The Smith-Kettlewell Technical File

A Quarterly Publication of
The Smith-Kettlewell Eye Research Institute’s
Rehabilitation Engineering Research Center

William Gerrey, Editor

Issue: [current-page:title]

Original support provided by:
The Smith-Kettlewell Eye Research Institute
and the National Institute on Disability and Rehabilitation Research

Note: This archive is provided as a historical resource. Details regarding products, suppliers, and other contact information are original and may be outdated.

Questions about this archive can be sent to







The disappearance of the TSI Mini Speech Board and the National DigiTalker put an end to limited-vocabulary devices requiring only straightforward control logic and standard binary codes; current products have complicated instruction sets that force designers to use microprocessor-based interfaces to even simple numeric displays. The RAM-Talker is a field-recordable speech device whose input addresses can be made logical--the words "zero" through "nine" can be organized in the binary sequence of 0000 through 1001, for example. Since you do the "recording" yourself, the Nattering RAM can speak any language and store the words of your choice.

Note: A new product for Apple computers is the "RAM-Talker Plus"; it is being marketed by Electronic Learning Systems in Florida. More information about this product can be found in "Technology Update" (the newsletter of the Sensory Aids Foundation, 399 Sherman Ave., Palo Alto, CA 94306). Apparently, that RAM-Talker is a piece of hardware containing extra memory plus a speech synthesizer (for which there is no screen review software). We feel the pressing need to change the name of our speech recorder so as not to conflict with their trademarked product name.

Description and Features

A Nippon Electric Corporation 256-kilobit RAM, the uPD43256-15L, is used to store digitized speech. "Entered" by the user via a microphone, the speech is digitized by a "delta modulator," collected in 8-bit bytes in a serial-to-parallel shift register, and stored in the RAM. In playback, 8-bit bytes are taken from the RAM by a parallel-to-serial shift register, fed in a serial string to a delta demodulator (which is just an integrator), fed through a filter to attenuate the "sampling" frequency, and fed to a loudspeaker by a power amplifier.

For an elementary explanation of digital recording and the logic behind this system--including a nice description of delta modulation--see the original RAM-Talker article by Susan and Tom Fowle (SKTF, Summer 1987).

The quality of speech depends on the sampling rate. The clock frequency is adjustable. You should select a recording length just long enough for your longest word (which may be "seven," "siete" or "nueve"). The shorter the message length, the better the quality (and the less impatient you will be when listening to a string of words). With the clock circuit shown, the frequency of the master clock (which is twice the sampling rate) ranges from 32kHz to about 125kHz; the corresponding recording times go from 1 second to about 1/4th second. (The frequency adjustment, a variable resistor with a range of 1K to 6K, has a loading effect on the inverter used as the clock oscillator; hence, the ratio of high to low frequency empirically turns out to be only 4 to 1.)

A DIP switch (S6) allows setting up each address to be recorded (independent of outside control logic). An optional DIP switch (S7) gives the user a presettable word that can be "addressed" through a single input line (perhaps for a "low-battery" message). (By overriding the "Setup/Operate" switch, a second preset word can be had, selected on S6; thus, two preset words with single-line operation are actually possible.)

The board has some fun features: A "pitch-shift" input allows you to accentuate words, or to present decimal digits in a different pitch than more significant ones, thereby eliminating the need to send the address lines over to a "point" word address. The most important feature of any speech thing is a "shut-up" capability; an "Abort" input allows you to immediately truncate its speech.

"Customizing" this addressable speech board for a project is very simple; this can be done after the whole project is complete. A toggle switch marked "Setup/Operate" is thrown to the "Setup" position; this connects the Nattering RAM's address lines to DIP switch S6. One at a time, the desired addresses are "simulated" by settings of S6 and these individual recordings made via a microphone.

To do a block, the "Record" switch is closed. The "Start" button is then depressed; as long as it is held down, the board's counters are "reset" and held in abeyance. As soon as the button is released, information is recorded from the microphone. Therefore, you start speaking immediately as you let up on the button. (You will get feedback unless the volume is turned down. Better yet, though, if an earphone is used during recording, you can hear your message live, and you will know when the "word-complete" pulse has come along to cut you off.) To listen to the recording, just open the "Record" switch and tap the button.

You can make as many tries as you like. Each time you "record," you merely reload the selected block in the RAM, so you can try over and over until you get something you like. In fact, you may find in using a completed talking instrument that you want to modify the pronunciation of a word; just open the cabinet, close the "Setup/Operate" switch, find the desired address on S6, close the "Record" switch, plug in a microphone and re-record that block.

The RAM in the Nattering RAM is "battery-backed" so that the recordings are preserved when the main power is shut off. Any 4.5V battery will do for backup. Nicad batteries are perhaps not ideal for this service because of their internal leakage and spontaneous discharge. You may find, after leaving a nicad-backed device on the shelf for nine months, that the Nicad has died and dumped your recordings. A "primary" type of battery with good shelf life is better--perhaps three AA cells, or even AAA cells.

As described here, changing the backup battery must be done with power on so that temporary absence of the backup battery won't be missed. A large capacitor could be installed to back up the backup battery, but the leakage of such a capacitor would probably cut the battery life in half, so we show no such capacitor. A consumer product (for which the owner is likely to forget to keep the power on) should probably contain a 470uF good-quality electrolytic across the 4.5V battery holder.

Counting the voltage regulator, the Nattering RAM contains 14 chips. No denying it, the Nattering RAM is a board full of stuff, but it uses off-the-shelf parts and doesn't cost very much. (The most expensive item is the RAM, and that is now just about $22.)

Although a more crowded version has been built by Jay Williams, the editor likes a square board of 6 inches on a side. A suggested layout is described immediately following the wire-wrap table. Except for the clock-oscillator components, a couple of diodes and a series resistor in the word-complete line, the digital portion of the circuit (containing ten chips, two resistor packs and some switches) can be wire-wrapped. The delta modulator is another story; since that section is fraught with resistors and capacitors, point-to-point wiring and a hot soldering iron makes the most sense.

Making Talking Instruments With the Nattering RAM

Now, if you have a source of BCD (binary-coded decimal) data (a circuit which can also pulse the "Start" line and wait for the finished "Word-Active" signal), you can make a talking instrument. But wait, it's never that easy, of course.

All too commonly, where you have a source of BCD data, it cannot be caused to wait around for speech. Not only is it usually changing all the time (due to "strobing" of the display), but hand-shake lines are seldom provided for the talking instrument designer.

This is where the FIFO comes in (see the FIFO article in SKTF, Winter 1988). Through a FIFO (first-in, first-out) device, you can accept data from your source as fast as it is available (at a rate of 35MHZ in the case of the ITD72401 FIFO), and then retrieve data from the FIFO's four output lines and make the FIFO feed data to the talker on command of the "Word-Active" Nattering RAM.

With the inclusion of the FIFO, you create a fully latched 16 word-BCD addressable speech board. The only task left is to make provisions for controlling this thing from the device to be made audible; the finalized circuit will depend upon the way the FIFO interfaces with the data source.

Comparisons With the Old RAM-Talker

The chief difference in this new circuit is that the "address counter" (formerly comprised of the CD4040 in cascade with one half of the CD4520) has been shortened, and address lines of the RAM have been brought out (through a multiplexer) to be addressed by external circuitry. Another main difference is a 4-fold increase in sampling rate; made possible by choosing a larger RAM, this assures excellent speech quality.

(Battery backup of this new RAM has to be done differently, since this new one has pin 26--formerly a "NOT Chip Select"--taken over by an address line. Thus, an additional transistor was needed to get a disabling signal to go high when power is dropped. This led to a bunch of other troubles. For example, if the backup battery goes into effect too soon, all the Nattering RAM's current gets drawn from it once in a while. To prevent this, there is a 5.6-volt zener holding this transistor on as long as the main battery is above 6 volts. It all seems complicated, but we think we have a scheme that works.)

Rather than using a mechanical 4-pole record/play switch, CD4053 electronic switches are now employed. This simplifies fabrication, and affords use of a simple SPST toggle for "Record."

Changes have also been made in the delta modulator circuit to make its adjustments more effective. [By the way, the capacitors in the filter (two 0.047's associated with 22K resistors in the original article) were mistakenly large, and should have been 0.0047uF instead.]

Technical Specifications

  • Supply Voltage--7.5V to 12V
  • Battery Backup--4.5V (three AA cells would last a couple of years)
  • Recording Time--16 messages; each block adjustable from 1/4th second to 1 second


  • Logic Inputs--5-volt logic with 47K pull-up or pull-down resistances
  • Microphone Input Impedance--10K
  • "Start Line" (pin 6 of the 4013)--bringing this high resets (goes back to beginning of word); Nattering RAM speaks when this line is allowed to go low again.
  • "Abort Line" (pin 4 of the 4013)--Bringing this high interrupts and "aborts" the word.
  • "Pitch-Shift Line" (pin 9 on S2C, the second 4053)--Bringing this low speeds up the clock by 18%.
  • "Preselected Word 1" (pin 15 of the 74HC257)--Bringing this high speaks the word addressed by DIP switch block S7.
  • "Preselected Word 2" (pin 1 of the 74HC257, in parallel with the "Setup/Operate" switch, S5)--Bringing this low speaks the word addressed by DIP switch block S6.
  • Binary Inputs "1" "2" "4" "8"--74HC257 pins 3, 6, 10 and 13, respectively.


  • "Word-Active" Output (Pin 1 of the 4013)--goes high when the "Start Line" is brought high, stays high while the word is being spoken (which happens after the "Start Line" is brought low again), and goes low upon receipt of the internal "word-complete" or an external "abort" signal.
  • "NOT Word-Active" (Pin 2 of the 4013)--The complement of the above, it goes low with operation of the "Start Line" and goes high again upon receipt of the "word-complete" or "Abort" signal.
  • Audio Output--125 milliwatts into an 8-ohm load.

Logical Modifications

The sense of the various inputs can be inverted by modifying various parts of the circuit. For starters, there is a spare inverter on the hex Schmitt trigger (pin 9 In; pin 8 Out) that can be used to invert any of the lines. The "Pitch Shift" input can be made to raise the pitch when this input goes high. This is done by choosing pin 5 of the CD4053(S2) instead of pin 3 for the connection to the system clock.

(Note: This circuit description comes in two halves. The second half, called "Digital Circuitry," is duplicated in a wire-wrap table; wiring this part will be most easily done from that table instead of the description. The first half of this presentation contains the on-board power supply arrangements, the delta modulator, and the demodulator and amplifier. These are not included in the wire-wrap table; you must build them from the paragraphs immediately following.)

Power to the entire circuit is an unregulated voltage of from 7.5 to 12 volts. The LM386 power amplifier runs off this voltage; other than that, most of the circuit runs off 5 volts. However, the RAM's 5 volts must be "diode-isolated" from that of the rest of the circuit so that it can be "battery backed." The regulator is set up to supply 5.6V. One silicon diode, whose forward drop is 0.6 volts, supplies the main 5-volt line. Another diode supplies the RAM's VDD pin (pin 28 of the uPD43256-15L). (Setting the regulator for 5.6 volts accommodates the diodes' 0.6 volts.) A 3- to 4.5-volt battery supplies the RAM when power is down; it does so through a third diode (which could be a garden-variety one, but is preferably a Schotky diode--such as the 1N5820--chosen for its low voltage drop). Two bias points are also established--one adjustable (for the comparator's offset), and a fixed 2.5V for the preamps and integrators.

Power Supply

The negative terminal of the unregulated supply (probably a 9-volt battery) is grounded. The positive of this goes to the "unregulated VCC line" (the main supply point of the RAM-Talker).

An LM317 is arranged to supply 5.6V. The unregulated voltage goes to the 317's "Input" terminal; this terminal is bypassed to ground by 220uF in parallel with 0.1uF (negative of the electrolytic at ground). The "Output" pin of this regulator goes through 160 ohms, then through 510 ohms to ground; the junction of these resistors goes to the "Adjust" pin of the LM317. The "Output" pin goes to the anode of a 1N4001 diode (D1), with the cathode of this going to the plus 5V line of the project. This plus 5V line is bypassed by 0.1uF in parallel with 470uF (negative of the electrolytic at ground).

