MembersPage/KarlBuchka/Trigger (2008-09-29 03:16:19)

Inline 5-cylinder, sequential ignition with camsync

Primary trigger: 60-2 w/ VR sensor.

Secondary trigger: Single tooth w/ Hall sensor.

The rising edge of the cam wheel hits the sensor after cylinder three fires and before cylinder one does.

Firing order is 1-2-4-5-3.


MembersPage/RonnieH has set up VEMS on a very similar motor also with seq. COP. After some additional reading, his settings seem obsolete due to code changes.

Primary trigger settings

Field name Variable name Value [decimal] (Value [hex])Comment
Primary Trigger primary_trigger 1 (0x01) For multi-tooth trigger without filtering (I think). 0x09 is with filtering? Which should I use?
Edge   Rising
Type   Multitooth
Filtering...Bit7   Disable
TDC after the trigger ign_tdcdelay 60 (0x3C) The GenBoard needs to know ahead of time when a cylinder is going to be at TDC. This is to allow for a certain amount of ignition advance. It needs about 60 degrees of 'lead time'. ign_tdcdelay is this lead time. Look at the tooth under the sensor at TDC of cylinder one (16 in my case), count backwards toward the missing teeth until you are ~60 degrees BTDC. In my case this is exactly 10 teeth and 60 degrees.
Number of teeth on wheel tooth_wheel 58 (0x3A) 60 "total teeth" - 2 missing teeth = 58 actual teeth
Trigger tooth trigger_tooth 5 (0x05) TDC of cylinder 1 (tooth 16) occurs 15 teeth after the first actual tooth following the missing teeth. 15 teeth is 90 degrees. ign_tdcdelay is already 60 degrees, leaving 30 degrees (or 5 teeth). This is trigger_tooth. Confusing as shit, I know. I'll post an illustration. More concise explanation here: InputTrigger/MultitoothTriggerWheel
Next trigger tooth another_trigger_tooth 24 (0x18) (from GenBoard/UnderDevelopment/FirmWare/TriggerRelated, 60 (teeth on the wheel) * 2 (wheel revolutions per cycle) / 5 (ignition events) = 24 teeth between events
Crank min. period crank_minper 188 (0xBC) 60 (seconds) / 8000 (max rpm) / 5 (number of engine events) * 2 = 0.003 (seconds) = 3000 (microseconds). In units of 16usec, so 3000usec / 16 = 188 [decimal] = BC [hex]. Can someone confirm this is the correct formula? Or even what this value means?
Tooth normal relative min %    25%
Tooth normal relative max %    150%
Tooth missing relative min %    200%
Tooth missing relative max %    413%

Secondary trigger settings (these are not used in 1.1.x, so only here for future use)

Field name Variable name Value [decimal] (Value [hex])Comment
 secondary_trigger (0xFE) For rising edge trigger
 tooth_wheel_twidth1 24 (0x18) Width of a single tooth on wheel, in quarter degrees (360/60 = 6 degrees per tooth, 6*4 = 24 quarter degrees per tooth, 24dec = 18hex)
 tooth_wheel_twidth2 0 (0x00) Not sure on this one either, changelog says set this to 0, but why?
Rising Edge cam_sync_r_edge_phase 2 See: Note
Falling Edge cam_sync_f_edge_phase 6 See: Note
Engine phase, when to reset reset_engphase_after (0xF0) See: Note

Note: According to GenBoard/Manual/InputTriggerCamSync


Re-iteration for this section: The rising edge of the cam wheel hits the sensor after cylinder three fires and before cylinder one does.

Firing order is 1-2-4-5-3.

Ign. outputs will connect as follows (for 2+6 configuration, according to PhatBob's User Guide):

Cylinder Ign. channel ECU Pin
1 00 EC36/35
2 10 EC36/33
3 40 EC36/11
4 50 EC36/12
5 60 EC36/24

h[2] is ignchannel array, h[2] = 00 40 60 50 10 00 00 00. So the array is traversed backwards starting at '10' (cyl. 2) (since "Ignition outputs to use in h[2]" will be '04..00'). Because the ignchannel traversal is reset after the ignition event following the cam-sync, cylinder one needs to fire last in the order.

Differently stated: h[2] = 00 <cam sync occurs here> 40 60 50 10 <traversal begins here> 00 00 00

I drew a pretty picture:

60-2.png

According to the changelog for 1.1.x: h[1] is the reference tooth array (trigger_tooth and ign_tdcdelay is added to this to get TDC), for each output specified in respective h[2] entry.

h[1] = 01 61 49 31 19 00 00 00

Confirmation on the above values would be greatly appreciated.