Smith-Kettlewell TECHNICAL FILE
A QUARTERLY TECHNICAL JOURNAL FOR BLIND AND VISUALLY IMPAIRED READERS
Sponsored by REHABILITATION ENGINEERING CENTER SMITH-KETTLEWELL INSTITUTE OF VISUAL SCIENCES
2218 Fillmore Street San Francisco, CA 94115
VOL. 6, NO. 2, SPRING 1985
TABLE OF CONTENTS
Choosing Appropriate Speech Technology
BELS AND DECIBELS or: "WITH DECIBELS, EVEN ADDERS CAN MULTIPLY"
Abstract
As labeled, this is a discussion of dB, one of the most commonly misunderstood concepts in science. Also incorporated into this paper is background (history) which you will find nowhere else (it comes from obscure engineering papers of the 1920's). For the beginner, the appendices should be read first; they describe logarithms in the general sense.
A Bit of History
Around the beginning of this century, a unit of measure called the "mile of standard cable" was used in the telephone industry to express transmission losses. This unit represented the ratio of the power received to the power transmitted over a one-mile pair of number 19 telephone wires. More specifically, this standard cable was defined as having a resistance of 88 ohms and a distributed capacitance of 0.054 microfarads per mile, and a test frequency of 796 Hz was used. This came to be called the "800-cycle mile," and represented a power ratio of approximately 0.75 to 1. Two "800-cycle mile" units equaled a power ratio of 0.75 times 0.75 or about 0.56 to 1.
This is a logarithmic scale such that adding "miles" is equivalent to multiplying power ratios. Losses or gains in a multi-component system could then be calculated, simply by adding (and subtracting) the number of "800-cycle mile" units which defined the loss (or gain) of power in each component.
In 1923, the Bell System replaced the "mile of standard cable" with a new unit -- initially called the "transmission unit" (TU). This new unit was defined by the statement that: two amounts of power differ by one transmission unit when they are in the ratio of 10 raised to the 0.1 power, and any two amounts of power differ by N transmission units when they are in the ratio of 10 raised to the 0.1 N power.
Beginning in 1924, the International Advisory Committee on Long Distance Telephony, in Europe, studied standardization of a unit for telephone transmission work. Some countries wished to continue their use of a unit, based on natural logarithms, for which the basic power ratio is e squared (about 7.4). The Bell system and others advocated the adoption of the "transmission unit" (TU). Two different units were proposed; one was defined as the power ratio of e squared, while the other was based on the power ratio of 10 (i.e., equal to ten transmission units). The former was called the "Neper" (after Napier, the inventor of natural logarithms), while the latter was to be called the "Bel" (after Alexander Graham Bell). You guessed it; a "deciBel" is one-tenth of a Bel.
Decibels
The Bell system subsequently adopted the name decibel
(1/10 of a bel) for the transmission unit. Thus the equation
defining the number of decibels is:
dB = 10 log (P1/P2)
where the log is to the base 10, and P1 and P2 are the two power levels
being compared. If P1 is smaller than P2 (i.e., P1/P2 is less than
unity), the log calculation gives a negative number of dB. This
indicates an attenuation, or loss, in the system; however, it is
often customary to use positive dB for losses when labeled as
such. For example, "there is a 10dB loss in the network," means
that P1 equals 0.1 times P2.
We will now construct a table of dB and power ratios. If P1/P2 = 2, then 10 times the log of 2 equals 3.01 dB. This is usually rounded off to 3dB. Therefore 3dB represents a power ratio of 2 to 1. Minus 3dB (often called "down 3dB") indicates a power ratio of one-half. (Note: This is used to define the bandwidth of an amplifier or filter. That is, the bandwidth extends between the pair of frequencies where the power output is 1/2 of the "flat" region.)
Each additional 3dB equals another multiplication of 2 (or 1/2 for minus 3dB increments). 10dB corresponds to a factor of 10 (or 1/10 for minus 10dB). One dB equals about 1.26 (or 0.8 for minus one dB).
From this information, we can complete the table.
Table I
| Approximate Power Ratios | ||
|---|---|---|
| Decibels | Gain | Loss |
| 0 | 1.0 | 1.0 |
| 1 | 1.26 | 0.8 |
| 2 | 1.6 | 0.63 |
| 3 | 2.0 | 0.5 |
| 4 | 2.5 | 0.4 |
| 5 | 3.2 | 0.32 |
| 6 | 4.0 | 0.25 |
| 7 | 5.0 | 0.2 |
| 8 | 6.3 | 0.16 |
| 9 | 8.0 | 0.126 |
| 10 | 10.0 | 0.1 |
| 10 | 10 | 0.1 |
| 20 | 100 | 0.01 |
| 30 | 1000 | 0.001 |
| 40 | 10,000 | 0.0001 |
| 50 | 100,000 | 0.00001 |
| 60 | 1,000,000 | 0.000001 |
Example:
Express 52dB as a power ratio. The table shows that 50dB represents 100,000, and 2dB is 1.6. Adding dB signifies a multiplication of the corresponding power ratios. Therefore, we have 52dB = 100,000 times 1.6, or 160,000.
To use the table to obtain the number of dB from a given power ratio, proceed as follows. First, divide the power ratio by the proper power of 10 to get a quotient between 1 and 10, and note the number of dB from the lower part of the table. Then find the nearest power ratio on the upper part of the table corresponding to the quotient. Add these two numbers together.
Example:
Express a power ratio of 5,000 in dB. Divide by 1,000 (30dB) to obtain a quotient of 5 (7dB). Adding, we get 37dB for a power ratio of 5,000.
In practice, we usually don't measure power directly, but
measure the voltage or current of a signal. Since power = V2/R
or I2R, the defining equation for dB, when given voltage or
current ratios, becomes:
dB = 20 log (V1/V2) or 20 log (I1/I2)
Strictly speaking, when using current or voltage ratios to calculate dB, V1 and V2 must be at the same impedance. It is incorrect, for instance, to refer to an amplifier which has a voltage gain of 1,000 as having a 60 dB gain unless the input and output impedances are equal. Even though wrong, a number of op-amp manufacturers specify open loop voltage gains in dB, without accounting for differences in input and output impedances.
We can construct a table showing the relationship between dB and voltage (or current) ratios from the previous table by doubling the dB number for each ratio. Use of this table is similar to using the power ratio table.
Table II
| Approximate Voltage (Or Current) Ratios | ||
|---|---|---|
| Decibels | Gain | Loss |
| 0 | 1.0 | 1.0 |
| 1 | 1.1 | 0.9 |
| 2 | 1.26 | 0.8 |
| 3 | 1.414 | 0.707 |
| 4 | 1.6 | 0.63 |
| 6 | 2.0 | 0.5 |
| 8 | 2.5 | 0.4 |
| 10 | 3.2 | 0.32 |
| 12 | 4.0 | 0.25 |
| 14 | 5.0 | 0.2 |
| 16 | 6.3 | 0.16 |
| 18 | 8.0 | 0.126 |
| 20 | 10.0 | 0.1 |
| 20 | 10 | 0.1 |
| 40 | 100 | 0.01 |
| 60 | 1000 | 0.001 |
| 80 | 10,000 | 0.0001 |
| 100 | 100,000 | 0.000001 |
| 120 | 1,000,000 | 0.0000001 |
Example:
Express 74dB as a voltage ratio. From the lower portion of the table we get 60dB equal to a ratio of 1000, and from the upper part, 14dB (74-60) gets us 5.0. Thus 74dB equals a voltage ratio of 5.0 times 1000 or 5000. Remember the requirement for equal impedances.