The "Output" of the regulator also goes to the anode of a 1N914 diode (D2); the cathode of this diode goes to pin 28 of the RAM. Pin 28 of the RAM also goes to the cathode of a 1N5820 Schotky diode (D3); the anode of this diode goes to the positive of a 4.5V battery, the negative of this battery being grounded. Finally, the RAM's pin 28 is bypassed to ground by 0.1uF.

When power is down, bringing pin 20 high puts the RAM into standby mode. To accomplish this, pin 20 goes to the collector of a 2N2222, the emitter of which is grounded. The junction of the collector and pin 20 also goes through 47K to pin 28, the RAM's VDD pin. The base of this transistor goes through 47K to ground. This base also goes through a 4.7K resistor to the anode of a 5.6V zener diode, D4 (1N752); the cathode of this zener goes to the input of the LM317 (the 7.5- to 12-volt line).

For a fixed 2.5V bias point, the 5V line goes through two 47K resistors to ground. The junction of these resistors is bypassed by a 1uF electrolytic capacitor (negative end at ground).

For the comparator's adjustable bias, a 50K trim pot has its bottom end going through 100K to ground; its top end goes through 100K to plus 5V. The arm of this pot is bypassed to ground by the parallel combination of 0.01uF and a 1uF electrolytic (negative end at ground.

Delta Modulator Circuit

An LM324 quad op-amp chip has its pin 11 grounded, while pin 4 goes to plus 5V. All the non-inverting inputs go to the fixed bias point; pins 3, 5, 10 and 12 are tied together and go to the junction of the 47K resistors and their 1uF bypass.

A closed-circuit microphone jack, wired as a "shorting jack," has its switch contact jumpered to the sleeve. The sleeve of this jack is grounded. The tip contact goes through 0.1uF, then through a 10K metal-film resistor to an inverting input of an op-amp (pin 13) in the LM324 quad op-amp chip. A metal-film feedback resistor of 221K goes between pins 14 and 13 of the LM324. Pin 14, the output of the first stage, goes through a 10K carbon-composition resistor to pin 2, another inverting input. A low-pass feedback network, the parallel combination of 220K and 100pF goes between pins 1 and 2. (Actually, the values of the metal-film resistors are not critical; they are used for their low-noise characteristic. The resistors in the second stage can be any type; carbon-composition will do fine.)

An LM311 comparator has its pins 1 and 4 grounded. Pin 8 of the 311 goes to plus 5V. Pin 3, the 311's inverting input, goes to the adjustable bias point--the arm of the 50K trimmer. Pin 2, the 311's non-inverting input, goes through 47K in series with 0.1uF to the output of the mike preamplifier (pin 1 of the LM324).

Pin 7, the open-collector output of the 311, goes through a 4.7K pull-up resistor to plus 5V. Pin 7 also goes to the "D" input, pin 9, of a CD4013 dual-D flip-flop.

On the 4013, pin 14 goes to plus 5V; pin 7, along with "Set" and "Reset" pins (8 and 10) are grounded.

The "Q" output of this flip-flop, pin 13, goes through 68K to the inverting input (pin 6) of an op-amp (used as an integrator) in the LM324. Between pins 6 and 7 of the LM324 is the parallel combination of 0.047uF and 100K. The output of this integrator, pin 7, then goes to the free end of another 47K resistor off the non-inverting input of the comparator (pin 2 of the 311), thus summing the integrated signal with the microphone input.

The "NOT Q" output of the flip-flop, pin 12 of the 4013, is the digital signal output of the delta modulation encoder.

Delta Demodulator

The input of this demodulator looks at either of two signals: One is the output of the above modulator, so that you can "monitor" the record signal. The other is the serial output of the shift register which is being fed bytes from the RAM.

The demodulator input point goes through 68K to the inverting input (pin 9) of an integrator in the LM324. The feedback network between pins 8 and 9 of the 324 consists of 0.047uF in parallel with 100K. The output of this integrator goes into the low-pass filter:

Pin 8 of the LM324 goes through two 22K resistors in series to the top of a 10K volume control; the bottom of this control is grounded. The junction of the fixed resistors goes through 0.0047uF to ground. The volume control is shunted by another 0.0047uF capacitor. The arm of the volume control goes through 0.1uF to pin 3 of an LM386 audio amplifier. To suppress oscillations in the amplifier, pin 3 also goes through 0.01uF to pin 2.

To boost the gain of this amplifier, pin 8 of the 386 goes to the negative end of a 10uF capacitor. The positive end of this cap goes through 470 ohms to pin 1.

Pins 2 and 4 of the LM386 are grounded. Pin 6 goes to the unregulated positive supply line. To suppress oscillations in the amplifier, it will probably be necessary to bypass pin 6 close to the chip, perhaps with 100uF (negative at ground). Pin 7 of the 386 is bypassed to ground by 22uF (negative at ground). Between pins 5 and 4, located close to the chip, is 0.1uF. Pin 5 goes to the positive end of a 100uF capacitor, with the negative end of this going through the speaker to ground.

(The delta modulator and demodulator can be tested at this point, but you have to apply a clock signal to do so. For details, see the section "Testing and Calibrating the Delta Modulator."

Digital Circuit Description

[It is recommended that the builder read this section for basic understanding, then build the device by following the wire-wrap table listed later.]

The system clock is built around one of the inverters in a hex Schmitt trigger chip (CD40106). An input (pin 1) goes through 0.0022uF to ground; pin 1 also goes through a 1k resistor in series with a 5k rheostat to the inverter's output (pin 2).

The "C" section of a CD4053 (S2, to be introduced later) can be used to raise the pitch for highlighting words. The arm of S2C, pin 4 of this 4053, goes through 390pF to ground. The "Y" position, pin 3, goes to the 40106 pin 1. The 4053 control pin, pin 9, goes through 47K to plus 5V; pulling it low raises the pitch a couple of semitones or so.

The system clock goes to one input (pin 13) of a CD4073 triple 3-input AND gate. The second and third inputs of this gate (pins 12 and 11) are tied together, and will be dealt with later. The output of the AND gate, pin 10, goes to the clock input (pin 1) of a CD4520 dual 4-bit "up counter." The Q1 output of the CD4520 bears the clock signal at the sampling rate.

The Q1 output of the 4520 (pin 3) goes to the clock input of the CD4034 (pin 15). It then goes to an inverter (pin 11 of the 40106), with the inverter's output (pin 10) going to the clock input of the delta modulation circuit's flip-flop (the CD4013's pin 11).

One half of the CD4013 is made into an RS flip-flop by grounding both the "Clock" and "D" inputs, pins 3 and 5. Pins 11 and 12 of the 4073 AND gate go to the "Q" output of the CD4013 (pin 1).

The "Set," pin 6, of this start-stop flip-flop goes through a 47K resistor to ground, as well as going through an SPST pushbutton "start switch" (S3) to plus 5V. This point is also connected to 4520 reset (pin 7), and to the reset of a CD4040 (pin 11). The junction of these pins with the pull-down resistor and the "Start" button also goes to the cathode of a diode (1N914); the anode of this diode is the "Start" input of the Nattering RAM (an input to be fed by control logic of outside circuitry).

The "Reset" terminal of this RS flip-flop gets a "Word-Complete" signal from the highest bit in the address counter; pin 1 of the 4040 goes through 47K to pin 4 of the 4013. (This resistor permits an external "Abort" signal to be imposed on the 4013's pin 4.) The 4013 pin 4 also goes to the cathode of a 1N914. The anode of this diode is the "Abort" input terminal; bringing it high "aborts" the word. (The article to follow suggests three uses for this "abort" input; to get three inputs, you would need three diodes, all with their cathodes going to pin 4.)

The "Q" output on this half of the 4013, pin 1, is a "Word-Active" line; it goes high when the start line is operated and stays high until the flip-flop is "Reset," either by the internal word-complete signal or an external "Word-Abort" signal. A complementary "NOT Word-Active" output exists on pin 2 of the 4013.

Eleven outputs of the CD4040 12-bit counter sequentially address the RAM. (Rather than restate those connections, we refer you to the wire-wrap table where they can be easily seen.) The 4040 is clocked from the inversion of Q4 off the first half of the 4520. Thus, the 4040's clock input, pin 10, goes to pin 4 of the 40106, with pin 3 of this Schmitt trigger going to pin 6 of the 4520.

The first half of the CD4520, in conjunction with a 5-input AND gate, marks the 8-bit "samples." The needed 5-input AND gate is made using two 3-input ones in the CD4073; pin 6, the output of one gate, goes to pin 8, the input of a gate whose other inputs are pins 1 and 2. The output of the resultant gate is pin 9, with its five inputs being pins 1, 2, 3, 4, and 5. The first four of these inputs go to Q1 through Q4 on the 4520 (pins 3, 4, 5 and 6). (The final input of this gate, pin 5 of the 4073, gets one of two versions of the system clock via a switching arrangement which will be described later.)

Pin 9 of the 4073, the output of the combined AND gate, is inverted; it goes to pin 5 of the 40106. The inverter's output, pin 6, bears the "Write-Enable" signal whose associated switching arrangement will be dealt with shortly.

A 5-pole "Record/Play" switch is required. This is comprised of two analog switch chips, CD4053's. These 4053's each have three single-pole double-throw switches--termed switch "A," "B" and "C." (The "C" section in the second package can be used to change the pitch, as described earlier.)

On both 4053's, pin 6 (their "Inhibit" pin) is grounded. Also on both, pin 7 is grounded. (This pin 7 is a VEE pin which can be placed below ground for analog applications.)

The "A," "B" and "C" control pins of the first chip (pins 11, 10, and 9), along with "A" and "B" controls of the second (pins 11 and 10), are tied together and go through 47K to plus 5V. (This resistor pulls them into the "Play" mode.) These control pins also go through an SPST toggle "Record" switch (S4) to ground. When this switch is closed, the RAM-Talker is ready to "Record.") Hereafter, the switches will be labeled S1A for section A of the first 4053, and so on through switch S2B on the second 4053.

S1A determines the direction of data in the 4034 shift register; it also enables the tristate outputs of the RAM. The arm of S1A (pin 14) is grounded. The A/B pin of the CD4034, pin 11, goes through 47K to plus 5V; pin 11 also goes to the "record" side of S1A (pin 12 of the first 4053). Pin 22 of the RAM (the output "NOT-Enable") goes through 47K to plus 5V; pin 22 also goes to the "play" side of S1A (pin 13).

S1B is associated with the "Write-Enable" pin of the RAM (pin 27). This pin 27 is taken through 47K to plus 5V; it also goes to the "record" side of S1B (pin 2 of the first 4053). The "play" side of S1B (pin 1) has no connection. The arm of S1B (pin 15) goes to the output of Schmitt trigger inverter pin 6. The input of this inverter, pin 5, goes to the aforementioned 5-input AND gate, pin 9 of the 4073.

This 4073 pin 9 also goes to the "play" side of switch S1C (pin 3). The record side of this switch (pin 5) is grounded. The arm of this pole (pin 4) goes to both pins 13 and 14 of the CD4034 shift register. (These pins are the "parallel/serial" and "asynchronous/synchronous" terminals of the shift register, respectively.)

The final input of the 5-input AND gate, pin 5 of the 4073, goes to the arm of S2A (pin 14 of the second 4053). The "play" side of this pole (pin 13) goes directly to the clock input of the 4520, pin 1 (a non-inverted version of the system clock). The "record" side of this pole (pin 12) goes to the output of one of the inverters (pin 12 of the 40106), with the input of this inverter (pin 13) going to the system clock--pin 1 of the 4520.

