA3);
	while (size){size--;_delay_us(100);}
}
void sendVal(unsigned long tosend){ // dont touch
	pulse(5); // send a space
	while (tosend){
		if (tosend&1){pulse(3);} // send ONE
		else {pulse(1);} // send ZERO
		tosend=tosend>>1;
	}
}

int readADC(char adcNum){
	_delay_ms(1);
	ADMUX=adcNum; // select which ADC to read, VCC as ref.
	ADCSRA=0b11000111; // enable, start, 128 prescale
    while (ADCSRA&( 1<<ADSC)) {}; // wait for measurement
	return ADC;
}

void takeReadings(){
        data1=readADC(0); // ADC0
        data2=readADC(1); // ADC1
        data3=readADC(2); // ADC2
		data4++; // incriment just because we want to
}

void sendStuff(){ // EDIT to send what you want
	solid(); //required
	sendVal(12345); //required
	sendVal(12345); //required
	sendVal(54321); //required

	sendVal(data1);
	sendVal(data2);
	sendVal(data3);
	sendVal(data4);

	pulse(1); //required
}

int main(){
	DDRA|=_BV(PA2)|_BV(PA3);
	for (;;){
		_delay_ms(1000);
		takeReadings();
		sendStuff();
	}
	return 0;
}

"""
file name: listenOnly.py

This is the PC code to listen to the microphone and display
and log the data. It probably does NOT need adjustment!
 Make sure the correct sound card is selected (in the code)
 and make sure microphone input is turned up in volume control.

This code is what was used on my PC for the demonstration
video. This is the listenOnly.py file which will turn any audio
 detected from a sound card into data, optionally logging it
(if the last few lines are uncommented). This also works to
capture data for the bidirectional communication method,
described below on this website.

If this is running but no data is coming through, make sure the
microphone is selected as a recording device, the correct sound
card is selected, and the microphone volume is turned to high.

REQUIRED: To run this, you need to have the following installed:
-- Python 2.6
-- numpy for python 2.6
-- matplotlib for python 2.6
-- pyaudio for python 2.6
(other versions may work, but this is what I'm using)
"""
import numpy
import pyaudio
import matplotlib.pyplot as plt
import wave
import time

def listCards(dontAsk=True):
    p=pyaudio.PyAudio()
    print "SOUND CARDS:"
    for i in range(p.get_default_host_api_info()["deviceCount"]):
        if p.get_device_info_by_index(i)["maxInputChannels"]>0:
                cardName = p.get_device_info_by_index(i)["name"]
                cardIndex = p.get_device_info_by_index(i)["index"]
                print "[%d] %s"%(cardIndex,cardName)
    if dontAsk: return
    return int(raw_input("CARD NUMBER TO USE:"))

cardID=1
listCards()
print "USING CARD:",cardID

rate=44100.0
sampleSize=1024

def data2vals(data):
    vals=numpy.array([])
    lastPeak=0
    for i in range(1,len(data)):
        if data[i]==True and data[i-1]==False:
            if lastPeak>0: vals=numpy.append(vals,i-lastPeak)
            lastPeak=i
    return vals

def binary2dec(binary):
    binary=binary[:-1]
    dec=0
    s=""
    for i in range(len(binary)):
        dec=dec*2
        dec+=binary[i]
        s="%d"%binary[i]+s
    #print s,"=",dec #11111100101100000 = 3391
    return dec

def readVals(vals):
    if len(vals)<7: return False
    vals2=[]
    aLow = min(vals[0:3])
    aMed = min(vals[3:6])
    aHigh = vals[6]
    thresh1=sum([aLow,aMed])/2+2
    thresh2=sum([aMed,aHigh])/2+2
    #print "tresholds:",thresh1,thresh2
    #print vals
    vals=vals[8:]
    binary=[]
    for i in range(len(vals)):
        if vals[i]>thresh2:
            vals2.append(binary2dec(binary))
            binary=[]
        if vals[i]>thresh1:binary=[1]+binary
        else:binary=[0]+binary
    vals2.append(binary2dec(binary))
    for i in range(len(vals2)):
        if vals2[i]==54321: return vals2[i+1:]
    return False

