The Cassette Connection

With this issue, we start a “mini-series” on improving cassette reliability. I’m sure you’ll agree that unreliable LOADs are perhaps one of the most frustrating aspects of personal computing. With a little knowledge about the SAVE/LOAD process, we can take steps to save ourselves a lot of grief.

We’ll start the series with a discussion of the nature of the cassette signal, and derive various hints (and perhaps get rid of some misinformation) that will help. Next issue we’ll pursue a quest for the “ultimate load aid,” and will cover various ways of monitoring & conditioning the cassette signal.

First, let’s have a discussion of what the LOAD signal actually is. By its nature, the cassette interface is a SERIAL device – each bit is sent one at a time. The simplest way of transmitting serial data is to define a pulse as binary “1,” and the absence of a pulse as binary “0.” This is not very reliable, and timing is critical. The slightest drop-out will cause false data.

Now, a statement that may surprise you; the ZX/TS SAVE/LOAD approach is one of the most reliable among the “bottom-end” computers. If you think you’ve got trouble with your ZX loading, you haven’t tried loading CoCo or other computers under less than ideal conditions. One reason for this is that ZX loading is relatively slow. Data is transmitted at an average of about 300 bits per second (or a little more than 2 Kbytes per minute), compared to say 600 baud (bits/sec) or more for most other machines. As a result the ZX system is far more immune to bad tape, crummy recorders, etc. Another reason is that exact timing is not crucial for the ZX; our guess, based on actual experience, is that speed fluctuations up to 30% are ignored.

How does the ZX do this? Take a look at the diagram. Instead of “pulse=1, no pulse=0,” our machine sends each bit as a SERIES of pulses. The rate of these pulses is about 3300 Hertz. Five such pulses represent a “zero,” nine pulses is a “one.” Each bit is separated from the next by a silence equal to the length of a “zero.” This implies that actual LOAD/SAVE time varies with the ratio of zeroes to ones! IE, an empty array (all zeroes) saves faster than the same array filled with data.

Try this experiment:

10 DIM A$(10000)
20 FOR A=1 TO 10000
30 LET A$(A)=CHR$ 0
40 NEXT A
50 SAVE "X"

After DIMensioning array A$, all values are already zero. So you can RUN, then BREAK, then GOTO 50. (Loop only needed later when filling the array with all binary ones.) Time how long it takes to SAVE. Then replace line 30 with LET A$(A)=CHR$ 255 (all bits = 1), RUN it again, and start timing when the SAVE begins. Fascinating, eh?

This approach is highly immune to differences in recorder/player speed. Even speed fluctuations are virtually ignored, shooting down “myth No. 1” – “your recorder’s speed must be dead-on, with no flutter or wow.”

“OK, if all is so rosy, how come I’m having trouble?” Well, let’s take a look at some factors which ARE important. First among these is amplitude (output level). The ZX needs to see pulses with a level of about 2 volts peak (4V P-P) for proper loading. The maximum voltage your recorder is capable of delivering is limited by supply voltage. With a 6V supply (4 1.5V cells), you can get up to about 5V P-P – plenty enough. Recorders of the “Walkman” variety typically use 3 cells (4.5 V) or even 2 (3 volts) – not enough for reliable loads. Another factor is high frequency response; if poor, the pulses are excessively “rounded,” and won’t convey the data properly. If you have a “tone control,” turn it to full treble. Worn or dirty heads, bad tape, or (very rarely – more about this later) azimuth misalignment can also cause poor HF response. Low frequency response (below 1 kHz. or so) we couldn’t care less about; “poor” LF response is actually a boon, as LF “garbage” is ignored.

Various hints have appeared over the last couple years; some are valid, others aren’t. Let’s take a look at some of these.

1) “Set LOAD level so you have 4, 6, or N bars on the screen.” I don’t know about you, but I find counting bars a little like counting the stripes on a frightened zebra. Even if you do your count during an “all-zero” part of the program, how many bars you see depends more on your TY than on the recorder. A slightly better criterion is the relative WIDTH of the bars; if the white portions are “fat,” level is too low; if the white bars are considerably narrower than the black portions, level is probably too high.

