There have been lots of threads on the early turn signal cancellation problem when turning left. The common remedy was take the cancellation unit out and clean the contacts on the attached switch which is used to monitor the rotation of the steering wheel. I did that on my car back around 2012 and it didn't really make any improvement. After putting up with the problem for close to 8 years, I decided to have a closer look at the switch and the cancellation unit to see if there was something fixable. I removed the cancel unit and the attached switch assembly so that I could bench test it.
The switch assembly consists of two switches; A and B operated by a cam on the steering shaft. When you turn the steering wheel the switches operate in a sequence. The following is the sequence for turning left (turning right just switch A and B). The service manual describes the switch operating sequence indicating that when you go all the way through this sequence and reach the 8th step in the sequence (0X), this is when cancellation occurs. The implied operation is that cancellation will not occur until you have completed the full sequence which requires that you turn the steering wheel more than 120 deg and then return to less than 90 deg so that you have more than 30 deg of dead rotation (more than 120 to less than 90) where steering corrections will not cause a cancel (the angles are approximate).
A B
0 0 (straight ahead / less than 60 deg rot)
0 X (60 deg rot)
X X (90 deg rot)
X 0
0 0 (120 deg rot)
X 0
X X
0 X (90 deg - this is when cancel occurs)
0 0
The cancel unit has a relay in it which briefly closes to energize the release solenoid in the turn signal switch. On the work bench with the cancel unit powered up I ran through the switch operation sequence above where X indicates contact closed. The cancel unit operated exactly as described with the relay not operating until the 8th step. I tested in both directions and operation was exactly as advertised. This suggested that the electronics in the cancel unit was not suffering from a partial failure (visual inspection of the board showed no soldering / trace / capacitor issues). I then attempted to emulate the early cancellation problem that occurs on a large radius sweeping left hand turn where the steering wheel may never get much past 90 deg. The switch sequence was:
A B
0 0 (straight ahead / less than 60 deg rot)
0 X (60 deg rot)
X X (90 deg rot)
0 X (60 deg rot)
0 0
and sure enough the cancel unit would initiate a cancellation when it hits the 4th step. It does this for both the left and right hand turn sequences, its just that the early cancel problem is less obvious on a right hand turn. On initial investigation, it appears that whenever the cancel unit encounters the following sequence
X X (90 deg rot)
0 X (60 deg rot)
0 0
it will initiate cancellation. So, you turn the wheel to 90 deg and in steering if you rotate back to 60 deg during a correction it will cancel - the early cancel problem.
I don't know whether this is as designed and Honda did not elaborate on this in the service manual or there is a glitch in the logic table in the cancel unit which allows this operation. The cancel unit is constructed from very conventional Toshiba CMOS logic gates (TC4584, TC4538, TC4013). I wanted this to be a quick turn around project since I didn't want an extended down time on the NSX. I removed the switch and did the bench test one evening and had it back in the car the next day when it became apparent that this was a design problem, not a repair problem. This didn't allow for the time to reverse engineer the cancel circuit by tracing out the circuit board and fabricating a schematic which would then allow the creation of a truth table to determine where the logic glitch was that allowed the early cancel (or whether it was an obvious designed in feature).
Since this appears to be more of a design issue, if you want to emulate the full 120 deg of rotation sequence described in the service manual I think the better solution would be to replace the guts of the turn signal unit with a micro controller. Something like a Teensy would definitely work (but be computational over kill). An ATTiny would definitely have the power to implement the algorithm and is available as a one board controller for about $5. The only limitation is that the Tiny only has 4 dedicated A/D I/O unless you repurpose the USB I/O (which is permissible) that gives you a total of 6. The cancel unit has 4 inputs (A and B switches and right and left turn indication) and one output (relay). You would need some additional bits, diodes, couple of resistors, a relay and a circuit board for everything; but, cost wise it should be around $20. For beta testing it would be nice to have the USB port so that you could reprogram the controller without having to take thins apart. One of the things you could do with a controller is put in a minimum timer over ride so that one the contacts entered the XX state the unit could not initiate a cancel for X seconds which might help with things.
I haven't had to work with logic gates and truth tables since 1981 (notice the term had not opportunity - not my favorite activity). This winter if I am so inclined I might take a run at writing the code to work with the complete turn signal sequence in a Tiny or a Teensy.
Note that if your turn signal switch is dirty or the traces on the board in the switch that form the contacts are starting to wear at the tips this could make the problem worse. However, my switch is 20 years old / 75+ thousand miles and the contacts were showing no visible wear. Unless your operation has become really bad I would not be yanking the unit out to do a test clean. If your signals are failing to cancel completely it is possible that the switches have failed; but, its more likely that the actual relay in the cancel unit has failed (which is replaceable) or that the cancel solenoid in the turn signal switch has failed (mega $ to fix since the turn signal switch is only available as part of the combination switch on later cars).