On the 4034 shift register, the "A" bus lines, A1 through A8 (pins 16 through 23), are connected to the RAM's data pins (pins 11 through 13 and 15 through 19). Also on the 4034, the "A-Enable" (pin 9) is tied to plus 5V.

Delta modulation is accepted on the 4034's serial data input (pin 10). Digital (undemodulated) playback information is taken from the 4034's B8 line (pin 1). Thus, this pin 1 goes to the "playback" side of switch S2B (pin 1 of the second 4053). The record side of S2B (pin 2) goes directly to the delta modulator pin 12 of the 4013. The arm of S2B (pin 15) goes to the input of the demodulator (it goes through 68K to pin 9 of the LM324).

The RAM has a total of fifteen address lines. As stated earlier, eleven of these are sequentially addressed by the CD4040. The four highest-order ones go to outputs of a quad 2-input tristate multiplexer; pins 4, 7, 9 and 12 of a 74HC257 multiplexer go to pins 23, 2, 26, and 1 of the RAM, respectively. This multiplexer allows the RAM's address lines to be "disconnected" from outside control logic for "setup" and "recording" purposes.

If desired, a package of four SPDT DIP switches (S7) can be installed for setting up a preset word, controllable by the tristate terminal of the multiplexer. Positions 1 of the switches are grounded; position 2 of them all goes to plus 5V. Each of the multiplexer's outputs goes through a 47K resistor to the arm of its assigned DIP switch. The multiplexer's tristate terminal, pin 15, goes through 47K to ground; pulling pin 15 high will defer the RAM's addresses to the number selected on S7. If this presettable feature is not used, pin 15 of the 74HC257 can be grounded directly.

All the 74HC257's inputs--pins 2, 3, 5, 6, 10, 11, 13, and 14--go through 47K resistors to ground. The 74HC257 "Select" pin, pin 1, goes through a 47K resistor to plus 5V; this pin also goes through an SPST switch (S5) to ground. (This switch is called "Setup/Operate.")

In order to select desired "addresses" for the recorded words, the package of four SPST DIP switches (S6) is used. One side of all four switches goes to 5V. The free ends of the switches go to their respective inputs of the multiplexer--pins 2, 5, 11, and 14, which are so-called "A" inputs of the 74HC257. When S5, the "Setup/Operate" switch, is closed, the RAM's address lines are looking at these DIP switches.

The "B" pins of the 74HC257 are the actual address lines of the RAM-Talker. In respective order: pins 3, 6, 10 and 13 of the 74HC257 are the "1," "2," "4" and "8" binary inputs. S5 has to be open for these inputs to work.

Note that by pulling down on the multiplexer's pin 1, the Select pin, any setting on DIP switch S6 can be used as another preset word. However, this is in conflict with the setup/operate switch. Pulling down on pin 1 with an open collector would be okay; running pin 1 through a diode (anode to pin 1) would also work.

Testing and Calibrating the Delta Modulator

If you like, these circuits can be tested before the main digital portion is built. To protect the start-stop flip-flop side of the 4013 (used elsewhere in the main digital section), it is good practice to temporarily ground pins 4 and 6 (along with 4013 pins 3, 5, 7, 8 and 10, which are permanently grounded). Then, a 5-volt-logic clock signal (from 15kHz to 130kHz) is applied to pin 11, the "clock input."

The clock you apply may be the "master clock" of the project. (On the 40106, pin 1 goes through 0.0022uF to ground. Between pins 1 and 2 is a 5K rheostat in series with a 1K resistor. Pin 2 of the 40106 is the master clock output, and this is temporarily connected to pin 11 of the 4013.)

Whether partially built or a finished Nattering RAM, it is most convenient to use a free-running clock for delta modulator calibration so that the "word-complete" signal doesn't keep stopping the modulator. Direct use of the master clock permits continuous operation of the modulator. On the other hand, it is easy to defeat the word-complete signal in the finished RAM-Talker; temporarily grounding pin 4 of the 4013 (not the "Abort" input, but pin 4 directly) causes the whole system to run continuously.

Arrangements should now be made so that you can listen to the modulator with the demodulator. In the completed Nattering RAM, this is automatically done when the "Record" switch is closed. Otherwise, pin 12 of the 4013 must be temporarily connected to the free end of the 68K resistor off pin 9 of the LM324.

If the offset for pin 3 of the comparator is not just right, the output of the flip-flop will not go "one up, one down, one up ..." as desired for no input signal. Instead, the flip-flop could easily go "three up, two down," etc. The effect of misadjustment is the presence of background noise. While asymmetry of the digital output can easily be displayed on an oscilloscope, adjustment of this bias trimmer can also be made by ear.

If an oscilloscope is at hand, connect its probe to pin 12 of the 4013; also, "externally sync" the scope to this signal at pin 12. With the mike input shorted (which is accomplished by unplugging the mike), slowly run the bias trimmer through its range until a squarewave output is attained. (Actually, since some noise is inevitable, you will see unwanted highs and lows becoming more and more faint as an indication of proper adjustment (this "dimming" of unwanted lines happens as they are refreshed less and less often.)

Luckily, misadjustment is quite audible. We think that what you hear is the second harmonic of the flip-flop's output beating with the master clock. Anyhow, the sort of windy-sounding whistle can be tuned for "zero beat" (the point where the beat note's frequency goes to 0Hz and the system gets quiet).

Nattering RAM Wire-Wrap Table

(Note: This only applies to the digital portion of the circuit. Moreover, the part of the 4013 involved in the delta modulator isn't fully covered here either, so you must build the delta modulator from the above verbal description.)

Each chip has its own "mini-table" within this section. The "mini-table" is not all-inclusive; i.e., if a string of connections has previously involved a pin of the chip being worked on, this connection will not be relisted in that chip's own mini-table. That's a good thing, since any other procedure would lead to possible doubling of wires. (There is an exception to this rule of nonrepetition; the backup battery and standby transistor appear again here, so that you can trouble-shoot all the RAM connections from this table.)

Reading the table is done as follows:

"4520-1 to 4073-10; 40106-13; 4053(S2)-13"

means that "the 4520's pin 1 goes to the 4073's pin 10, as well as going to the 40106's pin 13, as well as to S2's pin 13 (S2 being the 2nd CD4053). Occasionally, a pin of the chip which is the main subject of the table goes to another pin of its own; that's why you'll see things like 4073-6 to 4073-8. Good luck!

CD4520 Dual 4-bit "Up Counter":

  • 4520-1 to 4073-10; 40106-13; 4053(S2)-13
  • 2,15,16 to plus 5V
  • 3 to 4073-1; 4034-15; 40106-11
  • 4 to 4073-2
  • 5 to 4073-3
  • 6 to 4073-4; 40106-3
  • 7 to 4040-11; 4013-6; through 47K to ground; through S3 ("Start" button) to plus 5V; and to the cathode of a 1N914 diode whose anode is the "Start" input terminal
  • 8,9,10 to ground
  • 11,12,13,14 NC (no connection)
  • CD4034 Shift Register:
  • 4034-1 to 4053(S2)-1
  • 2 through 8 NC
  • 9,24 to plus 5V
  • 10 to 4013-12; 4053(S2)-2
  • 11 through 47K to plus 5V; 4053(S1)-12
  • 12 to ground
  • 13,14 to 4053(S1)-4
  • 16 to RAM-11
  • 17 to RAM-12
  • 18 to RAM-13
  • 19 to RAM-15
  • 20 to RAM-16
  • 21 to RAM-17
  • 22 to RAM-18
  • 23 to RAM-19
  • CD4040 12-Bit Counter:
  • 4040-2 to RAM-5
  • 3 to RAM-6
  • 4 to RAM-4
  • 5 to RAM-7
  • 6 to RAM-8
  • 7 to RAM-9
  • 8 to Ground
  • 9 to RAM-10
  • 10 to 40106-4
  • 12 to RAM-25
  • 13 to RAM-3
  • 14 to RAM-24
  • 15 to RAM-21
  • 16 to Plus 5V
  • RAM, NEC brand uPD43256-15L:
  • RAM-1 to 74HC257-12
  • 2 to 74HC257-7
  • 14 to Ground
  • 20 to the collector of the 2N2222; through 47K to RAM-28
  • 22 through 47K to plus 5V; 4053(S1)-13
  • 23 to 74HC257-4
  • 26 to 74HC257-9
  • 27 through 47K to plus 5V; 4053(S1)-2
  • 28 through 0.1uF to ground; to cathode of D2 (1N914, the anode of which goes to plus 5.6V); also to cathode of D3 (1N5820, the anode of which goes to the plus terminal of the 4.5V battery).

Miscellaneous Standby Circuit Items:

The negative side of the 4.5V battery is grounded.

The emitter of the 2N2222 is grounded. The 2222 base goes through 47K to ground. This base also goes through 4.7K to the anode of a 1N752 zener; the cathode goes to the input of the LM317, the unregulated supply.

CD40106 Hex Inverting Schmitt Trigger:

  • 40106-1 to 4053(S2)-3; through 0.0022uF to ground
  • 2 to 4073-13; Through 1K in series with 5K rheostat to 40106-1
  • 5 to 4073-9; 4053(S1)-3
  • 6 to 4053(S1)-15
  • 7,9 to Ground
  • 8 NC
  • 10 to 4013-11
  • 12 to 4053(S2)-12
  • 14 to Plus 5V
  • CD4073 Triple 3-Input AND:
  • 4073-5 to 4053(S2)-14
  • 6 to 4073-8
  • 7 to Ground
  • 11,12 to 4013-1
  • 14 to Plus 5V
  • CD4013 Dual D Flip-Flop:
  • 4013-3,5,7,8,10 to Ground
  • 4 through 47K to 4040-1; also to the cathode of a diode (1N914) whose anode is the "Abort" input terminal (in fact several diodes so connected will give you more "Abort" inputs)
  • 9 to LM311-7
  • 14 to Plus 5V
  • CD4053 (S1) Triple Double-Throw Switch:
  • 4053(S1)-1 NC
  • 5,6,7,8,14 to ground
  • 9,10,11 to 4053(S2)-10,11; through 47K to plus 5V; through S4 ("Record") to ground
  • 16 to plus 5V
  • CD4053 (S2):
  • 4053(S2)-4 through 390pF to ground
  • 5 NC
  • 6,7,8 to ground
  • 9 through 47K to plus 5V; NOT "Pitch shift" Input terminal
  • 15 through 68K to LM324 pin 9
  • 16 to plus 5V

74HC257 Quad 2-Input Multiplexer:

  • 74HC257-1 through 47K to plus 5V; through S5 ("Setup/Operate") to ground; to anode of D4 (1N914) whose cathode is the NOT "Word 1 Select" Input Terminal
  • 2 through 47K to ground; through S6A to plus 5V
  • 3 through 47K to ground; to address "1" terminal
  • 5 through 47K to ground: through S6B to plus 5V
  • 6 through 47K to ground; to address "2" terminal
  • 8 to ground
  • 10 through 47K to ground; to address "4" terminal
  • 11 through 47K to ground; through S6C to plus 5V
  • 13 through 47K to ground; to address "8" terminal
  • 14 through 47K to ground; through S6D to plus 5V
  • 15 through 47K to ground; to "Word 1 Select" Terminal
  • 16 to plus 5V

Suggested Layout

Counting the voltage regulator, the Nattering RAM contains 14 chips. Besides these, there is the DIP switch S6 and an optional DIP switch S7. A good many pull-up and pull-down resistors are required, so two DIP resistor packs, one for pull-up and one for pull-down, were used on the board (these packages looking just like ICs). (The packages chosen were Bourns No. 162-473. Various brands are similar; containing 15 resistors, the common connection--which goes to 5V for pull-ups and to ground for pull-downs--is usually pin 16. On the other hand, because they are a ready supplier, we list some Mouser SIP resistors--single-in-line packages--two of which can be fit into a 16-pin socket.)

