Starting off with echo units, there are three basic types on the market today — the tape echo, the analogue delay and the digital delay. Although they achieve their echo effects in different ways they all have several features in common. As such we can represent them all using a single block diagram to show the basic principles of generating one or more echoes from an input signal. (Figure one)

The main difference between the three types is in the mysterious box labelled 'delay box'. Okay, less of the waffle, let's look at the various ways it can be done.

Tape Echo

This was the start of it all, just a modified tape recorder when all's said and done and like most effects we've looked at it started life in the studio. The idea is simply to record a signal onto magnetic tape and then play it back again a short time later using a separate playback head. I emphasise a separate playback head because grandad's old reel to reel tape recorder almost certainly uses a single head for both record and playback — this means that it's not that easy to convert it into a tape echo, so you can put it back together! Now the echo delay time is dependent on two things — how far apart the record and playback heads are and how fast the tape is being transported past them. For example, to get a 100mS (1/10sec) delay with the tape running at 15" per second the heads would need to be 1.5" apart. Repeats are then got by feeding back some of the echo signal into the input mixer to be processed again. As it stands the only delay available from this set-up is 100mS and different manufacturers overcome the problem in different ways. Really it comes down to three choices:-

1) Have several playback heads laid out along the tape path. This is the method used in the WEM Copycat. It is much more versatile if the heads are different distances apart to allow much more complex echos. One of the great things about a Copycat is that with a small drill and a half hour of free time you can place the heads anywhere you like — particularly useful for extended delays. (Figure two). A final word about the Copycat — (sorry about this but I was brought up on them!) — if you disable the erase head via a push to break footswitch you can get a lovely continuous 'hold' facility.

2) Use a single playback head on some sort of sliding rail system so that you can quickly and easily alter its distance from the record head.

3) Have a speed control on the tape transport motor.

Some of the most versatile units have a combination of multiple playback heads and variable motor speed.

The main drawback with tape echos is the wear on the tape, which is usually in some form of loop. There are several ways of using looped tape ranging from a short length which tends to wear quickly, a long free loop (they tend to snag and have tension problems) or continuous tape cartridges (often expensive and difficult to find replacements — remember eight-track cartridges?)

Being electromechanical a tape echo also requires frequent and loving servicing to keep it in good nick and to give the best possible echo quality. Continuous tape loops will need frequent replacement and the record and playback heads, rollers and tape guides all need regular cleaning. About once a month it is also a good idea to use a tape head demagnetiser to restore the top end frequency response which becomes muddied due to a build up of residual magnetism on the heads and metal guides.

Analogue Delays

The shortcomings of tape echo units are almost exclusively due to there being a mass of moving parts involved — enter the solid state delay. Analogue delays are based around an integrated circuit chip called a Bucket Brigade Device (BBD), so called because its action is very reminiscent of the old fire fighting bucket chain. To understand how it works we can stick with the bucket chain idea. Imagine five blokes standing in a row, each with an empty bucket. The first one fills his bucket from a tap and then transfers the water into bucket number two. As he refills his bucket from the tap, number two empties his bucket into number three. Number one empties into bucket two as number three empties into bucket four and so on, and so on. As you can see each sample of water from the tap will work its way separately down the line until it gets to number five who then throws it onto the fire, the point being that this all takes time.

A BBD chip does a similar job for electrical signals. A simplified block diagram of a BBD circuit is shown in Figure three. The clock pulse is produced by an oscillator which provides alternate high and low pulses. As the first high pulse is received by the BBD a short duration electronic'snapshot' is taken of the audio signal at that instant — it is sampled and the particular voltage of the audio signal at that instant is stored in bucket one. As the clock pulse goes low this voltage is transferred into bucket two. At the next high pulse bucket one receives another (later) sampled voltage as the original sample voltage is transferred from two to three and so on until the samples come out of the end of the chain one after another to form a copy of the original input signal but delayed by however long it took for the samples to transfer along the line.

It all seems very simple but in practice we run into problems. To accurately reproduce the original signal you have to take lots of samples every second — the more samples, the more accurate the copy. As an example let's take 50,000 samples a second. This means that the clock will have to have a frequency of 50KHz and the delay along our five blocks will be 5/50000 or 0.0001 of a second — that's silly! In practice BBDs tend to have 1000 blocks (actually it's 1024 but let's not quibble) which would give a delay of 20mS or 0.02 sec. This is still too short to be classified as echo so in practice the clock frequency is reduced to have a maximum of 20kHz at which frequency you still get a pretty accurate copy of the input signal, delayed by 50mS. This is pretty quick 'slapback' echo. To get longer delays you simply slow down the clock but you then run into other problems. Clock frequencies of lower than 20kHz come into the audio range and will appear at the output as a high pitched whistle on top of your audio signal. To get rid of this a low pass filter is used to track the clock frequency and cut it out — unfortunately it also chops the top end off your audio signal at the sametime. In addition slowing down the clock means less samples per second and hence a lower quality copy of the input signal. In our example to get a delay of 100mS (1/10 sec) we would end up with a bandwidth of only 0 — 10kHz and longer delays make things even worse. This makes analogue delays rather restrictive for use with signals containing high frequencies such as a synthesizer and, to some extent, vocals. The obvious solution, adding more and more BBDs in the chain to give longer delays at higher sampling frequencies, tends to be limited as the more blocks you have the more noise and distortion you introduce into the signal. All analogue delays end up as a sort of compromise between low noise, long delays and bandwidth. The moral of the story is look at the technical data when you go shopping. In particular look for:

a. Reasonable delay times
b. Large bandwidth — this will be shown as something like
10Hz — 20kHz @ 50mS
10Hz — 7kHz @ 250mS
You should look for the highest possible figure at the longest delay (in this case 7kHz — pretty low!)
c. High S/N ratio.
d. Listen to the quality of the echo signal at the longest delay signal (preferably with no straight signal mixed in) and compare it for 'brightness' and 'edge' with the straight signal.

Digital Delays

One of the problems with analogue delay lines is the noise and distortion generated as the signal passes along the delay line. As the signal is in the form of 'packets', each one having a particular voltage, the noise and distortion show up as small variations in the packet voltage. The digital delay overcomes this problem by sampling the input voltage (at a very high frequency) and converting it into a digital code in which the voltage is represented by a series of on and off pulses (in fact a binary number) in the same way that a computer stores a number in its memory. This can then be transferred from one location to another and so on within a computer memory chip until it comes out the end when it is decoded back to the original voltage. As even a bog standard computer memory chip is capable of storing 16000 bits of information you can see that long delays and high sampling frequencies (thus high quality reproduction) become possible. Any spurious noise is not recognised by the decoder because it is not in the form of digital pulses and so it is ignored and lost in the decoding process. With the ultra clean signal longer delays are possible by incorporating more memory. With home computers booming the cost of memory is getting less and less which explains the reduction in price of a device that could previously only be afforded by a studio.