>But none that I have found actually explains which bit of the system is
>making use of the preamble and why it makes a difference.
>If something is being synchronised, then what is that something?
No, the preamble DOES NOT synchronize anything. What is does is the following: These cheap, dumb 433 MHz receivers have a variable gain input stage. If there is no actual signal (only random noise) coming in then this stage will have the gain cranked up to maximum, and you will see a random sequence of ones and zeros (i.e. NOT just e.g. a solid low) coming out from the data output. A data byte getting received at this moment would not be processed correctly because the receiver would still react to both signal and baseline noise. The preamble now "conditions" the input stage so it reduces its gain to a proper value and thus ignores the noise.
How many preamble bytes you need depends on a variety of factors - actual data rate, type of receiver, noise level and so on - in my experiments I saw 2-3 when running at 4800 baud, so I normally use 5 bytes to be safe. A scope or logic analyzer is a big help for such investigations.
Second, the receiver chain is AC coupled so it can't transmit a long sequence of all 1's or all 0's - your data must switch polarity sufficiently often and have approximately equal numbers of 1's and 0's over every not-to-long stretch of bits. Manchester coding achieves that naturally since it encodes each bit in a single 1 plus a single 0. There are other schemes that achieve the same thing (e.g. 8 bit / 10 bit encoding, 8b/10b, used in modern Gbps SerDes channels) with less overhead, but Manchester coding is by far the simplest way.
Finally, a marker byte (or bytes) tells the receiving Picaxe where the actual data starts.
To sum up, the preamble is NOT optional, and it is NOT just some "snake oil" miracle cure - there are solid reasons why it has to be used. Second, preamble and Manchester coding (or some other way of achieving DC balance) are two independent conditions to achieve correct data transmission. Please also read carefully through the old thread of mine that Westaust already mentioned,
http://www.picaxeforum.co.uk/showthread.php?11531-433-MHz-wireless-link-trick&highlight=Preamble, as it explains some more intricacies (pause after preamble to assure reliable sync to byte boundaries).
So a reliable transmission protocol - tested in practise many times - looks as follows (and NONE of the four steps is optional!):
(1) preamble - conditions the receiver's input amplifier
(2) short (~15 bit periods) pause - assures receiving Picaxe will correctly sync on start of serial byte
(3) marker byte(s) - lets Picaxe detect start of actual data Packet
(4) data packet (Manchester encoded if it is longer than a few bytes so receiver's gain stage can't fall out of lock)
Wolfgang