The square board--6 inches on a side--is divided up into two "zones." A 1.4-inch wide horizontal area across the top is reserved for items connected by point-to-point wiring (components on which soldering is required). The bottom two-thirds of the board--4.2 inches high by 6 inches wide--is reserved for digital circuitry, plus an LM311 comparator. Except for the comparator, items in this lower section are connected by the wirewrap technique.

The bus lines--ground and plus 5V--are L-shaped; each has a leg on one vertical edge, and then makes a horizontal run along the boundary between the two "zones." On this editor's board, the 5V bus runs along the 15th row from the top, then turns downward and runs vertically for 3.9 inches (terminating 0.6 inches up from the corner so as to afford room for a mounting hole). The ground bus makes its horizontal run along the 17th row from the top, then descends about 3-1/2 inches down the opposing edge. The result creates parallel bus lines at the boundary of the two zones; the larger zone is bordered on the left and right edges by ground and plus 5V. Across the very top, along the first row of holes, is the "main supply line" (7.5V to 12V); again, this falls 0.6 inches short of the corners to make room for mounting holes.

The editor prefers that all chips in a project face the same way. Except for the DIP switches, all packages have their pin-1 ends away from you.

Viewed with the component side up, the LM386 amplifier is near the upper-right corner. To its left is the LM324 op-amp package. To the left of the 324 is the trim pot for the delta-modulator adjustment. Finally, near the upper-left corner is the LM317 voltage regulator; somewhere around this regulator is also the "standby transistor," the 2N2222.

All resistors in the upper zone are discrete ones; although the packaged resistors might work okay here, it makes more sense to locate resistors nearby so as to eliminate long leads.

The lower section is divided up into five vertical "subzones" or columns. (By adjusting the spacing, the columns of vertically oriented chips can all be made three inches long. The resultant uniformity is visually appealing; people come for miles to see it.) Below these columns, a 0.6-inch strip across the bottom is used for the DIP switches--these packages being mounted horizontally along the lower edge of the board.

Viewed with the component side up, the left column contains two resistor packs and the 74HC257 multiplexer (three vertically oriented 16-pin packages with about three blank holes between them). Leaving a blank alley of 0.7 inches, the 24-pin shift register and the 28-pin RAM share a 3-inch column (with the shift register above the RAM). Leaving another 0.7-inch blank alley, the third column, from top to bottom, contains the 4013 dual D flip-flop, the 4520 dual 4-bit counter, and the 4040 12-bit counter. Leaving another blank alley and moving farther to the right, the next column, from top to bottom, contains the 40106 hex inverter, the 4073 AND gate chip, and the first 4053 (the one designated "S1"). Beyond yet another blank alley, at the lower end of the "fifth column" is the 4053 designated "S2." Above this, still in the fifth column, are the S4 and S5 toggle switches and the S3 Start button.

The LM311 comparator can be positioned just under the horizontal ground bus, possibly centered between columns for variety's sake.

Pin Identifications

National LM317T "Adjustable" Voltage Regulator:

With the mounting surface toward you and the leads pointing upward, the three leads are, from left to right: Adjust, Output, Input.

National LM324 Quad Op-Amp:

  • 11--VCC
  • 4--VSS
  • 1--Output A1
  • 2--Inverting Input A1
  • 3--Non-Inverting Input A1
  • 7--Output A2
  • 6--Inverting Input A2
  • 5--Non-Inverting Input A2
  • 8--Output A3
  • 9--Inverting Input A3
  • 10--Non-Inverting Input A3
  • 14--Output A4
  • 13--Inverting Input A4
  • 12--Non-Inverting Input A4

National LM311 Comparator:

  • 4--Minus V
  • 8--VCC
  • 1--Minus V Reference for the Output Swing
  • 7--Output (open collector)
  • 2--Non-Inverting Input
  • 3--Inverting Input
  • 5 and 6--Offset Correction Pins (not used)
  • National LM386 Audio Power Amplifier:
  • 4--Ground
  • 6--VCC
  • 7--Preamp Bypassing
  • 2--Inverting Input
  • 3--Non-Inverting Input
  • 1 and 8--Gain Boost Pins (not used)
  • RCA CD4013 Dual D Flip-Flop:
  • 7--VSS
  • 14--VDD
  • 1--Q1
  • 2--NOT-Q1
  • 3--Clock1
  • 4--Reset1
  • 5--D1
  • 6--Set1
  • 13--Q2
  • 12--NOT-Q2
  • 11--Clock2
  • 10--Reset2
  • 9--D2
  • 8--Set2

RCA CD4040 12-Bit Counter:

  • 8--VSS
  • 16--VDD
  • 11--Reset
  • 10--Clock
  • 9--Q1
  • 7--Q2
  • 6--Q3
  • 5--Q4
  • 3--Q5
  • 2--Q6
  • 4--Q7
  • 13--Q8
  • 12--Q9
  • 14--Q10
  • 15--Q11
  • 1--Q12

RCA CD4073 Triple 3-Input AND:

  • 7--VSS
  • 14--VDD
  • 1, 2, 8--In1
  • 9--Out1
  • 3, 4, 5--In2
  • 6--Out2
  • 13, 12, 11--In3
  • 10--Out3

RCA CD4034 Shift Register:

  • 12--VSS
  • 24--VDD
  • 16 through 23--A Bus, A1 through A8, respectively
  • 8 through 1--B Bus, B1 through B8, respectively
  • 9--A-Enable
  • 10--D Serial Input
  • 11--A/B (When high, A pins are inputs and B pins are outputs.) 13--P/S (high for parallel)
  • 14--A/S (high for asynchronous)
  • 15--Clock
  • RCA CD4520 Dual 4-Bit Counter:
  • 8--VSS
  • 16--VDD
  • First Counter:
  • 1--Clock
  • 2--Enable
  • 3--Q1
  • 4--Q2
  • 5--Q3
  • 6--Q4
  • 7--Reset

Second Counter:

  • 9--Clock
  • 10--Enable
  • 11--Q1
  • 12--Q2
  • 13--Q3
  • 14--Q4
  • 15--Reset

RCA CD40106 Hex Inverting Schmitt Triggers:

  • 7--VSS
  • 14--VDD
  • 1--In1
  • 2--Out1
  • 3--In2
  • 4--Out2
  • 5--In3
  • 6--Out3
  • 9--In4
  • 8--Out4
  • 11--In5
  • 10--Out5
  • 13--In6
  • 12--Out6

74HC257 Quad Tri-state 2-Input Multiplexer:

  • 8--Ground
  • 16--VCC
  • 1--Select
  • 2--In 1 A
  • 3--In 1 B
  • 4--Out 1
  • 5--In 2 A
  • 6--In 2 B
  • 7--Out 2
  • 9--Out 3
  • 10--In 3 B
  • 11--In 3 A
  • 12--Out 4
  • 13--In 4 B
  • 14--In 4 A
  • 15--Tristate (bringing this high puts the outputs into tristate mode.)

RCA CD4053 Triple Analog Switch (3 spdt's)

Note: The three switches are controlled by pins labeled "A," "B" and "C." When a control pin is high, its switch is in the "y" position; a logic 0 on the control pin throws the switch in the "x" position. The terminals of the switches are labeled with lower-case letters; for example, the "y" position of the first switch is called "ay."

  • 8--Ground
  • 16--VCC
  • 7--VEE (Can be taken below ground so that AC signals can be accommodated.)
  • 6--Inhibit (When high, no switch is closed.)
  • 11--A (control)
  • 12--ax
  • 13--ay
  • 14--Arm of a
  • 10--B (Control)
  • 2--bx
  • 1--by
  • 15--Arm of b
  • 9--C (Control)
  • 5--cx
  • 3--cy
  • 4--Arm of c
  • NEC uPD43256-15L 32K by 8-BIT Static RAM (Note that the prefix actually starts with the Greek letter "mu"; i.e., muPD43256-15L):

Note: It is absolutely necessary that you get a RAM with "L" tacked on the end of the number; this means "low-power," which is necessary for battery backup.) You can get suffix numerals of 10, 12, or 15; these stand for access delay times of 100ns, 120ns, and 150ns, respectively. The 15s are cheapest, and with speech being intrinsically slow, we're not in a hurry over a few nanoseconds.

  • 14--VSS
  • 28--VDD
  • 10, 9, 8, 7, 6, 5, 4, 3, 25, 24, 21, 23, 2, 26, and 1--Address Lines A0 through A14, respectively
  • 11, 12, 13, 15, 16, 17, 18 and 19--Data Lines 0 through 7, respectively
  • 20--NOT Chip Select
  • 22--NOT Output Enable
  • 27--NOT Write Enable

Parts List

Resistors (Low-Wattage Garden Variety 5%, unless otherwise stated):

  • 1--160 ohms
  • 1--470 ohms
  • 1--510 ohms
  • 1--1K
  • 2--4.7K
  • 1--10K
  • 1--10K Metal Film
  • 2--22K
  • 23--47K *
  • 2--68K
  • 4--100K
  • 1--220K
  • 1--221K Metal Film

Small Capacitors (Low-Voltage):

  • 1--100pF
  • 1--390pF
  • 1--0.0022uF
  • 2--0.0047uF
  • 2--0.01
  • 2--0.047uF
  • 7--0.1uF

Electrolytic Capacitors (6V, unless otherwise stated):

  • 2--1uF
  • 1--10uF
  • 1--22uF
  • 2--100uF
  • 1--220uF 16V
  • 1--470uF


  • 1--LM317 Voltage Regulator
  • 1--LM311 Comparator
  • 1--LM324 Quad Op-Amp 1--LM386 Audio Amplifier
  • 1--CD4013 Dual D One-Shot
  • 1--CD40106 Hex Schmitt Trigger
  • 1--CD4034 Shift Register
  • 1--CD4040 12-bit counter
  • 1--CD4520 Dual 4-bit Binary Counter
  • 1--CD4073 Triple 3-Input AND Gate
  • 2--CD4053 Triple Analog Switch
  • 1--74HC257 Multiplexer
  • 1--NEC uPD43256-15L Static RAM
  • 1--1N752 5.6V zener
  • 4 (or more)--1N914 General-Purpose Small-Signal Diodes
  • 1--1N4001 General-Purpose Low-PIV Rectifier
  • 1--1N5820 Schotky Diode (or, a 1N34 germanium will do)
  • 1--2N2222 General-Purpose Silicon Transistor


  • 1--5K 10 turn trim pot
  • 1--10K panel mount
  • 1--50K 10 turn trim pot


  • 1--SPST normally open push button for S3
  • 1--SPST normally open for S4
  • 1--SPST normally open for S5
  • 4--SPST dip switches for S6a-d (JAMECO 206-4) 4--SPDT dip switches for S7a-d (Grayhill 76STC04)


  • 1--4.5V battery, made from 3 AA cells
  • 1--7.5V to 12V power supply or battery
  • 1--1/8th inch closed-circuit mini phone jack for microphone
  • 1--1/8th mini phone jack for earphone (optional)
  • 1--up-scale loudspeaker of your choice

*The many 47K pull-up and pull-down resistors can most easily be included in wire wrap work by getting them in DIP or SIP packages. These can be plugged into wire-wrap sockets, saving a lot of misery soldering wrap pins onto normal resistors. These units come in 2 varieties:

First, you can get several individual resistors (with no common connection) like the Mouser No. 268-47K; this contains four isolated 47K units in a single 8-pin SIP. Second, there are SIP or DIP packages containing seven or fifteen resistors with one end of each common to a single pin. The Mouser No. 265-47K has 7 resistors common to pin 1 in a single in-line pack. The Bourns No. 16-2-47k is in a DIP configuration which looks just like a chip; this has fifteen identical resistors all common to pin 16.

For double-throw DIP switches for S7, contact Grayhill for a local distributor: Grayhill, P.O. Box 10376, La Grange, IL 60525, (312) 354-1040.


Susan Fowle gets the credit for designing this interface circuit. It's cleverly done. Our first test of it was to build a Spanish-language talking blood-pressure meter, and it worked right off. What you get at the end of this paper is a 200mV any-language meter. For now, we leave it to you to build pre-scaling resistor networks and to find uses for the Nattering Voltmeter. Thanks, Susan.

[Editor's Note: A detailed description of the National ADC3511 voltmeter chip appeared in SKTF, Winter 1988. However, most of its features will be recapped here. We have found a couple of errors in the original article, and they are corrected as follows:

It was stated that an overflow condition not only brought pin 5 (the overflow pin) high, but that all data lines would also go high. Not so. When the overflow condition exists, the data lines read 1110 ("E" in the hexadecimal system). This is good; it saves us the trouble of having to detect the overflow pin. Also, it appears that when the output (data) lines are latched, the different digits cannot be addressed; thus, pin 19 (the Data-Latch Enable) needs to be operated to scan the digits. Finally, the original article omitted an important fact about the "Sign Out" pin. If you use the V-IN pins, pin 8 goes high when the voltage at pin 11 is positive with respect to pin 10; it goes high to indicate "plus." More important is that if the single-ended hookup is used--with pin 9 used as the input--the sign output (pin 8) is sometimes up and sometimes down; it must be summarily ignored.]

Review of the Voltmeter Chip

Basically, the input of this chip is a comparator. The "plus" input of the comparator is looking at a filtered version of the input voltage--seen at pin 9, the "V Filter" pin. This voltage is being compared with an internally generated voltage on pin 12--the "V Feedback" pin. A single-ended signal (positive only) can be applied through 100K to pin 9 (which is bypassed to ground by 0.47uF). Or, there is a DPDT switch inside the chip that permits a voltage of either polarity to be applied between pins 10 and 11 ("V In plus" and "V In minus"); each of these pins is fed through a 51K resistor.

The input resistance of the voltmeter is not 100K as might be thought from looking at the input resistors. The comparator's input current is typically only 1 nanoamp; this would imply an input impedance of 200 megohms. However, no matter which input scheme you use, you are running your signal into a stiff low-pass filter consisting of 100K (or 102K) and 0.47uF. If you have pure DC, the input impedance is essentially the leakage of the capacitor; if there is an AC component on your unknown, that component of your input signal is essentially looking at 100K.

The ADC3511 was used. We could have used the ADC3711 (3-1/2 digit model), but since this latter counts to 3999 for an input of 200 millivolts, it requires input scaling, and we wanted to get on with testing the interface. (The 3511 counts to 1999 for 200mV, which only requires definition of the decimal point.)

Much is said in the literature about the need for an isolated power supply if the differential inputs are used. We got around this by powering the Nattering RAM and this voltmeter from a single 9V battery, and then bringing the "V-IN plus" and "V-IN minus" wires out to insulated binding posts.

The digital pins of the voltmeter chip are reviewed as follows:

  • Pin 5--Overflow (whose information we get another way)
  • Pin 6--Conversion Complete (going high when the A/D converter has finished taking a reading)
  • Pin 7--Start Conversion (an input we trigger every time we want a new reading)
  • Pin 8--Sign Output
  • Pin 19--Data-Latch Enable
  • Pins 20 and 21--Address inputs (which we operate with a counter to fish for the four digits)
  • Pins 4, 3, 24, and 23--BCD Outputs ("8", "4", "2" and "1", respectively)

On the face of it, the Nattering RAM wants to see those BCD outputs; it wants a new 4-bit number every time it speaks a digit of the voltmeter reading. The digit it speaks depends on the operation of pins 20 and 21 of the 3511. When these are both high, the first "partial" digit (the "1" of 1999, if you will) appears on the BCD lines; when pin 20 is brought low, the second digit appears on the BCD lines, and so on. Thus, to "scan a reading," we would present pins 21 and 20 with a binary counter that goes: "11", "10", "01", "00". You will notice that this constitutes counting (in binary) backwards--"down counting."

Theory Behind the Interface

The 74C193 (or 74HC193 and 40193 as well) is a so-called "presettable up/down counter"; it can be triggered so as to count either way. Thus, it provides the way to step through the digits by addressing the voltmeter chip's pins 21 and 20 with a descending binary count.

The 74C193 has another feature we need: Before the Nattering RAM says any digits, we want to know the polarity of the input signal; we want information from the 3511's "Sign Output." Not only do we have to divert the speech board's attention to that sign bit, but we need to include a cycle ahead of the number reading during which the Nattering RAM can speak the "plus" or the "minus." Fortunately, the 74C193 can be loaded with a preset number. If we start its down-counting at 0100, it will waste a cycle before it gets to a meaningful address for the voltmeter's pins 21 and 20.

If you let the down-counter count below 0000, it will just wrap around again to a full binary count of 1111. This is good. The way the system knows that a voltmeter reading is over is by looking at the most-significant bit of the 74C193; when this bit goes high, a flip-flop (the second half of the first CD4013) is "set" so as to stop the cycling of the interface circuit. This will be called the "stop flip-flop."

The four Nattering RAM address inputs must be sent somewhere else to read plus/minus information, since that does not appear on the data lines of the voltmeter chip. The device accomplishing this is a multiplexer (a 74HC157 or 74HC257, it doesn't matter which). One set of the multiplexer's inputs--the "B" inputs--goes to the data lines of the voltmeter. Three of the four "A" multiplexer inputs are tied high; the "A" input (pin 5) belonging to the "2" bit goes to the "Sign Output" of the 3511. Bringing the multiplexer's "Select Input" (pin 1) low directs the speech board to this sign information.

As far as the Nattering RAM is concerned, whenever it sees an address of 1111, it says "plus"; when it sees an address of 1101, it will say "minus."

As mentioned, the first cycle of the system is devoted to reading and speaking the sign. A D flip-flop (one half of the first CD4013) has its Q output going to the multiplexer's "Select Line." One of the first things done is to clear this flip-flop, thus causing the "A" inputs of the multiplexer to be selected. The D terminal of this flip-flop is tied high; thus, any subsequent clock signal will flip Q high and select the multiplexer's "B" inputs for all clock signals to follow. This flip-flop is clocked by the "NOT Word Active" output of the Nattering RAM.

Four one-shots are used in the system; mostly, they just make sure that timing of the various events is all right. These are contained in two dual one-shot packages--74C221 or 74HC221 non-retriggerable one-shots. These have three features of interest here: First, each one-shot has two inputs--a "B" input which must go high for triggering, and an "A" input which must be brought low for triggering. Either input can be used to inhibit further triggering; specifically, the "A" pin must rest low in order for "B" to work, and the "B" pin must rest high in order for "A" to work. Second, each one-shot has a "Clear" pin; when brought low, the one-shot is disabled and reset to the pretriggered state. The third feature is that both "Q" and "NOT Q" outputs are available.

The first half of the first 74C221 is the "read" one-shot. It is triggered by either of two signals: Its "A" input is connected to the "read" button (the button you press when you want a readout of the voltmeter). This button pulls the "A" input high, then drops it upon release of the button; it is the negative-going signal which properly operates this input. On the other hand, the "B" input is operated from the Q output of the fourth one-shot (the second half of the second 74C221). This latter one-shot, triggered by the aforementioned stop flip-flop, provides an "auto-repeat" function so that the voltmeter will, if desired, read over and over again.

Both "Q "and "NOT Q" outputs of the first one-shot are used. The NOT Q goes to a "Load Pin" on the down counter which sets that counter up for 0100--one count higher than we need to address the voltmeter, but giving us a wasted cycle for announcing the sign. The Q output of the first one-shot clears three flip-flops and operates the "Start Conversion" line of the 3511 voltmeter.

The second one-shot (the second half of the first 74C221) operates the "Start Input" of the Nattering RAM with its Q output. The first trigger this one-shot sees is the "Conversion Complete" signal from the 3511; this signal brings the "B" input high when the voltmeter has converted its analog input voltage to a digital reading. Thereafter, triggering of the A" input is done by negative-going transitions of the speech board's "Word Active" line.

The third one-shot (the first half of the second 74C221) uses its "NOT Q" to clock the "Count-Down Input" of the 74C193 down counter. (The "Count-Up" pin of this counter is tied high.) Triggering the "B" input of this third one-shot comes from positive-going transitions of the "Word Active" output of the Nattering RAM. (The "A" trigger is grounded.)

[The fourth one-shot, the second half of the second 74C221, has been discussed as being responsible for the auto-read function.]

The second half of the first 4013, the "stop flip-flop," is cleared at the outset. It has its "set" pin going to the highest-order bit of the down counter. Its "NOT Q" output clears the second one-shot so that no more triggering of the speech board can occur. Since the subsequent one-shot is triggered by positive-going transitions of the "Word Active" line, it too has seen the last of its triggering signals; since this third one-shot clocks the down counter, operation of the counter ceases as well.

[That much of the circuit is all you need to get it working. Though we haven't yet addressed the issue of the decimal point, nor included fancy features like a "shut-up" button and suppression of unwanted figures, at this point you see just what we had for our talking blood-pressure meter. However, the embellishments are nice, so we'll press on.]

Providing Decimal-Fraction Information

Sometimes, a digital readout system has a decimal point that you can scan for--interrupt the reading--and present in turn. This voltmeter chip provides no such decimal indicator; it is assumed that instrument designers would account for it in their pre-scaling and input circuitry. How you treat the decimal point will depend on what you use this voltmeter for. The position of the point may never change, in which case you would just hard-wire it in one place. Or, if you were to make a voltmeter that has a range-selector switch, you could add another deck on the switch to give this circuit "point" information at the right time. The circuit to follow will allow you to do either.

We decided not to interrupt the reading and insert the word "point"; that's harder to accomplish, and it slows down the readings. (Of course, there must be ways by which that could be done, and we'll set about doing that if necessary.) A simpler thing to do is to raise the pitch of the Nattering RAM's voice for all figures that make up the decimal fraction.

To accomplish shifting the pitch, we installed another flip-flop (using the first half of a second 4013 dual-D package). Both D and clock inputs are grounded to make an RS flip-flop. Its NOT Q output goes to the "Pitch Shift" input of the Nattering RAM (the pitch goes up 18% when this input is brought low). This flip-flop gets cleared, along with the previous ones, from the Q output of the first one-shot.

A CD4017 sequential decimal counter was installed; this chip has 11 outputs (including the "Carry") which go high--one after another--as the count of spoken words progresses. You can look at it as if each successive pin, as it goes high, "flags" one of the spoken words. On the other hand, a "leading-zero suppression" circuit to be described needs the "1" output to go high after the sign (so that the circuit won't suppress the sign information). Therefore, you will please excuse the strange clocking arrangement that follows:

Like a lot of counters, the CD4017's clocking arrangement has two inputs (operating an OR gate) one of which is fed through an inverter. Thus, you can either clock it with positive-going transitions of the proper clock pin --the one with direct access to the OR gate--or with negative transitions on the so-called "Clock Enable"--the input containing the inverter.

The actual "Clock" pin goes to the clock of the down counter (although this is a negative-going pulse, and we are thus clocking the 4017 on the termination of the pulses). However, the "1" output is delayed so that it occurs after the sign; this is accomplished by tying the "Clock Enable" pin to the "NOT Q" of the sign flip-flop. The first "clock" signal the 4017 sees is actually the negative transition of this flip-flop's output, not the down counter's clock signal.

The "0" output of the 4017, which comes up with occurrence of the "clear" signal, is thrown away. The "1" output comes up between the sign and the first digit. Thereafter, the "2" output coincides with the second word (which is the first digit). The last digit--which is the fifth word--is flagged by the "5" output of the 4017.

Having eight outputs (each going high in succession), we now have a sequence of pins, any one of which can be used to "set" the pitch-shifting flip-flop. If the outputs--"2" through "5"--are taken to positions on a deck of a range switch, and the "Set" terminal of the flip-flop is hooked to the arm, the position of the switch will determine at which "point" in the reading the pitch goes up.

In our blood-pressure meter, this system is used to read from 0 to 199.9 millimeters of mercury. By tying the "Set" pin of this flip-flop to the "5" output of the 4017, the gauge reads numbers like, "Plus, one, three, seven, four!" and "plus, zero, seven, six, five!"

Suppression of Unwanted Digits

The Nattering RAM can be made to abort on command from any of the 4017 outputs. For our blood-pressure meter, for example, we could get rid of the "last place" entirely by connecting the "5" output of the 4017 to the "Abort" line.

A more formal approach would be to connect any number of the outputs of the CD4017 to inputs on an "OR" gate (such as the CD4072 dual 4-bit OR gate). The output of this gate would go to the "Abort" line. Then, switches could be installed so that any digit you wish to suppress could be selected.

Suppression of least-significant digits might be handy when trying to make rough adjustments to a circuit; you could add precision as you got closer to the desired reading. On the other hand, in cases where you are making only fine adjustments--where only the last one or two digits may change--you would save time by suppressing leading digits that never change.

Suppression of leading zeroes

A scheme for suppressing leading zeros works as follows: A 4-input OR gate (the other half of the above 4072) has its inputs going to the Nattering RAM's address lines (the outputs of the multiplexer in this circuit). The output of this OR gate is low for 0000, but goes high for any other number. The trick is, though, to abort only leading "0"s, not every one that comes along. Thus, the output of this OR gate is used to "reset" a flip-flop whose output aborts the speech board. When toggled by the OR gate, this flip-flop stays in the "non-abort" position after the first non-zero signal comes along.

The "1" output of the 4017 always "sets" this flip-flop to the "abort" position immediately after the sign is spoken. It will let the speech board run without speaking until the first non-zero BCD reading, detected by the OR gate, causes it to be cleared.


The most blessed button on any talking device is the "Shut-up!" button. I suppose it's not so important on a 5-word voltmeter, but any more digits and you'd demand this control--an anti-natter button. This is easy to accomplish with the Nattering RAM's "Abort" line. An RC circuit with a 1-second decaying time constant is quickly charged by the shut-up button; the voltage on this RC circuit holds the "Abort" line high via a diode. (The "Abort Input" already has this diode, but if you're going to use this input for unwanted digit suppression, a separate diode must be used.)

* * *

Why all these embellishments? Well, you haven't lived until you've tried using a talking meter for a while. Speech is slow and electrons are fast, and you often get mad at the disconnection between what you know and what is instantaneously going on with the circuit. In its purest form (without short-cutting features) your impression is that such a meter talks too much. Features like not having to say the word "point," or suppressing some of its digits, are feeble attempts to tighten the coupling between you and the tests you are making.

What talking meters are really good for is making static measurements. If you have a nice quiet unknown and you want to know the value of it, there's nothing better than a meter with digital output. If you are trying to make adjustments to a circuit, or trying to calibrate something whose adjustment is fussy, you frequently engage in heated debates with a device who is unequipped to reason with you.

On the other hand, with all of these possibilities invoked, one would have a rather formidable mess of switches and configurations. It is improbable that all these features would be added to a single unit, but they're all there to be included as you like.

Calibration and Other Circuit Adjustments

If you follow the manufacturer's literature, there are three adjustments to be made around the ADC3511 voltmeter:

First, the LM336 reference diode must be "trimmed"; this is done by adjusting the 10K pot on its "Adjust" pin so as to get 2.49 volts on its cathode. When that is accomplished, the 20-ohm pot on pin 16 of the 3511 is adjusted for 2 volts; this is the "calibration" pot of the voltmeter.

I haven't tried my simplified reference circuit, but I hope it will be possible to trim the regulator so that the desired 2 volts appears at pin 16. What might get in the way of this is that the trimming adjustment is very sensitive. No doubt the 20-ohm pot gives the technician the power of "fine tuning" for 2.000V. A clear advantage of my system, if it works, would be that you can just calibrate the instrument to a known voltage; with a 10-to-1 voltage divider on a flashlight cell, you could adjust the thing to say "1550" and be done with it.

The 50K "offset" pot (which feeds pin 12 through a 22meg resistor) seems to require that you apply a test voltage to the differential inputs--first one way, then the other. By adjusting this pot, it is hoped that you can get the readings to be the same for either polarity.

Finally, the 1meg pot in the "Auto-Read" circuit is adjusted for an appropriate space between readings. If the one-shot's time is too short, this feature may not work at all. The setting will depend on the sampling rate of the Nattering RAM. If your mother tongue contains long words, and the speech board must be slowed accordingly, the Auto-Read one-shot time will have to be long enough to overshoot the last word and trigger the system after the full reading has been spoken.

Table of Word Addresses

The following are the words needed for this voltmeter, together with the addresses to put them in. The only one we had trouble with is finding short indications for "overflow." In Spanish, this is most properly "saturacion"; we had to settle for "mucho." (You can slow the Nattering RAM down to accommodate "saturacion," but that will make your normal readings go by at a snail's pace.)

Note: By experiment, we found that the pitch-shifted numbers will not be particularly obvious if you don't say the entries all at one pitch. If you don't require decimal information, say the numbers as you like. If decimal information is required, pick a comfortable note and sing these entries.

Important! If you are just using the single-ended input with positive voltages, the floating differential input will cause the Sign Out pin to aimlessly jump between plus and minus. Thus, if the voltmeter is used in that hookup, record the sign addresses as blank.

  • 0000--Zero or Oh
  • 0001--One
  • 0010--Two
  • 0011--Three
  • 0100--Four
  • 0101--Five
  • 0110--Six
  • 0111--Seven
  • 1000--Eight
  • 1001--Nine
  • 1010--*
  • 1011--*
  • 1100--*
  • 1101--Minus
  • 1110--Over (overflow)
  • 1111--Plus (or blank)

* Expletive inserted.

The addresses tagged with asterisks are not used by this circuit and may be used, with the Nattering RAM board in the "setup" mode, for demonstration purposes. In our prototype model, we left the setup address switches set to one of these locations so that we could record snatches of the passing world without fear of destroying our carefully timed words of wisdom.

Circuit Description for the Nattering Voltmeter

Power Supply and Reference Hoopla

These voltmeter chips have both a digital and analog ground (pins 22 and 13, respectively); these are both grounded. The same ground is shared by this circuit and the Nattering RAM. Pin 1, the VCC pin, goes to a 5V line which was just borrowed from the Nattering RAM. Pin 2 is a decoupling point for the analog circuitry (a mis-named "analog VCC" pin), and this is bypassed by 10uF (negative of the capacitor at ground).

A precision 2-volt reference is needed. The literature suggests use of the LM336 2.49V "reference diode" as follows:

The anode of the LM336 is grounded. Its cathode goes through 680 ohms to the 5V line. Also going to 5V is the anode of a 1N914 diode; the cathode of this goes to the top of a 10K multi-turn trim pot. The bottom of this pot goes to the anode of another 1N914; the cathode of this diode is grounded. The arm of this pot goes to the "Adjust" pin of the LM336. (This pot is adjusted to get 2.49V at the cathode of the 336.)

The output of the LM336, its cathode, goes through a 232-ohm 1% resistor to the top of a 20-ohm pot; the bottom of this pot goes through a 1K 1% resistor to ground. The arm of this pot goes to the V-REF input, pin 16 of the 3511. (This pot is adjusted to get exactly 2 volts on this pin 16.)

Finally, an offset pot is added. The cathode of the LM336 goes through 100K to the top of a 50K pot; the bottom of this pot is grounded. The arm goes through 22 megohms to the V-Feedback pin, pin 12 of the 3511. (This pot is adjusted so that, when a known voltage is presented to the differential inputs first one way and then the other, the magnitude comes out the same.)

Clocking the 3511

The "f Out" pin (pin 18 of the 3511) goes through 7.5K to "f In" (pin 17). Pin 17 also goes through 250pF to ground. (The frequency of this clock is 320kHz; it is not critical, so if you want to use 220pF and 10K, go ahead.)

Input Setup for 200mV

Note that a slightly simpler input setup can be had to give you a full-scale sensitivity of 2 volts (see the original article on the 3511). A sensitivity of 200mV is more useful, so here 'tis:

V-Feedback goes through the parallel combination of a 0.47uF good-quality (Mylar) capacitor and a 100K 1% resistor to ground. Between V-Feedback and "Switch 1" (between pins 12 and 15) is a 900K resistor. (This resistor was created by shunting a 1 megohm 1% unit with 9.1 megohms. The 9.1meg resistor can have a tolerance of 5%.)

Between pins 14 and 15 ("Switch 2" and "Switch 1") is a 200-ohm resistor (they don't say with what tolerance).

The V-Filter pin goes through a good-quality (Mylar) 0.47uF capacitor to ground. (This should be of identical brand to that on pin 12; it's not that the values should be the same, but that their leakages must match as closely as possible.) This pin 9 also goes through a 100K resistor to the hot terminal of a single-ended input (optional); the cold side of this input is grounded.

The V-In plus and V-In minus pins (pins 11 and 10, respectively) each go through 51K to their respective binding posts. (These pins can just be left alone if the above single-ended input is used. We provided them both, just for ducks.)

* * *

[The editor has it in mind to simplify all that adjustable drivel for the 2-volt reference. Not only are two adjustments just too many, but the circuit shown asks for some odd-ball parts--you need a 232-ohm resistor and a 20-ohm pot. (That pot is fairly hard to get.) On the other hand, the circuit they show will be the best. If you trim the LM336 output to 2.49V, it will stay within a millivolt over a temperature range of from 5 to 45 degrees C. If you don't trim your LM336, and if its natural breakdown voltage is as deviant as 2.41 volts, you will get a plus/minus 0.2% tolerance over the temperature range of 5 to 45 degrees C; this will mean that the last place of 199.9" may be off by 4 or 5.

It will be my experiment to try the following:

Bill Gerrey's Reference Circuit

As before, the anode of the LM336 is grounded; its cathode goes through 680 ohms to the 5V line. This 5V line goes to the anode of a 1N914, with the cathode of this diode going to the top of a 10K multi-turn pot. The bottom of this pot goes to the anode of another 1N914; the cathode of this diode is grounded. The wiper of the 10K pot goes to the "Adjust" pin of the LM336. The LM336 has a voltage divider across it: The cathode of the 336 goes through a 2.21K 1% resistor, thence through the parallel combination of 10K 1% and 82K 5% to ground (this parallel combination forming an 8.913K resistor). The take-off point, the junction of the 2.21K and the other resistors, will give you 2 volts. (Instead of ordering a proper 2.21K 1% value, I'd use a 2.2K 5% metal film unit; metal film is pretty stable, and it's "close enough for jazz.")

An op-amp is wired as a follower; its output is tied to its inverting input. The non-inverting input goes to the take-off point on the above voltage divider. The output of the op-amp goes directly to the V-REF pin, pin 16 of the 3511 voltmeter chip. Now, since this reference voltage is low impedance, there need be no 200 ohms between the "Switch 2" and "Switch 1" pins; pins 14 and 15 of the 3511 can now be tied together (I think). The 10K pot is adjusted to get 2 volts at the op-amp's output.

The above circuit doesn't eliminate the need for the 50K offset pot whose arm goes through 22 megs to V-Feedback, pin 12.]

Digital Interface Circuit

[Dear Reader: You don't have to read this. It exists for two reasons: One is so that if you have to trouble-shoot the circuit, you have somewhere else to go besides the wire-wrap table. The second reason is so that a print schematic may spring from the design.]

Pins 23, 24, 3 and 4 of the 3511 voltmeter go to 74C257 inputs 3, 6, 10 and 13, respectively (these are the "B" inputs of the multiplexer, and pin 3 is the least-significant bit's input). Three of the "A" inputs, pins 2, 11 and 14, are tied high. One "A" input, pin 5 of the multiplexer, goes to the 3511's "Sign Out," pin 8.

The outputs of this multiplexer address the speech board. The multiplexer's pins 4, 7, 9 and 12 go to the Nattering RAM's 74HC257 inputs 3, 6, 10 and 13, respectively. The tristate feature of the interfacing multiplexer is not used; pin 15 is grounded.

