ASSEMBLY & OPERATING INSTRUCTIONS

for the

W9GR DSP-3 DIGITAL SIGNAL PROCESSOR


Return to Quantics Home Page

Clicking on one of the following links will take you to the corresponding place in the DSP-3 Kit Assembly Instructions


        ASSEMBLY & OPERATING INSTRUCTIONS (version 1.71)
                             for the
               W9GR DSP-3 DIGITAL SIGNAL PROCESSOR
                    (GR = "Great Reception")

     The W9GR DSP-3 is an improved low cost digital signal
processor for radio communications use. It features 18 DSP
functions, a LED display, and 13 bit audio precision.
     To use these instructions and build this kit, you should be
capable of constructing a PC board from a parts list and
schematic diagram, recognize electronic component parts, identify
integrated circuit pin numbers, and solder. Potential kit
builders lacking these skills are referred to The Radio Amateur's
Handbook. This is not a kit for beginners!
     The PC board is double sided with silk screen and solder
mask. The "silk screen" is the set of white markings (such as
"C14" and "R5") on the PC board which guides you in inserting the
parts at the correct location. The components mount on the side
with the silk screen.
     The dark colored solder mask will help to avoid "solder
bridges" between adjacent leads. It also provides insulation
between the traces on the board and the exposed metal parts on
the bottom of some of the IC sockets. Because of this, be careful
not to scratch the solder mask; otherwise, there could be shorts
or intermittent shorts between the IC sockets and the traces
which run underneath the IC sockets.
     The installation of parts progresses from the shortest
parts, such as IC sockets and resistors, to the tallest parts
such as the electrolytic capacitors and heatsink.
     This double sided board has plated-through holes. At various
places on the board, the traces on the top side need to connect
to traces on the bottom side. A plated-through hole called a
"via" performs this function. If you look at the board, you can
see several examples of "vias" under IC1, the largest IC on the
board. Notice that the vias have somewhat smaller tinned copper
pads than the pads where the parts mount. When you install parts,
do not install them in these "via" holes, which may be close to
the hole where the part is supposed to go. Most of the parts have
silk screened lines which will help you avoid making this error.
     If you find you made a mistake and need to unsolder and
remove a part, be careful so as to not remove the plating or the
pads from the plated-through hole. Use a minimum amount of heat,
and either a vacuum solder removal device or wire braid to remove
the solder from the hole.
     The electrolytic capacitors are polarized, and must be
installed in a particular direction. There is a "+" mark on the
PC board where the "+" or positive end of the capacitor must be
installed. Sometimes capacitors have the "-" end marked instead
of the "+" end. In this case, insert the unmarked end, which will
be the "+" end, into the "+" hole on the PC board. As an
additional check on polarity, the "+" lead on the polarized
capacitors will usually be longer than the "-" lead.
     Similarly, the two diodes have a circular band at one end;
this marked end should match up with the markings on the silk
screen.
     The integrated circuits must also be inserted in the correct
direction. Each IC will have either a notch or a "dimple" at one
end. This end of the chip must match the silk screen markings on
the board.
     The integrated circuits used in this kit are susceptible to
electrostatic damage. When handling the ICs, use standard
procedures to avoid static damage. In brief, ground the circuit,
the ICs, and most importantly yourself before handling static
sensitive parts.
     When you insert the ICs into their sockets, it is very
important to check that the pins are all straight and that the IC
is pushed into its socket straight. Otherwise, an IC pin can
easily be bent under the chip as you insert it, and visual
inspection will not reveal the problem!
     If you need to remove a chip from its socket, insert a small
screwdriver between the socket and the chip alternately at one
end then the other, and back the chip out straight so as not to
bend its pins.
     Review the following parts list and identify all parts
before soldering them into the board. Notice that one of the 20
pin IC sockets resembles a wire-wrap type, with its pins bent 90
degrees to accommodate the LED bargraph display.
     Some parts may be marked with a number that indicates the
value indirectly. According to this convention, the first two
digits are followed by a number of zeros which is indicated by
the third digit. "103" on a disc ceramic capacitor would indicate
10 followed by 3 zeros: 10000 pF or equivalently, 0.01 uF. The 22
pF capacitors may be marked "220J03." The "220" part of the
marking means "22 followed by no zeros." So this means 22 pF, not
220 pF! The 22 pF capacitors may also be marked simply "22."


Parts List
     Unless otherwise specified, all resistors are 1/4 watt 5%
and capacitors are 10% tolerance. Some of the capacitors may be
marked with a three digit number where the last digit represents
the number of zeros to follow the first two digits. For example,
"103" would translate to 10000 picofarads, which would be the
same as 0.01 uF (0.01 microfarad). Sometimes these numbers are
combined with other digits or letters.
     Any of the 74XX series TTL parts (marked with a * below) may
be supplied in any the following variations: 74XX, 74LSXX, 74SXX,
74FXX, 74ALSXX, 74HCXX, 74HCTXX, 74AHCTXX, 74HCTLSXX, etc. For
IC5, a HCMOS type is recommended (e. g. HCT, HCTLS, etc.).

Quantity  Reference Designator(s)  Description
2         C13, C14            22 pF silver mica or disk ceramic capacitor
                              (may be marked "22" or "220J03" or "22PJ3")
1         C7                  0.001 uF ceramic 20% (may be marked "102" or
                              "102Z")
2         C5, C8              0.01 uF ceramic 20% (may be marked "103" or
                              "103Z")
12        C6, C10, C11, C15,  0.1 uF monolithic ceramic, 20%
          C17, C18, C20, C21, (may be marked "104," "104M," ".1Z," ".1M," or
                              "104Z")
          C22, C23, C24, C25
3         C2, C3, C4          1.0 uF electrolytic (16V or greater)
2         C1, C9              10 uF electrolytic (16V or greater)
3         C12, C16, C19       220 uF electrolytic (16V or greater)
1         R14                 2.7 ohms (red-purple-gold-gold)
1         R16                 10 ohms (brown-black-black-gold)
1         R13                 39 ohms (orange-white-black-gold)
11        R8, R21, R22, R23,  330 ohms (orange-orange-brown-gold)
          R24, R25, R26, R27
          R28, R29, R30
1         R7                  820 ohms (gray-red-brown-gold)
1         R15                 1000 ohms (brown-black-red-gold)
1         R2                  1800 ohms (brown-gray-red-gold)
2         R10, R12            3000 ohms (orange-black-red-gold)
2         R4, R5              6.2 K ohms (blue-red-red-gold)
6         R1, R3, R17, R18,   10 K ohms (brown-black-orange-gold)
          R19, R20
2         R6, R9              100 K ohms (brown-black-yellow-gold)
1         R11                 100 K ohm pot
2         D1, D2              1N4001 silicon diode (diodes with higher
                              voltage ratings such as 1N4002, 1N4003, 1N4004
                              etc. may be supplied)
1         X1                  20 MHz fundamental crystal
1         IC1                 TMS320P15 DSP CPU
1         IC2                 TCM320AC39 audio codec (TCM320AC38 may be
                              substituted)
2         IC3, IC4            74299 shift register (*)
1         IC5                 7400 quad NAND gate (HCMOS type recommended)
                              (*)
3         IC6, IC7, IC11      7474 dual flip flop (*)
1         IC8                 74368 hex inverter (*)
1         IC9                 74139 decoder (*)
1         IC10                74374 octal flip flop (*)
1         IC12                LM380 audio power amplifier
1         VR1                 7805 or LM340T-5 TO-220 voltage regulator
2         L1, L2              10 uH RF chokes
1         LED1                10 segment LED bargraph, Radio Shack 276-081B
                              or equivalent
1         J1                  Coaxial DC power jack (2.1 mm)
2         J2, J3              1/8" audio jack
1         J4                  1/8" stereo headphone jack
2         S1, S2              DPDT push-push switch
2                             Switch caps for S1 & S2 (1 red; 1 gray)
1         S3                  16 position binary rotary encoder switch
1         S4                  SPDT toggle switch (for BIO function)
1                             TO220 heatsink
5                             14 pin IC socket
2                             16 pin IC socket
4                             20 pin IC socket
1                             40 pin IC socket
1                             20 pin right angle socket (for LED bargraph)
2                             knob
1                             PC board
1                             set of #4 hardware for VR1 & heatsink

The W9GR DSP-3 is a digital signal processor which operates on audio signals.
It connects between your receiver's audio output and your loudspeaker. It
provides filtering functions for noise reduction, automatic notch filtering,
FSK filters, a SSTV filter, narrow SSB filters, CW filters, a DTMF decoder,
and a CTCSS decoder. A Product Review of this DSP appeared in the August 1995
issue of QST (pages 73-75). This DSP is also featured as a construction
project in the 1996 ARRL Handbook.