Up to this point, we have talked about expressing ratios in dB. We can say the gain of an amplifier is 20dB, but we cannot say that its output is 20dB. However, if the reference level is specified, we can use dB as an absolute measure.
bdBm is a common absolute measurement unit. By definition dBm is a power level referenced to 1 milliwatt in a 600-ohm circuit. That is, 0dBm equals one milliwatt, 10dBm equals 10 milliwatts, etc. When they appear on analog volt-ohm meters, dB scales indicate 0dB for 0.773 volts rms. This is equal to one milliwatt in a 600-ohm circuit.
In video work, dBmv is widely used. In this case, 0dBmv is one millivolt across 75 ohms. Then 20dBmv would be ten millivolts etc.
Probably the most familiar absolute dB measure is sound pressure level (SPL). Here, the reference was chosen to approximate the threshold of hearing. In terms of power, the reference, 0dB SPL, equals an acoustic power flow of one pico watt per square meter. Perhaps in a subsequent article, we will delve into the subject of dB SPL in greater detail.
In summary, the decibel is a logarithmic measure of the ratio of two quantities; this allows for expressing large ratios as small numbers, and it also simplifies calculations by the use of addition rather than multiplication.
Appendix A
Basic Discussion of Logarithms
The idea behind logarithms is that any positive number can be expressed as a power of 10. For example, 100 is 10 to the second power, 1000 is 10 to the third power, 750 is 10 to the 1.875 power, and so forth. The power of 10 that is needed to obtain a given number is called the logarithm (or log) of that number to the base 10. Three is the log of 1,000, 1.875 is the log of 750, etc. Numbers between zero and one have a negative log, the log of one is zero, and numbers greater than one have positive logs, as seen in the above examples. Logs of negative numbers do not exist.
One feature of logs that makes them very useful is the fact that mathematical operations between numbers may be accomplished by performing simpler operations between the logs of the same numbers. Multiplication in the "real world" becomes addition in "log land," division becomes subtraction, power operations become multiplications, and roots become divisions. Prior to the widespread use of calculators, log tables were an essential aid to engineers and scientists in performing calculations.
Example:
Calculate 5 times the cube root of 2 using logs, given that the log of 2 is 0.3 and the log of 5 is 0.7.
The cube root operation becomes a division by 3 to give
0.1; 0.3 divided by 3 equals 0.1.
0.3/3 = 0.1
The multiplication of the cube root by 5 becomes adding;
the log of 5 is added to 0.1 to obtain 0.8.
0.7 + 0.1 = 0.8
We now need to find the so-called antilog of 0.8; i.e., we want the number whose log is 0.8. The answer is 6.3. This was found by using the log table "in reverse," as you might say.
The formula for calculating decibels, when voltage or
current ratios are given, is derived from the fact that the power
ratio is proportional to the square of the voltage or current ratio.
This becomes a multiplication, by 2, of the log of the ratio.
Thus, where dB equals 10 times the log of P1 over P2, the
equivalent expression for voltage ratios -- with the necessary
multiplication by 2 -- is: dB equals 20 times the log of V1 over
V2.
db = 10 log (P1/P2) = 10 log (V1/V2)2
= 20 log (V1/V2)
In the above discussion, 10 was used as the base for the log
calculations. Any positive number which is greater than 1 may
be used. Ten is the most widely used and most convenient for
numerical calculations. Logs to the base 10 are referred to as
"common logs." The other frequently used base is the number e,
(approximately 2.7183). Logs to the base e are called "natural"
or "Napierian" logs, and are often distinguished from common
logs by using the notation "ln." The relationship between the
common and natural log of a number (a) is: ln of a equals 2.303
times log of a.
ln a = 2.303 log a
Now that this background has been set forth, the reader should consider the following illustrative fable:
Appendix B
Educational Fable
This is an anecdote about Noah, and the reconstruction after the Great Flood.
After the rain had stopped, and the dove had come back with the olive branch, Noah docked the ark and lowered the gangplank. He got up on deck and bade the animals go out -- two by two -- and multiply.
You can imagine, with all the species present, it took upwards of two days for them all to disembark. When it seemed that everyone had left, Noah went back to give the vessel a "once over." There, in the shelter of one of the cabins below deck, he found two snakes huddled in the corner. Gently, he picked them up and escorted them to land, saying, "As per divine instructions, go out into the world and multiply."
One of the snakes looked at him sadly and said, "We cannot, Sir -- we're adders." [Pained laughter from the audience, thank you.]
Noah was only temporarily perplexed, since he was a student of science. He set to work cutting down trees, and constructing a platform from their trunks. (Not another raft, I hope.)
He summoned the snakes, laid them on the platform, and declared that their problems were over. "There now, on a log table, even adders can multiply."
CHOOSING APPROPRIATE SPEECH TECHNOLOGY
by Tom Fowle
Abstract
A general discussion of electronically reproduced speech is given; its types, uses and a bit of the how's and wherefore's are also discussed. Emphasis is placed on the selection of systems for specific purposes.
Introduction
Now that we have all been assaulted for several years with a multitude of types of electronically reproduced speech, it is certainly time to have a gabbing session about what it all means, especially discussing where different types are -- and are not -- useful. I have been, and will continue to be, careful not to use the term "synthesized speech." The term "synthesized speech" is often abused to describe the entire mess of electronic speech of the kinds where the only moving part is the speaker.
Basically, we are dealing with two totally different processes for electronically reproducing speech. (These do not include tape and other recordings; these involve mechanical or magnetic recording processes.)
True synthesized speech is the process of imitating human speech by the use of electronically created noises. This means that you take a couple of oscillators, feed 'em through waveform shapers, filters -- and more filters; you then mix 'em together a couple of times and run the whole mess with a computer. Then you get some guy at M.I.T. who really understands both human speech and programming; he writes you some fancy software for the hardware described above.
This software, in its most basic form, takes a set of symbols representing the speech phonemes and makes the oscillators beep, the shapers shape, and filters filter in such a way as to produce something that usually sounds like a weary Scandinavian baritone with a Martian head cold. You then stuff all this junk into one chip and call it a "VOTRAX," "CompuTalker," or other such familiar names.
For your next entrepreneurial trick, you add some more software which takes ASCII data -- like that which would be sent from a computer -- to a printer, and converts it into a fair approximation of the phonemes needed to say the words. Now you shove all this stuff into a box with a speaker and an RS232C port, and call it a speech synthesizer. (Examples would be: "Type-and-Talk," "DEC-Talk," the "Echo GP," and the like.)
Now that you have it, what on earth is it good for? Since the average person needs a couple of hours of experience to begin understanding the speech, you can't yet put this thing in talking bank machines. The "gimmick market" is much of what sustains these products in the mainstream. Perhaps in the hands of blind people, these devices could make the world of computers accessible to them. Well, even this is not quite as simple as it seems, and further discussion on this point will be entertained later.
The other kind of electronically reproduced speech is called "digitized speech." In this method, a real human voice is simply recorded in digital memory. This is most simply done by sampling the speech, perhaps at ten thousand times per second, then assigning a relative digital value to the voltage of each sample, depending upon the quality you need and the memory you have to waste.
This digital number is stored in memory, on disk, or on magnetic tape in long strings. The playback process is the reverse of recording; you take a digital number and change it into an analog equivalent voltage. What you get is a reproduction of the human speech, varying in quality from that of the old TSI speechboards to the best "hi-fi" digital recordings you have heard.
If you store the digital representations of individual words in read-only memories (ROM's), and assign a location in the ROM to the beginning of each word, you can then have a computer select these words at random and play them back for you. The computer is "grown" into a chip, whereupon it is called a "controller;" there you have such devices as the National DigiTalker (SKTF, Winter 1985). This is, of course, the method used in all our talking calculators, clocks and so forth.
Buzz Words
You have no doubt heard terms which are touted as great advances in speech technology. This list should include: linear predictive coding (LPC), parametrically controlled filters, and "delta modulation." The question is, "Are these earth-shaking developments?" They are, indeed, advancements of interest to makers of digitized speech systems -- however, they make little or no difference to the user. These are names of techniques by which digitized speech is recorded and/or reproduced. The purpose is to reduce the amount of on- board memory (and the sampling rate) required in their manufacture. Some brief definitions are given here for your curiosity.
Direct digital coding has already been discussed; frequent samples of the amplitude are taken, and corresponding digital numbers are stored in memory. This is the most extravagant system, as far as memory is concerned, but it is capable of high- quality recording.
Delta modulation encodes only the difference in magnitude of adjacent samples; changes in the amplitude are stored. For direct recording and playback, this is more efficient in terms of memory used, compared with straightforward digital sampling.
Linear predictive coding encodes parameters of a mathematical model of speech. Like a synthesizer, it has tone and noise generators which play through an adjustable filter. At perhaps 50 times a second, the sound sources and the filter are updated by instructions stored in memory. This scheme is very efficient, although the speech produced is more mechanical- sounding than direct recording.
Filters of some systems need to simulate the vocal tract. Their "parameters" can be controlled digitally. Mathematicians who work for speech systems companies can have a field day with models that make these things perform.
The DigiTalker (see SKTF, Winter 1985) uses tricks to create redundancy in the speech sounds, thereby simplifying the information to be stored. For example, they alter the phase of harmonics in the speech sounds so as to create zero crossings about which the waves are symmetrical; then they digitally record only one-half of each cycle, using an inverter to generate the other half.
What this means to the builder and the user is practically nothing -- except that DigiTalker chip sets are cheap, and that your talking watch doesn't have much memory in it. These are clever inventors -- bully for them. We are left with the task of controlling the speech products from the outside, and how we do so is just as complicated regardless of the works of the speech products.
Comparison of Speech Types
So now that we have a very simplistic idea of these two methods (synthesized vs. digitized speech), why do we care? It turns out that the two systems have their own sets of advantages and disadvantages; there are times when making the wrong decision about which one to use would go a long way towards making a resulting device useless. Let's take each system from the listener's point of view and show the good and bad points of each. Then we may be in a position to figure out what we need -- and maybe even why.
Again being described as speech made from purely electronic sounds, synthesized speech has the major advantage that, at least with most systems, it will say almost anything. The "text to speech" conversion program does not know anything about words as individual structures. It simply takes the data from the input port and subjects it to several hundred rules of English -- or whatever language -- pronunciation.
The beast will attempt to say anything which doesn't blatantly violate its rules. Most modern synthesizers will default to spelling out strings of letters which have too few vowels, or which contain numbers, but this default varies from case to case. Now, if we lived in a world of really fast chips and really cheap memory (not quite yet), and if English were more predictable than it is, we might have a synthesizer which could say most everything correctly. In the real world, however, one quickly learns to recognize the particular idiosyncrasies of the synthesizer "up with which one must put."
Aside from these mispronunciations, the fact that the speech is electronic in origin means, at least until very recently, that the speech sounds strange, and one takes a couple of hours, at least, to get used to it.
This leads to the conclusion that, where one expects to have speech from some machine understood by the naive listener, synthesized speech -- except for the new and very expensive units -- is not likely to be the best choice. Therefore, in such devices as talking clocks, which are most certainly made for the naive listener, using a speech synthesizer would rapidly assure that the speech gimmick would not lead to very many sales.
The choice of speech types boils down to a choice between high intelligibility with a limited vocabulary, or an unlimited vocabulary with poor intelligibility -- at least, with a heck of an accent. An interesting case that illustrates this dilemma follows:
Suppose you were handicapped in such a way that you were nonvocal. Wouldn't it be nice to have a portable gadget that you could cause to speak for you. You'd want the advantages of a synthesizer, which can be made to say anything. But you also want the thing to be kind of understandable -- to a stranger on the street or in a store. So far, most such devices -- and there are several -- have chosen to use synthesizers, and the public has been left to cope with the weird sounds. This is probably the appropriate choice -- and, for sure, is a whole lot better than nothing.
Digitized speech, on the other hand, will say only what has been "recorded" in the digital memory. There are any number of ways to encode data, and therefore any number of considerations are involved in the calculation of how much memory is needed for a word. Suffice it to say that it takes quite a chunk of memory. As memory gets cheaper, vocabularies in digitized speech will greatly increase; sometime we may have a DigiTalker with a huge vocabulary sufficient for reading the output of your computer. Don't hold your breath.
The great thing about digitized speech is its quality. If you've heard an audio "compact disc" player yet, you know how good digitization gets. In a future article, I hope to describe how to build a little digitizer just to play with. You will see that a fair chunk of memory is needed to record just a few words with moderately good fidelity. (A distinction here is implied between speech quality and the fidelity of reproduction.)
From all this, then, you can see that a talking frequency counter or voltmeter is far better done with "digitized speech" as opposed to "synthesized speech." You can get chip sets or speech boards which probably have most of the words you need for such applications -- you might as well enjoy the sound of good-quality speech.
Now, from the number of talking products which have flooded the "gimmick market" recently, you would think that making things talk is easy. Oh, it is easy -- if you start out from scratch -- if you have the bucks to grow your own chips. On the other hand, a review of the Talking Meter article (SKTF, Fall 1981) will have you saying things unprintable. Here is a very brief summary of what needs to happen when you want to "convert" an existing visual system to talk:
First, you have to get at the data on the visual display. If you are dealing with an LCD (liquid-crystal display), you may find that most of the machine, including the display, is all on one chip; you just can't get at the data at all. If you can gain access to points to which you can solder wires, this may have to be done under a microscope.
The speech device will require appropriate signals; you will most certainly have to convert a multiplexed 7-segment coded mess into these appropriate signals. At this point, it is an additionally good idea to store these digits in some sort of memory (latches). Holding each reading in latches is necessary because of the slowness of speech; the digits on the display may change during the very long time that the speech is blabbing, a condition which would lead to erroneous readings.
Now that you have some kind of ordered memory full of digits, you have to send this information to the speech controller chip. I hope this sounds hard -- it is. It can be done with "hard- wired logic," and an example of this is in the Talking Meter of SKTF, Fall 1981. In a way, it is easier to let a microprocessor do most of that work for you; several such designs have been created (see QST Magazine, March 1985). The problem with this approach is that the program for the "micro" has to be stored in a permanent memory chip (burned in a PROM and wired to the microprocessor). Creating the program in memory is a process which is beyond most small builders. (In the QST article, the authors offer to supply the burned PROM's.)
Another level of complexity is introduced when you consider using "bargain speech chips" whose original purpose was very specific. (Example: talk chips would include the innards of the "Speak-and-Spell" toy and the chips from talking clocks.) All the vocabulary is there. You will have made sure that the information from the converted instrument is there. But now you must figure out what causes the speech chip to say appropriate numbers; the controlling signals will certainly not be BCD (binary-coded decimal) or other sensible code (as it is for the DigiTalker, for instance). The two-fold problem is that the visual display you are reading presents information in its own code, which you must convert to something sensible, and the "bargain talk chip" wants its own insensible code in order to say the right thing. Processing of this complexity must certainly be done using your own microprocessor.
Now that we've found out where you want to use digitized speech, and how hard it is, let's do the same thing with synthesized speech:
We left before with a synthesizer in a box and an RS232 port on the back. This is one way in which they come, and they can be compared directly to a printer (a talking printer, as it were). When you send a bunch of stuff to a speech synthesizer, it doesn't say a thing until it gets a carriage return. This creates all kinds of problems if you want to use the speech box as a computer-access device.