The switch assembly consists of two switches; A and B operated by a cam on the steering shaft. When you turn the steering wheel the switches operate in a sequence. The following is the sequence for turning left (turning right just switch A and B). The service manual describes the switch operating sequence indicating that when you go all the way through this sequence and reach the 8th step in the sequence (0X), this is when cancellation occurs. The implied operation is that cancellation will not occur until you have completed the full sequence which requires that you turn the steering wheel more than 120 deg and then return to less than 90 deg so that you have more than 30 deg of dead rotation (more than 120 to less than 90) where steering corrections will not cause a cancel (the angles are approximate).
A B
0 0 (straight ahead / less than 60 deg rot)
0 X (60 deg rot)
X X (90 deg rot)
X 0
0 0 (120 deg rot)
X 0
X X
0 X (90 deg - this is when cancel occurs)
0 0
The cancel unit has a relay in it which briefly closes to energize the release solenoid in the turn signal switch. On the work bench with the cancel unit powered up I ran through the switch operation sequence above where X indicates contact closed. The cancel unit operated exactly as described with the relay not operating until the 8th step. I tested in both directions and operation was exactly as advertised. This suggested that the electronics in the cancel unit was not suffering from a partial failure (visual inspection of the board showed no soldering / trace / capacitor issues). I then attempted to emulate the early cancellation problem that occurs on a large radius sweeping left hand turn where the steering wheel may never get much past 90 deg. The switch sequence was:
A B
0 0 (straight ahead / less than 60 deg rot)
0 X (60 deg rot)
X X (90 deg rot)
0 X (60 deg rot)
0 0
and sure enough the cancel unit would initiate a cancellation when it hits the 4th step. It does this for both the left and right hand turn sequences, its just that the early cancel problem is less obvious on a right hand turn. On initial investigation, it appears that whenever the cancel unit encounters the following sequence
X X (90 deg rot)
0 X (60 deg rot)
0 0
it will initiate cancellation. So, you turn the wheel to 90 deg and in steering if you rotate back to 60 deg during a correction it will cancel - the early cancel problem.
I don't know whether this is as designed and Honda did not elaborate on this in the service manual or there is a glitch in the logic table in the cancel unit which allows this operation. The cancel unit is constructed from very conventional Toshiba CMOS logic gates (TC4584, TC4538, TC4013). I wanted this to be a quick turn around project since I didn't want an extended down time on the NSX. I removed the switch and did the bench test one evening and had it back in the car the next day when it became apparent that this was a design problem, not a repair problem. This didn't allow for the time to reverse engineer the cancel circuit by tracing out the circuit board and fabricating a schematic which would then allow the creation of a truth table to determine where the logic glitch was that allowed the early cancel (or whether it was an obvious designed in feature).
Since this appears to be more of a design issue, if you want to emulate the full 120 deg of rotation sequence described in the service manual I think the better solution would be to replace the guts of the turn signal unit with a micro controller. Something like a Teensy would definitely work (but be computational over kill). An ATTiny would definitely have the power to implement the algorithm and is available as a one board controller for about $5. The only limitation is that the Tiny only has 4 dedicated A/D I/O unless you repurpose the USB I/O (which is permissible) that gives you a total of 6. The cancel unit has 4 inputs (A and B switches and right and left turn indication) and one output (relay). You would need some additional bits, diodes, couple of resistors, a relay and a circuit board for everything; but, cost wise it should be around $20. For beta testing it would be nice to have the USB port so that you could reprogram the controller without having to take thins apart. One of the things you could do with a controller is put in a minimum timer over ride so that one the contacts entered the XX state the unit could not initiate a cancel for X seconds which might help with things.
I haven't had to work with logic gates and truth tables since 1981 (notice the term had not opportunity - not my favorite activity). This winter if I am so inclined I might take a run at writing the code to work with the complete turn signal sequence in a Tiny or a Teensy.
Note that if your turn signal switch is dirty or the traces on the board in the switch that form the contacts are starting to wear at the tips this could make the problem worse. However, my switch is 20 years old / 75+ thousand miles and the contacts were showing no visible wear. Unless your operation has become really bad I would not be yanking the unit out to do a test clean. If your signals are failing to cancel completely it is possible that the switches have failed; but, its more likely that the actual relay in the cancel unit has failed (which is replaceable) or that the cancel solenoid in the turn signal switch has failed (mega $ to fix since the turn signal switch is only available as part of the combination switch on later cars).
Last edited: