What is true bypass?

The first guitar effects were designed to be hooked directly from your guitar to your amplifier, and often sit on top of your amplifier. A common example would be the Rangemaster treble booster. The earliest guitar effects with a bypass used a switch that could be throw two ways with a single push, the SPDT (single pole, double throw). Your guitar signal would always be fed into the effect input and the switch would then select between the effect's output or your guitar signal at the front of the effect. This lead to signal degrading as both of the effect and the amplifier would be sharing an already weak guitar signal. If you only used the one or two effects and almost always had one on then this would not be a problem.

To combat this problem, most modern manufactured pedals use transistor switching. Instead of a big expensive mechanical switch, a small momentary button switch is used to trigger a flip-flop circuit to switch the signal using transistors. This has a few advantages. It's much, much cheaper. It's more reliable as they have eliminated a mechanical part, and you get the buffering effect to combat loss of high end. However, for every one of these pedals you have in series, you now have a complete electronic circuit re-processing your signal. This adds to the noise to signal ratio! Also, most of them don't give unity gain, so you will usually loose a tiny amount of volume for each one.

In comes the DPDT (double pole, double throw). This can be thought of as two SPDT switches in the same casing sharing the same actuator. By mechanically switching both the input and the output of the effect we have true bypass. The effect has been physically isolated as much as possible from your clean signal.

There still is no LED (indication light) though. Again, this wasn't a problem if you had the one or two effects and you didn't switch them often. You should be able to hear the difference anyway. An LED was sometimes added using a DPDT by only switching the effect output like before, and using the other switch to turn the light on. This of course can not be classified as true bypass.

True bypass + LED?

Presenting the 3PDT (three pole, two throw)... Same deal as above, except now we have a switch to switch the input and the output and the led all in one push of a button. Alright! With some thoughtful wiring you can also have the input of the effect grounded to prevent it from self oscillating in high gain situations.

True bypass DPDT + LED?

A subject of many arguments. Achieving true bypass with a DPDT + LED is uncommon, but it is technically possible. A good example of this is The ProCo Rat distortion. With some clever electronics, a sensor is put on the input of the effect to see if there is a connection to your guitar or if the switch has broken the connection and the input has been left floating. This can then switch the LED. Technically, the clean signal is still completely isolated from the circuit in bypass mode, so there should be no complaints about degrading the signal.

Stereo true bypass and relays?

Obviously there are twice as many audio connections with a stereo signal, so neither the SPDT, DPDT nor 3PDT has sufficient capability. Introducing the 4PDT! Okay, now this is getting crazy! It's like the latest razor where they keep introducing one more blade for closer shaving! From what I understand, so far the 4PDT switches in production are fairly unreliable and have a high failure rate. The best way to achieve stereo true bypass is with relays. These are electronic powered switches which have an internal mechanical operation. They are probably the last method that you could count as still technically being true bypass.

Optical bypass?

This technically is not true bypass, because the signal is not physically isolated from the effect. This uses optocouplers (which go by many names). To make an optocoupler you have a light shining against a light dependant resistor (LDR). The LDR works like a tap valve, opening further as the amount of light shown on it increases. When the light turns on the resistance value drops very low, to the point where it is practically almost a closed circuit. When the light is off the resistance increases and the circuit is almost open. This can be a useful method of switching stereo.

Let's get technical

A switch can be described in terms of its poles and throws. The poles connect to the throws. When the position of the switch is changed, the pole's connection changes to a different throw. The simplest true bypass switch is a DPDT (double pole, double throw). It has 6 terminals and toggles in two positions. In the diagrams below, the connections are represented for each position in red:

Using this information, we can construct this simple true bypass circuit:

There you have it. But there is no way of knowing if the effect is on other than aurally. Most of the time this will be okay, but it could be a lot better if you added an LED. We can use a switch with an additional pole, the 3PDT (three pole, double throw) switch. In Figure A we see the same circuit as above, but with an additional pole on the right to switch a single LED on while the effect is toggled. Use a resistor in series on either side of the LED to limit the current to prevent the LED from burning out. 1.5K should be safe for a 9 volt source. Increasing resistance decreases brightness.

We can improve the design further, as in Figure B. The ground is used to both switch the LED, as well as stop the effect's input from floating while in bypass. This can prevent oscillating effects in high gain circuits which can, by capacitance, cause unwanted effects on your bypassed tone, as well as discharge the effect's coupling capacitor to prevent some popping noise. The connection circled in orange eliminates an unnecessary switch. This theoretically gives better continuity and reliability. If you are using a DPDT without an LED, this connection should also be made.

Pop?

One other reason why commercial manufacturers don't usually use true bypass is because they have the potential to pop when toggled. If good design practices are met, then this shouldn't be a problem. It's more noticeable at bedroom levels and less noticeable at performance levels and in a live mix. It's also more noticeable when you toggle without being playing anything at the time. The cause of this is usually the decoupling capacitors on the input and output of the effect. While in true bypass mode, the outside of the capacitors are left floating and do not connect to your amp or guitar anymore. During this time, DC can leak through and build up on them. When you go to toggle the effect on, BAM, that DC goes to your amp and it doesn't sound pleasant. To prevent this, you would use a pull-down resistor of around 1MΩ or higher. This is a high enough value to not effect the tone when the effect is toggled on, but low enough to not allow the capacitor to build up a DC charge. Grounding the input of the effect while it is bypassed is also a good idea.

In the end it becomes a matter of quality of tone versus consumer convenience.