2) “Set recorder volume to full blast.” Works with many machines, but not a cure-all. What happens at full volume is that the cassette amp “bottoms out” and clips the tops and bottoms of the wave. Such clipping is actually beneficial, as it makes the signal look more “digital,” ie, more like an ideal square wave. Some recorders, however, show a pronounced overshoot on the last cycle of every wave-train under severe overload; this can be interpreted by computer as an extra pulse. This results in a bad load.

3) “Adjust azimuth alignment frequently.” Balderdash. Once it’s set, it’s set, and it’s VERY rare that you’ll have to mess with this. You’re likely to have more trouble fooling with this, unless you have a KNOWN GOOD azimuth test tape. I have yet to see a machine that needs to have this factory adjustment tampered with. [See also “More Tips and Myths” later on for more info on this ed.]

4) “When SAVEing, start the recorder after the screen has gone blank to avoid the pre-SAVE buzz.” An article in a well-known magazine suggested that the buzz is an accident in design; maybe so, but in fact, this “null signal” has an important function: to set the record level on machines with automatic volume control (AVC). If you start the recorder after the buzz, chances are good the gain will be “wide open”. The first burst of data will “splatter” until the AVC adjusts itself; too late, you’ve got a bad recording. The article in question did make a valid point – the pre-SAVE buzz is derived from the composite video signal to your TV. What you’re hearing is essentially the vertical retrace pulse. Being a single pulse (and not a train of 5 or 9 pulses) the computer completely ignores it during LOADing. Time for another experiment. Enter this short program:

10 SLOW
20 FOR A=0 TO 21 STEP 2
30 PRINT AT A,0;"\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:"
40 NEXT A
50 PAUSE 60
60 CLS
70 PAUSE 70
80 RUN 

Put your cassette recorder in RECORD and RUN the program for a minute or so. Rewind the tape, and listen to the playback. You’ll hear a definite change in the buzz as the screen is alternately filled and cleared. Rewind the tape again, enter LOAD “X” (or some other nonsense) and start the tape. You’ll SEE the difference between the “clear screen” and the “full-screen” buzz. You have actually used your cassette machine as a crude video recorder. Moral – if the messy-looking pre-SAVE buzz annoys you, SAVE from a clear screen. The computer doesn’t seem to care either way, I’ve yet to have trouble loading with a file name (LOAD “” should be used with caution) no matter how whiny the buzz sounds. By the way, the above test can be used as a level check criterion; adjust volume so that the clear part looks clean, and the “full-screen” part looks good and garbagey.

Mention should be made of how the various available speed-load routines fit in. (See Z XLR-8 review later this issue), These can be a marvelous refinement for your system, but they require a good recorder and very good tape for reliability. If you’re having trouble with the regular tape routines it will only be worse at higher speed.

Look for these points:

1. Does it operate from a 6 Volt or greater supply?
2. Does it have a MIC jack? (Sounds silly, but at least a few models don’t)
3. Does it have a tape counter? VERY handy.
4. AC adaptor included?
5. Does it allow cue/review (F.Fwd & Rewind permitted in PLAY mode for finding selections); quite useful but will wear head faster if used a lot.

I’ve been using Radio Shack Minisette IX’s and have been happy with them. (At that price I’d better be!) The first one still works fine even though the head is about shot after 18 months of daily use.

I’ll close this installment with a few words about tape. The tape itself is rarely the cause of bad loads, unless it has bends or kinks in it. More often, the cassette shell is at fault, causing binding which results in the tape being pulled across the head unevenly; the poor contact gives drop-outs (and bomb-outs). I used to use Maxell tapes, but recently I’ve found that their shells aren’t up to snuff. Memorex, forget it. Good tape, rotten shells, TDK, Agfa, Sony, 3M tapes seem to be fine. Bargain “3 ferabuck” tapes are ok sometimes, but don’t trust them. Some tapes record “hotter” than others; mark your ideal volume setting with a spot of paint on the thumbwheel, vary it from this setting to accommodate different tapes. Commercial software is a generation removed from the master, and can present special problems. Reputable manufacturers provide at least two copies of each program, under the premise that at least one of them will be LOADable. But ALWAYS make a back-up copy the first time you get a successful load, then put the original away. You’ll probably regret it if you don’t.