SOME CONSTRUCTION HINTS

o Choose a place to work where you can leave the kit sit for a while if
necessary. The workplace should be uncluttered, and stray parts from other
projects should not wander into the area.

o Use a 15-50 watt soldering pencil with a clean tip.

o Use only rosin core solder intended for electronic use.

o Use bright lighting. A magnifying lamp or bench-style magnifier may be
helpful.

o Do your work in stages, taking frequent breaks to check your work.

o Carefully remove wire cuttings so they do not create shorts.

o For the best appearance and finesse, locate parts so that the component
marking is visible when viewed from the direction that the silk screen
marking is read. In the case of resistors, insert them so that the color
bands all read in the same direction; left to right when viewed from the
direction that the silk screen marking is read.

o One source of further information on kit building is the article Kit
Builder's Primer (Yeah, I built that) by Mike Bryce, WB8VGE which appeared in
the December 1994 issue of 73 magazine.


PC BOARD ASSEMBLY

o 1. First install and solder each of the fixed resistors.

o 2. There are three pads at jumper position JP1 which are used to configure
the DSP for the right CPU chip. Take one of the clipped-off resistor leads
and solder it between the center pad at jumper JP1 (MC/MP-) and the "1" pad.
There is a silk screened line between the two pads. The other position is
used only for special purposes - if you intend to provide your own firmware
in the 57C43 EPROMs at IC13 and IC14, and your own TMS320C10 or TMS320C15 CPU
chip at IC1 (The "0" position tells the TMS320C10/15 processor to look for
firmware in the external EPROMs at IC13 and IC14. The "1" position tells the
TMS320P15 processor to use its internal program.)

o 3. If you do not want the headphone jack to mute speaker audio when you
plug in a pair of headphones, then install a jumper wire at JP3. If you leave
the jumper wire out, then when you plug in a pair of headphones, the speaker
audio output at J3 will mute.

o 4. Install and solder the two 1N4001 silicon diodes (D1 & D2). Be sure to
match the polarity bands on the diodes to the ones on the silk screen.

o 5. Install and solder the IC sockets. Do not install sockets at IC13 and
IC14. These ICs are only used for special applications.. Be sure to align the
notch at the notched end of the socket with the notch on the PC board silk
screen. Sometimes sockets will have a beveled mark, a dot, a "1," or some
other distinguishing characteristic at pin 1 instead of a notch. (Pin 1 is
the lower left hand corner viewed from the top with the notch on the left;
pin numbers increase counterclockwise from pin 1.) (Note: if you look at the
board from the "front" end, where the LED bargraph is located, all the
notched ends go to the left.)

o 6. Install and solder the two RF chokes, L1 and L2.

o 7. Install and solder the 0.001 uF ceramic disc capacitor at C7.

o 8. Install and solder the 0.01 uF ceramic disc capacitors at C5 and C8.

o 9. Install and solder the twelve 0.1 uF monolithic ceramic capacitors.

o 10. Install and solder the two 22 pF capacitors at C13 and C14. These may
be either silver mica or ceramic types.

o 11. Install and solder the rest of the electrolytic capacitors, being
careful to match the polarity markings on the board with those on the
capacitors. The electrolytic capacitors are polarized, and must be installed
in a particular direction. There is a "+" mark on the PC board where the "+"
or positive end of the capacitor must be installed. Sometimes capacitors have
the "-" end marked instead of the "+" end. In this case, insert the unmarked
end, which will be the "+" end, into the "+" hole on the PC board. As an
additional check on polarity, the "+" lead on the polarized capacitors will
usually be longer than the "-" lead.

o 12. Install and solder the 20 MHz crystal at X1. If you leave a little lead
length on this component, it will not break off if inadvertently bent over.
The crystal supplied may have a third wire soldered to the metal case. If it
does, cut the wire off.

o 13. In this step, the voltage regulator (VR1) is mounted together with a
heatsink to the PC board, and then the leads are soldered to the board. Use a
small amount of thermal grease (not supplied) between VR1 and the heat sink.
(Thermal grease is not absolutely essential. If you do not have any, just
make sure that the voltage regulator and heatsink are clean and free of
debris so that the best thermal contact can be made without thermal grease.)
If the heatsink has solder tabs which would interfere with the voltage
regulator leads, cut them off. Use the #4 hardware supplied to mount VR1 and
the heatsink to the board.

o 14. In this step you will install the two audio connectors J2 and J3 at the
rear of the PC board. But first you will need to bend the socket pins
slightly. If you look at each connector you will see that there is a "kink"
in each of the three pins. Using a needle nose pliers, squeeze each pin so
that it becomes flat instead of kinked. There are two plastic pins that go
through corresponding holes on the PC board. Insert the two plastic pins and
the three de-kinked metal leads through the holes on the PC board. The
plastic pins establish the socket's height above the PC board. Adjust the
sockets so that they are parallel to the PC board surface before soldering
them in place. This will avoid mechanical interference problems later when
the board is installed in the enclosure.

o 15. Install the coaxial DC power jack at J1 on the rear of the board.

o 16. Install the DPDT push-push switches at S1 and S2 on the front of the
board. Be sure to orient the actuators towards the edge of the board, and
make sure that the switches are seated firmly on the board before soldering,
to avoid mechanical problems later. Install the red switch cap on S1 (ON/OFF)
and the gray switch cap on S2 (IN/OUT).

o 17. A right angle 20 pin IC socket is supplied for the LED bargraph
display. In most applications the PC board will mount horizontally and the
LED bargraph will poke through an opening in your front panel. Make sure that
the socket is firmly seated against the board before soldering, so that
mechanical problems can be avoided later.

o 18. Install the stereo headphone jack at J4. Let the plastic pins stay on
the board surface and establish the socket's height above the PC board.

o 19. Locate the 100 k ohm AF GAIN control. Look at the control with the
shaft side facing you. If there is a metal locating tab beside the shaft,
bend it outward so that it will not interfere with the front panel. Install
the AF GAIN control at R11 on the front of the board.

o 20. Locate the 16 position rotary encoder switch. Look at the encoder with
the shaft side facing you. If there is a metal locating tab beside the shaft,
break it off with a pair of pliers. Install the encoder at S3 on the front of
the board. Make sure that the encoder is firmly seated against the board
before soldering, so that mechanical problems can be avoided later.

o 21. Using three short pieces of hookup wire (not supplied) (about 2 inches)
connect the SPDT toggle switch S4 to the three pads at JP2. The middle pad
goes to the middle terminal of the switch, the "1" pad goes to the bottom
switch terminal, and the "0" pad goes to the top switch terminal. This way,
when the switch is up, BIO will be set to a logic one.

o 22. Do not insert the integrated circuits or the LED display into their
sockets just yet! A "smoke test" will be performed to make sure that the
supply voltage applied to the logic integrated circuits is not excessive. For
the "smoke test" it is best to use a 12 volt power supply which is current
limited to approximately 1 ampere. If you do not use a current limited power
supply, there is the danger of burning traces off the board if there is a
short. With none of the integrated circuits installed in their sockets
(except for VR1, the soldered-in voltage regulator), apply power and turn on
the switch S1. You can apply power either at the pad E2, or via the 2.1 mm
coaxial DC power connector at J1. The center post is positive, and the
outside shell is negative at J1. Using a voltmeter, connect its negative lead
to ground. While viewing the board with the LED bargraph at the right side,
at VR1 the leftmost pin should have approximately 12 volts present.
(Depending on your power supply, the voltage may be as low as 8 volts or as
high as 16 volts.) There should be zero volts at the middle pin. The
rightmost pin must have 5 volts +/- 0.5 volts. If there is more than 5.5
volts at the rightmost pin of VR1, do not install the socketed integrated
circuits. If you do, they may be destroyed and the warranty will be voided.
Check the installation of components and correct the problem before
proceeding. After this test is completed, disconnect the power.