def playFile():
    chunk = 1024
    wf = wave.open("short_onenum.wav", 'rb')
    p = pyaudio.PyAudio()
    stream = p.open(format =
                    p.get_format_from_width(wf.getsampwidth()),
                    channels = wf.getnchannels(),
                    rate = wf.getframerate(),
                    output = True)
    data = wf.readframes(chunk)
    while data != '':
        stream.write(data)
        data = wf.readframes(chunk)
    stream.close()

def captureData():
    pyaud = pyaudio.PyAudio()
    stream = pyaud.open(format=pyaudio.paInt16,channels=1,
        rate = 44100,input_device_index=cardID,input=True,output=True)
    sample=numpy.array([])
    while True:
        sampleNew=numpy.fromstring(stream.read(sampleSize),dtype=numpy.int16)
        sampleNew=(sampleNew<-25000)*1
        if True in sampleNew: sample=numpy.append(sample,sampleNew)
        else:
            if len(sample):
                stream.close()
                return sample
    stream.close()

tone_quiet=0

def buildNumber(num=123):

    if num>255: print "NUMBER TOO HIGH!!!"
    #print num,'=',
    num+=1
    for i in [7,6,5,4,3,2,1,0]:
        if num>2**i:one();num=num-2**i;#print"1",
        else: zero();#print"0",
    #print
    space()

def pulse():
    global data
    data+=[-30000]*10

def space():
    global data
    data+=[tone_quiet]*900
    pulse()

def one():
    global data
    data+=[tone_quiet]*600
    pulse()

def zero():
    global data
    data+=[tone_quiet]*300
    pulse()

def silence(msec=1000):
    global data
    data+=[tone_quiet]*int(41.1*msec)

data=[]
def sendAudio(numbers=[11,66,77]):
    global data
    data=[]
    silence(100)
    buildNumber(250)
    print "SENDING",
    for numba in numbers:
        buildNumber(numba)
        print numba,
    buildNumber(250)
    silence(100)
    data=numpy.array(data)
    data=-data
    data=data.tostring()
    print

    p = pyaudio.PyAudio()
    stream = p.open(rate=44100, channels=1, format=pyaudio.paInt16,
                    input_device_index=cardID, output=True)
    stream.write(data)
    stream.close()
    p.terminate()

i=0
while True:
    i+=1
    val=readVals(data2vals(captureData()))
    if val == False: continue
    line=""
    for item in val: line+=str(item)+","
    print i,line
    #f=open('log.csv','a')
    #f.write("%s\n"%line)
    #f.close()

~ BIDIRECTIONAL SOLUTION ~

What if we want to send data TO the microcontroller? The solution is a little more complex, but quite doable. Just add an extra wire to the sound card’s speaker output and attach it to PCINT0 (the highest level internal interrupt). This is intended for advanced users, and if you’re doing this you probably are better off with USB or serial anyway! … but heck, why not do it as a proof of concept!

Note that the USB sound card speaker output was not powerful enough to trigger the digital input pin of the AVR, so an inverting buffer was made from a single NPN transistor (2n3904). The hardware interrupt was attacked to the collector, and the collector was attached through +5V through a 220 ohm resistor. The emitter was grounded. The base was attached directly to the sound card output. I also tried running the sound card output through a small series capacitor (0.1uF) and biasing the base to ground through a 1Mohm resistor and it worked the same. Hardware, simple. Chip-side software… a little more complex.

### VIDEO ###

"""
This code is what was used on my PC for the
 demonstration video. The listenonly.py file
 (above on site) was also used without modification.
"""
import pyaudio
from struct import pack
from math import sin, pi
import wave
import random
import numpy
import time

RATE=44100
maxVol=2**15-1.0 #maximum amplitude
p = pyaudio.PyAudio()
stream = p.open(rate=44100, channels=1, format=pyaudio.paInt16,
		input_device_index=1, output=True)

def pulseZero():
    global wvData
    wvData+=pack('h', 0)*30
    wvData+=pack('h', maxVol)

def pulseOne():
    global wvData
    wvData+=pack('h', 0)*40
    wvData+=pack('h', maxVol)

def pulseSpace():
    global wvData
    wvData+=pack('h', 0)*50
    wvData+=pack('h', maxVol)

def buildNumber(num=123):
    if num>255: print "NUMBER TOO HIGH!!!"
    num+=1
    for i in [7,6,5,4,3,2,1,0]:
        if num>2**i:
            pulseOne()
            num=num-2**i
        else:
            pulseZero()

wvData=""
wvData+=pack('h', 0)*2000
pulseOne() #required before sending data

buildNumber(55)
buildNumber(66)
buildNumber(77)
buildNumber(123)

wvData+=pack('h', 0)*2000

while True:
	print "SENDING",
	stream.write(wvData)
	raw_input()

/*
This code is what was used on my AVR
microcontroller for the demonstration video
*/
#include <avr/io.h>
#include <avr/delay.h>
#include <avr/interrupt.h>

volatile long commandIncoming=0;
volatile char command1=0;
volatile char command2=0;
volatile char command3=0;
volatile char command4=0;
volatile char bitsGotten=0;

// timing thresholds are critical! Send pulses to the chip
// and have it report the time between them. Use this to
// determine the best threshold value for your application.
// The ones here must be changed if you run at a speed other
// than 1mhz or if you use different timings in PC software
#define thresh_low 100 // between this and the next
#define thresh_high 130 // is the range for a logical 'one'

// ######## OUTGOING AUDIO DATA #########
void solid(){
	_delay_ms(1); //LONG LOW
	pulse(1);pulse(1);pulse(1);pulse(3);pulse(3);
	pulse(3);pulse(5);pulse(5);// CALIBRATION PULSES
}
void pulse(char size){
	PORTA|=_BV(PA3);
	_delay_us(100);
	PORTA&=~_BV(PA3);
	while (size){size--;_delay_us(100);}
}
void sendVal(unsigned long tosend){
	pulse(5); // send a space
	while (tosend){
		if (tosend&1){pulse(3);} // send ONE
		else {pulse(1);} // send ZERO
		tosend=tosend>>1;
	}
}

// ######## INCOMING AUDIO DATA #########
// NOTE THAT INPUTS ARE NORMALLY *HIGH* AND DROP *LOW* FOR SIGNAL
SIGNAL (PCINT0_vect) { // audio input trigger
	TIMSK0|=(1<<TOIE1); //Overflow Interrupt Enable
	if (TCNT0<10){return;} // seem too fast? ignore it!
	// Enable the following line to test custom timings
	//command1=command2;command2=command3;
	//command3=command4;command4=TCNT0;
	bitsGotten++;
	commandIncoming=commandIncoming*2; // shift left
	if (TCNT0>thresh_low){commandIncoming++;} // make 1
	TCNT0=0;
}

ISR(TIM0_OVF_vect){ // TIMER OVERFLOW
	if (bitsGotten){sendStuff();}
}

void fillCommands(){
	command1=(char*)(commandIncoming>>24);
	command2=(char*)(commandIncoming>>16);
	command3=(char*)(commandIncoming>>8);
	command4=(char*)(commandIncoming);
}

void sendStuff(){
	TIMSK0=0; //Overflow Interrupt
	cli(); // disable interrupts!
	fillCommands();
	solid(); // start data transmissions with this
	sendVal(12345);
	sendVal(12345);
	sendVal(54321);
	sendVal(command1);
	sendVal(command2);
	sendVal(command3);
	sendVal(command4);
	sendVal(1234567890);
	pulse(1);
	bitsGotten=0;
	sei(); // enable interrupts again!
	TIMSK0|=(1<<TOIE1); //Overflow Interrupt
}

// ######## MAIN PROGRAM #########
int main(){

	DDRA|=_BV(PA2)|_BV(PA3);

	// SET UP FOR SOUND CARD INTERRUPT
	MCUCR = 0b00000010; // trigger interrupt on falling edge
	GIMSK = 0b00010000; // pin change interrupt enable 0
	GIFR =  0b00010000; // flag register, same as above
	PCMSK0 = (1<<PCINT0); // Set Pin to use (PCINT0)
	sei(); // enable global interrupts

	// SET UP 8-bit COUNTER
	TCCR0B|=0b00000010;
	//TCCR1B|=(1<<CS12)|(1<<CS10); // prescaler 1024
	TIMSK0|=(1<<TOIE1); //Enable Overflow Interrupt Enable
	TCNT0=0;//Initialize our varriable (set for 1/15th second?)

	// MAIN PROGRAM
	for (;;){}
	return 0;

}

In closing, I’m tickled this works so well. It’s funny to me that no one’s really done this before in the hobby field. I’m sure I’m not the only one who wished there were an easy way to do this. I’m sure the process could be greatly improved, but this is a fun start. Wow, it’s late, I should get to bed. I have to treat patients tomorrow morning!