Before we get on with our discussion on loading aids, here are some additional comments on recorders and the tape medium in general.

Ray Kingsley, proprietor of SINWARE and the author of the already famous HOT Z and Z EXTRA programs, sent in a letter sharing some of his experiences with the vagaries of the tape medium.

After receiving Ray’s letter, I did some more asking around, as it appeared that computer users fall into two camps: those with chronic head azimuth problems, and those who never touch the adjustment and don’t have trouble. From the limited sample of people I know with computers, there seems to be less trouble with the miniature “pocket notebook” type recorder than with the “table-model” size machines. This may be because of the physically smaller linkages in the little machines, resulting in less play as the head mechanism wears. Trouble tends to develop as the machines get older and the head and head carriage wears; so take Ray’s suggestion (from HOT-Z documentation) and “please do not make things difficult for all of us by using the oldest machine in the house and sticking with it.”

If your system is working fine most of the time, don’t go adjusting the head azimuth for the heck of it. My reason for recommending against head alignment in the first installment was based on these considerations: 1) if you don’t have a known good test tape, it is difficult if not impossible to determine exactly where “dead-center” is, 2) Frequently adjusting the azimuth can introduce added play, as the assembly is rarely very stable to start with. Especially true of machines that use lock-paint to help keep the assembly “tight.” The more you adjust it, the more it will need adjustment.

If you suspect head aligment trouble, you can adjust it by ear if you have to; using a data tape, adjust the little setscrew until the sound is the shrillest you can get it; most affected are the high frequencies, so listen for the maximum “edge” in the sound rather than maximum volume, which can be hard to judge using a computer data tape.

If you received a new tape that you just can’t load, the problem may be head azimuth on the recording end; in this case, you can do as Ray suggests and adjust your machine to that tape if necessary. But don’t forget to set it back before making your backup and other saves, or you’ll perpetuate the error and really mess yourself up.

Finally, in case you’re not already doing it, always avoid the first 30 seconds or so of any tape; almost all of them have some degree of bumpiness in this area, due to the little pin in the hub which holds the end of the tape. This is also the area most likely to get damaged, so consider the first 30 seconds “just more leader” and your problems will decrease. In a real emergency, you can sometimes get a balky tape to load by increasing the head pad pressure by bending the spring outward slightly with a non-magnetic tool. This on occasion will improve head contact enough so you make it over the “whoopde-do’s”. If it now loads successfully, make a backup and throw the modified one away, as bad tapes only get worse, never better. Lastly, in case you’re brand new to this,

MAKE BACK-UPS, MAKE BACK-UPS, MAKE BACK-UPS!

Even with some knowledge of how the tape process works, you still might have to resort to external loading “aids” to get the 99%+ reliability you need. A variety of products are commercially available to do this, ranging in price from about $15 to over $40. At least one fast-load system (Powell Hargrave’s SDS) comes with its own loading “filter” as part of the package price. What are these, and what do they do? Is there an “ideal” load aid that will solve all your loading problems? In this discussion I’ll try to cover the various approaches that have been successfully used, starting with the simplest and working up to the more sophisticated.

If your hearing is relatively unimpaired, then you already have a highly effective audio-signal analysis tool. This is what I’ve found to be the “ultimate load aid” — with a little practice anyone can learn to use their ears to judge the setting of the tape recorder and the overall integrity of the tape signal. But first you have to be able to hear the signal. Sure, you can unplug the ear jack and play the tape; but as supplied, you can’t listen to the tape while it is loading. You could buy or make a “Y” connector to the ear jack, with one end of the “Y” going to the computer as usual, and the other end going to one of those cheapo earphones that usually come with cassette recorders. This will allow you to hear the signal while it is loading into the computer. But since the required voltage for successful loads is quite high, chances are you won’t be able to live with that nasty screech even coming from an earphone. Solution: place a resistor in series with the earphone to attenuate the signal without affecting the signal to the computer. Typically, a value in the range of 22 to 75 ohms will cut down the sound to where it’s still audible but won’t turn the entire household against you. Or you can install a 100 ohm rheostat (potentiometer with only two connections used) to adjust the relative volume of the earphone.

