All About Clocks By Lauren A. Colby Frederick, MD. March 21, 1992 For years, I've been fascinated by clocks and, when computers came out, I became very interested in computer clocks. As everybody knows, XT machines don't have a built-in clock. AT machines do. It is built into their CMOS ram. Unfortunately, the CMOS clocks don't keep very good time. I read somewhere that an AT CMOS clock can gain or lose as much as five minutes per month. Whoever wrote that was awfully charitable. Some of our AT machines in the office have been known to gain or lose five minutes in a week! Most high-priced AT clones will have a little clock trimmer on the mother board, to adjust the clock. Fiddling with that trimmer is not, however, a good idea! A couple of years ago, we had a machine which kept particularly poor time, so I resolved to adjust the trimmer. Everyday, I'd turn on the machine, move the trimmer clockwise or counter-clockwise, and make a record of what I did, to see whether I needed to slow down or speed up the clock. One day, I turned on the machine and found that it had gained 12 hours, just overnight. I went into Setup and found, to my horror that the CMOS clock was running at twice the normal speed. Evidently, I hadn't properly grounded myself when I adjusted the trimmer, and had wiped out a frequency divider in the clock, so that it no longer divided by two, and was running straight through, instead. The only way to save the machine was to install an auxiliary clock board, which I did. In fact, I now install an XT type clock board in all my machines. The 40 mhz 486 machine which I'm using right now has a little $10.00 XT clock board, because it keeps good time, while the CMOS clock does not. In that connection, don't believe the nonsense you may heard, that an XT clock board won't work in an AT machine without disabling the CMOS clock. The board will work just fine! All you have to do is plug it in, and read the time with the software that came with the board. The best ones use a program called TIMER which you invoke from your autoexec.bat file. If you have DOS 3.3 or better, invoking TIMER /S will set your DOS clock and the CMOS clock as well! Also, you don't necessarily have to use up a slot to install the board. Probably, you have an I/O board in your machine already (parallel, serial, game port) and it most likely has some cables coming out of it which go to sockets, installed in slotcovers. If so, you can remove the slot cover from your clock board and plug the board into the bus, behind one of these socketed slot covers. That way, a slot which is normally useless becomes useful. Purists, and I am one, will naturally want to set any clock so that it keeps perfect time. The obvious way to do that is to count the clock oscillator with a frequency counter. That, however, is easier said than done. The output from the oscillator, whether it be from a CMOS clock (which, as noted, you should not touch, lest you damage an expensive system board) or an XT timer board, is so weak, that the input of an ordinary frequency counter may swamp the oscillator and make it quit. You can, however, make an impedance converter to allow you to read the oscillator with the counter. Go to Radio Shack and purchase a VMOS power FET. Connect the source to ground through a 1,000 ohm resistor. The drain is connected to plus nine volts (ordinary nine-volt battery), through a 10,000 ohm resistor. Connect a 2 meg resistor from the gate to the ground, and a 10 meg resistor from the gate to the plus nine volt line. Connect a .1 ufd capacitor to the gate; the signal from the clock oscillator is picked up from the hot side of the clock crystal and fed into the converter through the capacitor. Connect another .1 ufd capacitor from the FET drain to the input of your frequency counter. This setup will let you observe the output from the clock on your counter, However, the signal from the clock is so tiny that you may still have to fiddle with the bias some (the 10 meg resistor), to get a good, clean output. The clocks all use 32,768 HZ crystals, so you should adjust the trimmer for a 32,768 count. If you are really finicky, you'll want to use a frequency divider; and divide the output down to the point where you can use the "period" mode on your counter, to get a super-accurate reading. All of this may, however, be a waste of time. Recently, I bought a couple of clock boards by mail. They had a place for a trimmer, but no trimmer was installed. I figured the manufacturer was just trying to save money, so I installed the missing trimmers. Then , I hooked up the frequency counter, and began adjusting the trimmers, to see how much I could "rubber" the crystals. To my surprise, twisting the trimmers did nothing at all! The oscillators just kept ticking away at 32,768, no matter what I did. If I added too much capacitance, the oscillators just quit! Finally, I realized that the manufacturer had eliminated the trimmers because they were not needed. Apparently, the boards have an improved type of clock crystal, which won't allow itself to be "rubbered". As a matter of fact, the boards seem to keep nearly perfect time, but I still miss the ability to fiddle withthe trimmer. Finally, a few words should be said about the clock oscillators themselves. Some use low powered 7400 series IC's. Others are constructed from varicap diodes, and sometimes, you will see the oscillator built into the clock chip. In all cases, however, the best place to pick up the signal for testing is right at the crystal itself. Use a micro-clip lead from Radio Shack, and always isolate the lead from the frequency counter, scope, or other test instrument, with an isolation capacitor.