PS: If you replicate this concept, let me know about it! I’d love to see your project!

from struct import pack
from math import sin, pi
import wave
import random

RATE=44100

## GENERATE MONO FILE ##
wv = wave.open('test_mono.wav', 'w')
wv.setparams((1, 2, RATE, 0, 'NONE', 'not compressed'))
maxVol=2**15-1.0 #maximum amplitude
wvData=""
for i in range(0, RATE*3):
	wvData+=pack('h', maxVol*sin(i*500.0/RATE)) #500Hz
wv.writeframes(wvData)
wv.close()

## GENERATE STERIO FILE ##
wv = wave.open('test_stereo.wav', 'w')
wv.setparams((2, 2, RATE, 0, 'NONE', 'not compressed'))
maxVol=2**15-1.0 #maximum amplitude
wvData=""
for i in range(0, RATE*3):
	wvData+=pack('h', maxVol*sin(i*500.0/RATE)) #500Hz left
	wvData+=pack('h', maxVol*sin(i*200.0/RATE)) #200Hz right
wv.writeframes(wvData)
wv.close()

The output is two sound files which look like this:

Notice the massive lag spikes around the time class begins. I think it’s caused by the retarded behavior of windows update and anti-virus software updates being downloaded on a gazillion computers all at the same time which are required to connect to the network on Windows machines. Class start times were 8:30am, 9:35am, and 10:40am. Let’s view it on a logarithmic scale:

Finally, the code. It’s two scripts. One pings a website (kernel.org) every few seconds and records the ping time to “pings.txt”, and the other graphs the data. Here are the two scripts:

import socket, time, os, sys, re

def getping():
	pingaling = os.popen("ping -q -c2 kernel.org")
	sys.stdout.flush()
	while 1:
		line = pingaling.readline()
		if not line: break
		line=line.split("\n")
		for part in line:
			if "rtt" in part:
				part=part.split(" = ")[1]
				part=part.split('/')[1]
				print part+"ms"
				return part

def add2log(stuff):
	f=open("pings.txt",'a')
	f.write(stuff+",")
	f.close()

while 1:
	print "pinging...",
	stuff="[%s,%s]"%(time.time(),getping())
	print stuff
	add2log(stuff)
	time.sleep(1)

import pylab, time, datetime, numpy

def smoothTriangle(data,degree,dropVals=False):
	triangle=numpy.array(range(degree)+[degree]+range(degree)[::-1])+1
	smoothed=[]
	for i in range(degree,len(data)-degree*2):
		point=data[i:i+len(triangle)]*triangle
		smoothed.append(sum(point)/sum(triangle))
	if dropVals:
		print "smoothlen:",len(smoothed)
		return smoothed
	#smoothed=[smoothed[0]]*(degree+degree/2)+smoothed
	#while len(smoothed)<len(data):smoothed.append(smoothed[-1])
	while len(smoothed)<len(data):smoothed=[None]+smoothed+[None]
	if len(smoothed)>len(data):smoothed.pop(-1)
	return smoothed

print "reading"
f=open("pings.txt")
raw=eval("[%s]"%f.read())
f.close()

xs,ys,big=[],[],[]
for item in raw:
	t=datetime.datetime.fromtimestamp(item[0])
	maxping=20000
	if item[1]>maxping or item[1]==None:
		item[1]=maxping
		big.append(t)
	ys.append(float(item[1]))
	xs.append(t)

#print xs
#raw_input("WAIT")

print "plotting"
fig=pylab.figure(figsize=(10,7))
pylab.plot(xs,ys,'k.',alpha=.1)
pylab.plot(xs,ys,'k-',alpha=.1)
pylab.plot(xs,smoothTriangle(ys,15),'b-')
pylab.grid(alpha=.3)
pylab.axis([None,None,None,2000])
#pylab.semilogy()
#pylab.xlabel("time")
pylab.ylabel("latency (ping kernel.org, ms)")
pylab.title("D3-3 Network Responsiveness")
fig.autofmt_xdate()
#pylab.show()
pylab.savefig('out.png')
pylab.semilogy()
pylab.savefig('out2.png')
fig.autofmt_xdate()
print "done"