History of MembersPage/MattiasSandgren/IgbtBurn
Older Newer
2011-08-23 21:12:01 . . . . 53d82fd0.adsl.enternet.hu [ign-dual is sensitive to dwell settingS, and even healthy coils can be damaged]
2011-08-23 11:09:40 . . . . h233n4-aahm-a11.ias.bredband.telia.com [Sketchy IGBT problem, more info requested]


Changes by last author:

Changed:
A guy (Alfred) sent me an ECU where I replaced two burnt out IGBTs. The ECU controls a BMW M50, 6 cyl engine, 60-2 trigger, no cam sync, with separate passive coils fired in dual-out mode.
A guy (Alfred) sent me an ECU where I replaced two burnt out IGBTs. The ECU controls a BMW M50, 6 cyl engine

* 60-2 trigger

* no cam sync

** (this engine should have camsync, right? just not connected ?)

* with separate passive coils fired in ign-dual-out mode.

Added:
* can you change it so 4A fuse is used for each branch ? (can be changed to 5A after the setup passes all tests)

What could cause the IGBTs to overheat?

* using ign-dual-out often ends up that way: usually coils fry first, than if fuse is >5A IGBTs fry immediately after (because a 5A fuse can allow 20..30A for a short time).

** ign-dualout is only recommended for starting the engine first (while sectrig is not available) or limphome. (or when someone really knows what he is doing and makes sure the coils and other components are not overloaded)

* often the coils are damaged as well

* recommended to use either wasted-spark coil or cam-sync.

* measure the coil current consumption by voltage measurement over a 0.05 Ohm resistor (2 x 0.1 Ohm in parallel, 1.6W each). Good idea in any case, especially if exceedingly high currents are suspected - as in this case.

Reasonable answers are excessive dwell or failure to remove fuses during an upgrade.

The coils connected to the outputs are also dead, hard to say which came first but the coils probably died from the IGBTs short circuiting. Shouldn't a 7.5A fuse blow and protect the coils from that?

* no, it's not fast enough. 4..5A often saves the IGBTs from a shorting coil. But not always.

Could "bad" coils cause this?

The coils that handle normal operation simply are not designed for twice the frequency, especially COPs that live on harsh temp cyl head. From double freq the coil lifetime can drop from 15-30 years to 1-10 months.

Changed:
What could cause the IGBTs to overheat? Reasonable answers are excessive dwell or failure to remove fuses during an upgrade.

The coils connected to the outputs are also dead, hard to say which came first but the coils probably died from the IGBTs short circuiting. Shouldn't a 7.5A fuse blow and protect the coils from that? Could "bad" coils cause this? Should one use an even smaller fuse?

Should one use an even smaller fuse?
Added:
* Ign-dualout is a bad idea even with the stock M50 harness.

** It's very rare that an installer measures coil current. But a proper install is not sensitive to that (2.5 - 3msec dwell is usually sufficient). Ign-dual causes so much stress on the ign-coils that it's very sensitive to dwell (and therefore also to dwell MAP scale setting, and large errors in battery calibration)

** solution should be easy: it sounds like this engine has camsync. Connect camsync. Upgrade firmware might be a good idea (save the old config. We'll run some bench output-compare-tests to 1.1.96 if you publish old config before 2011-08-28).

Changed:
Information about the install is sketchy at best and I've requested full details and story. For one, the harness is custom made and the ground for the coils was connected to the chassis but had come loose. I don't know what came first.
----

Information about the install is sketchy at best and I've requested full details and story. For one, the harness is custom made and the ground for the coils was connected to the chassis but had come loose. I don't know what came first.

* while the ECU is sensitive to GND coming off, passive coils are not.

** Active coils: can be... depending on type. We've seen at least one VW 2x2 skoda active ignition transformer died because of "negative" voltage on the input.