Electronic Construction -- Tips, Tricks, Gens, Traps, and Snares

updated 2003年08月08日

(what's a "Gen" ?)

This page has general electronic construction tips (applicable to ciruits build with solderless breadboards, wire-wrap boards, and printed wiring boards). Do we need another page for tips specifically for PWB design ? [FIXME: seperate "design-time" tips from "assembly-time" tips]

power supply tips
standard signal tips: series resistor, etc.
high-power and/or high-voltage tips
assembly tips
other tips

power supply tips

Protect against incorrect (reversed) insertion of batterys

Appling Vcc to the GND and Gnd to the Vcc will fry just about anything. Use of a fuse (or one of the PTC thermal auto-resetting devices from Bourns or Littlefuse) and a diode is a minimum. You can take an enhancement power MOSFET and put in the negative return line from your circuit. The N-Chanel Enhancement MOSFET is used "upside down" compared to other circuits. That is to say, the Drain is connected to -ve, Source through the load to +ve, and gate to +ve. The gate is driven through a 1 Meg resistor from the positive supply terminal. A correct power connection drives the MOSFET fully ON, and everything works. Reversing the supply connections turns the MOSFET OFF, and no current flows and you're protected. Correct the power connections, and you're back to normal. Select a MOSFET with a low "R(on)" resistance, and there will be very little voltage drop.The better way to go about this is to protect reversal based on the mechanical properties of the battery. mostly this is done by a positive battery terminal that is set back in the holder so only the positive battery pole that sticks out can reach it.

Connect all power and ground pins

Many chips have several power supply and ground pins (sometimes labeled Vdd and Vcc). Make sure each and every one of them are connected appropriately (typically Vcc to ground, Vdd to +3V or in some cases +5V).

There are a long list of strange things that happen when you skip this step.

The PICC74A may not work properly with it's A/D Vref set to reference the voltage on AN3 instead of chip Vdd if pins 31 and 11 are left floating.
The Scenix SX52 may not start running if all of its supply pins are not connected.

Q: "What does "dd" stand for in "Vdd" ?" A: "dd" in "Vdd"

Make good power supply and ground connections.

A skinny, daisy-chained wire-wrap connection from chip to chip is probably not going to be enough. The more robust you can make power supply and ground leads, the better

Despike your power supply to each chip.

This means putting a .01 to .1 uFd ceramic capacitor from the +5 volt supply to ground right at the chip power line. Spikes occur when there are sudden current changes far from the power supply, and it's amazing how much trouble this can cause. If there are two or more Vdd pins, you need to put a bypass cap on each and every Vdd pin. It is critical to de-couple power supply lines. You want the caps to absorb well at the third harmonic of the clock. 0.1uF does well at 3 MHz, 0.01 at 30. 0.001 at 300. It's a broad response, so dont think that there's one specific value. However, if you use 0.1uF on a 20 MHz part, you won't get the supression that you could if you used 0.047uF. (In theory, an ideal 0.1 uF capacitor should always give better suppression than a 0.01 uF capacitor. But real capacitors have parasitic lead inductance and other non-ideal annoyances.)

Use the right kind of capacitors

Watch out for class 'X' and class 'Y' capacitors: which are designed for connection to mains voltages. These are fail safe capacitors which are self healing and must be replaced with the same type.
Replacing electrolytic caps with much higher voltage rated devices can upset circuit operation.: Electrolytics don't start behaving as proper capacitors until they reach a fraction of their rated voltage. Also look out for special low ESR (equivalent series resistance) capacitors found in switch mode power supplies.
Some circuits require capacitors capable of withstanding large current pulses.: Using the wrong sort of cap in this situation would lead to overheating and other nasty consequences.

Keep the distance between the Vdd and Vss pins on the same chip as short as possible. There should be one connection from a chip to the power supply. If two leads on the chip, the best method is to make a T with equal left/right sides if possible. Don't kill yourself for it, though. If you need to make a main feeder which splits into two short (unequal but not greatly disparate) lengths, so be it.

