The tech-puzzle part: AVC-LAN integrity vs. DICE iPod thingy

This is the gearhead portion of my DICE iPod-thingy-in-Classic saga. As we saw there, it seemed to give troubles not unlike what others have often reported in 2004+ cars (where it is officially supposed to work). Delayed/erratic response to controls, delayed/erratic on-screen displays, occasional disappearance as an available device (control response goes back to "The external system is not connected"), delayed/erratic behavior of other MFD functions like MPG display. I noticed codes in MFD LAN check mode; I haven't been able to find threads where people with DICE problems in 2004+ have checked that, but it would be worthwhile.

This all seems consistent with the idea that the DICE unit basically supports all the right command set to work with the stock G1 radio and MFD, but the AVC-LAN communications are going all flaky. That's how the DICE unit gets control input and reports status back, and also how the MFD gets energy and MPG data to display. Noise or a lot of collisions causing dropped packets or retransmission delays would account for the observed weirdness.

That leads to a few possible diagnoses that at the moment I have no way to choose between. A scope or some kind of AVC-LAN protocol analyzer might be in order. Has anyone got experience with this kind of thing?

1. Electrical integrity of the LAN.

Apparently, a star is the usual shape of an AVC-LAN, with 120 ohm terminations at the ends of segments. The G1 wiring is a little weird that way, with the gateway ECU and NAV (when present) having their own segments back to the MFD, while the radio and switch cluster share one, and I'm not sure where a terminating resistance goes. It may all work because the wire lengths are short, but the DICE harness extends that same shared segment by about 0.7 m and doesn't seem to be shielded or twisted. If electrical effects are increasing the error rate and causing more packets to be discarded and retransmitted, that could be the problem. This could perhaps be solved by attention to the wiring and not be a problem in the DICE box itself.

1a. Bug in DICE harness?

At the DICE-box end of the supplied harness, the blue wire (AVC-LAN TX-) and a green wire join together at one pin. The green wire goes to pin 11 at the radio, which is (so help me) the dedicated Eject signal. I draw a complete blank on why DICE would do that on purpose; the Toyota diagrams nowhere show AVC-LAN circuits being mixed with any other signal like that; it makes me think 'mistake'. Does anyone from DICE still read these forums who could comment? I could see AVC-LAN interference coming from this.

2. Protocol correctness in the DICE box.

The AVC-LAN in the G1 probably has a lot more constant traffic on it (because of the MFD and gateway ECU chattering away about powertrain status) than in a more conventional Toyota the DICE guys may have tested on (where the LAN would be pretty quiet except for audio status updates and control presses from the driver). If the DICE unit has any bugs in sharing the LAN with other traffic--for example, if it doesn't follow the whose-turn-to-talk arbitration rules quite exactly--it might seem to work ok in another car but have problems here. Garbling when the DICE transmits simultaneously with some other device would mean the loss of both packets.

Under the lid, I see chips I don't recognize but I don't see a NEC uPD72042 IEBus-on-a-chip, so it's possible DICE did an implementation of the protocol from scratch, which would boost the probability that they could have subtly botched it somewhere. That would, of course, be hard to work around. (An "IEBus bridge" might do the trick, if such a thing exists.)

3. Maybe it Just Can't Be Done.

Could the existing Prius traffic on the AVC-LAN be heavy enough that additional DICE MD-emulation traffic just can't be accommodated? I'm going to rule this one out because it appears Toyota sold an actual MD changer in some markets that's reported to work fine in the G1.

Cut the GREEN wire! GREEN!

Well, for once, the simplest problem possibility seems to be it.

It turned out to be easy to confirm that I wasn't crazy in thinking that pin 11 was the eject signal. All it took was trying to play a tape while the DICE harness was in place (which for some reason I never bothered to try when I last futzed with this a couple weeks ago). Nothing but immediate eject. There's one thing you'll get for tying half of your LAN pair to the Eject button.

So, I made one quick little snip in the green wire at the DICE end of the harness (just where it's shown in the pic above), reattached the DICE and the iPod, and took it for a spin.

Rock solid.

Did a bunch of playing and exercising the controls in ACC, in IG without READY, in READY, and then driving up to Five Points and back. No playing glitches, positive control response, no DICE disappearances, no MFD energy/consumption display hiccups, no errors shown in the MFD LAN-check mode, just a whole board of OKs.

Sheesh, I wonder how many of DICE's wire harnesses got built that way and how much it ultimately cost them in support headaches and reputation. It wouldn't surprise me if the reported similar problems in 2004+ cars have the same root cause (though I can't check that myself).

There's a part of me that figures as soon as I post this update, the next time I get in the car the DICE'll mess up again. But at the moment I'm pretty pleased; there seems to be a good chance this nailed it.

-Chap