Probably the most common error made by blind people buying computers is to think that by buying a speech synthesizer that plugs into the printer plug of their new computer, they will be able to get meaningful use of the system. This really doesn't work because the computer doesn't expect to have to tell you everything it wants you to know through the printer port. What you need to know is directed to the screen, and not to the printer port.
This means that you need an access system which tells you about the format of the screen in a quick and easily understandable way. Keeping track of the movement of the cursor and of such things as blank lines and paragraphs are essentials in using the machine. Just having a talking printer, even if you can get the computer to send everything meant for the screen to the printer port, is very rarely sufficient. In the best of cases you may hear what you really want to know, but only once. The synthesizers have no built-in facility for review, or for checking spelling or punctuation, etc. They can usually accept commands which can help give you the information you need, but that means that the computer needs to treat the speech system differently than it treats the screen. This means trouble!
Software-Adapted Computers
The most common way in which people have attempted to solve this problem is to reprogram the "operating system" of the computer (that is the set of programs which makes the computer able to run general programs). These software-adapted systems usually make the user remember a complex system of key commands in order to get around the screen. This means that you have to learn how to run the adaptation system, as well as the program you are trying to adapt to. The result is that it is twice as hard for the blind person to learn the computer as for the sighted person. Oh boy, the things they do to us while trying to make things better.
The second problem with modified operating systems is that the commercial programs you are trying to use do not expect to run into a weird environment, and when this happens, they tend to "crash" (spilling toxic waste all over you and your diskettes).
There are two good things about software adaptations. One, there are several of them for many of the common computers, especially the IBM PC; and two, they are usually cheaper than hardware-modified systems (see below). In many cases you can spend a few hundred dollars for the software adaptation and another few hundred on a speech synthesizer, and be at least somewhat on the air. If you aren't rich, and you don't have the backing of an agency, you may have to be satisfied with such a system. Many blind computer users (in fact, at this time probably most) are using such and getting along O.K.
Hardware-Modified Systems
This is probably not the best name for this type of adaptive system; perhaps they should be called "transparent adaptations." However, there are defenders of software adaptations who claim that their systems are transparent; I ain't seen one yet and don't expect to. What the heck do we mean by transparent systems? In this case, we mean a way for a blind person to use a computer such that the computer doesn't know anything is different, just that the user is a bit slow. The easiest way to see this is to think of a person using an Optacon to read the screen of a completely unmodified computer. There are certainly folks doing this with some degree of success, and more power to them.
Another such situation might be the one in which I operate when doing most of my programming here at Smith-Kettlewell. The computer I use for programming is such that it does all of its talking to the operator through the use of a "terminal." This is an external box connected through an RS232c serial interface. Normally the sighted user hooks up a terminal which has a keyboard and a video screen on it. However, I have merely hooked up a VersaBraille which has a braille keyboard and a 20- cell braille line that acts sort of like a screen. This works fine if the software I am using on the computer is not too sophisticated. The VersaBraille thinks that everything coming in is supposed to go below the last thing it saw. In other words, it works like a teletype, except that the output is not paper (it has little pins that pop up to form 20 cells of braille).
In modern software such as word processors, the computer may very well move the "cursor" (the point at which it wants to write on the screen) all over the place. The cursor may go up a few lines, up to the top of the screen and then back down again. The VersaBraille is totally incapable of following this kind of movement. This setup is really a hardware modification but is not "transparent," in that I am limited as to the software I can use. I stick to it simply because I find braille vastly more accurate than speech output for programming -- where every letter counts.
On the other hand, the machine on which I am writing this article is a truly hardware-modified system. It is so nearly transparent as to place very few, if any, limitations on what kind of software I can use. No, that is a filthy lie; I cannot use anything with graphics output -- but then, neither can anybody else as far as I know.
This magazine is now being written on an Osborne I computer which is adapted with a system known as the "Frank AudioData System" (also available for the IBM PC). This system has a whole new computer -- plus memory and speech synthesizer chip -- in the keyboard. This new hardware does all the work of controlling the speech.
The controls are very simple, consisting of two "sliders" mounted one to the left, and one below the keyboard. These sliders are for moving your viewpoint around the equivalent of the visual screen. As the sliders are moved, the speaker produces tones which indicate what types of information are being passed over; i.e., upper or lower case letters, numbers, punctuation marks, blank spaces, or the cursor. There are pushbuttons in the tops of the sliders, and when you've found out about where you want to be on the screen, you push the sliders and it talks starting from the sliders' position. Also, and this is very important, the beast shuts up when the buttons are released -- immediately. This means that you only have to listen to the speech when you want to, not when the system thinks you should.
There is more to it than that, but let's cut a story slightly less long by saying that you don't spend so much time learning how to use the adaptive system, just learning how to use the computer.
There are other hardware-modified transparent systems, but this was not supposed to be a primer on computer adaptations for the blind, but a discussion of speech systems.
Conclusion and Review
There are two basic ways of electronically reproducing speech. Digitized speech is a process of recording human speech in electronic memory, and selectively playing it back. It features very high quality speech, but has a limited vocabulary. Commercially available speech chip sets are fairly easily controlled in applications where these requirements are important.
Synthesized speech is the process of producing purely electronic sounds which mimic speech. In most cases, the vocabulary is unlimited; however, the speech quality is usually not so high as with digitized speech, and some mispronunciation is to be expected. In applications where the user can take the small amount of time necessary to get used to the strangeness of the speech, and where the flexibility of practically unlimited vocabulary is vital, synthesized speech is the method of choice. It is vital to restate here that the usefulness of a speech synthesizer system, especially when being used to access a computer, is not dependent so much on the speech quality as on the controllability of the system. You can get used to funny speech, but a control system which is ineffective and gets in the way will ruin your efficiency.
In the future, the quality of synthesized speech is going to get a great deal better -- witness TSI's Prose 2000, and DEC- Talk. Right now they are very expensive, and for most applications do not justify the extra money. We can hope, as seems likely, that the prices will fall, but we must remember that they are still not being made for our market, and to make them useful we will need specifically designed control systems.
As mentioned above, it is not inconceivable that a digitized speech system with a really huge vocabulary may someday be available, but this would not really give us anything new -- we still have to make it do what we want.
On the other hand, there really is something to worry about when it comes to speech devices. It is very probable that as the gimmick market is filled to overflowing with talking clocks, calculators, etc., the manufacturers will quit producing them, and we may find them only in garage sales. It may soon be the case that, to quote Dr. T. V. Cranmer, "Someday, the only way blind people will have talking clocks may be to build their own into cigar boxes."
FROM PAPER TO PROJECT
Part I
Abstract
This series attempts to bridge the gap between the shop-worn experienced builder and the inexperienced one. Acquisition of basic tools and project layout will be discussed here; the beginner may get his start from this material.
Yet the beginner is not our only target. Discussion of modern building techniques will be integral to this series, and many "hints and kinks" will have more general appeal. It is hoped that a wide range of readers will find this series lively reading. At the same time, this will provide Smith-Kettlewell with experimental "seed material" from which descriptive techniques in our "kit projects" can be shaped.
[This series will not comprise a comprehensive course in electronics. It would be well to read text material along with this. The editor is ill-equipped to recommend the specific text; I haven't read an introductory-level textbook in eighteen years. Then, too, I cannot judge at what level -- or levels -- a group as diverse as you are should be reading. The editor suggests that you look through the column called "Lexicography" in back issues of this magazine (mostly found during the years of 1982 and 1983). Perhaps, if you order print catalogs from the book suppliers listed, you can get a local high-school or college teacher to recommend one of the available texts.]