o 23. After you have verified that there is not excessive voltage on the
nominally 5 volt output of the voltage regulator IC VR1, disconnect power and
insert the integrated circuits into their sockets, being careful to insert
them straight and with the proper orientation. On most ICs it is usually
necessary to bend the leads slightly inward so that they will go into the
socket straight. Bend the leads as a group against a flat surface. Your kit
may be supplied with "LS" (low power Schottky) parts, "HCT" (high speed CMOS)
parts, or some other compatible version. So for example, a 74LS139 may be
installed at IC8 where the board is labeled "74139." Similarly, a 74HCTLS00
may be installed at IC5 where the board is labeled "7400."

o 24. Your DSP-3 kit uses three dual D flip flop ICs of the 74xx74 family.
You will find included with your DSP kit two 74AHCT74 variants of this
device, and one 74LS74 variant (bipolar). To make sure that the DSP works
properly over temperature extremes, be sure to install the bipolar 74LS74
device at IC6, and install the two 74AHCT74 devices at IC7 and IC11.

o 25. Insert the LED bargraph into its socket. The markings, if any, are
usually along the pin 1-10 edge, which goes down in this application. If the
bargraph does not light up you can try reversing it; we have found a few
bargraphs with the lettering along the pin 11-20 (top) edge.


INSTALLATION IN YOUR
CIRCUIT & ENCLOSURE

     If you purchased the custom made box for the DSP-3, it is installed by
first removing the panel nuts and washers from the AF GAIN control (R11) and
from the rotary encoder (S3). Insert four #4x1/2 screws through the holes in
the cabinet bottom. Put a 1/4" aluminum spacer on each screw. Slide the front
of the DSP board into the cabinet so that the LED bargraph, switches,
headphone jack, and controls poke through the appropriate holes. Attach the
board with four #4 lockwashers and nuts. S4, the "BIO" switch connected to
the pads at JP2, installs on the back panel of the enclosure. Make sure that
the switch does not interfere with JP1, the MC/MP- wire jumper, or C13. If
your kit was supplied with silver mica capacitors, you may have to bend C13
out of the way. Put the control washers and nuts on the AF GAIN and mode
select controls. Do not over tighten the nut on the AF GAIN pot or it may
bind. Put a knob on both of these controls. Do not push the knob all the way
on to the AF gain pot or it may bind.
     The lid is attached to the bottom with four #6 sheet metal screws. If
your enclosure is supplied with screws that are 3/8" long, you will also find
four #6 flat washers and finishing washers. The sheet metal screws should
each go through a finishing washer, then a flat washer. This will stop the
screw from protruding too far into the cabinet and damaging the PC board.
     If you did not purchase the custom made box, install the PC board in
your enclosure. (If there is room to spare, you may want to install it in
your speaker cabinet.) Any of the switches, jacks, LED bargraph display and
volume control may be removed from the PC board and installed on your front
or rear panels. If you remove parts, make sure that the wires you use to
attach the components make the right connections.
     In case you choose not to use the DC and audio connectors supplied,
there is a second set of I/O (input/output) connections next to the heatsink
at VR1. The four PC board pad connections are GND, +12 volts, audio in, and
audio out.
     High speed CMOS, as used in this project, produces EMI in prodigious
quantities. As with most any project designed for use in a ham environment,
it is highly recommended that you shield the PC board by installing it in a
metal cabinet and bypass signals going into and out of the cabinet. Some of
this EMI bypassing is already provided on the PC board itself. If you do not
shield the digital signal processor, at least its autonotcher firmware will
help clean up some of its own radiated birdies which you might tune across!


TESTING THE DSP

     To "bring up" the DSP, you will need the following:

     1. A source of audio, which may be taken from your receiver.
     2. A loudspeaker.
     3. A 12 volt DC power supply.

     Do not apply power yet. First, make sure that the on/off switch S1 is
off, which is the "out" position. Apply the audio from your receiver or any
other source which is capable of driving a speaker to the 1/8" input
connector at the rear labeled J2 and "AF IN." Alternatively, you may apply
the audio at pad E3. Similarly, connect your speaker to the 1/8" rear output
connector at J3 labeled "AF OUT." You should now be hearing "straight
through" audio passing through the turned-off DSP. Next, set the mode switch
S3 to the 12 o'clock position (knob pointer straight up or flat on shaft
straight down). Next, apply DC power. If you use the coaxial DC power
connector at J1, the center pin is positive. Turn the power on by pushing S1
in. When you first turn the power on, you should see a pattern of LEDs light
up and quickly sweep back and forth across the LED bargraph. Then you should
be seeing some LEDs light up in response to the input audio. Try increasing
and decreasing the audio to see how the bargraph display responds. If you
push S2 in, you will be listening to the digitally processed audio. With S2
out, the DSP is bypassed and the signal only passes through the AF gain
control and analog audio amplifier stage. Finally, with the in/out switch S2
in, rotate the mode switch through its various modes, and you will be able to
hear the effect of the various DSP modes.


OPERATION

     The DSP will normally connect between your receiver and your
loudspeaker. The DSP requires a source of 12 volt DC power to operate. There
are three connections to be made:

     1. Audio in - from your receiver
     2. Audio out - to your speaker
     3. DC power - a source of reasonably clean 12 volt DC power

     The front panel controls, from left to right, are:

o Mode switch (S3). This 16 position rotary switch selects the DSP's
operating mode.

o On/off switch (S1). This switch turns the power on and off. When this
switch is pushed in, the power is on; when the switch is out, the power is
off. When the DSP is turned off, the audio input is connected to the audio
output. This will bypass the DSP when its power is off.

o In/out switch (S2). This switch enables or bypasses the digital signal
processing function. When this switch is pushed in, the output signal has
been digitally processed; when the switch is out, the digital processing is
bypassed and the signal only passes through the analog speaker amplifier.
This switch does not affect the LED display. In other words, when the switch
is out, bypassing the DSP, the LED display will continue to indicate audio
level or decoded tones (depending on the setting of the mode switch S3).

o AF GAIN control (R11). This controls the audio output level from the DSP.

     The DSP includes a 10 segment LED bargraph display. In most of the
operating modes, the LED bargraph displays audio level. The audio level
display is a peak reading type with a delayed recovery characteristic. Each
LED represents a 3 dB change in audio level. The maximum input level without
clipping is 5 volts peak to peak, which is about 1.8 volts rms.
     Although the hardware and software have been designed to accept and
process audio across a wide range of input levels, it is best to operate the
DSP so that the audio level makes most of the LEDs on the bargraph display
illuminate on peaks. The recommended method of operation is to set the
receiver audio gain so that on the strongest signals, most or all of the LEDs
occasionally come on. Once the receiver gain is properly adjusted, do not
touch it; instead use the AF gain control on the processor to adjust speaker
volume. This procedure will keep the A/D and D/A precision highest, and the
quantizing noise lowest. Most modern receivers have good AGC characteristics
with a flat AGC slope, the result being consistent audio level for both
strong and weak signals. If your receiver's audio level is inconsistent, you
may experience occasional overloads of the digital signal processor,
producing clipping distortion. If this happens, simply turn down the audio
input a tad.
     The "BIO" switch (S4) selects DSP configuration options and activates
special functions. It may be thought of as a "mode" switch. "BIO" stands for
"binary I/O."
     The DSP functions included in the DSP-3 are arranged with function #1
being at the "12 noon" position of mode selector switch S3. From the straight
up #1 position, the functions are numbered clockwise. The 16 positions and
the 18 DSP functions are as follows:


o 1. Combined automatic notch and denoiser. This mode simultaneously
automatically notches carriers and reduces noise. This mode is recommended
for most HF SSB operation. When the BIO switch is set to "1" in this mode,
audio AGC is enabled, and audio level will be digitally brought up to full
level if the input level lights at least one LED of the bargraph display.
     These functions both use the Widrow-Hoff LMS adaptive filtering
algorithm. The noise reducer mode is most effective against hiss and thermal
noise but also reduces impulse noise and static crashes. Noise reduction
reduces listener fatigue and is recommended for long-term monitoring. The
automatic notch function eliminates multiple carriers very quickly, within a
few milliseconds. Tuner-uppers, CW interference, carriers, and other forms of
undesired audio tones are quickly eliminated. If a carrier comes on your
frequency, all you will hear will be a subtle "click" as the automatic notch
acquires.