So now what? You’ve hooked up your “computer to ear interface”, now how do you use it? Well, first off you can use it to set level. The best way to do this is to play a short section of the tape to be loaded. Listen to the sound from the earphone as you turn up the volume on the recorder. You will find a point, usually between 1/2 and 3/4 of full volume, at which point the character of the sound starts to change, becoming raspier and even more offensive than the load signal already is at low levels. This is the point at which the amplifier is starting to bottom out. The best volume setting is usually right after this bottoming starts to occur, but before it has gotten so bad that the machine “splatters.” Once you’ve set the volume, rewind the tape and LOAD it into your computer. While it’s loading, you’ll be able to detect changes in the sound resulting from drop-outs or other irregularities. An earphone monitor of this sort will also help you in finding desired selections on a long tape, and is helpful in trimming up azimuth alignment if you have to. It won’t take you long to differentiate between tapes that sound crisp and steady, and those that sound muddy, or fluctuate in level, or (most common) have “highs” that weave in and out (the “crispy” sound getting louder and quieter), Granted, simply monitoring your signal won’t make a bad tape loadable, but at least you’ll have a better idea of what’s going on, Since on most recorders the amplifier output signal is available at the EAR jack during recording (for monitoring purposes), you’ll be able to hear when you SAVE as well as LOAD, helping to insure that the tape recorder is indeed connected to the computer. The unfortunate side effect is that many systems will have trouble if the EAR jack is connected ‘during a SAVE; a positive feedback loop is estabiished and the system “howls.”

It is good practice to disconnect the cable to the EAR jack during a SAVE.

Instead of using an external earphone, you can install two resistors in your deck to use the internal speaker as an audible monitor. Connect flying leads to two 150 ohm resistors and cover with tape or shrink-tubing. Connect one end of each to the speaker lugs. Place an unconnected plug (open circuit, not a shorting plug) in the EAR jack. Now with a tape in the machine and in the PLAY mode, touch the two free ends of the resistor wires to the three lugs on the EAR jack; there are exactly six permutations of two wires and three lugs; try each one. In at least one position you will hear a quieted-down version of the signal. Now, try to record. SAVE a program and put the player in RECORD mode. Once again, touch the resistor wires to the ear jack and try the various possibilities. On nine out of ten machines there will be a connection which leaks a little signal to the speaker in both PLAY and RECORD modes; connection to the EAR jack will usually be the center lug and the lug closest to the edge of the board. Since there are many variants in exactly how the switching is done, the exact connection will vary with different machines. Often (as on the Minisette IX) you can simply bridge the two resistors across the board from EAR jack to where the speaker lines go on the board (blue wires on Minisette IX).

In case you’re worried about loading – the 150 ohm resistors are so much greater than the 8 ohm speaker impedance that the “leakage” load is negligible. Since the speaker is partially engaged during RECORD mode, some machines might give acoustic feedback if you record with the built-in mike. If this is important, increase the value of the resistors until the howling stops. Diagrams 1 and 2 below summarize these acoustic monitors.

There are folks who just can’t bear to listen to those computer screeches, attenuated or not. For you, and for those with impaired hearing, an optical approach may be better. Circuit 1 shows a way to do this with LEDs. This circuit also gives some clipping action and helps flatten out the tops of the load pulses. The 10-ohm resistor is not really vital, but some machines have enough power to blow out the LED at high volume. this resistor helps limit maximum current and extend the lamp’s life, (It also tends to enhance the clipping action somewhat.) For further LED protection, shunt 3 1N4001 type diodes in series across the LED. Unlike the acoustic monitor, this circuit usually won’t show the monitor signal during recording. This is because the level at the ear jack is usually lower in record mode, and cannot be changed with the volume control. It therefore never reaches the 2V required to light the LED.

Circuit 2 shows a recommended switch addition which will save you from having to pull the EAR plug before a SAVE. Points of insertion for circuit 1 and circuit 3 (discussed next) are indicated. Switch pole “A” in circuit 2 may be omitted on most machines (allows DPDT switch).

