T87A Bins to Pins - Who's In ?

[kidturbo]

Straight CANbus only on the 8 and 10 speeds now.
Besides no P/N or Reverse indicator wires, they also dropped the TOS/VSS line.

Updates make the stand-alone swaps easier that earlier GM platforms.

[zack4200]

Anyone got some decent schematics of the various years to compare say 8 speed to 10?

https://lemon-manuals.org.ua/Chevrolet/ ... nsmission/

[Tre-Cool]

So i posed a question to Claude the other week after feeding it a defined T87A bin file in ghidra on what was needed to make it work with the E38.

The summary was essentially:
The CAN link between your E38 (270 or 930) and the 10-speed T87A should need calibration changes only, almost all of them in the TCM, with no code injection. The T87A still contains GM’s older engine-torque interface, the generation the E38 uses; it’s just switched off by calibration.

What’s already there

0x191, E38 to TCM (engine torque): the T87A has a complete receive handler for it. It decodes the Global-A DBC layout exactly: actual, airflow and non-trans torque, pedal, kickdown and the torque-request failure flags. Both 270 and 930 already send 0x191 every 12.5 ms.
0x199, TCM to E38 (shift torque request): the T87A has a complete builder for it, also an exact DBC match. It’s disabled because its transmit object is 255; its 25 ms period is already set.
Other frames the E38 reads: the T87A already sends 0x0F9, 0x19D (which carries the engine-speed request your rev-match uses), 0x1F5 and 0x4C9 in the layouts the E38 reads.

Calibration-only: the receive mailboxes are programmed straight from KaCANG_NormReceiveMsgObj, and transmit is scheduled from KaCANG_NormTransmitMsgObj plus KaCANG_TxMsgTimes.

So with that good news i ordered all plugs & pins i needed to make myself a new harness...

[gixxer92]

So i posed a question to Claude the other week after feeding it a defined T87A bin file in ghidra on what was needed to make it work with the E38.

The summary was essentially:
The CAN link between your E38 (270 or 930) and the 10-speed T87A should need calibration changes only, almost all of them in the TCM, with no code injection. The T87A still contains GM’s older engine-torque interface, the generation the E38 uses; it’s just switched off by calibration.

What’s already there

0x191, E38 to TCM (engine torque): the T87A has a complete receive handler for it. It decodes the Global-A DBC layout exactly: actual, airflow and non-trans torque, pedal, kickdown and the torque-request failure flags. Both 270 and 930 already send 0x191 every 12.5 ms.
0x199, TCM to E38 (shift torque request): the T87A has a complete builder for it, also an exact DBC match. It’s disabled because its transmit object is 255; its 25 ms period is already set.
Other frames the E38 reads: the T87A already sends 0x0F9, 0x19D (which carries the engine-speed request your rev-match uses), 0x1F5 and 0x4C9 in the layouts the E38 reads.

Calibration-only: the receive mailboxes are programmed straight from KaCANG_NormReceiveMsgObj, and transmit is scheduled from KaCANG_NormTransmitMsgObj plus KaCANG_TxMsgTimes.

So with that good news i ordered all plugs & pins i needed to make myself a new harness...

You and I are on the same path(s)!!

[Tre-Cool]

well i've only had a rwd output shaft and components for over a year sitting in a box in my living room. im in no hurry.

[greaseonmyhands]

Dunno if it would be helpful but I've had AI build this reference based on alot of bins and can logs from my vehicles. 2015 Sierra, and 2014 Silverado thats got 2018 e92a/t87a and 10 speed swapped. I'm new to alot of this but have been enjoying breaking it down and learning from it.

[gixxer92]

greaseonmyhands - Thanks for posting, it's a lot of careful work, and your captures are useful. I checked it against my own firmware decodes and raw HS-CAN logs from two trucks: a 2014 Sierra (E92 12663390 + T43 24270599) and a 2017 Silverado with a 2018 OS (E92A 12711165 + T87A 24288259, 10-speed). About 2 hours of driving.

What holds up

• Your wire results match mine: same missing frames on the T87A truck (0x191, 0x199, 0x2D1, 0x4A3, 0x0AB, 0x4A5), same wire DLCs, and the 0x189, 0x19D and 0x287 checks pass on every frame here too. Your 0x287 formula, (-counter) mod 4, held on about 200k frames.
• The TX-length array at 0x40940 matches what we have.
• Your 0x199 finding is right. I checked it independently on 24288259: the five request inputs (0x400111B4, 0x40010A9E, 0x40010714, 0x4001049E, 0x400117D6) are mirrored cells read through 0x239E90, and none of the 163 mirror-write call sites in the OS ever writes them. Enabling 0x199 in the cal would send frozen request fields.
• The 0x189 -> 0x18E loop works on both trucks. The ECM answers within about 5-9 ms, and when it reports "honored" its torque value sits on the request (median about 0 Nm, roughly +/-10-20 Nm spread). When it reports "lost arbitration", torque is below the request, meaning something else is holding it lower.