Pin 1 of the 74HC257, its A/B "Select" line, goes to pin 1, the Q output of the first half of the first CD4013. The D input (pin 5) of this 4013 flip-flop is tied to VCC. The "Clock" input, pin 3, goes to "NOT Word Active." Its "Reset" input (pin 4) goes to the first Q output of the first 74C221, the one-shot's pin 13. (This pin 13 of the 74C221 "Clears" a lot of stuff.) Pin 6 of the 4013 ("Set") is grounded.

The other flip-flop in this 4013 has its D and clock pins, pins 9 and 11, grounded. Pin 10, the reset, goes to pin 13 of the first 74C221 (the one-shot that clears everything). Pin 8, the "Set" pin of the 4013, goes to the highest-order bit on the down counter, pin 7 of the 74C193. The Q output of this flip-flop goes to the "B" input of the Auto-Read one-shot, pin 10 of the second 74C221. The NOT Q output "clears" the one-shot that starts the Nattering RAM; thus, the 4013's pin 12 goes to the first 74C221's pin 11. (This half of the 4013 is the "stop flip-flop.")

The first one-shot's Q output, pin 13 of the first 74C221, goes to the "Start Conversion" input of the voltmeter, pin 7 of the 3511. The 3511's "Conversion Complete" signal, pin 6, is one of two signals fed to the second one-shot in this package; the voltmeter's pin 6 goes to pin 10 of this 74C221, while the "Word Active" line feeds the one-shot's pin 9. Its Q output, pin 5, goes to the "Start" input of the Nattering RAM.

The timing of the first one-shot (94 milliseconds) is set by a 0.2uF capacitor between pins 14 and 15, and a 470K resistor going from pin 15 to VCC. The 47-millisecond timing of the second in this package, the one-shot that starts the speech board, is set by 0.1uF between pins 6 and 7, with pin 7 going through 470K to VCC. [Don't be fooled--like I was--that the RC network on one side of the chip applies to the one-shot whose inputs are on that side; these pin assignments are in much disorder.]

The first one-shot's NOT Q output, pin 4 of the first 74C221, goes to the "Load" line of the down counter, the 74C193's pin 11. The number 0100 is hard-wired to the down counter's "Data" lines; thus, the counter's pins 1, 9 and 15 are grounded, while pin 10 goes to VCC.

The last two bits of the down counter go to the address pins of the voltmeter; the 3511's pin 20 goes to the counter's pin 3 (least-significant bit), and the 3511's pin 21 goes to the 74C193's pin 2. The voltmeter's "Data Latch Enable," pin 19, goes to the "Word Active" line (where it is alternately latched and unlatched).

The down counter is clocked by a one-shot--the first half of the second 74C221. The NOT Q of this one-shot, pin 4, goes to the count-down pin of the counter, its pin 4. The count-up pin, pin 5 of the 74C193, goes to VCC. On the 74C221, the "A" input (pin 1) is grounded. The "B" input is pin 2, and this goes to the Nattering RAM's "Word Active" line. Both oneshots in the second package have their clear terminals (Pins 3 and 11) tied to VCC (these are more properly "NOT Clear").

The timing (47 milliseconds) of the first half of this latter chip is set by a 0.1uF capacitor between pins 14 and 15, with pin 15 going through 470K to VCC.

The fourth one-shot is for the Auto-Read feature. As already stated, its "B" input, pin 10, goes to the Q output of the stop flip-flop, pin 13 of the first 4013. Its "A" input, pin 9, goes through 47K to VCC, as well as going through an SPST toggle "repeat" switch to ground. Its timing is variable (from 1/3 to 1 second); pin 7 goes through 1uF to pin 6 (negative end of the cap at pin 6), while pin 7 also goes through 330K in series with a 1 megohm rheostat to VCC.

The very first one-shot gets triggered in two ways: Its "B" input, pin 2, goes to the NOT Q output of the Auto-Read one-shot, pin 12 of the second 74C221. Pin 1 of the first 74C221, its "A" input, goes through 47K to ground, but also through the all important "Read" button to VCC. [Note that, if you wanted to reverse the sense of the "Read" line (so as to operate it from a switch with one wire common to ground, for example), you could accomplish this by putting the "A" input, pin 1, to the Q output of the Auto-Read one-shot (pin 5 of the second 74C221); then the Read switch could go from pin 2 to ground, with 47K going from pin 2 to VCC.]

Optional Features

Digit Counter

A CD4017 decade counter is used. Mainly, its pin 14 is clocked by positive-going transitions of the down counter's clock signal (pin 4 of the 74C193). However, it triggers on one negative-going transition; the so-called "Clock Enable," pin 13, goes to the "NOT Q" output of the sign flip-flop, pin 2 of the 4013. Like lots of other stuff, its "Clear" terminal, pin 15, goes to the Q output, pin 13, of the first one-shot.

The first output of the 4017 is not used. The important outputs are: pin 2 ("1" Out), pin 4 ("2" Out), pin 7 ("3" Out), pin 10 ("4" Out), and pin 1 ("5" Out).

Pitch Shifter for the Decimal Point

The first half of a second CD4013 dual D flip-flop has its "Reset" (Pin 4) tied to the "Reset" of the 4017 (Pin 15), with both of these going to pin 13 of the first 74C221 (the Q output that clears everything). To make this an RS flip-flop, pins 3 and 5 are grounded (Clock and D). (If the other half of this 4013 is not used for leading-zero suppression, pins 8, 9, 10 and 11 are grounded.) The "NOT Q" output (pin 2) of this flip-flop goes to the "Pitch Shift" input of the Nattering RAM.

The pitch-shifting flip-flop's "Set" pin, pin 6, goes through 47K to ground; this pin can then be taken to the appropriate output of the 4017. For our blood-pressure meter, the last (fourth) digit is always a decimal fraction; thus, pin 6 of the 4013 is hard-wired to the "5" output of the 4017 (pin 1). On the other hand, a separate deck on a range switch could be used. In that case, pin 6 of the 4013 would go to the arm. The 0.1999 volt position would go to the "2" output, pin 4 of the 4017; the 1.999 volt position would go to the 4017's pin 7; the 19.99 volt position would go to pin 10; the 199.9 volt position would go to pin 1 of the 4017; finally, the 1,999 volt position would be left blank.

Suppression of Unwanted Digits

A CD4072 dual 4-bit OR gate is installed; the first half is used for this feature. Each of its inputs, pins 2, 3, 4 and 5, go through 47K to ground. Four SPST toggle switches allow you to suppress any of the digits. The OR gate's inputs each go to one side of a switch. The free terminal of the least-significant switch goes to pin 1 of the 4017. The free end of the next switch goes to the 4017's pin 10. The next in line goes to 4017 pin 7. The most-significant switch goes to pin 4 of the 4017.

The output of the OR gate, pin 1, goes to the "Abort" line of the Nattering RAM.

Suppression of Leading Zeros

A 4-input OR gate (the second in the CD4072) is used to detect 0000 on the outputs of the multiplexer; pins 9, 10, 11 and 12 of the 4072 go to the 74HC257's pins 4, 7, 9 and 12. The second half of the optional 4013 is employed as an RS flip-flop. Its pins 9 and 11 (D and Clock) are grounded. Pin 10 of the flip-flop ("Reset") goes to the output of the 4072, pin 13. The "Set," pin 8, goes to the "1" output of the 4017, pin 2.

The "NOT Q" output of the flip-flop, pin 12 of the 4013, goes to the "Abort" line of the Nattering RAM. Note that if there are other features sharing use of the abort line (such as the "shut-up" button), each feature's output must have its own diode to pin 4 of the speech board's 4013 (the cathodes of all diodes going to this pin 4).

Shut-up Button

One side of an SPST normally open pushbutton goes to VCC. The other side goes through 100 ohms to the top of a parallel RC circuit comprised of 10uF and 100K; the bottom of this parallel combination is grounded (negative end of the cap at ground). The junction of the RC circuit and the 100 ohms goes to the anode of a 1N914; the cathode goes to pin 4 (Reset) of the 4013 on the Nattering RAM Board. (Note that the speech board circuit of the previous article already shows this diode. However, if other features share the "Abort" line, separate diodes must be used for each.)

Wire-Wrap Table for the Nattering RAM Voltmeter Interface

Note: This does not include the analog-to-digital circuitry around the 3511 voltmeter chip; those external components would logically be soldered together rather than wire-wrapped.

For instructions as to how to read this table, see those introducing the wire-wrap table in the preceding article on the Nattering RAM. However, so that you won't trip over them, notice the RC timing circuits on the 74C221 one-shots. Considering the second 74C221 as an example: The wire-wrap table states, "7 through 1uF to second 74C221-6 (negative at pin 6); through series combination of 330K and 1meg rheostat to VCC." Translating that table entry into words, we see that the second 74C221's pin 7 goes two places (those two places being separated by a semicolon). Pin 7 goes through 1uF to its own pin 6 (the negative end of this cap at pin 6); pin 7 also goes through a 330K resistor, in series with a 1 megohm rheostat, to VCC.

Certain add-ons are optional. These will be so designated in the headings.

Note also that the "terminals" of the Nattering RAM are referred to by name, not by pin numbers of chips in that circuit. To aid in connecting these circuits together, the actual pins correlating with those terminals are listed at the end.

ADC3511 or ADC3711 BCD Voltmeter Chip:

  • 3 to 74HC257-10
  • 4 to 74HC257-13
  • 5 NC
  • 6 to first 74C221-10
  • 7 to first 74C221-13; 4013-4-10
  • 8 to 74HC257-5
  • 20 to 74C193-3
  • 21 to 74C193-2
  • 23 to 74HC257-3
  • 24 to 74HC257-6

First 74C221 Dual One-shot (each having triggers of opposite sense as well as complementary outputs):

  • 1 through 47K to ground; through "Read" button to VCC
  • 2 to second 74C221-12
  • 3 to 4013-3; "NOT Word Active"
  • 4 to 74C193-11
  • 5 to the "Start Input" (on Nattering RAM)
  • 7 through 0.1uF to first 74C221-6; through 470K to VCC
  • 8 to ground
  • 9 to second 74C221-2; 3511-19; "Word Active"
  • 11 to 4013-12
  • 12 NC
  • 15 through 0.2uF to first 74C221-14: through 470K to VCC
  • 16 to VCC

Second 74C221 Dual One-shot:

  • 1-8 to ground
  • 3-11-16 to VCC
  • 4 to 74C193-4
  • 5-13 NC
  • 7 through 1uF to second 74C221-6 (negative end at pin 6); through series combination of 330K and 1meg rheostat to VCC
  • 9 through 47K to VCC; through the "Auto-Read" toggle to ground
  • 10 to 4013-13
  • 15 through 0.1uF to second 74C221-14; through 470K to VCC

74C193 or CD40193 4-Bit Up/Down Counter:

  • 1-8-9-14-15 to ground
  • 5-10-16 to VCC
  • 6-12-13 NC
  • 7 to 4013-8

CD4013 Dual D Flip-Flop (designated "first 4013 in tables of optional features):

  • 1 to 74HC257-1
  • 2 NC (for now)
  • 5-14 to VCC
  • 6-7-9-11 to ground

74HC257 or 74HC157 Multiplexer:

  • 2-11-14-16 to VCC
  • 3 to 3511-23
  • 4 to "'1' Address"
  • 6 to 3511-24
  • 7 to "'2' Address"
  • 9 to "'4' Address"
  • 10 to 3511-3
  • 12 to "'8' Address"
  • 13 to 3511-4
  • 8-15 to ground

Wire-Wrap Table for Optional Features

Note: Unless all these options are included, there will, no doubt, be free inputs to unused portions of the CMOS devices. These unused inputs must be grounded. By looking through the "Pin Assignments" section, you can find these inputs and deal with them properly--ground 'em.