The importance of this depends on frequency. Remember, a DC signal is an AC signal with an infinite number of frequencies all out of phase with one another. Also, you _always_ have some ripple on a trace. As the trace lengths change, the relative position of the chip on the ripple changes. Think of a long boat on the ocean. The keel is always "ground" but if one end of the boat is higher than the other (i.e., there is a potential on one ground pin vs. the other), then you're asking for problems. As the lead lengths change, the differences become more pronounced. Worst case would be when one lead was 1/2 wavelength from the other.

standard signal tips

Don't "despike" your signal lines, add a resistor instead.

David Vanhorn says:

[The worst thing is ] putting a cap to ground (some arbitrary point in the ground system) from a signal line, to "smooth it out"... Grrr. Now the driver needs much more current on each transition, and the problem is worse.
Resistance in series solves this problem, but it's not so convenient to implement usually. (Tip: high speed clock lines should always have a resistor at the source. Even if later you choose the value to be zero, though likely 100 ohms will be better)

(see "Spice provides signal-integrity clues" article by Ken Boorom, EDN, 2/18/1999 http://www.reed-electronics.com/ednmag/article/CA56674?pubdate=2%2F18%2F1999 for more details. )

Cirrus (2006) recommends:

"... In addition to standard mixed-signal design techniques, system performance can be maximized by following several guidelines during design. ...
- Place a buffer on the serial data output very near the A/D converter. Minimizing the stray capacitance of the printed circuit board trace and the loading presented by other devices on the serial data line will minimize the transient current.
- Place a resistor, near the converter, beween the A/D serial data output and the buffer. This resistor will reduce the instantaneous switching currents into the capacitive loads on the nets, resulting in a slower edge rate. The value of the resistor should be as high as possible without causing timing problems elsewhere in the system."

-- from the 2006 datasheet for the CS5368 24-bit A/D converter

Keep your digital and analog circuitry physically separate whenever you can.

Digital switching, especially at microprocessor buss or video card speeds, can throw all sorts of noise and trash into analog or audio circuitry via the power supply. Adaquate Isolation is a must.

Add Grounded Shields or Ferrite Beads to long cables carrying high frequency signals

it won't solve all problems, but it's better than nothing.

A few more tips at Massmind: Avoiding Noise.

high-power and/or high-voltage tips

Cut the PC board to seperate positive and negative power terminals in high current products: Opening and closeing a power switch can cause a pico second arc which slowly deposits a metal film across the board. There may also be an impurity in the laminate that occasionally gets you. Then when you close the switch after a number of depositions of metal, there is enough of a path for current, and its downhill from there once there is a carbonized path. ...this is why slots are often cut in boards around these sorts of areas, to stop potential carbonization tracks with dirt and moisture build up.

"PCB Trace Width Calculator" gives the IPC-2221 (formerly IPC-D-275) recommended trace width vs. current. (Nearby is a "PCB Via Calculator")
"PCB Layout: The Impact of Lightning and Power-Cross Transients" by Milton Hilliard describes an experiment to determine the minimum trace and spacing recommended for conductors that connect to outdoor wires, to withstand the sort of high voltage and high current transients that cheap "lightning protection" devices let through. summary:
- minimum 40 mil spacing on surface layer traces connected to outdoor wires to avoid 2500 V arcing
- minimum 7 mil spacing on internal layer traces connected to outdoor wires to avoid 2500 V arcing
- minimum 22 mil trace width to ensure that a trace connected to a 10 A fuse does not become an open before the fuse blows.
Trace Spacing Guidelines at http://www.pcb-designer.com/ and Trace Spacing Guidelines at Micro Technology Services and "trace spacing based on voltage levels" at the PCB Wizards FAQ, all apparently have data from the "Design Standard for Rigid Printed Boards and Rigid Printed Board Assemblies IPC-D-275". summary: to avoid arcing,
- minimum 0.008 inch spacing for internal conductors up to 300 V(peak)
- minimum 0.015 inch spacing for conformal coated external conductors up to 300 V(peak)
- minimum 0.005 inch spacing for conformal coated external conductors up to 15 V(peak) or bare component terminals
- minimum 0.025 inch spacing for bare external conductors up to 15 V(peak)
Although IPC-D-275 standard has been superseded by "IPC 2221A - Generic Standard on Printed Board Design" and "IPC 2222 - Sectional Standard on Rigid Organic Printed Boards" , as far as I can tell the new standards have the same recommendations as the old standards.

assembly tips

Don't use silicon sealant to mount wire-wrap sockets or to seal/insulate circuitry.