o 2. Denoiser. This mode simultaneously reduces noise with somewhat greater
effectiveness than the denoiser included in mode #1. This mode is recommended
for weak voice signals, including HF SSB, VHF SSB, and FM. When the BIO
switch is set to "1" in this mode, audio AGC is enabled, and audio level will
be digitally brought up to full level if the input level lights at least one
LED of the bargraph display.

o 3. Automatic notch. This mode automatically notches carriers with somewhat
greater effectiveness than the automatic notch included in mode #1. This mode
is recommended for HF SSB operation where interfering carriers are the most
objectionable problem. When the BIO switch is set to "1" in this mode, audio
AGC is enabled, and audio level will be digitally brought up to full level if
the input level lights at least one LED of the bargraph display.

o 4. 2.1 kHz narrow voice FIR filter. This mode is a fixed "brick wall" 100th
order narrowband voice FIR filter, intended for rejecting adjacent
overlapping SSB signals. There is no adaptive noise reduction or automatic
notching included in this mode. The theoretical filter specifications are as
follows:

passband: 300-2100 Hz +/- 0.3 dB ripple
<150 Hz attenuated >50 dB; >2300 Hz attenuated >70 dB

o 5. 1.8 kHz narrow voice FIR filter. This mode is a fixed "brick wall" 100th
order narrowband voice FIR filter, intended for rejecting adjacent
overlapping SSB signals. There is no adaptive noise reduction or automatic
notching included in this mode. The theoretical filter specifications are as
follows:

passband: 300-1800 Hz +/- 0.3 dB ripple
<150 Hz attenuated >50 dB; >2100 Hz attenuated >70 dB

o 6. RTTY filter. This is a 100th order linear phase bandpass filter. This
mode is intended for 170 Hz shift frequency shift keying (FSK) signals such
as Baudot, AMTOR, etc. When the BIO switch is in the "0" position, North
American RTTY tones are selected (2125/2295 Hz). When the BIO switch is in
the "1" position, European RTTY tones are selected (1275/1445 Hz).
Theoretical specifications are as follows:

BIO=0 (North American RTTY tones)
passband 2075-2345 Hz +/- 0.3 dB ripple
<1875 Hz attenuated >60 dB; >2545 Hz attenuated >60 dB

BIO=1 (European RTTY tones)
passband 1225-1495 Hz +/- 0.3 dB ripple
<1025 Hz attenuated >60 dB; >1695 Hz attenuated >60 dB

After changing the position of the BIO switch to change between high and low
tones, this mode must be reselected with the mode switch S3 before the new
mode option takes effect.

o 7. HF packet filter or SSTV filter. This is a 100th order linear phase
bandpass filter. This mode is intended for 200 Hz shift 1200 baud HF packet
FSK signals or SSTV signals. When the BIO switch is in the "0" position, the
HF packet mode is selected (1600/1800 Hz). When the BIO switch is in the "1"
position, this mode becomes a SSTV filter (1200-2300 Hz). Theoretical
specifications are as follows:

BIO=0 (HF packet)
passband 1550-1850 Hz +/- 0.4 dB ripple
<1350 Hz attenuated >65 dB; >2050 Hz attenuated >65 dB

BIO=1 (SSTV)
passband 1075-2350 Hz +/- 0.3 dB ripple
<875 Hz attenuated >70 dB; >2550 Hz attenuated >60 dB

After changing the position of the BIO switch to switch between HF packet and
SSTV, this mode must be reselected with the mode switch S3 before the new
mode option takes effect.

o 8. DTMF Decoder. This mode will decode "dual tone multi frequency (DTMF;
also known as "Touch-Tone") signals, using the LED bargraph to indicate the
decoded tone pairs. You can use this mode to test DTMF encoders, troubleshoot
autopatches, monitor patch access, etc. The LED bargraph does not indicate
audio level in this mode because it must be used to indicate decoded DTMF
tones. To use this mode, first use one of the other modes to set the input
level so that at least three LEDs light up when a DTMF tone pair is received.
When you first select this mode, a unique pattern of two illuminated LEDs at
each end indicates that no valid DTMF tones have yet been received. When a
valid DTMF tone pair is received, the corresponding LED pattern will
illuminate and stay illuminated until another valid DTMF tone pair is
received. The LED patterns are as follows:

DTMF signal    LED pattern
Unknown   **------**
1         *---------
2         -*--------
3         --*-------
4         ---*------
5         ----*-----
6         -----*----
7         ------*---
8         -------*--
9         --------*-
0         ---------*
*         *****-----
#         -----*****
A         **--------
B         --**------
C         ------**--
D         --------**

* = LED illuminated
- = LED not illuminated

     NOTE: if you set the IN/OUT switch to the IN position, you may notice a
chopping noise in the DTMF mode. This is a normal artifact of the DSP's DFT
algorithm which is used to search for DTMF tones. If you want to listen to
the received audio in this mode, simply set the IN/OUT switch to OUT.
     The DTMF decoder may occasionally "trip" on voice audio. This is normal
and due to the momentary presence of audio energy at the DTMF tone
frequencies.
     You can also use the DTMF mode to "play back" DTMF sequences. The DTMF
mode has a 16 tone memory. To play back the last 16 tones heard, simply
toggle the BIO switch (in other words, change the switch's position, either
from "1" to "0" or "0" to "1"). The previous tones (up to 16 maximum) will be
played back in the order that they were received. This form of memory is
called "first-in first-out" (FIFO). The LED patterns will light up one by one
on the LED display at the rate of about one per second. The display will
blank between patterns to make it easy to discern repeated digits. NOTE: if
you switch the DSP out of the DTMF mode, the DTMF playback memory will be
erased.
     If a DTMF sequence is received and you want to make sure it is not
overwritten in playback memory, simply disconnect the audio input to the DSP
or turn off the receiver feeding the DSP. Otherwise, any new DTMF tones
received will be added to the memory, and the sequence you want to review may
be pushed off the end of the FIFO memory.

o 9. CTCSS Decoder & Squelch. This mode is useful for decoding subaudible
frequency "continuous tone-coded squelch system" (CTCSS) tones. This mode
will decode any of the 38 subaudible tones (also known as "PL" tones) used
for repeater access, tone squelch, etc. This mode is useful when testing your
own CTCSS transmissions, for determining the CTCSS tone frequency by
monitoring stations on a repeater's input frequency, etc.
     IMPORTANT NOTE: this mode may not work if your receiver cuts off the low
frequency range below 300 Hz! If your receiver removes the tones, then the
DSP cannot detect them!
     You can often get around this problem by (a) increasing the coupling
capacitor sizes within the audio section of your receiver, or (b) driving the
DSP from the "discriminator output" of the receiver. In any event Quantics
will not be liable for any damage done incidental to making modifications to
any receiver for this purpose.
     Another way to ensure the integrity of the subaudible tones is to use a
receiver which includes a "wide FM" mode such as the ICOM R7000. Often the
"narrow FM" mode will roll off low audio frequencies, and the "wide FM" mode
will not.
     Intermodulation distortion can create low frequency signal components
which can be falsely interpreted as CTCSS. If the signal you are monitoring
has too much deviation for the receiver you are using, then distortion will
result. Distortion can also be caused by overdeviation, off-frequency
transmitters and/or receivers, speech clipping, etc.
     Proper operation of the CTCSS mode depends on having a low distortion
signal path with adequate low frequency response. Remember the adage:
"garbage in, garbage out!"
     The LED bargraph does not indicate audio level in this mode because it
must be used to indicate decoded CTCSS tones. To use this mode, first use one
of the other modes to set the input level so that at least three LEDs light
up on signal peaks. A pattern of one illuminated LED at each end indicates
that no valid CTCSS tone is being received. When a valid CTCSS tone is
received, the corresponding LED pattern will illuminate and stay illuminated
only as long as the CTCSS tone is present. The LED patterns are as follows:

CTCSS signal   LED pattern
NO CTCSS  *--------*
67.0      **--------
71.9      -**-------
74.4      --**------
77.0      ---**-----
79.7      ----**----
82.5      -----**---
85.4      ------**--
88.5      -------**-
91.5      --------**
94.8      ***-------
97.4      -***------
100.0     --***-----
103.5     ---***----
107.2     ----***---
110.9     -----***--
114.8     ------***-
118.8     -------***
123.0     ****------
127.3     -****-----
131.8     --****----
136.5     ---****---
141.3     ----****--
146.2     -----****-
151.4     ------****
156.7     *****-----
162.2     -*****----
167.9     --*****---
173.8     ---*****--
179.9     ----*****-
186.2     -----*****
192.8     ******----
203.5     -******---
210.7     --******--
218.1     ---******-
225.7     ----******
233.6     *******---
241.8     -*******--
250.3     --*******-

* = LED illuminated
- = LED not illuminated

     When the IN/OUT switch is set to OUT, the DSP passes the input audio to
its output. But when the IN/OUT switch is set to IN, this mode can be used as
a simple tone squelch. When the switch is set to IN, then the audio will be
muted unless any CTCSS tone is detected. In other words, this mode will pass
the audio if any of the 38 CTCSS tones is received.
     You can also use the CTCSS mode to "play back" detected CTCSS tones. The
CTCSS mode has a 16 tone memory. The CTCSS memory will only record a new tone
if it is different from the previous detected tone. To play back the last 16
different tones heard, simply toggle the BIO switch (in other words, change
the switch's position, either from "1" to "0" or "0" to "1"). The previous
tones (up to 16 maximum) will be played back in the inverse order that they
were received (this is the opposite of the DTMF playback order). In other
words, the last tone received will be the first one displayed. This form of
memory is called "last-in first-out" (LIFO). The LED patterns will light up
one by one on the LED display at the rate of about one per second. NOTE: if
you switch the DSP out of the CTCSS mode, the CTCSS playback memory will be
erased.
     If a series of CTCSS tones is received and you want to make sure they is
not overwritten in playback memory, simply disconnect the audio input to the
DSP or turn off the receiver feeding the DSP. Otherwise, any new CTCSS tones
received will be added to the memory, and the sequence you want to review may
be pushed off the end of the LIFO memory.

o 10. 400 Hz CW filter, 100 Hz bandwidth. This is a fixed linear phase FIR CW
filter of 229th order. Theoretical specifications are as follows:

passband 350-450 Hz @ -3 dB
<270 Hz attenuated >60 dB; >530 Hz attenuated >60 dB

A short 400 Hz "dit" is sent when this filter is selected, to indicate the
center of the passband. This filter may be tuned anywhere from 400 Hz down to
280 Hz.

o 11. 400 Hz CW filter, 50 Hz bandwidth. This is a fixed linear phase FIR CW
filter of 229th order. Theoretical specifications are as follows:

passband 375-425 Hz @ -3 dB
<320 Hz attenuated >50 dB; >480 Hz attenuated >50 dB

A short 400 Hz "dit" is sent when this filter is selected, to indicate the
center of the passband. This filter may be tuned anywhere from 400 Hz down to
280 Hz.

o 12. 600 Hz CW filter, 100 Hz bandwidth. This is a fixed linear phase FIR CW
filter of 229th order. Theoretical specifications are as follows:

passband 550-650 Hz @ -3 dB
<470 Hz attenuated >60 dB; >730 Hz attenuated >60 dB

A short 600 Hz "dit" is sent when this filter is selected, to indicate the
center of the passband. This filter may be tuned anywhere from 600 Hz down to
420 Hz.

o 13. 750 Hz CW filter, 200 Hz bandwidth. This is a fixed linear phase FIR CW
filter of 229th order. Theoretical specifications are as follows:

passband 650-850 Hz @ -3 dB
<575 Hz attenuated >60 dB; >925 Hz attenuated >60 dB

A short 750 Hz "dit" is sent when this filter is selected, to indicate the
center of the passband. This filter may be tuned anywhere from 750 Hz down to
525 Hz.

o 14. 750 Hz CW filter, 100 Hz bandwidth. This is a fixed linear phase FIR CW
filter of 229th order. Theoretical specifications are as follows:

passband 700-800 Hz @ -3 dB
<620 Hz attenuated >60 dB; >880 Hz attenuated >60 dB

A short 750 Hz "dit" is sent when this filter is selected, to indicate the
center of the passband. This filter may be tuned anywhere from 750 Hz down to
525 Hz.

o 15. 750 Hz CW filter, 50 Hz bandwidth. This is a fixed linear phase FIR CW
filter of 229th order. Theoretical specifications are as follows:

passband 725-775 Hz @ -3 dB
<670 Hz attenuated >50 dB; >830 Hz attenuated >50 dB

A short 750 Hz "dit" is sent when this filter is selected, to indicate the
center of the passband. This filter may be tuned anywhere from 750 Hz down to
525 Hz.

o 16. 1000 Hz CW filter, 100 Hz bandwidth. This is a fixed linear phase FIR
CW filter of 229th order. Theoretical specifications are as follows:

passband 950-1050 Hz @ -3 dB
<870 Hz attenuated >60 dB; >1130 Hz attenuated >60 dB

A short 1000 Hz "dit" is sent when this filter is selected, to indicate the
center of the passband. This filter may be tuned anywhere from 1000 Hz down
to 700 Hz.


o Tuning the CW filters. To tune any of the CW filters, perform the following
steps:

     1. Turn the power off.
     2. Set the BIO switch S4 or jumper wire connected to the pads at JP2 to
"1."
     3. Set the mode selector switch S3 to the CW filter position you want to
tune.
     4. Turn on the power. The DSP will produce a slowly declining audio tone
at the center of the CW filter bandpass. When the audio tone reaches the
frequency you want, turn the mode selector switch S3 to any other position.
The filter is now tuned to the note that you heard when you turned the mode
selector switch. If you do not switch the mode selector during the declining
tone, the CW filter tuning will terminate at a limit of -30% detuning.

     After this process, all of the CW filters will be tuned down by the same
percentage. So if you tuned the 1000 Hz filter to 900 Hz, for example (-10%),
then the 750 Hz filters will be at 675 Hz, the 600 Hz filter will be at
540 Hz, etc. This condition will remain until you turn the power off. Even if
you select one of the voice modes and then return to the CW filters, the
filters will remain tuned to the same frequency unless the power is turned
off. When you select one of the CW filters after the tuning process, the
"dit" sent will be at the new center frequency of the filter.
     This process does not affect or tune the other filters (RTTY, SSTV,
narrow SSB, etc.).

o BIO switch (S4) function summary.