Where it goes wrong: the ECM receive side on 12711165

• The table you called T1 mostly isn't the receive path (you found that for 0x189). The real handlers are in the same cluster as 0x34C17C: 0x0F9 0x34C098, 0x197 0x34C29C, 0x199 0x34C2EC, 0x19D 0x34C408, 0x1A6 0x34C5AC, 0x1AF 0x34C764, 0x1F5 0x34C9BC. Through those, 0x0F9, 0x19D, 0x1A6, 0x1AF and 0x1F5 are all read properly, and the counters/protection on 0x19D and 0x1A6 are checked with fault counting. That reverses several of the "no data path" and "hard disagreement" verdicts.
• Your "Unit C" at 0x34C2EC is the 0x199 receive handler, and it's cal-disabled. 0x197's receiver is cal-disabled too.
• 0x34C578 isn't the 0x1F5 unpacker. It's the 0x19D timeout handler (clears nine cells, sets one flag).
• The ECM does have per-message enables: 0x40AF0 (75 TX + 96 RX u16, 0xFF = off), periods at 0x40634, RX min DLC at 0x408E0, timeouts at 0x406CA. 0x191 isn't on your bus because its period is 0 and its enable is 0xFF in the cal. Same for 0x4A3 and 0x2D1. So that one isn't an open question.
• In 0x19D the speed request is the low 10 bits. Byte 0 bits 7-2 are the counter and the two channel fields, so (b0 & 0xFC) is mostly counter.
• On 24288259, 0x1A6 byte 3 carries a 2-bit counter plus a 2-bit check (valid on every frame here). Your byte map is the 8-speed one.

Your 2015 Sierra

• A 2014 Sierra on T43 24270599 shows exactly your wire pattern: 0x1A6 as one constant byte, 0x189 and 0x4C9 at 8 bytes, same missing frames.
  On 24270599, 0x18E and 0x1BA are receive-enabled and the 7-byte 0x1A1 is accepted, so on that OS the ECM's torque broadcasts do land. If you can read your T43's OS number (HP Tuners shows it), we'll know whether that applies to your truck.
• 0x287 3-byte form: on 12663390, byte 1 bit 6 carries a drive-mode switch state to the T43. If your drive never used that mode, a "heartbeat only" frame is exactly what you'd see. A drive with it switched on would show it.

A tip for the AI-assisted side: have it check each "ECM ignores this" verdict against the actual receive handler before trusting it. That's where most of the wrong calls came from.

[greaseonmyhands]

Thanks for the feedback. I'll be sure to start doing that to get better results. I usually feed everything through a couple different agents too. I made that reference with Muse and had chat and claude both look it over. I've actually been using muse alot more than I expected too.

Also, nice to see another K2XX with the 6.6/10spd. Heck of a combo. You had it easy being a 2017. The 14/15s dont play as nice. Cruise doesn't work out of the box but thankfully a newer EBCM is enough to get everything in order. You should see my CTS wheel with working paddles and my C7 cluster. And its got ratchet console shifter.

Most of my motivation in diggin into this, is if I can get an e78 to work with the t87a without a gateway. It would be easy to put a 2500 harness into a 14-18 truck to run an LS and 6spd. But LS and a 10spd would be even better. I like LTs but LS sure is cheaper and simpler for bigger power.

[gixxer92]

You’re welcome! I’ve been using Claude and ChatGPT too, mostly to help me work through the firmware and CAN data I’ve collected. I’ve also been developing some tools to help map and analyze it. Some platforms are further along than others, but it’s been interesting digging into how the ECM and TCM communicate.

I’m working on the BCM in my truck too, hoping to get remote start working again before the snow flies here in eastern Iowa.

Your CTS wheel with paddles and Corvette cluster sounds like a nice setup! I saw customECM’s cluster too—that’s a cool mod.

Your E78/T87A is right along the lines of what I’ve been working through a similar integration with an E38 and an 8HP in a Bruiser converted Jeep, so I’m interested in the communication side of these swaps. Different controllers and transmissions, of course, but there are similar questions about what the ECM and TCM need from each other.
I've also still got my 6.2L and the 8 speed, along with a 1993 2WD dually my dad ordered new and am going to start working on integrating the engine trans into it. The working on changing it over to the newer2019+10 speed.
There is a future 8HP I need to get to go with the MaxxECU Race 2 sitting by my desk here going in my IROC at some point, but that is going to have to be built to handle a bit more HP than the Hellcat.

An LS and a 10-speed without a gateway would make a great swap. Do you already have an E78 BIN/HPT or any CAN/HPL logs from one that could be compared with the T87A traffic?

Sorry for steeling the thread! But if anyone needs help with anything I have lots of data around the T87