Lukasz Szamota-Nagy has interfaced a Triple Info Display to a 8051 microprocessor. Using the following schematic:

He also shared his source and a ".hex" file, which can be downloaded here: szamota-nagy.zip. An excel file was included containing a sampled waveform, produced by the 8051. A small part of the graph is also shown below, click the image to see the whole graph from sample 0 to sample 27000!!!!

In this picture you can see the start of a communication frame. It is slightly different than I have sampled it myself from the CAR600 Radio. Lukasz writes:

Hi Eelke!

Eventually it works!
I think, the main gist was, that my timing was wrong. As I change the timing, it works or not. I think you were just lucky if it worked for the first time. The communication can be neither too slow nor too fast. I send you my timings, unfortunately there are no time-labels, but just sample numbers. Nevertheless if you watch the timing carefully you can see that the let-us-say power-up-reset looks otherwise as on your timings. After the MRQ goes low, the SDA changes to the low too, then after a longer while the SCL goes low too. Then I release MRQ and it goes high. At this moment the TID "blinks" one time. After that power-up-reset still I couldn't write to the display, because there is a need for a "small reset". The "small reset" is when you put the MRQ low and after this the SDA follows it, then you release the MRQ. 
You can see this "small reset" between 4100 and 4200 sample. Now you can start the protocol sentence. You put the MRQ low - and as you can see it, the SDA doesn't follow it - then release MRQ, and so on. It goes smoothly. After I put the first message on the display, I tried to put more messages one after the other, and if the timing is too fast, then it will hang after 1,2 or maybe 5 minutes. I use the acknowledgment, that's why it hangs, without the ack it doesn't hang (it hangs too, but just for a while, e.g. till the next "data-packet"). 
With the 8051 I had no problem, it has actually quasi bidirectional ports. It also worked through the 7407 (Buffer with OC output), but this way I couldn't check the ACK.
You were right, that the MRQ goes low every minute. And there is also something like "time-out" on the TID. Every minute, about 7 seconds after the number of minutes changes, the display is cleared. 
I send you the waveforms as a self-extracting Excel file, the schematics and the asm source code. The asm source code just shows my name.
I wanted to make the TID driver quite smart, so if the string, you want to print, is too short, then it puts additional spaces. Otherwise, if the string is too long, it makes a "marquee' of it (turns it from right to the left). Of course it counts the parity too, and capitalizes the small letters.

Unfortunately, I'm not capable of sending you photos right now. But I will do it in the future. (I've got no digital camera :-( yet)
Nevertheless, I hope you can make a good feedback of the info I send you now.

Best regards, Lukasz