Modes 1,2,3:   BIO=1: digital AGC enabled
               BIO=0: digital AGC disabled
Modes 4,5:     BIO has no effect
Mode 6:        BIO=1 selects
               European (low) RTTY tones
               BIO=0 selects
               North American (high) RTTY tones
Mode 7:        BIO=1 selects SSTV filter
               BIO=0 selects HF packet filter
Modes 8:       BIO toggling initiates DTMF tone playback
Modes 9:       BIO toggling initiates CTCSS tone playback
Modes 10-16    BIO=1 enables CW filter tuning
               BIO=0 disables CW filter tuning

     Note: in modes 6 and 7, the BIO switch does not have an immediate
effect. You must re-select the desired mode after changing the BIO switch.
If, for example, you are using the RTTY filter and you want to switch from
high tones to low tones, you must re-select the RTTY mode after changing the
BIO switch position. This is because the firmware only reads the BIO switch
when mode 6 or 7 is first selected.
     In modes 1, 2, and 3, the BIO switch does have an immediate effect
(selecting or de-selecting digital audio AGC).
     When the DSP is powered up with the BIO switch in the "0" position, a
quick LED test sweeps back and forth across the display four times (two
"round trips") with one LED at a time illuminated. When the BIO switch is in
the "1" position at power up, the same LED test is performed, followed by
another two "round trips" of LED sweeps where multiple LEDs light up. This
enables you to tell the position of the BIO switch, which may be located on
the rear panel, without having to look at it.

o Digital AGC and Reading the LED bargraph audio level display. In most of
the DSP's modes, the LED bargraph display indicates audio level. In these
modes, each LED corresponds to a 3 dB change in input level. Since there are
10 LEDs, the range displayed is 30 dB. The "meter" is peak reading.
     When the digital AGC function is selected (BIO=1 in modes 1, 2, and 3),
the AGC range is limited to 30 dB so that noise, hum, etc. are not brought up
to full output amplitude in the absence of a signal. So if you use AGC in
modes 1, 2, or 3, as long as at least one LED comes on, the audio output
level will be brought up to full amplitude.
     The AGC function is entirely DSP based, and it exists in software. In
other words, there is not a separate analog AGC chip in the DSP hardware. The
recovery time constant is very slow, taking about 30 seconds to adapt to a
new (lower) signal level. The time constant was made deliberately slow to
avoid "pumping" and "breathing" artifacts. Attack time is fast, virtually
instantaneous. If the input level suddenly drops, and you do not want to wait
up to 30 seconds for the AGC to adapt, you can make it re-initialize by
turning the mode switch S3 away from the mode you are using, then moving it
back. Then the AGC will instantly re-initialize to the new signal level, and
the output of the DSP will be at full amplitude.


o The LMS "denoiser" algorithm and different types of noise. The LMS
"denoiser" algorithm discriminates for or against signals depending on their
degree of autocorrelation or repetitiveness. Thermal or atmospheric noise
generally has no repetitiveness whatsoever. An interfering carrier, on the
other hand, has each cycle identical to the next, and therefore has a high
degree of repetitiveness. Speech lies somewhere between these two extremes.
     The denoiser works by automatically forming bandpass filters around the
most significant spectral lines in the signal, allowing the repetitive
components of voice signals to pass, while greatly reducing random signals
like thermal or atmospheric noise.
     The automatic notch works by rejecting those parts of the signal which
are highly repetitive. The adaptive filter automatically forms narrow
bandpass filters around the undesired audio tones, which are then subtracted
from the signal.
     If the signal you are trying to listen to is occasionally buried by
strong impulse noise or static crash QRN, the denoiser will not work any
miracles. But if you have background QRN, such as essentially "white" thermal
noise, it will help a lot. It is most effective with noise that is more or
less constant (like hiss) rather than spiky noise (the kind conventional
noise blankers take out). But whereas a noise blanker works best if the noise
is very strong, the LMS algorithm works best if the noise is moderate to
weak. If the noise you are experiencing comes in bursts of an amplitude such
that during the noise bursts the instantaneous signal to noise ratio is still
at least a small positive number (6 dB or preferably more) then the LMS
algorithm DSP filter will probably help. But if the impulse noise is strong,
then a conventional noise blanker would work better, because it has access to
the wideband RF signal prior to IF filtering.
     When people say "noise" or QRN it can mean a lot of things. If the noise
consists of strong impulses related to 50/60 Hz power line frequencies or to
an automotive ignition rate, then the DSP will not do much, but a
conventional noise blanker will. On the other hand, if the noise is closer to
white noise or maybe a background "frying" noise, as opposed to a "buzz,"
then the DSP will provide some relief.


o A few notes on the LMS automatic notch algorithm. The LMS automatic notch
algorithm forms extremely narrow notch filters at frequencies where it finds
continuous audio tones. In order to stop the automatic notch from attacking
voice signals, it is necessary to make it sensitive only to pure audio tones,
and to make the notch quite narrow. This means that if the tone has any sort
of "wobble" or frequency modulation on it, it will not receive as much
attenuation. If your receiver has AC hum on its local oscillator or beat
frequency oscillator (BFO) which makes any FM, you may notice that the
effectiveness of the notch is reduced. Similarly, particularly on 40 meters,
where a tone may be due to several co-channel foreign broadcast stations'
carriers, the resulting complex phase modulation may produce enough "wobble"
in the audio tone that the effectiveness of the notch filter is reduced.

o Bandwidth limiting in the DSP modes. When the DSP is switched in, you may
notice a reduced audio bandwidth. This should not be noticeable when you are
driving the DSP with SSB type audio, which normally has about a 3 kHz
bandwidth. But if you are driving the DSP with a wideband audio source, such
as a broadcast FM radio, you will hear the bandwidth reduction. This is
normal. U2, the audio codec chip, includes switched capacitor lowpass filters
for antialiasing and reconstruction. The filters in U2 cut off at about
3.2 kHz. These filters ensure that the Nyquist criterion is not exceeded,
which requires that the highest audio frequency must be less than half of the
sampling rate. The sampling rate in this DSP is approximately 7700 Hz.


o A few words about the CW filters. Although some amateurs use the noise
reduction mode for CW, the dedicated CW filters are better in most
applications. The noise reduction modes work by automatically forming little
bandpass filters around the most significant spectral lines in the signal,
and on CW, the bandpass filters must recreate themselves with each dit and
dah. But in the case of CW, we have a priori knowledge of the desired
signal's spectrum, so we can design a fixed filter which is already optimal
and which does not need to adapt itself. Furthermore, the fixed CW filters
are linear phase (no time delay distortion) but the adaptive filters formed
by the LMS algorithm will not in general be linear phase. Where it would make
sense to use the noise reduction mode for CW would be in the case where you
are tuning the band and want a wide (SSB type) bandwidth to hear more
signals. But when you want to hear only a single signal, it is probably best
to use the dedicated CW filters.
     On the subject of "ringing" there is a lot of misinformation. All
bandpass filters "ring" - they must do so in order to work! But there are two
causes of ringing. The first cause of ringing is the fact that the bandwidth
has been reduced. All bandpass filters ring due to this mechanism. The second
cause of ringing is nonlinear phase, or phase distortion. Conventional
filters incur this additional ringing due to nonlinear phase. A FIR filter is
linear phase, and so it does not ring because of this mechanism. In other
words, for a given bandwidth, a FIR filter will ring less than a conventional
analog filter, but it will still ring. The CW filters in this DSP are all FIR
types, so there will be only minimal ringing.
     Most radios set the sidetone so that if you are transmitting on the same
frequency as the CW station you are working, the note of the received signal
will be at the same frequency as your sidetone. But if your radio's sidetone
does not match the received signal's audio frequency, then the CW sidetone
will be outside the passband of the digital CW filter, it will not be heard!
There are several ways around this problem.
     First, it may be possible to modify the transceiver so that the sidetone
is centered within the DSP CW filter bandpass you use (400, 600, 750, or 1000
Hz). This may be a matter of changing parts values in an RC oscillator timing
circuit. But if the sidetone is derived by dividing down from a crystal
timebase, then it might not be possible to change its frequency.
     Another way is to tune the CW filter to the right frequency as outlined
in these instructions.
     Yet another way would be to bypass the DSP when transmitting, either by
switching it out manually or by using a relay in place of the switch at S2.

o Power supply considerations. The DC power required is 12 volts at 400 ma
maximum. It should be fairly well filtered, meaning not more than a few
tenths of a volt of AC ripple. Many "wall transformer" supplies do not have
adequate filtering even though they may be rated to supply the necessary
400 ma. However, they can be bolstered with the addition of about 5000 uF of
additional filtering capacitance. So, if you experience AC hum, try adding a
4700 uF, 5000 uF, or larger capacitor across the DC power supply.

o Receiver AGC effects. Keep in mind that your receiver's AGC takes place
prior to the DSP. So if the DSP is filtering out a strong carrier, or strong
QRM on CW, the audio output level of the DSP may drop significantly. However,
this will be due to the receiver's AGC gain reduction action because the QRM
is present in the receiver.


o Further information. An explanation of the algorithms used in this digital
signal processor and a description of the hardware are in the September 1992
QST magazine article "Low Cost Digital Signal Processing for the Radio
Amateur" by Dave Hershberger, W9GR. For those looking for a thorough
mathematical treatment of the firmware noise reduction and autonotcher
algorithm, refer to "Using  the  LMS Algorithm for QRM and QRN Reduction" by
Dr. Steven E. Reyer, WA9VNJ, and David L. Hershberger, W9GR, in September
1992 QEX magazine.


