0.4 The bugs  (Page 2/4)

• It cannot be done.
• It hasn't been done yet.

I have learned to always assume the second definition as correct until proven wrong. It continues to amaze me how many engineers and scientist always assume the first definition until proven otherwise.

Now back to the story. After thinking about it a bit, I discovered a tricky little circuit using a resistor, capacitor and a silicon controlled rectifier (SCR) that not only solved the problem, but could be used on the existing printed circuit boards (PCB) with minor modification. As in any production environment, the ability to rework a populated PCB rather than throwing it away was always a winner. Even with a repair procedure in place, a second version of the PCB was created with the fix in place. And, before you ask, the fix was still cheaper than an on/off switch. It also was only needed as a repair to a small portion of the products. Figure 1 shows the permanent fix to the bug.

The idea behind the SCR circuit was to have the SCR turn off the power supply once the bit line was turned off by the TMC0271. Once turned off, the SCR circuit could not turn on again until the on button was pressed. If you ever get a chance to tear one of the first year Speak N Spells apart you might see an example of the repair.

Garbled speech (software)

We had a phenomena on the Speak N Spell that we called garbled speech. Garbled speech was just as it sounds, the Speak N Spell would put out an utterance which was imperceptible, mostly. The reason I say “mostly” is that the software version of the phenomena would actually, every once in a while, actually say something that was recognizable.

There was a module key on the Speak N Spell, when pushed would send the product out to the module for the vocabulary words rather than using the ones on the product itself. What we failed to do was perform a simple handshake between the controller chip and the module to see if there was actually a module there. Without the handshake, the Speak N spell, when in the module mode, would try to pronounce whatever came over the open bus from the module connector. To add to the problem, sometimes it would magically pick up a bit of data from the on board vocabulary memories and pronounce it.

We explained to the early users of the Speak N Spell what the problem was and how to fix it (i.e., press the Module key again). It was resolved simply by adding a handshake to the program.

But, as usual, not all is that simple. We received a letter from one of our customers. In the letter was the explanation of how the “garbled” speech sounded in his Speak N Spell. Also enclosed was an audio cassette on which he had recorded the Speak N Spell saying a dirty word during the sequence of garbled speech. The letter’s explanation of the cassette recording was simple. It went something like this.