Now, we can at least monitor what’s going on. How about conditioning the signal somehow to improve reliability? Various approaches are possible. As mentioned above, even the simple LED circuit helps out a little by clipping off anything over about 2V peak (or 4V peak-to-peak), The best solution is to use any one of a class of circuits called “comparators,” which basically “square up” an analog signal to make it more acceptable to digital equipment.

Comparators are essentially high-gain amplifiers that are designed to be over-driven, and usually have some positive feedback from output to input which latches the output in either of the two “overdrive” conditions, full ON or full OFF. The output stays OFF until the input climbs above a certain “upper trip point,” and then snap ON and stay there until the input drops below the “lower trip point,” at which time it snaps OFF, Result nice, constant level square waves no matter how the input signal meanders up and down.

The span from upper to lower trip points is called the amount of hysteresis of the circuit. With no positive feedback, the upper and lower trip points are the same, and the output state changes whenever the input crosses the reference line. The disadvantage with this is that it can result in extra pulses if there is a glitch riding on the leading or trailing edge of the input signal. The wider you make the hysteresis band (or “dead zone”) with positive feedback, the less likely it is that the circuit will transmit glitches. On the other hand, too wide a hysteresis band may result in skipped pulses during a drop-out. The circuit shown in Circuit 3 is set up as a compromise, with the upper trip point at .3 volt and a hysteresis of about .1 volt. It uses a commonly available LM311 comparator, though many others can work; see the “Databooks” for suggested circuits. The circuit also includes a LOAD/SAVE switch and LED monitoring, and represents my “ultimate” load aid. A great many tapes that were balky before will be loadable now, and you can run your recorder at a lower volume setting.

If you’re using a fast-load system like Z-XLR8, SDS, Q-SAVE, Fastloader, etc., you’ll get better results if you add a .01 uF. capacitor in series with R1 (between RCDR “EAR” and RI) and a 1K resitor from the junction of the cap and R1 to ground.

What if all this is too much trouble for you, and you’d rather just buy a loading aid? Which one is the “ultimate” in this case? Well, whichever one ultimately solves your problem. The least expensive option is G. Russell Electronics‘ “Winky Board II,” which basically gives you visual and acoustic monitoring, a LOAD clipper and a SAVE filter to reduce RAM pack noise. It also allows you to patch two tape-recorders together for multiple saves or direct copies. (Tape-to-tape copies are MUCH more prone to trouble than tapes saved from the computer, although the direct patch may at times prove useful.) The price is not unreasonable ($19.95 A&T, 14.95 kit) and includes an in-depth manual. I have one here for testing, and it does everything advertised. I should mention that it appears to attenuate (reduce) the signal slightly, with the result that some extremely “thin” tapes won’t load even at “full blast.” However, you’re better off without tapes like that around anyway. If you get an unusually quiet tape from a manufacturer or dealer, consider sending it back for replacement if reasonable efforts to get it loaded fail.

The accessory I personally recommend for a tape-cleanup device is the VOTEM. Unlike Winky, which is a passive device, VOTEM uses the computer’s supply to power a comparator much like our Figure 4. I have found VOTEM’s tape clean-up circuitry to be extremely effective; even on fast-load tapes where some other devices fail. (I do, however, suggest the addition of a high-pass filter before VOTEM when used with quick-load routines; i.e. .01 uF. in series, and 1K from VOTEM input to ground.) With the modification shown earlier this issue, the only minor “bug” (in the LED monitor circuit) is corrected. The result; even if data acquisition is not your favorite subject, your VOTEM will not sit about idly if you buy one, It’s almost worth its price for the tape conditioner alone.

If after all this you STILL have trouble, face it, pal; you need a new tape recorder. Consider the Timex data recorder, which appears to be a virtual clone of the Minisette IX at about $30 less cost. (Though after getting one and tearing into it, I see why it’s cheaper; it’s not quite as solid as the Minisette, but it is very good.) While I suppose there could be obscure computer problems that could cause save/load difficulties, it is far more likely that trouble will lie with the magnetic/ electronic/ mechanical contraptions we call “tape recorders.”

If you have further data on any of this, or if there’s something you feel we should have covered, drop me a line. We always can use more knowledge.