Schematic
+---> +(Voice > Accumulator)
|
+5Volts | 100UF bypass
o----------+-----------------------+----------------------|(-----GND
| | | diode
| | 4.7K +--|<-- 2.2K---Most Sig Bit
4.7K | | |
| e---- | -----+ (DAC) +--|<-- 4.3K---Next Most Sig
+----+----bQpnp | | |
| | c | | +----+--|<-- 9.1K-----
| 10K |___ | | | |
c | | | | +--+ +--|<-- 18K -----
npn Qb---+----+ 1K | | | | |
+-------e | | | | c | c +--|<-- 36K -----
| | | | c npnQb-*-bQnpn |
\----(+)------+ 2.2K 100UF +-bQnpn e e +--|<-- 75K -----
|Loudspeaker | | | | e | | |
|as Microphone| | | 10K | 100ohm 100ohm +--|<-- 150K ---
/----(-)------+ Ground | | | | |
| Ground Ground +--|<-- 300K--- Least Sig Bit
Ground
The diodes serve to make current contribution unidirectional, thus avoiding modulation of the sum-of-current by exactly how good a logic-0 is provided.
There are NO ANTIALIASING FILTERS, and no reconstruction filters. You can experiment with tones greater than one-half the sampling rate, and observe where they are placed in the sampled data spectrum. By the way, running without a 80287, my PC (12MHz AT) needs about 7 seconds to perform a 256-point FFT. On my PC, acceptable ports are: #1=0x0378, #2=0x0278, #3=0x03bc
Driver in C
/* This routine implements, in conjunction with an external
Digital-to-Analog converter and analog comparator, a TRACKING
ANALOG-TO-DIGITAL CONVERTER. This routine has in software an
accumulator that sums the input signal---a single bit. The bit
indicates whether the accumulator value, the current guess as
to the actual analog signal, is GREATER or LESS than the analog
signal.
Turns out that a "C" routine on a 12MHz PC-AT is more than
fast enough to sample a voice signal. The loop spins about
60,000 times per second. This allows 20 samples--of 1 bit
change per sample---for 3KHz signals, or 10 samples peak-peak
at 3KHz. The normal male voice at 500 Hz can be 60 samples
peak-peak, or 1/4 of full scale.
At one time, variable-size steps were used, but the benefit
was not observable since there are NO ANTI-ALIASING filters,
and the distortion of the voice as judged by playback quality
was not improved by dynamically varying the step size.
But the electronic parts cost is only $2.00 */
/* will bind global 'sample_rate' */
void grab_voice()
{
unsigned char tracking_value;
unsigned int vndx;
char interval;
unsigned char input_signal;
int control_port;
char slew_accum; /* will vary between +5 and -5 ? */
unsigned char delta; /* how much to step 'tracking_value' */
clock_t start_t, stop_t;
printf("\nStarting to grab voice\n");
control_port = active_port + 2;
outportb(control_port, 0x04); /* free up the 4 OpenCollector I/O pins */
/* We'll can read data from Pin 1 */
tracking_value = 0x80; /* Init voltage to half-scale */
interval = 0;
delta = 1;
slew_accum = 0;
start_t = clock();
for ( vndx = 0 ; vndx < VOICE_SIZE; ) /* until VOICE is full */
{
input_signal = 0x01 & inportb(control_port); /* PIN 1, by chance */
if ( !input_signal ) {
/* tracking_value = min(0xfb, tracking_value + delta); */
tracking_value++;
slew_accum++;
}
else {
/* tracking_value = max(4, tracking_value - delta); */
tracking_value--;
slew_accum--;
}
/* FASTER if port is constant */
outportb(active_port, tracking_value);
if ( interval++ EQUALS 4 ) /* 10 gives about 5 seconds in 0x02ff */
{
voice[vndx++] = tracking_value;
interval = 0;
if ((slew_accum > 3) || (slew_accum < -3)) /* Has this any use? */
delta++;
else if (delta > 2)
delta--;
slew_accum = 0;
}
} /* end FOR ( VNDX ; VNDX ; ) */
stop_t = clock();
sample_rate = /* bind global var */
(VOICE_SIZE * CLK_TCK) / ( stop_t - start_t) ;
printf("Grabbed voice. Took %f seconds. Rate is %d samples/sec\n",
((stop_t - start_t) / CLK_TCK), sample_rate );
}
/* ------------------------------------------------------------------- */
/* This function supports exploring the behavior of your printer port.