This stuff is handy and common but it is not an insulator. It will leak small currents, which may not matter in logic circuits but can wreak havoc in high-impedance analog circuits. Lawrence Lile [lilel at toastmaster.com] says:

Also, don't use it because some silicone sealants contain acetic acid (smell 'em if you don't believe me) which will react with copper or iron, causing corrosion. Corroded copper and rust are famously poor conductors.

Use IC sockets on prototypes.

After you have things debugged, you might consider soldering chips directly to boards to save cost and eliminate connections that can oxidize or come loose, but during the design/testing phase you need to be able to swap chips without repeated desoldering.

Cover the window on any chip with EPROM even if you don't care about the EPROM part of the chip

PIC processors for example are sometimes affected by non-uv light entering the EPROM window and shineing on the silicon. This can actually inject signals into the non EPROM parts of the chip!

Always provide mechanical strain relief for wires soldered to a PCB

Wagner Lipenharski says:

Another possibility is drilling two extra holes at the board (little bit bigger than the wire diameter), just to zig-zag the wire (snake path) through them after soldering. If you don't use any mechanical lock to the wire, sooner or later it *will* break up at the solder point.

Jinx wrote:

If you DO end up just soldering the three wires to the board, the back presumably, put a small dab of silicon sealer over the wire adjacent to [but not in contact with] the solder point. This will hold the wires if you're worried about movment loosening them but will still give access to the joint. Silicon holds fast but is very easily cut off with a scalpel.

unsorted other tips

Make sure you can get the parts before basing a design on it.: You may find the ideal integrated circuit for your application in a data book, but it may not be in production, may be unavailable from distributors, or may be too expensive. Especially if you are creating a design you hope to produce for a while, it's wise to choose devices that are widely available and that (you hope) won't be discontinued.
NEVER plug untested circuitry into a computer backplane slot!!!: Never never never, if you love your PC. All it takes is a simple little wiring error and your motherboard and disk controllers will be toast. While IBM-PC/AT buss interfacing isn't rocket science, it's all too easy to do expensive damage to your machine. Consider other interfacing methods (parallel/serial ports, commercial control/data acquisition boards) first. There are buffer cards that transparently permit prototypes cards to be used while protecting the computer buss, but they are not cheap.
Pick chip types with the correct switching speed for your design: Digital logic chips having very fast edge speeds (dV/dt), such as 74S, 74AS, 74F, and 74FACT can cause RF crosstalk and interference (especially in poorly designed or laid out cicuits) that will frustrate the most brilliant engineer. See Signal Levels@. Avoid them unless you need a super high frequecy, low propagation design.
Don't trust data sheets labeled "Preliminary Information".: This means that the data sheet was written before the chip was actually in production, and the device may change significantly by the time it is actually released. The device performance may be different; still more important, the pinout may be completely rearranged in the final design. Preliminary data can help you choose a device but don't set your design in stone based on it.
Don't rely on simulators: Simulations are doomed to succeed. Reversed biased Transisters are a good example.
Watch out for temperature variation in component specs
Shottky's aren't better, just different: Chris Eddy says:
Just a quick look at a Diodes Inc. catalog shows that the shottky's are typically 1mA reverse current, and the regular diodes are typically 5uA reverse current. I found out the hard way when I built a circuit that needed protection from negative going signals. I put the Shottky across the diode in an RC filter between two op amps, so the resistor served two purposes. I didn't look real hard when I tested it, and the customer complained about temperature dependency. We quickly discovered that when the cover was removed and a heat gun was used, an easily observed error at ambient became a nearly zero signal level out at high temperatures. The figures shown above for leakage are at 25C! Needless to say, it was a rude awakening to the drawbacks of Shottky.
Make sure you know how your connectors actually connect: Some DIN connectors internally connect pin 2 to the shield
Don't assume that a reverse biased bipolar transistor will not conduct through a reverse biased junction: Adding a series diode in series with the transistor's collector is a good idea when reverse biased C-E junction will occur if well defined operation is desired,