Also, so much background on soldering has already been published here, that I cannot and would not attempt to restate everything in this series. I will be careful to refer you to the pertinent issue of SKTF, where an issue of soldering comes up. For starters, I recommend that you glance through "Soldering, Parts II and III," (SKTF's, Winter and Spring of 1981), as there are things which directly apply.
Hints and kinks will usually comprise the final section of each of these installments. I recommend that you read this before building something described in an earlier section. Seasoned builders have my permission to jump right to this "Hints and Kinks" section, then glance at the earlier sections in search of projects.
Introductory Remarks from the Editor
We are individuals; we each possess a selection of aptitudes which we may exploit for pleasure or profit. How we exploit our aptitudes depends on other factors -- our interests, our time constraints, as well as our prejudices and how we were brought up. Our skills are combined results of all pertinent factors.
The complexity involved in the acquisition of our skills is so elusive as to defy analysis. For example, in the dark ages of my youth, women were thought not to be mechanically inclined; studies have been done to find out why, and to prove that this is so. Yet, in the same dark ages, all intricate industrial processes were put in the capable hands of women -- they are so "patient that their soldering and work under a microscope is less flawed." I must say, for being mechanically disadvantaged, they sure built some fine aircraft, radios, medical equipment, and talking book machines -- I've appreciated all of these. And they did so at bargain prices, I'll bet. The point is, when it came down to it, the skills of women can be changed at a moment's notice, even though their "aptitudes?" may have been entombed in concrete (or whatever they used back then to entomb things).
It is a fact that a minority of blind people have crafting of electronic instruments as their avocation, and precious few of us do such work as a living. Why? The interest is certainly there -- after all, unless we do make things which address our own needs, these devices will either not be available at all, or their limited market will bear directly on their price. My own theories lead me to consider the following:
The Boy Scouts have a "merit badge" for technical achievement. (In my day, this meant that you built a radio receiver, or something.) The sighted Boy Scout buys a set of parts to solder together to make this merit badge. His interests, the time available as he strives for other achievements, are all a balance for him (or for her, now thank heaven). This "Scout" (he or she) tries for fifteen merit badges, hoping that enough will come through to make the grade for the next class.
Our Scout gets all the right parts (unmistakably from the pictures). He uses the pictures to put them together (unmistakably). Yet, the chances of this project working are small: His soldering iron may be all wrong. His handling of the components may be too rough. Things may not be getting hot enough (thereby leading to "cold solder connections"). His reading of the diagrams may be incorrect (I have seen more than one sighted person try to "pad" a variable capacitor by soldering a fixed capacitor between posts of the stator, not between the stator and the rotor as it should have been.) For our sighted Scout, this leads to one, and only one, syndrome: "that thing I bought from "silky Surplus was so cheap . . . and the cord of the soldering iron was intermittent."
Our Scout curses his tools. He curses the book. He curses the "cheap parts" and the store that sells them. Only vaguely does he wonder if his "aptitude" encompasses electronics. This same person may get into the field. He -- or she -- may find that electronics addresses an interest, or that it serves the needs of other sciences.
Contrast this with the experience of the blind adolescent. Suppose he asks his parents for the money to build a thing, the same radio circuit as the Scout used, for example. A parent asks him:
"How can you see the color-codes of the parts? How can you see the diagrams that show how to put them together? Can you solder? Who's going to run the wires along the fence?"
Being an individualist, he sticks to his guns and gets the needed materials.
Actual parts are delicate, and they are sensitive to heat; it takes practice to handle them properly. Soldering irons are given to corrosion and contamination; it takes experience to know how to repair them. Tools only work when they are used properly. These are aspects of apprenticeship, and every aspect is learned through apprenticeship.
With little experience (even less than his sighted counterpart, as is the usual case), he builds something. The soldering is all wrong, the parts are sloppily placed, some of the parts are damaged by heavy-handedness, etc. Who gets the blame?
Blindness gets the blame every time. A "parent" says, "Those are the last parts you get; you ought to know better." A counselor may say, "Your mechanical aptitude came back as . . ." The blind person finds his "lowest energy path" in accepting his failure. "The parts are so delicate nowadays." He ruined this part, he ruined that part. He curses himself, not the materials and not the book.
The first three or four things anyone builds do not work -- at least on the first try. A main difference is in the interpretation of the results. Our Scout accepts a failure now and then; he can afford it. His blind counterpart often adopts this failure, and does so at great expense.
We have all heard the statement, "We take chances every day; you can't afford to worry about every one." Quite true. Why not take this one -- getting our "mits" on tools and gadgets of the electronic age. If it doesn't pan out, we'll study history, poetry, or first aid. At least we will end up being conversant with how chips look, and how things fit together. We'll at least have an honest trial, subject to the "due process" of apprenticeship.
Genesis
In the beginning, you must know what parts and tools look like. Get thee to a friendly TV repair shop and ask to see:
Needle-nosed pliers (the ones used in electronics are about half the size of those found in hardware stores), diagonal cutters (again appropriate to electronics), soldering irons (cold, of course), and the basic multimeter.
Have him show you which resistors are: 1/4-watt, 1/2-watt, 1-watt, 2-watt, and which are "power resistors" (include a couple of wire-wound power resistors, the rest being "carbon composition"). Note that the size bears absolutely no relation to resistors' ohmic value; this size merely tells the builder how much power a unit will dissipate. (Actually, you can use wire- wound resistors at their rated power; you must use carbon- composition units at only half their rated power.)
Have him show you capacitors: include types of disc ceramic, tubular ceramic, dipped mica, "paper" units (most certainly plastic nowadays), and electrolytic capacitors. Note that the physical size of all capacitor types increases with two factors -- capacitance value and "working voltage."
Get your hands on the following transistors: a small-signal unit in a plastic package, two sizes of small-signal units in metal cans (TO18 and TO5 cans being standard), and power transistors in packages of the TO3 can and of the TO220 style.
Finally, have a look at a couple of chips of different sizes -- the number of pins being determined by the complexity of these animals. Chips in a metal can are no longer as common as they once were; it is the "dual-in-line packaging (DIP) that you'll want to see. Note that all of these chips have a standard lead spacing -- 0.1 inches from the center of one pin to the center of the next.
A Radio Shack store simply will not do for this "taste treat," since most of their stuff is in plastic packages that prevent you from getting your mits on things. The shop you visit, whether in the home of a friend or at Friendley's Fixitry, must be a well-equipped one with lots of parts drawers that allow you to handle these animals. Why not tape record this session, including your own description of what you are handling, so that this experience won't drift into the fog of over-exposure.
Next, look at things which are already built -- junked radios and TV's, a dead computer board (call around at local computer stores and ask to see one); contact the local "radio club" and ask to see what someone has built, or contact schools and ask to see a student's project. You need only follow a couple of these suggestions -- enough is enough. The point is to see parts in place, just as you will be placing them when you build things. Try to identify the parts as you explore these items. Note also that they are rigidly fixed in some way; they should not be left to dance atop long leads (their wires are cut short enough to support them), and heavy items are held close to their circuit board, or are sometimes bolted down. Finally, note that they are usually placed in some purposeful order, much as if they were lying on the floor of an orderly room. Only when necessary are parts laid out at odd angles, having a strewn appearance.
Your familiarity with the parts is crucial to your "visualization" as you read and comprehend circuits. The picture of "two resistors in series" (connected in a line -- end to end) should bring to mind two of those little capsules that have wires out their ends arranged in a line and having a solder connection between them. If the junction of these resistors then goes to the base of a transistor, a tiny three-legged stool with one of its leads going to the solder connection between the resistors should come to mind. If materials of the trade are meaningless abstractions to you, then reading circuits will be no fun -- the words will just be words, and putting them into practice in an orderly fashion will be unlikely.