You specify an input pin of that port- 1, 14, 16 17.
Pin 2 will output a signal that alternates between low and high. The
rate is determined by how fast your computer can scroll. You are
shown the state of Pin 2 as output, and the state of the input pin.
This continues until you PRESS ANY KEY to EXIT.
Needs some HELP words, to explain what is good and bad display */
void input_diddle()
{
unsigned char diddle_out, diddle_in, bitmask;
int pinspec, which_io_port;
printf("\nWill alternately turn Pin 2 ON and OFF, as test output.\n");
printf("Select an input Pin that has pullup: 1 14 16 17, as 1 4 6 7 ? ");
printf("You must CONNECT a WIRE BETWEEN pin 2 AND your test input.\n");
pinspec = getche();
which_io_port = active_port + 2;
outportb(which_io_port, 0x04); /* 0100 frees the pullups */
/* of I/O bits 17, 16, 14, 1 */
switch( pinspec )
{
case '1': bitmask = 0x01; break;
case '4': bitmask = 0x02; break;
case '6': bitmask = 0x04; break;
case '7': bitmask = 0x08; break;
default: ;
}
diddle_out = 1;
for ( ; ; )
{
if ( kbhit() )
{
getch(); /* eat char so won't exit the outer loop */
break;
}
outportb(active_port, diddle_out);
diddle_in = inportb(which_io_port);
diddle_in = diddle_in & bitmask;
printf("Output > %x <, Input value > %x < Phase-%s Default-%s\n",
diddle_out, diddle_in,
((pinspec=='6') ? "NonInvert" : "Invert"),
((pinspec=='6') ? "High" : "Low")
);
diddle_out = ((diddle_out == 1) ? 0 : 1);
}
}
/* ------------------- end function INPUT_DIDDLE -------------------- */
/* This routine allows keyboard/cursor-key definition of the relative
output analog voltage. Thus you can test the threshold characteristics
of the comparator. Early on, the comparator had a non-voice-in
threshold of level 174; this was moved to about 135.
The size of the transition is 3 bits, so a continuous dither occurs.
*/
void define_level()
{
unsigned char level;
int action;
int which_io_port;
unsigned char diddle_in;
level = 128;
which_io_port = active_port + 2;
outportb(which_io_port, 0x04); /* 0100 frees the pullups */
/* of I/O bits 17, 16, 14, 1 */
printf("\nStarting at level of 128/midscale, move (u)p or (d)own.\n");
printf("The comparator digital output is reported. SPACE to abort.\n");
printf("\nHardware MUST be in DIGITIZE mode, not PLAYBACK.\n");
for ( ; ; )
{
action = getch();
if ( action EQUALS ' ')
break;
switch(action)
{
case 'u': level = level + 1; break;
case 'd': level = level - 1; break;
default: ;
}
outportb(active_port, level);
diddle_in = inportb(which_io_port);
diddle_in = diddle_in & 0x01;
printf("Level %d Comparator %d\n", level, diddle_in);
} /* end FOR */
}
Here's a TRIAC-based solid-state relay circuit:
+5VDC
| 180 180 2.2k
+---/\/\/\----+-----+ +----/\/\/-+--/\/\/\---+-------> 120V
| 1| |6 | | Hot
| +=====+ | | MT1
| | MC | TRIAC | +-+
| | 3032| Driver | G | | TRIAC
| +=====+ | /| |
\ 2| |4 | / +-+
2N3904 |----+ | | | | MT2
/ | +--------- | -------+ |
V \ | | |
| / | \ |
| \ 43 .01u --- 10k / |
| / 500V --- \ |
| | | / |
+------+ | | | Neutral
| +--------+--+---o o--> 120V
/ load
>-/\/\--| 2N3904
\
V
|
---
///
The MC3032 is an optoisolator TRIAC driver. The 180-ohm resistor sets the current for the LED emitter in the optoisolator. Change the value of this resistor - if necessary - to get reasonable current (e.g., 15 mA).
Note that you cannot test this circuit without a load. The TRIAC will not switch unless connected to an AC voltage source, so you can't test it for simple switching w/o applying AC and a load. Note the 500V rating on the .01 cap.