HARDWARE DESCRIPTION

     +5 volt power and ground connections are not shown on the schematic.
They are:

     ground    +5 volts            ground    +5 volts
IC1  10        30             IC8  8         16
IC2  16        5              IC9  8         16
IC3  10        20             IC10 10        20
IC4  10        20             IC11 7         14
IC5  7         14             IC13 10        20
IC6  7         14             IC14 10        20
IC7  7         14

     The DSP-3 hardware starts with a 13 bit digital audio codec chip, IC2.
13 bits of precision is theoretically equivalent to an 80 dB audio dynamic
range. This is 30 dB more range than the original W9GR DSP which only used 8
bits of audio precision. The 13 bit dynamic range, together with improved DSP
software, makes setting audio levels much less critical. It also makes it
possible to do much better CW filtering, since a wide dynamic range is
necessary to filter out a weak CW signal amidst much stronger QRM. For voice
applications, given that the signal to noise ratio on a relatively strong SSB
signal might be 30 dB, there is 50 dB of S/N "headroom" left with 13 bits! In
many cases, the S/N ratio will be quite a bit worse than 30 dB. Since 80 dB
is much greater than 30 dB, 13 bits is more than adequate for amateur
applications.
     The audio codec chip IC2 uses switched capacitor technology for
antialias lowpass filtering. It also includes a switched capacitor highpass
filter. The lowpass filter cuts off at about 3.2 kHz. IC2 accepts the analog
input at pin 18, and it outputs a serial digital audio bit stream at pin 13.
Shift register chips IC3 and IC4 convert the serial output from IC2 to a
parallel word which is can be read via the parallel data bus by the DSP CPU
chip IC1. The sampling rate is approximately 7700 Hz.
     The CPU chip IC1 contains a OTP (one time programmable; not erasable)
PROM memory which has been pre-programmed with the DSP software or firmware.
The CPU executes the instructions in its internal 8k bytes of program memory
to perform the various DSP functions. IC1 is a 16 bit digital signal
processor CPU with a 32 bit accumulator. This processor executes DSP type
instructions at a rate of 5 MIPS.
     For special applications, the CPU (IC1) can be replaced with one that
does not contain internal program memory, such as a TMS320C10 or TMS320C15.
These processors require an external program memory, which can be provided by
using 57C43 high speed EPROMs at IC13 and IC14. This capability is provided
for advanced experimenters who want to do their own DSP software development.
Normally, IC13 and IC14 are not used in this DSP.
     The rotary encoder switch at S3 outputs a 4 bit binary code which is
periodically read by the CPU via the tristate inverter IC8. The setting of S3
determines which DSP function is executed by IC1.
     IC5D is connected as an inverter; this gate produces a power-on reset
pulse. When the power is turned on, this short pulse initializes the CPU
(IC1), the shift registers (IC3 and IC4) and some of the various remaining
flip flops. The output of IC5D should go low for a fraction of a second when
the power is turned on.
     IC7A accepts the 5 MHz clock output from IC1 via IC8 and produces a
2.5 MHz clock which is necessary to run the audio codec chip IC2. IC5A, IC5B,
IC5C, IC7B, and IC6 all work to produce the proper timing signals to
interface the audio codec chip and shift registers to the CPU.
     IC9 is a port decoder, which produces active low outputs to read the
binary encoder mode switch S3, to read the A/D converter's output from the
shift registers, to write parallel digital audio words to the shift
registers, and to write data to the LED bargraph display.
     When the CPU writes to the LED display, the data word controlling the
individual LEDs is written to IC10 and IC11. IC10 holds the "bottom"
(leftmost) 8 bits of the LED display, and IC11 holds the top (rightmost) 2
bits. A logic 0 at any output turns the corresponding LED on, and a logic 1
turns the associated LED off.
     After the CPU performs the desired filtering operation on the input
audio, it outputs parallel audio words to the shift register chips at IC3 and
IC4. These shift registers are bidirectional; in addition to converting the
serial A/D data to parallel, they also convert the parallel digital audio
data back to serial again so that it can be sent to the D/A in IC2.
     The processed serial digital audio data enters IC2 at pin 8. Within IC2,
it is converted back to analog again. IC2 also includes a switched capacitor
lowpass filter for analog reconstruction and sin(x)/x correction. The
filtered analog output is taken from IC2 pin 2.
     At this point the only thing that is needed is AF GAIN control and audio
amplification. R11 is the AF GAIN control, and IC12 provides over 1 watt of
audio output power.
     The 12 volt DC power supply powers the audio power amplifier IC12. The
12 volt supply is dropped to 5 volts to power the audio codec IC2 and the
digital circuitry by linear voltage regulator VR1.
     In the interests of tradition, a small PC board "cartoon" is also
included with the DSP hardware at no additional charge.


IN CASE OF DIFFICULTY

     o 1. Check your soldering for unconnected pins, cold solder joints,
and/or solder bridges. This is a very important troubleshooting step! Almost
every kit which has been returned to us for repair has not worked because of
simple soldering defects! The odds are that you can save yourself a lot of
time and possible embarrassment if you carefully check your soldering work
for unsoldered connections, "cold" joints, solder bridges, and scratched off
PC board traces. Use a magnifying glass to find suspicious connections, and
re-solder them. Don't rely on the solder mask to be an infallible insulator
against sloppy soldering. Remove solder blobs which encroach over the solder
mask or onto nearby "vias." Use a soft brush to remove metallic and
conductive solder dust and debris. In several cases, the soldering problem
was found underneath an IC socket, necessitating removal and replacement of
the socket! Sometimes capillary flow can draw solder up through vias and
holes to the other side of the board. If this happens underneath a socket, it
could be very hard to find the problem. It's really true what Heathkit used
to say in their manuals: 90% of the problems with Heathkits were due to bad
soldering.
     o 2. Check all of the integrated circuits to see if pins are bent
underneath the chips. With the power off, use an ohmmeter to check for
continuity from the IC pin to the pad underneath the board. Or, you can
remove each chip for visual inspection and then reinstall them.
     o 3. Use an oscilloscope if you have one available to look for "sick"
looking (i. e. lack of full logic swings) digital waveforms. If you find a
sick looking logic waveform, it may be shorted to a neighboring trace. Look
for bad soldering somewhere on that node. Bear in mind that many of the logic
signals on the board are "tri-state" meaning that the drivers go high
impedance at certain times. You may see what looks like "sagging" voltages;
this is normal on some nodes such as the data bus. Look for the logic swings
to go all the way to a valid logic 1 and a valid logic 0 at least some of the
time. With the power off, use an ohmmeter to check for
     o 4. From our statistical experience so far, there is at least a 95%
probability that any given non-working kit has soldering defects. To search
for soldering defects, begin by first going around each chip with an
ohmmeter. Check to see if pin 1 is shorted to pin 2, pin 2 to 3, etc. on each
chip. If you find pins shorted together, refer to the schematic to see if
they are supposed to be connected; sometimes that will be the case.
     o 5. If you do not find a short that way, then take your ohmmeter and go
around each chip and check not only adjacent pins on each IC, but every other
pad on the PC board. Each time you find continuity, check against the
schematic to see if the two points which you found to be connected are in
fact supposed to be connected. Although this may sound time consuming, it may
be the fastest way to get your kit working.
     o 6. Some of the shorts we have found have been due to very small
dendrites of solder which worked their way under the solder mask. You might
have to look very closely, with a magnifying glass, to find them!
     o 7. We have learned of several failures which have occurred after weeks
or months of normal operation. These failures were due to residual debris and
flux from soldering working through the solder mask and contacting nearby
traces. These failures may not be "zero ohm" shorts but may measure in the
tens, hundreds, or thousands of ohms. If you experience such a delayed
failure, it may be cured by using a small brush and chemical flux remover.
Alternatively, you might try carefully and gently scraping off the flux and
soldering residue from around each pad with a small screwdriver or awl, being
careful not to scrape off the traces.
     o 8. Make sure that all grounds are connected together among the devices
that connect to the DSP: the 12 volt power supply, the audio source
(receiver), the DSP unit itself, and the speaker. Make sure that the "hot"
and "ground" wires in cable connectors are not reversed.
     o 9. Check the two diodes at D1 and D2; the banded ends should be
oriented towards the front (LED bargraph) end of the board.
     o 10. Does the mode switch (S3) appear to be intermittent or inoperative?
If so, the bipolar 74LS74 chip is probably not installed at IC6. Check to
make sure that a bipolar type 74LS74 is installed at IC6. (Not a CMOS 74HCT74
device!) (Refer to assembly step #24.)
     o 11. Do the noise reduction modes seem to increase the noise level
rather than reducing it? Check the setting of S4. It should be set to "0"
(AGC off). When S4 is set to "1," then AGC is enabled in the noise reduction
modes. The AGC can increase the signal level up to 30 dB. When switching the
DSP in and out with S2, it can sometimes seem like the noise is increased,
but this is only because the overall gain is higher when the DSP is in when
the AGC is on. When comparing the output to the input, always make sure that
S4 is set to "0," so that the "DSP in" and "DSP out" settings of S2 will have
the same gain.
     Also, when evaluating the noise reduction capability of the DSP-3, be
sure to use a noisy signal - as opposed to pure noise, with no desired signal
present. Make sure that there is at least some signal present, to give the
algorithm a correlated signal to acquire.
     o 12. Check the DC voltages at various points throughout the circuit.
With power on but no signal applied, you should measure the following DC
voltages:

IC5 pin 11                    5.0
IC2 pins 2,3,4,17,18,19,20    2.5
IC12 pins 1,8                 approximately half of the +12 volt power supply
                              (about 6 volts)
IC12 pin 14                   12 volts (supply voltage)
IC10 pin 20                   5 volts (Vcc)

     o 13. Short out the 10 k ohm resistor at R1 with a clip lead. With the
resistor shorted, the voltage at IC5 pin 11 should go to zero. When you
remove the clip lead, the voltage should return to 5 volts. This tests the
reset circuit. When IC5 pin 11 is at 0 volts, the DSP is being reset. If IC5
pin 11 remains at a low voltage, the DSP cannot operate.
     o 14. If you have an oscilloscope, or audio signal tracer, you can trace
audio the signal through the unit. Input audio is applied to the junction of
R5 and R7. Input audio will also appear at pin 19 of IC2, riding on a
2.5 volt DC bias. At this point the audio is digitized inside IC2 leaves in a
serial format from IC2 pin 13. The digital audio is sent to the DSP CPU (IC1)
via shift registers IC3 and IC4. Serial digital audio returns from the CPU
via shift registers IC3 and IC4, and arrives at IC2 pin 8. After being
converted to analog, the audio appears at IC2 pins 2, 3, and 4. From this
point the audio is applied to the AF GAIN control R11 and then to the audio
power amplifier IC12. Finally, the speaker output is driven from IC12 pin 8.
     o 15. Does the bargraph display seem to indicate audio presence? If so,
audio is getting to the A/D converter (IC2) and into the CPU (IC1). The
problem is probably at the D/A converter IC2, S2, or IC12. If the bargraph
does not indicate anything when audio is applied, or if it flashes randomly,
look for problems in the digital part of the circuitry.
     o 16. Of the units that have been returned for "warranty service," about
95% of the problems have been defective soldering. One non-soldering problem
that has occurred several times has been PC board defects. So in the great
majority of cases, the problems have been attributable to defects in kit
building. We say this not to impugn anybody's building skills but to point
out that there have been only about 3 defective electrical components
discovered (out of over 200,000 components shipped as parts of kits).
     o 17. Although the great majority of kit builders have completed the
W9GR DSP kit successfully,  there have been significant numbers of "warranty"
returns, the vast majority of which are anything but that. Over 95% of the
kits returned for "warranty service" do not work because of bad soldering.
The warranty covers "defects in materials and workmanship" of Quantics, and
does not cover the workmanship of the kit builder, over which we have no
control.
     o 18. We are spending so much time correcting soldering defects that it
is slowing down kit shipments. Therefore, effective immediately, any kit
returned for "warranty repair" must be accompanied by a check for $40.00
which is our "flat rate" for getting defective kits to work and returning
them to you (abuse excepted). If in our judgment the failure is due to a
defect in our components or workmanship, your $40.00 payment will be returned
to you with the repaired kit. On the other hand, if the failure is due to
improper assembly or soldering defects, the flat rate fee will be retained.
     o 19. In summary, the warranty covers defects in components supplied by
Quantics (for example PC board manufacturing flaws). The debugging service,
which is not free, covers kit building errors (for example soldering
defects). This policy supersedes and clarifies any previous statements.
     o 20. As of November 1994, the statistics regarding kit problems are as
follows:

     Defective soldering:                    96
     Chips plugged into sockets incorrectly: 2
     PC board flaws:                         3
     Bad parts supplied by Quantics:         2


     o 21. If you are still unable to resolve the difficulty, you may either
write for assistance or return the unit for repair. Any returned kit must be
accompanied by a check for $40.00. If in our opinion the problem is covered
by the warranty, then your check will be returned to you with the kit. The
warranty is not intended to cover defective soldering, so please check your
soldering before returning kits for repair. Support is handled by
correspondence; at the present time we are not able to provide support via
telephone. However, problem reports, suggestions, comments, and questions are
welcomed and should be directed to:

                                  QUANTICS
                               P. O. Box 2163
                     Nevada City, California 95959-2163

If you need a reply to a question, a self-addressed stamped envelope would be
appreciated.
     Internet users can send questions to w9gr@oro.net - this is the fastest
way to get a response. E-mail is normally answered the same day unless we are
out of town. Compuserve users can also send us e-mail by addressing it to
INTERNET:w9gr@oro.net.

UPS shipping address for returned kits:

                         David L. Hershberger, W9GR
                             10373 Pine Flat Way
                     Nevada City, California 95959-9136

Parcel post shipments (U. S. mail) should be directed to the P. O. Box
address above.

COPYRIGHT NOTICE

The program software/firmware (contained within IC1) supplied with this
digital signal processor is copyright (c) 1994-1997 and all rights are
reserved. Software/firmware supplied may not be copied or distributed to
third parties. A license is required for commercial use of this
software/firmware. Licensing inquiries may be directed to Quantics, P. O. Box
2163, Nevada City, California 95959-2163.


LIMITED WARRANTY

This product is warranted to be free of defects in materials and workmanship
for a period of thirty days from the date of purchase. In the event of
notification within the warranty period of defects in materials or
workmanship, the seller will, upon return of the product, repair or replace
(at its option) the defective parts. The remedy for breach of this warranty
shall be limited to repair or replacement and shall not encompass any other
damages, including but limited to loss of profits, special, incidental,
consequential or other similar claims. This warranty does not cover any
damages due to accident, misuse, abuse or negligence on the part of the
purchaser. This warranty does not cover other equipment or components that a
customer uses in conjunction with this product.
     THE SELLER SPECIFICALLY DISCLAIMS ALL OTHER WARRANTIES, EXPRESSED OR
IMPLIED, INCLUDING BUT NOT LIMITED TO IMPLIED WARRANTIES OF MERCHANTABILITY
AND FITNESS FOR A PARTICULAR PURPOSE. Some states do not allow the limitation
or exclusion of liability for incidental or consequential damages, or the
exclusion of implied warranties, so the above limitations and exclusions may
not apply to you. This warranty gives you specific legal rights and you may
also have other rights which vary from state to state.

Thank you for your purchase of the W9GR Digital Signal Processor!
73,  Dave Hershberger,  W 9 G R


Return to Quantics Home Page