[There are, perhaps, those in the study of "perception" who could refute the above simplistic statements about "visualization." There may be people who do not need to tie their abstractions to hardware of the physical world. Nevertheless, I am willing to elbow them aside for the moment, close the window on science, and declare that most of us "see" concepts in terms of real pieces and Tinker Toys.]
You are mistaken if you think a sighted person, with his schematic drawings, has it any easier. The drawings bear little relation to how parts of a project should be arranged. To begin with, the parts drawn in schematic diagrams are "symbols" and not pictures of the actual items; the symbols can look entirely different from the components, and their connecting leads often do not emerge in corresponding places. Therefore, when laying out projects, your sighted counterpart, too, must picture little capsules (resistors), stools (transistors), wafers (disc capacitors), and benches with fourteen legs (14-pin chips).
You can, as well, use schematic symbols in your mental picture of a circuit; this is sometimes to great advantage. For example, NOR gates and NAND gates come in multiples on 14- pin chips. Viewing these as pins on a centrally located chip can be distracting, especially in circuits where gates occur all over the place. There are times when it is easier to picture these gates in their individual schematic form -- cigar butts with the output coming off the round end, and inputs going into the other end. Of course, this leads you to your next assignment -- getting a helpful high school student to draw the schematic symbols for you.
In learning the visual schematic symbols, get drawings of enormous size, not of actual size. One of the reasons that schematic drawings are of little use to blind people is that the symbols, like the symbols of print, are complex graphic structures, and their fine points do not come over unless the "scale" of the drawing is distorted to your advantage. Whether you have your pal make them out of pipe cleaners or thin noodles of modeling clay, draw them heavily over a soft napkin (so you can feel the other side of the paper as raised lines), or if you use one of several products for making raised drawings (see catalogs of AFB and APH), have him make 'em big!
Wiring Systems
Modern electronics products are all done on "printed-circuit boards." Instead of connecting parts together with wires, as was done back in the 1950's, components are now arranged with their leads poking through holes in phenolic or glass-epoxy boards -- these boards having thin lines of copper foil running from hole to hole as connecting "wires." This technique is wonderful for industry; the printed-circuit (PC) boards are made and etched photographically, whereupon the parts are inserted by machines, and the whole mess is then dipped in (or splashed by) a bath of solder. Instead of products creeping down assembly lines with people making individual solder connections, whole circuit boards come cascading off the line at a rate of several a second. This is why a small radio cost ten bucks in 1950, while a much better radio can now be gotten for under five dollars.
Some home builders (and some prototype designers) find the orderliness and durability of PC boards attractive, and they commit their projects to a PC-board design. From the schematic, they go through the steps of "layout," they carefully draw the actual layout on paper (where it must fit the component dimensions exactly), copy this drawing by photographic means onto a copper-clad board, etch the board (removing all the copper but that required for the "printed circuit"), and drill holes for the components. Then, they insert the components and solder them in place individually.
Not all builders go through all that rigmarole. Those who do are making several steps of "production engineering," and it is the intent of less ambitious builders to defer these "engineering steps" to such time as the need arises for building several of one product. I am one of the less formal school for two reasons: my designs change as I evaluate the workings of a project, and no one yet has figured out a time-competitive system by which blind board designers can independently make these precise drawings.
I am not alone, though, in avoiding the work involved in PC-board designs, especially for "one-off" projects. It is for this reason that many "prototyping techniques" have been developed and marketed for small-scale building.
Several kinds of "universal PC boards" have been marketed. These have patterns of holes which match various components (so-called "chip-socket spaces," for example) which have connecting lines going to rows of holes for less specific parts arrangements. One solders in the chips or transistors first, and then ties their "traces" (the little foil strips) together with jumper wires and so-called "external components" (resistors, capacitors, and the like).
Universal PC boards have the advantage of PC-board durability, and they impose some degree of orderliness on a project (both of which are sought after by the ambitious PC- board designers). The only disadvantage these pose for the blind person is that he needs extensive mobility lessons around each such universal PC board; he must learn where socket spaces are and how many pins they can accommodate, where long "bus strips" are -- as well as which of these are continuous busses, and which are segmented -- and the location (and shape) of every little stray pad provided for "tie points." This can be done -- I've done it in my early days as a technician at Smith- Kettlewell -- but I got so fond of having more freedom of layout than this system affords that I've dropped the technique.
Other schemes are actually versions of "hand wiring," as it was called back in the old days. Circuit boards are still used, but they are not "printed" on. These boards, either phenolic or of glass-epoxy, are full of holes in a regular pattern. (The hole pattern varies, depending on which kind you buy -- one kind for chips and another kind for transistor stuff.) The terminals of the major parts fit in the holes -- from here on, the variations in technique begin.
One technique of hand-wiring on perforated board is patterned after PC-board techniques. This involves using enamel-covered wire to connect protruding component leads as they emerge on the bottom side. The insulation on this wire, which is enamel, is of the "solder-soluble type"; the insulation breaks down and disappears when you solder this wire to something. I have found this technique to be a nightmare. The insulation chafes and is damaged as you handle the wire. The wire is thin, and it can go astray very easily. Then too, sometimes the insulation doesn't break down in the soldering process, thus leading to bad connections. This technique takes more finesse than I'm able to exercise.
Another scheme turns the perforated board into a terminal block, much like those used in military construction of World War II. Push pins, some of which are called "flea clips," are inserted into the perforated board. The heads of these pins now become terminal lugs to which component leads are soldered; the "tails" of the pins, which protrude through the bottom side of the board, are connected by bits of wire to major items (chip sockets and the like). I tried this for a while too; it looked like hand-wiring to me, and I could go ahead as if this newfangled PC-board stuff had never happened. I soon learned, though, that because there were twice as many solder connections -- using both the heads and tails of the flea clips -- I had twice as many cold solder joints to find in troubleshooting the finished item. Then too, the clutter of all this stuff makes it impossible to "feel the floor of the jungle" and trace the circuit for errors. I abandoned this technique also.
Then, I saw some boards in the lab equipment of my university that struck me senseless -- their clarity of layout and lack of cluttering hardware were such that you could trace the circuits as if they were drawings. The technique used to build these boards was "point-to-point wiring." All you have to do is arrange the components' leads so that when they are put through holes from the top side, the ends of their leads come through somewhere near the points they are to be soldered to on the bottom side.
Take the case of our two resistors in series, which was mentioned earlier. Imagine a strip of board lying on the table. Because one-tenth inch spacing of leads is becoming standard, we'll buy board with a non-staggered hole pattern, with the holes being "centered" every tenth of an inch. If our strip of board is 1 by 3 inches, there will be 300 holes in it. Now, for each of the two resistors, bend their leads at right angles to the resistor body. If the resistors you have are 1/2-watt, the centers of their leads will end up being about 1/2 inch apart; when you fit these leads into holes in the board, they will span five holes (0.5 inches, because of the one-tenth-inch spacing).
Now, with the long dimension of the board parallel to the edge of the table, insert the resistors -- one by one -- into the board so that they are in a direct line; leave a little space between them -- perhaps two or three blank spaces. (Naturally, you will have to pick the board up off the table to accomplish this, letting the ends of the leads dangle out the underside.) My picture of your board now has the resistors lying flat atop the upper surface of the board; they are in a line which is perhaps 1.3 inches in total length (1/2 inch for each resistor, and a 0.3 inch gap between them). Now, we must connect them in series by bending the adjacent leads down against the board so that they cross each other (forming a narrow X-shaped junction on the underside of the board). Bend the outer leads outward to get them out of the way, solder the X-shaped junction in the center, and cut off the excess lead length (where ends reach beyond the junction).