To "round" your square wave, you need to integrate it twice; once to produce a triangular wave, and a second time to produce a parabolic waveform. (A parabolic wave looks pretty much sinusoidal.) You can do this passively using a simple network such as the one shown below. Let R2 = 10*R1, and let R1*C1 = R2*C2 = 1/f, where 'f' is the frequency of the applied square wave. It's important that R2 be much (at least a factor of 5 or so) smaller than the input impedance of the amplifier; likewise, R1 should be a good bit larger than the microcontroller's output impedance.
R1 R2
from u-controller >---/\/\/\----*----/\/\/\----*----> to amplifier
| |
| |
C1 ----- C2 -----
----- -----
| |
| |
----- -----
--- ---
- -
Just create a low pass filter set a couple of khz above the frequency your square wave is running at. Seeing a squarewave is made up of an infinite amount (well supposed to be inifinite but we'll settle for a lot) of odd harmonics, cutting off all the harmonics after the first will usually result in a nice sinewave. Something like this will do:
R
o------/\/\/\/\-----+-------------o Sine wave out.
|
Square Wave in | fc=1/(2piRC)
=C where fc= -3dB drop (half power
| point). There may be a slight
| shift. Live with it.
o-------------------+-------------o
Here's a design (from the 27 May 1981 EDN) that I have built and have used in operational hardware; it works well:
AC---+---68k--+--------+--------+ +-------------+-------out
| | | | | |
| | | A C 10k
100k 100k 0.1uF 4N27 ==== 4N27 |
| | | C E |
| | | | | |
| | | 1.5k +---0.1uF-----+
| | | | | |
| | | A ground +5V
| +--------|------2N5062
| | | C
| C | |
+------2N3904 | |
| E | |
C | | |
1N914 | | |
A | | |
| | | |
AC---+--------+--------+--------+
The combination of the 68k resistor and the capacitor provides a phase-shifted version of the AC input, so there is still 30-40V across the capacitor when the AC line goes through zero. When the downward zero crossing does occur, the 2N3904 turns off, the SCR's gate goes positive, and the SCR fires, dumping the capacitor's charge through the 4N27's LED and pulling the output down to ground. The output pulse has a fairly fast fall but a slow rise, and even the fall is slow by digital-logic standards; I used a Schmitt trigger to clean it up.
The great virtue of this approach is that it is virtually immune to noise. The one-shot nature of the charge dumping ensures only one output pulse per downward zero crossing. If you want a stable 60Hz source with no phase jitter, you probably want to use a phase-locked loop with a slow response to clean things up further -- that's what I did, for a complex dimmer circuit that had to be phase-locked to the AC -- but if all you want to do is count zero crossings, it's perfect as is.
Many years ago I tried to build a decent A/D circuit with a switching supply and I had a similar experience. While good grounding is important, it may not be enough. Particularly, be aware that many switching converters, even those spec'd for low-noise output, still pass any INPUT noise through (that is, the spec really speaks to the amount of additional switching noise introduced). Thus, you're really at the mercy of the power supplied by the PC's switcher (which is usually pretty ugly, particularlly on a loaded bus).
Analog Devices suggests a circuit in their 1992 Amplifier Application Guide for filtering a 5V (digital) supply for use in single-supply analog applications. Should work just as well for dual-supply (with appropriate modifications):
-------/==\------*-------*------*------ +5 (filtered)
\__/ | | |
2 turns --- --- ---
around /-\ /-\ ---
Fair-Rite | | |
#2677006301 | | |
ferrite bead | | |
-------/==\------*-------*------*------ +5 return (filtered)
\__/ 100uF 10-22uF 0.1uF
elec. tant. cer.
An opto-triac is the easiest way. Digi-Key sells them for 1-2 bucks apiece. The sensitivity is in the 5-25mA range, but you could probably coax the 8255 to *sink* that much (ie, drive the internal LED between +5v and the IC pin; to turn on, write-out a zero). Put a 220 ohm resistor in series with the LED.
If you insist on using a relay, use a darlington driver (available at Radio Snack) to boost the gain to drive the relay (probably 12 volts):
+12 V
|
+-------------+
| |
RELAY COIL |
| |
+------|>|----+
| snubber diode
c
8255 >--1k-----b <-- this thing is a darlington
e
|
GND
The relay method is a bit more robust, and not susceptible to phony triggering caused by line-voltage spikes. The only drawback is the +12v supply.
