There is a link for the loco command format under the "Links & downloads" section. This link does not describe the turnout format, and the link is dead for the moment. An alternative source for the loco format description is http://home.arcor.de/dr.koenig/digital/motorola.htm.

The solenoid command format is a little bit different. The information I will describe here is from a book written by Rutger Friberg.

1. Voltage levels.
The commands are coded by alternating between positive and negative voltage. The approximate voltage level on the track is +/- 20 V.
(If you use the booster connection output from the 6021, "positive voltage" is +5V and "negative voltage" is 0 V (referenced to "brown") ).

2. Basic time unit.
The basic time unit for solenoid control commands is 12.5 us (some Control Units seem to use 13 us instead).

3. coding of bits.
One bit is coded in 8 time units, i.e. one bit is 100 us (or 104 us). A "0" is coded as 1 time unit of positive voltage followed by 7 time units of negative voltage. A "1" is coded as 7 time units of positive voltage followed by 1 unit of negative voltage.

4. Packet format
Each command consists of two identical packets of 18 bits each. The pause between the two packets, and between the packet pairs, is coded as negative voltage. I think the pause between the two packets is 762 us but I am not sure. (It is 1525 us for loco commands, and I think it is the half for solenoids). The pause between the two double packets is at least 500 us. A command should only be considered valid if both packets are received and are identical. Once a valid command is received, further identical packets to the same unit should be ignored.

5. Packet coding
Bit 0-7: Decoder address. The address coding is "trinary" (see the loco command reference). The same coding applies as for loco addresses 1-64.
Bit 8-9: 00
Bit 10-11: 11 vs 00 for red vs green for turnouts(I do not know which code that is "red" and which is "green").
Bit 12-15: Address within the decoder. This is coded as "double bits" i.e. bit 12-13 is either 00 or 11, and bit 14-15 is also either 00 or 11. This gives 4 possible alternatives.
Bit 16-17: 11 for activation, 00 for deactivation.

Mikael,
to get an idea of what you can expect when looking at your oscilloscope, you could have a look at this webpage: http://people.zeelandnet.../zondervan/digispan.html
It shows only the shape of the voltage, not how the message is composed.

Sure thing.
I will experiment with it tomorrow. I already think I have a plan for how to decode the packets with a microcontroller instead of the usual Motorola IC.

Yes, there is different kinds of traffic going on all the time on the line. It is very difficult to trigger specificly on one kind of traffic. You would need something more than a standard oscilloscope to do that.

Even more success.
I've wrote a proper decoding program, that can decode all solenoid addresses simultaneously, so I can now realize my original idea: The intelligent signal. One that you only send a stop or go signal to, and green or green/yellow is determined by the signal itself, based on one or more switchtracks ahead. Or the signal that refuses to go green if another signal (perhaps from a sidetrack) is already green.
Or one button to control several signals. Switchtrack positions then determine which signal is actually activated. The possibilities are endless.

But none of this would have been possible without your help. Thanks a million guys.

John.... you any idea what these guys are talking about..all I can do is fit a decoder to a loco and press the buttons
I can add some extra lighting etc but thats about it. as to how it works I leave to Marklin.
David

It really comes together for me now. Here is a small video of two signals. They are close on the video, but imagine them farther apart, and with a switchtrack somewhere between them. The two signals does not have any communication with each other. They only have two power lines from the tracks wherever they are placed.
The front signal reacts to the other signal and the (for now imaginary) switchtrack as well, just by listening for the commands to them.
http://www.ejberg.dk/external/marklinusers/signal2.mpg

I have ordered some parts to make a few of these signal modules for real. Until now it has been running on a development kit.
I am thinking of putting a brake module in it too. And perhaps some sort of feedback to make the signal go red by itself when a train has passed it.