Suppose that we now consider the case of parallel- connected resistors. Once again, imagine a small piece of perforated board -- square or strip, it doesn't matter. We take two more resistors and bend their leads at right angles to their body. This time, we want them lying side by side (so that they are "parallel" to each other, as the saying goes). Somewhere in the middle of the board, place a resistor so that its body lies flat against the upper surface, and so that its body is oriented vertically (spanning a column of five holes). Next to it, and perhaps two spaces to the right, place the other resistor so that it lies next to the first. Bend each pair of adjacent leads to form X- shaped junctions; rock the resistors toward each other and firmly flatten their leads against the underside, so that they tend to stay in contact while you solder them. Turn the board over on its back and solder the junctions with dispatch, Strudelmeyer.
How could there be anything simpler. The parts, and their resultant hookup, make the circuit take a logical shape as you fit them together. Because the clutter is minimal, you can usually trace the circuit tactually. The number of solder connections is minimal -- even less so than on PC boards -- so that time taken to make them (and find bad ones) is not out of line for individually built items. With the parts lying flat against the board, the durability of the project is similar to that gotten with printed-board techniques.
I use the technique of "point-to-point wiring" exclusively; I have done so for many years, covering a gross of projects, few of which have come back for repair. This is the main technique we use in the Smith-Kettlewell Training Program, and everyone who stuck with the program learned to use the method successfully. I am so stuck on this technique that you will be stuck with it throughout the first installments of this series, anyway.
The Point-to-Point Wiring Method
Let us design a game -- a "wiring game" which has the following rules. (This game is all the rage at West Poynt):
- The playing field is a board that is full of holes. The player gets to choose the size of the board; it must be large enough to accommodate the finished project, or the player loses. An easy way out of this risky business is that, for some small down-grading of your score, you can choose one dimension, make the other dimension too long, and cut the board to size afterward. In choosing this one dimension, however, it must be remembered that points will be taken off for component leads not being long enough to reach their destination, as well as for components riding atop each others' backs for lack of space.
- The components, and any jumper wires, all have bare leads which are put through the board from the top side (said to be the "component side"); components and loops of jumper wires all rest on this side of the board. All interconnection and soldering is done on the other side (called the "wiring side" of the board). Any forgotten component which must ride on the bottom side will cost you dearly, unless you can state that, "being a bypass capacitor, this was necessary to keep its leads short."
- Wire is very scarce in the land of West Poynt, and use of wire jumpers costs you points. The player tries to use the leads of the components as his main source of connecting wire, and jumpers should be used only to promote clarity and to avoid having unconnected leads crossing each other on the wiring side. Having spaghetti-covered leads crossing over each other on the wiring side is not illegal, but it costs you dearly on your final score.
- Reaching over components with leads of subsequent components is not illegal, but it costs you, since this makes replacement of the encircled items difficult. This carries a slightly higher penalty than does having wire jumpers reaching around the encircled items.
- Having components resting at odd angles makes your project look like beginners' work; two points shall be taken off for each offense.
- Obeying all rules is not possible, nor would this lead to a good sense of compromise. Perhaps the finished project can be graded "on a curve," as you might say. Perhaps a panel of experts should listen to the arguments of the builder, weighing each one and deciding if there really was no other way but to put the output capacitor on top of the chip socket (such that you cannot now plug the chip in). The main criteria are:
Are the leads short and direct? Might not items short out when the project is dropped? Is there a sense of order that would promote retracing and repair? Is there a flow to the thing, or does it look like a first draft of a composition -- knocked out on the typewriter, off the cuff? Of course, the important thing is, can it be made to work? No matter how it looks, this constitutes passing.
Sample Project
An Attenuator for Tape Machines
Who among us has not wanted to record high-level program material onto a small cassette recorder that only has a microphone input jack. With this simple project, you can do so from a "Tape Out" jack of hi-fi systems, or from the various outputs of talking book machines. (I suppose it's only fair to tell you that you can buy something like this already -- it's called an "attenuator cord," and it is made for just this purpose. However, the "attenuator cords" are often low-impedance, so as to provide a proper load for speaker and earphone outputs. This one, at least, is high-impedance, so you can use it on the aforementioned "Tape Out" jack.)
There are several ways to "say" what this thing is: You can call it a 40dB pad. You can call it a voltage divider. You can say that it is two resistors in series across the source, with the output taken off the "bottom one." I will commit a cardinal sin and describe how I would build it in a pictorial way. Afterward, I will describe the circuit in three ways, so that you can compare styles of presentation. How it is described to you, or what you call it, should not affect how you build it -- there is one sensible way to put it together and follow the rules of our game.
Building the Attenuator
Let us start with a 1- by 3-inch piece of board of the type having ten holes per inch. Next, we need two resistors -- 100K and 1K. (Their power rating could be anything -- 1/4-watt or 1/10-watt -- but 1/2-watt units are beefy enough to withstand the rough handling that we beginners might give them.) Finally, we need a patch cord to cut in half, so that we can put the attenuator circuit in the middle. A typical choice of cord might be one with a "mini phone plug" on one end for the tape recorder, and an "RCA phono plug" on the other for the jack in your hi-fi.
First, let us mount the 100K unit on the board. With a long side of the rectangular board facing your tummy, we'll put this 100K resistor near the far edge, say two holes down from the top, with the body of this resistor horizontal and centered with respect to the long dimension. Bend the resistor leads at right angles to its body so that they come off being parallel and spaced a half-inch apart. Count two holes down from the top, and plug the resistor in; one lead will be in the 17th hole from the left end, and the other will be in the 22nd hole from the left end. Pull the leads through so that the body of the resistor lies flat on the "component side" of the board; then bend the leads straight out towards the ends of the board so that they lie along the "wiring side."
Next, we'll mount the 1K unit. This one will lie vertically on the board, just to the right of the 100K unit; bend its leads just like the other one in preparation for this. We'll leave a little space to the right of the first resistor; perhaps we'll use the 24th column of holes, counting from the left end. (All I mean is to leave two blank columns to the right of the first resistor.) Put one lead of the 1K resistor in the third hole from the top, and its other lead in the second or third hole from the bottom (in this same vertical column). Pull the leads through so that its body lies flat on the component side. On the wiring side, bend its upper lead straight away from you, so that it crosses the horizontal one coming off of the 100K unit. Finally, bend the 1K unit's lower lead so that it lies along the bottom edge of the board and points to the left, crossing the center line.
When you flip the board over, the wiring side looks like a short piece of railroad track, with a "ground rail" (not running full length) along one edge, and an interrupted rail along the other. Get used to this railroad track analogy; you will find almost all of your projects resembling a track of two or three rails.
Next comes surgery on the patch cord. Cut it in two. Carefully remove an inch and a half of insulation from the free ends. With a braille stylus or other implement of destruction, gradually unbraid their shields, twist them into pigtails and "tin" their strands with solder. Strip off half an inch of insulation from the center conductors and tin these too.
Both shields go to the ground rail. You may have to enlarge holes in the board with about a No. 40 drill. Anyhow, put the shield of the RCA cable in a hole near the left end of the 1K resistor's lead, say four holes up and sixteen from the left end. Pull this shield lead about half way through, bend it over against the ground bus, and solder it there. Put the hot lead through a hole in the board near the top edge, say the third down and the sixteenth from the left end; bend it over to cross the free 100K resistor lead and solder it there.
Follow this procedure with the miniplug cable; put its shield into a hole just to the right of the 1K unit's bottom end. (Shall we say four up and twenty-sixth from the left.) Bend it over against the "ground bus" and solder it there. Plug in the center conductor next to the junction of the two resistors; bend this over onto the junction and solder all three items at once.