If you wanted this to be a solid state circuit then you could use a Triac Driver connected to a Triac driving you load. A suggested part would be something like a Motorola MOC301[012] driving a general Triac such as a SC141D
For a straight resitive load (incandesent lights) you would need the following:
270 1 +-------+ 6 180
+5v -VAVAVA-----+ +----VAVAVA-----+-------------- Line Hot
2 | MOC | |
TTL in ---------+ 3012 +nc VA SC141D
| | 4 / |
nc+ +------------/ |
+-------+ +----\/\/\/---- Line Neutral
LOAD
Subject: Re: HELP!!! Video amp questions
From: iisakkil@gamma.hut.fi (Mika Iisakkila)
Anyway, here's a simple discrete video output stage. Can't get much simpler than this. Note that there's a serial matching resistor on the output, so you'll have to feed 2Vp-p video into the buffer to get the usual 1Vp-p into the equipment you're driving. This is the way it's usually done. Sorry for the crude transistors, but I hate doing ASCII graphics.
o +5v
|
+---+
| |
1k R |
| |c BC108B
+-b
2Vpp | |e 75ohm
video |e +--------R-------> video out
in >---b | 1Vpp @ 75 ohm
|c R 1k +-->
BC178B| | |
+---+ ---
|
o -5v (yes, two-sided power supply, not ground)
o +5v
|
+----+-----------+
| | |
3k8 R R 680R R 56R
| | |
| | |e BC178B
Video in | +---------b
1Vpp/75R + | |c |c 100n plastic
>---+----||---+--b BC108B +-----||--------> to black level clamping
| 47u | |e | 2Vp-p
| | | R 220R
R 75R R +-----------+
| |1k8 R 150R
--- --- |
--- (single power supply this time)
Here is the design for some gates I made for a kid's lab show and tell which did all the different gates and the culminating two projects were a JK flip-flop and a four bit binary adder.
You use open-collector logic and thus the ouput is simply the collector of the transistor for each simple gate. An inverter and a NAND can be made quite easily, and then from those, any other gate you wish! You pull up the base with a 10K ohm resistor or even larger, and you also hang diodes to that connection as well which are the inputs. They isolate each input from the others, (if any), and allow the current to the normally ON transistor to pull the collector low except when one of those inputs with a diode is grounded by a previous collector or switched input. It might be a good idea to pull up the output (collectors) weakly as well with a 10K to the 5V source. The emitters always go to ground. Thus a typical inverter/NAND gate looks like this:
+5V
+5V |
| 10K
| |
10K +-------> output
| |/ c
input A -----|<|-----+---+------|b
| |\
| |e
input B -----|<|-----+ |
-----
---
-
Then set about reducing gates which are double inversions, and look up DeMorgan's Laws to manipulate the sense of AND and OR with inverters for your optimal use of transistors. If some book doesn't show you the "magic castle" means of making an XOR with 4 NANDS, 4 transistors, then here it is:
+--------------------+------+
in A | | NAND |---+
------+---+------+ +-----+------+ +-----+------+ out
| NAND |---| | NAND |--->
------+---+------+ +-----+------+ +-----+------+
in B | | NAND |---+
+--------------------+------+
If you really want to be true to form, you can replicate the standard old original TTL NAND gate with the original components except with discrete packages, using 3 of this transistor and the oft stated values for the components in an actual 7400 TTL QUAD 2-INPUT NAND gate, even as the plain 7400's are today! The voltage thresholds come out rather accurately, actually! You can find that design in many textbooks and old databooks on TTL logic. The logic I had you use is called open collector DTL logic, and it was used in the early 70's both on chips and out of discrete components. Experiment with fan-out and the power dissipation and compare it to the TTL emulation. Oh, and that funny multi-emitter thing used in the input end of the TTL can be accurately relicated with diodes, as its transistor nature is really not being exploited. It would be more correct to draw it as diodes, as it uses the bc junction for a diode as well!
Try using a simple transistor like a 2N4124 just feed the base from the emitter of the opto-isolator. Don't forget to put a clamping diode across the relay (reverse biased) and maybe a 1uF cap. Drive the photo transistor with an IR LED and couple it with some heat shrink tubing.
+12V +12V | | / C +-----+------relay----+ | <- photo trans. / C | \v.E________________B_| --- | \v.E /_\ 220K 2n4124 | | clamping diode. | | | GND GND GND