4. You may have heard from an analog course about" ground loops"and the necessity to avoid them. In digital systems it is completely impractical to avoid ground loops, and the opposite approach is taken, best summarized in a phrase which I will pass on: Let Ground Abound 5. If you take care with the ground distribution in your projects, few bypass capacitors will be necessary, but for conservatism you might consider placing a 0.01 to 0. 1 ceramic bypass capacitor between the ground and +5 power grid ever half a dozen packages or so. Your oscilloscope is your best friend here. Look at the +5 power runs while your project is running. Does it look like a DC power supply? Anything more than about 1 volt p-p noise needs to be fixed. Sometimes adding power supply bypassing can actually make a project not work. The way this happens is that the ground distribution system is high impedance, and the addition of the decoupling capacitors makes the current in the ground lines highe leading to ground noise. Since the noise margin in the low state is worse than in the high state this can be a bad tradeoff. Unused Inputs They have to be connected to either ground or to high". Ground is easy to come by th " can be the output of a grounded input inverter, or a resistor tied to the positive supply. The resistor value is non-critical, somewhere in the range of 1K. Many inputs can be wires in parallel to this resistor. Making the run extremely long can be bad from the standpoint of finding errors. It is common, and the source of difficult problems, for the high run to have some output connected to it. Since you often pull up sets and clears on widely distributed circuits, this can cause very bizarre behavior. Keeping the number of inputs on the high run to a reasonable number(5-10)will help isolate these problems Behavior of Ungrounded parts The behavior of parts which do not have a ground pin connected can be quite disconcerting. They seem to almost work. The reason for this is that the input signals provide a source of negative supply. So, if you have a NANd gate, for example, in a package with no other sections used, then it can behave correctly for the case when the inputs are both low, one or the other high, but incorrectly if both are high. This can be MOST confusing, and is even worse if you are using all four sections of some NAND gate, since you are(almost)certain to have some input low, and the gates will almost work Looking at your logic signals with a scope will discover this and other problems, since the signal levels will look very poor Tri-State logic signals Busses which have tri-state signals driving them are tricky to debug at times. Scoping them can give signals which appear to be middle"due to the parts not driving the bus at all times. One way to counteract this is to explicitly add pullup resistors to the runs, which will force the bus to be a logic1"even if it is undriven4. You may have heard from an analog course about ``ground loops" and the necessity to avoid them. In digital systems it is completely impractical to avoid ground loops, and the opposite approach is taken, best summarized in a phrase which I will pass on: ``Let Ground Abound". 5. If you take care with the ground distribution in your projects, few bypass capacitors will be necessary, but for conservatism you might consider placing a 0.01 to 0.1 ceramic bypass capacitor between the ground and +5 power grid every half a dozen packages or so. Your oscilloscope is your best friend here. Look at the +5 power runs while your project is running. Does it look like a DC power supply? Anything more than about 1 volt p-p noise needs to be fixed. Sometimes adding power supply bypassing can actually make a project not work. The way this happens is that the ground distribution system is high impedance, and the addition of the decoupling capacitors makes the current in the ground lines higher, leading to ground noise. Since the noise margin in the low state is worse than in the high state, this can be a bad tradeoff. Unused Inputs They have to be connected to either ground or to ``high". Ground is easy to come by. ``High" can be the output of a grounded input inverter, or a resistor tied to the positive supply. The resistor value is non-critical, somewhere in the range of 1K. Many inputs can be wires in parallel to this resistor. Making the run extremely long can be bad from the standpoint of finding errors. It is common, and the source of difficult problems, for the high run to have some output connected to it. Since you often pull up sets and clears on widely distributed circuits, this can cause very bizarre behavior. Keeping the number of inputs on the high run to a reasonable number (5-10) will help isolate these problems. Behavior of Ungrounded Parts The behavior of parts which do not have a ground pin connected can be quite disconcerting. They seem to almost work. The reason for this is that the input signals can provide a source of negative supply. So, if you have a NAND gate, for example, in a package with no other sections used, then it can behave correctly for the case when the inputs are both low, one or the other high, but incorrectly if both are high. This can be MOST confusing, and is even worse if you are using all four sections of some NAND gate, since you are (almost) certain to have some input low, and the gates will almost work. Looking at your logic signals with a scope will discover this and other problems, since the signal levels will look very poor. Tri-State Logic Signals Busses which have tri-state signals driving them are tricky to debug at times. Scoping them can give signals which appear to be ``middle" due to the parts not driving the bus at all times. One way to counteract this is to explicitly add pullup resistors to the runs, which will force the bus to be a logic ``1" even if it is undriven