Occasional Sync Loss Pulling Away and Changing Gear

[Murdoc]

Hi,

I have a Locost with a fireblade engine that I have converted over to injection. I've got it running OK however I have a problem with sync loss, when pulling up to a junction then trying to pull away as I reach ~2000-3000rpm the revs die and come back, TS shows the rpm dropping to zero and back up. It normally clears after a few seconds of holding traffic up, if I'm moving dipping the clutch or changing gear seems to help. Sync loss reason is 2. I adjusted R52 and R56 back from mostly fully clockwise to no improvement. I failed to get a tooth logger as my laptop froze. I have taken some scope plots of the input and output of the speed conditioning circuit with R52 and R56 set to various positions.

This is as the MS was setup when causing problems (all traces are cranking pulses)



R52 and R56 both fully clockwise

On 0.08V off -0.52V

R52 still fully clockwise, R56 at 83% from fully anticlockwise (5 turns back from full clockwise on a 30 turn pot)

On 0.6V off -0.02V

R52 still fully clockwise, R56 at 66% from fully anticlockwise (10 turns back from full clockwise on a 30 turn pot)

On 1.66V off 0.9V

R52 83% from anticlockwise, R56 at 66% from fully anticlockwise

On 1.7V off 0.88V

I've checked that I have VR sensor polarity right, also the I/P capture is set to rising edge. They seem to be triggering fine, which out of the above is the best combination of R52 and R56 setting?

Trigger wheel is on the crank, missing tooth 12-3.

MSQ and two logs showing the behaviour are attached.
I'm very grateful for any help. Thank you!

[Murdoc]

Link to log file with lots of sync issues, it wraps around there are that many!

https://drive.google.com/open?id=0B1Pz9 ... 19iWWJWQ28

[anastos]

Where are you connected when measuring the VR signal? It appears to me that the raw signal is corrupt/ missing teeth.

Sent from my XT1585 using Tapatalk

[Murdoc]

Hi anastos, thanks for the reply. They are plots of cranking pulses the apparent missing teeth are because I was only letting the engine turn over a couple of times so I didn't flatten the battery to much. I had let off the starter.

I tried a few things and confused myself with the trigger wheel and ended up fouling the plugs but didn't realise this was the problem and lost a few days trying to work out why the engine would even try to start. With that fixed I managed to get a scope plot of when the problem was occurring and save it as CSV, the resolution isn't great as I had to capture a decent amount of time but it was enough to work out what's happening.
I have uploaded a spreadsheet of the data plus the post processing to google sheets (although the graphs are now a mess), I put some formulas in to count the pulse durations and total up the number of pulses per cycle.
https://docs.google.com/spreadsheets/d/ ... sp=sharing

The VR circuit is frequently seeing 9 pulses instead of 10 (12-3 wheel) and losing sync.


Example of a good pulse train


Example of a bad pulse train


You can see the last pulse gets missed. I've tweaked R52 and R56 a lot without improvement so I'm going to put a signal generator on it today and follow through the VR circuit. I'm still getting quite a lot of hysteresis when R52 is at 100k so I need to look in to that. There may be a poor connection. I replaced R52, R56 and C31 yesterday, I had a coolant leak that filled the lidless MS up and took it out of action until it spent some time bathed in MEK and spent time in an oven. The daugherboard needed attention under a microscope. I may still be fighting issues from that, which is why I replaced R52 and R56.

EDIT: Properly timed and non clipping plots.



EDIT2: Looking at that I wouldn't be surprised if there is a low pass filter at the op-amp causing it to miss the last, shorter, pulse.

[Murdoc]

Fixed it!

C32 was giving a big phase shift, may be oscillating with the VR inductance as I was getting an additional small pulse (although there is a fair bit of resistance in the way...). It's far cleaner now, night and day difference.
I'm going to check the value just in case I fitted the wrong part.

Is C32 there to help overcome the capacitance of D24 of help with forward recovery perhaps?

Edit: Just to tie this up for anyone looking in the future, the capacitor was the correct value 10nF. All other components were the correct values to the latest VR circuit. I think it must be a peculiarity of the VR sensor. I have a spare i might put on an LCR meter if i get a chance. I tweaked the threshold to around 2V and 0.6V of hysteresis and it works great.