CD4017 Digit Counter (used for the decimal point and for unwanted digit suppression):

  • 1 ("5" Output) flags fourth digit
  • 5-6-9-11-12 NC (unless you need these for projects with more digits)
  • 3 ("0" Output) NC
  • 4 ("2" Output) flags first digit
  • 7 ("3" Output) flags second digit
  • 8 to ground
  • 10 ("4" Output) flags third digit
  • 13 to first 4013-2
  • 14 to second 74C221-4; 74C193-4
  • 15 to first 74C221-13 (and a lot of other "Clears")
  • CD4013 Dual D Flip-Flop:
  • 1-13 NC
  • 3-5-7-9-11 to ground
  • 14 to VCC
  • Pitch-Shift Flip-Flop (first half of added CD4013):
  • 2 to "Pitch Shift" Input
  • 4 to first 74C221-13 (and to several other "Clears")
  • 6 through 47K to ground; to desired 4017 output

Leading-Zero Flip-Flop (second half of added CD4013):

  • 8 to 4017-2
  • 10 to 4072-13
  • 12 to "Abort" line (using separate isolation diode if necessary)
  • CD4072 Dual 4-Input OR Gate:
  • 6-8 NC
  • 7 to ground
  • 14 to VCC

Selectable Digit-Suppression OR Gate (first half of added CD4072):

  • 1 to "Abort" line (using separate isolation diode if necessary)
  • 2 through 47K to ground; through SPST "least-significant" toggle to 4017-5
  • 3 through 47K to ground; through SPST toggle to 4017-10
  • 4 through 47K to ground; through SPST toggle to 4017-7
  • 5 through 47K to ground; through SPST "most-significant" toggle to 4017-4

Zero-Suppression OR Gate (second half of added CD4072):

  • 9 to 74HC257-4
  • 10 to 74HC257-7
  • 11 to 74HC257-9 12 to 74HC257-12 14 to VCC

Review of Nattering RAM Terminals

Note: the chip numbers referred to here are in the Nattering RAM circuit, not the circuit of the above interface. Both circuits have CD4013s and 74HC257s, so stay on your toes.

Nattering RAM Inputs:

  • "1" Address--74HC257-3
  • "2" Address--74HC257-6
  • "4" Address--74HC257-10
  • "8" Address--74HC257-13
  • Start Input--Anode of a diode whose cathode goes to 4013-6 (pull high to start; but speech begins after it goes low again)
  • Abort--Anode of a diode whose cathode goes to 4013-4 (pull high to abort)
  • Word 1 Select--74HC257-15 (pull high for selected word)
  • Word 2 Select--Cathode of a diode whose anode goes to 74HC257-1 (pull low for selected word)
  • Pitch Shift--Second 4053-9 (pull low to raise pitch)
  • Nattering RAM Outputs
  • Word Active--4013-1 (goes high when Start line goes high; thereafter, stays high until word is complete)
  • NOT Word Active--4013-2 (goes low when Start line is brought high; thereafter, stays low until word is complete)

Prototype Voltmeter Interface Layout

A rectangular board was used for our prototype--6 inches high by 4.5 inches wide. The chips were oriented with their columns of pins parallel to the board's long dimension. Viewed with the component side up, a square section (perhaps 3 by 3 inches) was set aside for the ADC3511 voltmeter and all its crazy regalia. The VCC bus runs down the left edge; it also has a 3-inch-long run to the right just beneath this voltmeter section. The ground bus runs across the top and down the right edge. (Although the positions of these buses are given viewing the component side, remember that they exist only on the underside--the "wiring side").

In a column down the right edge are, from top to bottom: The 74HC257 multiplexer, the 74C193 down counter, and the 4013 dual flip-flop (this column being just over 3 inches long). Then, 3 inches down from the top and positioned either side of the midline, the dual one-shots are placed (the "first" being nearest the 4013). There should be room along the bottom for subsequent Optional embellishments--the "point counter" and its flip-flop, for example.

Pin Assignments

ADC3511 or ADC3711 BCD Voltmeter Chip:

  • 1--VCC
  • 2--"Analog VCC" (which is merely bypassed)
  • 13--Analog Ground
  • 22--Digital Ground
  • 5--Overflow (goes high when overflowed)
  • 6--Conversion Complete (goes high when complete)
  • 7--Start Conversion (high to start)
  • 8--Sign Out (high for plus)
  • 9--V Filter (doubles as single-ended input)
  • 10--V In (minus)
  • 11--V In (plus)
  • 12--V Feedback
  • 14--"Switch 2"
  • 15--"Switch 1"
  • 16--V Ref Input
  • 17--f In
  • 18--f Out
  • 19--Digit-Latch Enable (high to latch)
  • 20--Least-Significant Address Line
  • 21--Most-Significant Address Line
  • 23--"1" Out
  • 24--"2" Out
  • 3--"4" Out
  • 4--"8" Out

74C257 Quad 2-Input Multiplexer:

  • 8--Ground
  • 16--VCC
  • 1--Select Line (low selects "A" inputs)
  • 2--A1 Input
  • 3--B1 Input
  • 4--Out1
  • 5--A2 Input
  • 6--B2 Input
  • 7--Out2
  • 9--Out3
  • 10--B3 Input
  • 11--A3 Input
  • 12--Out4
  • 13--B4 Input
  • 14--A4 Input
  • 15--Tristate Input (high for tristate)
  • Note: The 74C157 uses pin 15 to send all outputs to ground when it is held high; since we ground pin 15 here, you can use a 157 just fine.

74C193 Presettable Binary Up/Down Counter:

Note: The direction of the count depends on which "clock" pin you use--"Count Down" or "Count Up." The one you do not use is tied high. They trigger on positive-going transitions. The "Carry" pin goes low at count 1111, so that it can go high to trigger another up-counter with its positive-going edge when 0000 comes around again. Conversely, the "Borrow" pin goes low at a count of 0000, so that it can go high to trigger another down-counter with its positive-going edge when the 1111 comes around again.

  • 8--Ground
  • 16--VCC
  • 4--Count Down
  • 5--Count Up
  • 11--Load (bringing this low "presets" the count to the binary number impressed on the "Data" lines)
  • 12--Carry
  • 13--Borrow
  • 14--Clear (high to clear)
  • 15--Data "1"
  • 1--Data "2"
  • 10--Data "3"
  • 9--Data "4"
  • 3--Q "1"
  • 2--Q "2"
  • 6--Q "3"
  • 7--Q "4"

CD4017 One-of-Ten Decimal Counter:

  • 8--Ground
  • 16--VCC
  • 13--Clock Enable (grounded for positive-edge triggering on pin 14)
  • 14--Clock (tying this high permits negative-edge triggering of pin 13)
  • 15--Reset (high to reset)
  • 3--Out "0"
  • 2--Out "1"
  • 4--Out "2"
  • 7--Out "3"
  • 10--Out "4"
  • 1--Out "5"
  • 5--Out "6"
  • 6--Out "7"
  • 9--Out "8"
  • 11--Out "9"
  • 12--Carry

[What is it about the designers of counters that they are so "unsteady on their pins?" Maybe they spend too long at the counter?]

74C221 Dual Non-Retriggerable One-Shot:

  • 8--Ground
  • 16--VCC
  • 1--A1 Input (Bring low to trigger; resting high inhibits B1)
  • 2--B1 Input (Bring high to trigger; resting low inhibits A1)
  • 3--Clear1 (low for clear)
  • 4--NOT Q1
  • 13--Q1
  • 14--C1 external
  • 15--R1C1 External
  • 9--A2 Input
  • 10--B2 Input
  • 11--Clear2
  • 12--NOT Q2
  • 5--Q2
  • 6--C2 External
  • 7--R2C2 External
  • The formula for timing is R times C.

CD4013 Dual D Flip-flop:

  • 7--Ground
  • 14--VCC
  • 1--Q1
  • 2--NOT Q1
  • 3--Clock1
  • 4--Reset1
  • 5--D1
  • 6--Set1
  • 13--Q2
  • 12--NOT Q2
  • 11--Clock2
  • 10--Reset2
  • 9--D2
  • 8--Set2

CD4072 Quad 4-Input "OR" Gate:

  • 7--Ground
  • 14--VCC
  • 1--Output1
  • 2--In1
  • 3--In1
  • 4--In1
  • 5--In1
  • 6--Not Used
  • 13--Output2
  • 12--In2
  • 11--In2
  • 10--In2
  • 9--In2
  • 8--Not Used

LM336 Reference Diode:

  • With the flat side toward you and with the leads pointing upward, the terminals are, from left to right: Anode (ground), Cathode (output), and Adjust.

Parts List

Resistors (1/4-watt 5%):

  • 1--100 ohms (optional)
  • 1--200 ohms
  • 1--680 ohms
  • 1--7.5K (10K will probably work)
  • 7--47K
  • 2--51K
  • 3--100K (one is optional)
  • 1--330K
  • 3--470K
  • 1--9.1 megohms (used in making the 900K precision unit)
  • 1--22 megohms

Precision Resistors (low wattage 1%, unless otherwise stated):

  • 1--232 ohms
  • 1--1K
  • 1--100K
  • 1--900K (comprised of a 1meg 1% unit in parallel with 9.1meg 5%)

Trim Pots (multi-turn will work for all):

  • 1--20 ohms multi-turn
  • 1--10K multi-turn (must be of good mechanical quality)
  • 1--50K (single-turn will probably work)
  • 1--1meg single-turn
  • Capacitors (with a "working voltage" of 6V DC or higher):
  • 1--250pF (220pF will probably work)
  • 2--0.1uF
  • 1--0.2uF (0.22uF will probably work)
  • 2--0.47uF good-quality plastic, such as Mylar (These must be of identical brand, not so their values match, but so that their leakages match.)
  • 1--1uF electrolytic
  • 2--10uF electrolytic (one is optional)
  • Semiconductors:
  • 5--1N914 (three are optional)
  • 1--LM336
  • 1--ADC3511CCN (or ADC3711, although proper scaling must be done for the latter, see SKTF, Winter 1988)
  • 1--74C193 or 74HC193
  • 2--74C221 or 74HC221
  • 1--74C257, 74HC257, 74C157 or 74HC157 2--CD4013 (one is optional)
  • 1--CD4017 (optional)
  • 1--CD4072 (optional)


  • 5--SPST toggles (four are optional)
  • 2--SPST normally open pushbuttons (one is optional)

Note: The ADC3511CCN and the ADC3711CCN are available from Hamilton Avnet, 1175 Bordaux Drive, Sunnyvale, CA 94086, Tel: (408) 743-3350. A minimum order of $25 must be overcome; part of this can be taken care of by ordering H11F3 opto isolators which we use frequently in Smith-Kettlewell designs.


Hurrah! It's been a long time coming, but you can now get books on IBM diskettes. This has been troublesome in the past for two reasons: First, it used to take quite a bit of work to strip text files of commands used by publishers. Second, before copying machines made it easy to pirate copies of books (which, though it is illegal, is a snap nowadays), publishers used to be a little nervous about having printable disks out there.

However, through efforts of two people listed below, we've got our hands on diskette manuals. Your editor asks a favor of you. Whether or not you want what is immediately available from these two sources, would you please write them a letter of support and encouragement; you are technical people, and knowing that you're out there will be valuable data for them. (Also, grants are sometimes available for such activity, and letters of support are a required appendix in grant applications.)

Remember to specify whether you want 1.2 megabyte diskettes, or the less dense 360 kilobyte ones. Also, there are two physical disk sizes: 5-1/4-inch and 3-1/2-inch.

For $40, Word Perfect documentation is available from Grassroots Computing (Debee Norling), P.O. Box 460, Berkeley, CA 94701; (415) 644-1855.

A whole bunch of technical stuff is now available (and they're working on the "IC Master," for heaven's sake). An initial fee of $25 is required; after that, just send them disks. They are called: Computerized Books for the Blind (George Kerscher), 33 Corbin Hall, University of Montana, Missoula, MT 59812; (406) 243-5481.