First, all solder connections should be tested by wiggling the items. If connected materials "act as one piece of metal," you've earned the license to cut off excess wire -- careful now. With diagonal cutters, cut the excess lead of the 1K resistor's upper connection -- nip it off before it leaves the edge of the board. Cut the long ends of the 100K resistor, just past where they are joined by the center conductors of the cable. If there is any, cut off excess shield lead before it extends off the edge of the board. Pay attention, gang; if you miss and cut a lead between a part and its intended connection, you'll have to use a jumper wire to join the two pieces -- you know what that does to your score.
With strong tape of some kind, wrap the ends of the board so as to imprison the cables. This will keep the cables from twisting and breaking.
What's our score? Well, everything fit on the board, it was possible to plug everything in from the component side, there is a clear sense of where the signal is going, and not one extra piece of wire was used in the building. In fact, this may be the only thing I've designed which is free of demerits.
Now, let's look at the circuits; note that, while they're all very different, they all describe the very same thing that we just finished building.
Circuit Description I
The cold input is grounded. The hot input terminal goes through 100K, then through 1K to ground. The cold output is grounded, while the hot output lead goes to the junction of the two resistors.
Circuit Description II
The hot input lead goes through two resistors in series to ground -- 100K and 1K, respectively. The cold side of the input is grounded. The output is taken across the 1K resistor, with the cold side at ground.
Circuit Description III
The hot input goes into a voltage divider which is referenced to ground. This divider consists of a 100K resistor on top, and a 1K resistor to ground. The output of the voltage divider is the output of the attenuator. The cold signal lines are common to ground.
Note: The terms "input" and "output" refer to the circuit we're working on, not to where they plug in to. The input of this circuit is an RCA plug, and this gets plugged into the output of your hi-fi. The "output" of this circuit goes into the mike input of your tape machine.
The descriptions and/or drawings of circuits are not direct instructions for laying out projects. Whether circuits are drawn or described, they are done the way they are to promote "circuit sense," as ol' Mr. Hendriks used to say in my university classes. In the above example, only Circuit I vaguely suggests an order which resembles what we have constructed. As you gain experience in building things, you will learn to divorce yourself from the literalness of "schematics." Eventually, you will know exactly what they mean, and you will loosely interpret what they specifically state.
Yes, I'm afraid that the "project design" rests with you. As you read a circuit (either by way of a drawing or a description), you must: picture the parts in your mind as you mentally construct little sections within the circuit; orient the chips and transistors in your mind so that their input sides are near the input signal, and so that their output sides are nearest the output end of your board; then decide on the size of board with all these things in mind. I have long since learned to read a circuit a couple of times, letting my mental image get more and more refined each time. Even trial placement of items on the board is something I frequently resort to.
Beginners could well use some pictorial examples -- to watch someone else play the "wiring game" for a while. That's what this series, "From Paper to Project," is doing here.
Hints and Kinks
Cutting Perforated Board
I never buy small pieces of perforated board. I buy the largest chunk they've got, then knock off little pieces as I need them. My term "to knock off" is not far wrong; this board can be sawed with a hack saw or coping saw, but it is much faster (and often more accurate) to break pieces off along a row of holes.
I decide on the width I want; the length I get depends on the dimensions of the piece I'm starting with. (I break it to length after the project is done.) You can mark the row of holes you're breaking with a piece of bus wire; I very often just place a braille stylus somewhere in this row to keep track of it. Then, I arrange a clamp which spans the board. At work, this "clamp" is a large bench vise; at home, it consists of a slab of wood which I line up with the front edge of my work bench, and which has the perforated board protruding out from under it. C- clamps are used to firmly hold this sandwich together.
Arrange the board so that the row of holes you want "just peeks out from under" the clamp -- just above the top of the vise, or just showing at the edge of your wooden sandwich. Running a braille stylus along the clamp will tell you if your board is straight or not; if you don't get a constant rattle of its point along the holes, the board is crooked.
Next, with a scribe or the point of a knife, scribe several times along these holes, using the clamp or the vise as a guide. Do so on both sides. Use firmness, but not so much pressure that your tool will tend to veer. (There is a principle in drafting that, in drawing a heavy line along a straight edge, making one heavy stroke will be less straight than several passes using lighter pressure.) Now, grip the board firmly where it emerges from the clamp; spread your fingers out so as to distribute the force as best you can. Bend the board in one direction, then the other, until it breaks off. The edge will feel ragged because it is broken at the holes, but it will be straight (for a change, compared with my ability to saw a straight line). A file will quickly take some of the roughness away, although the serrations of the holes will always be there.
Preparing Connections for Soldering
Wires must be touching firmly if soldering is to take place. A common beginner's mistake is to allow components to flop around loosely, with the intent that solder will join them. Usually, this does not work.
In the above "sample project," I would blithely say, "Bend this wire down onto that wire and solder it there." However, metals do not stay where you bend them. In order to "bend" a piece of wire into some position, you must actually bend it past this position, in hopes that, when it springs back, it will be where you wanted it. "But how," you ask, "do you bend a wire further down than down against the board?" The answer is: you rock the component on the top side of the board so as to encourage the lead to leave the surface on the wiring side; you bend it down again while the component's body is under tension, that's how.
Let us consider the 1K resistor of our attenuator. Its bottom lead is supposed to lie along the bottom edge of the board. You notice, however, as you roll the resistor body from side to side, the lead flops up and down. Therefore, roll the resistor toward the center of the board, which will cause the lead to lift up off the wiring side; while holding the body thus, bend the lead down again. When you let go, the lead will be less likely to wave up and down. In the same way, try to slide the resistor -- along its column -- toward the top edge of the board, thus causing its upper lead to lift away slightly. Hold this tension on the body and rebend the lead.
The same is true for leads of the cable. In the case of the ground braid, however, I sometimes actually loop it over the edge of the board so as to clamp it in place.
Speaking of shielded cable, there are complicated ways of preparing it. I've chosen the simplest in this beginners' discussion -- unweaving the braid. Discussions of other procedures are well documented in "Soldering, Parts IV and V," SKTF, Winter and Spring of 1982.
I spoke of "tinning" the stranded leads of the cable. This is done by wrapping them in coils of solder, and then heating each such assembly until the solder melts. For a more complete discussion of this process, see "Soldering, Part III," SKTF, Spring 1981.
Soldering point-to-point wiring is really neat, because of the "landmarks" available to you. Often, on PC boards, there is no convenient high-profile item that you can find with the soldering iron on the way to the target. With point-to-point wiring, however, you are always creating more things to find with the tip of the iron -- rails and trails and things that can help guide you. As long as you sponge the tip free of excess solder just before you attack each solder connection, the number of droppings of solder you deposit in unwanted locations will be minimal. Remember, too, that because of the perforated board, you can easily make a special landmark where you need one: inserting a flea clip nearby, bending a wire up to make a "tree," or leaving the lead you're working on excessively long, are all easy landmarks.
Miscellaneous
If you wish to experiment with layout schemes on your board, set it up on a block of Styrofoam. This will allow you to poke the component leads straight down into the Styrofoam, thus helping to keep them in place while you ponder the affair. A coarse grade of Styrofoam is preferred, since the force needed to insert the leads to their full extent will be reasonable, not stressful. A good stable board clamp will hold the board off the table for you; this is the next best thing to Styrofoam.
You will be needing a good board clamp, and a small removable bench vise as well. These fixtures, and other electronics tools, are described in "Soldering, Part II," SKTF, Winter 1981. Addresses are also given in there -- places from which you should have catalogs for the purpose of ordering tools.