cold start issue, voltage related

[dave]

For some reason when the battery is cold, this thing will just not fire. There's enough juice to crank the motor, just for some reason low voltage is a no-go. As soon as I touch the charger on it's got enough extra to fire up.

At first I suspected the 5V regulator was the original LM7805 with 2V forward drop so ordered up a super low Vf regulator off digikey and went to put it in.. To my dismay it was already a 7834 in there. So the new one only improved things on that front by .1V and added some extra current handling capability. Needless to say that wasn't the issue.

Has anyone else had to tackle this issue? I'm looking for ideas. Does the crank sensor circuit run off 12V or the 5V?
Unfortunately I can't reproduce the problem until tomorrow morning after the battery sits for a cold night. I'll be sure to take a datalog.
Thanks in advance for any ideas on things to look at closely.

[dave]

http://msextra.com/doc/general/ms3v3schems.html
pg 3.
Looks like it all runs off VCC (5V) so that shouldn't be dipping on battery V.

[Matt Cramer]

[dave]

datalogs attached.

From what I can tell, my observations are correct in that there's something going on with the RPM and it's not sending spark. It appears when it doesn't work there's no RPM reading at all!

[dave]

pulled it out and turned both R52 and R56 full CCW. They both had a few turns to go. Car seems fine with it as far as I can tell. I'll see if it's any different in the morning after sitting all night.
I'll be out late and it's not THAT cold so this might take a few days to iron out.

[dave]

ok, had same issue this morning. it is NOT fixed.

Battery had plenty of juice to turn the motor but just would not fire.
Hooked charger up for all of 1 minute and it fired right up.

I considered just getting a beefier battery but that is just kicking the can down the road, this would still happen when it's below zero out there. It's not THAT cold yet. If there's enough juice to turn the motor it should just fire up.

I can't seem to find the voltage in the datalog? Could have sworn I've seen it in there before. Might have to drag a scope out there and see if the 5V is staying up.

Any ideas guys??

[Matt Cramer]

Try a tighter sensor gap.

[dave]

It's not adjustable. :/

[elaw]

The business with battery voltage not being logged is weird... are you running old firmware?

One thing you could do is put a multimeter on the power feed to the MS and see what happens when cranking - it's not as good as a log but may be better than nothing. If it stays consistently above 7 volts, the voltage probably is not the problem.

Also if your system supports it (I'm not that familiar with MS1), you could take a composite log while cranking and post it. That may help us diagnose the problem.

[dave]

I have not touched the firmware since I set it up years ago. Don't remember anything to be honest.
Maybe this is a tunerstudio thing? I installed megatune but it won't connect yet, have to remember all sorts of details it wants set up. I'll try to get that tonight just to see.

This problem may have been here the entire time, I just never drove this car in the winter before.

The multimeter is too slow to catch voltage fluctuations, I'll have to scope it. Will be tricky to get that set up to see this in the AM and still get to work! And my scope is ancient, no way to save and share what I see with it.

[DaveEFI]

pulled it out and turned both R52 and R56 full CCW. They both had a few turns to go. Car seems fine with it as far as I can tell. I'll see if it's any different in the morning after sitting all night.
I'll be out late and it's not THAT cold so this might take a few days to iron out.

The fully anti clock position of the pots is merely a starting point. Which may work for the majority of VR sensors - but not all. I've found R56 set to about 6-8 turns clockwise from zero usually works when the default doesn't. But I do check with a 'scope on Tsel.

[dave]

didn't scope last night, temperature dropped like crazy! This morning was 10ºF!! fingers were numb in 5 min just getting chargers hooked up to start it. This sucks!!!

It's definitely the same issue though. It's not reading ANY rpm's while cranking.. This time even the chargers didn't get it, had to jumper on a second battery! How strange.. Neither batt had much trouble turning over the motor, should just plain work. It's being so sensitive to low voltage.

Once it got going and the motor was even 1/3 warm it restarted no problem.

[DaveEFI]

Have you done a log of an attempted cold start? Will show if it's getting sync when cranking.

[dave]

Have you done a log of an attempted cold start? Will show if it's getting sync when cranking.

3rd post. ;)
I used the tooth logger once too but it wasn't very helpful. The crank signal is either 100% fine or 100% missing.

Riding with a spare battery in the trunk so I can double-up and start. Seems like a handy thing to keep in there even after I get this figured out..

[dave]

Already have a plan cooking to combat voltage dropout issue.. obviously treating the symptoms instead of the cause but when it comes to reliability I don't mind doubling-down.



The battery is in the trunk so there's probably some voltage drop there. I'll check to verify that there's a separate run for the electronics instead of sharing the high-current starter wire. Cause that may explainit.. Once it's back up above freezing. So probably tomorrow. Also need to get the probe in there and verify that this is indeed the problem. I'd be surprised if I was that stupid but it's possible! I'm more concerned that I DID wire it right but somewhere in the car I'm losing voltage. Might have to run battery voltage straight to the ECU and add a FET to switch, bypassing whatever other switches are causing voltage drop.

Sorry to keep posting so much speculation. I'm really hoping someone chimes in with similar experience and/or ideas to check and try.

[elaw]

I would try the voltage-drop angle first.

The MS won't cut out until its supply voltage drops below about 6 volts. 6 is an *awfully* low voltage... if the MS and the rest of the FI system are only seeing 6 volts, you're liable to have other problems like injectors or idle valve not opening.

I'd check carefully for drops both on the ground and positive side of things. And use a multimeter directly on the battery posts, not the connectors, to see what voltage the battery is putting out.

I understand your point about the multimeter not catching transients, but if the voltage is dropping due to starter drain as the engine comes up on each compression stroke (as well as drag in the motor due to thick oil) you probably will see that. Voltage dips due to starter load aren't usually in the microsecond range... in this case they're probably somewhere in the 10s if not 100s of milliseconds.

Speaking of... what oil are you running in your engine? If the electrical system is in top shape it shouldn't matter, but with cars as old as yours the electrics are not always like new. If you're running 20W-50 or something like that, switching to 0W-40 (or synthetic if you're running dino oil) could help by reducing the load on the starter.

[DaveEFI]

You need to do an ordinary log. That shows pretty well everything - including the battery voltage MS is seeing when cranking. But I doubt it is low voltage as such - because as you know the tach inputs and processor all run off 5v, so if the processor isn't seeing a tach signal when cranking it will likely be something else - like, say, noise of some description corrupting the tach signal.

Noise problems are often down to poor grounding, for example.

[elaw]

He actually attached logs to the 4th post in the thread. However for some unknown reason they're not showing battery voltage.

My first inclination in this kind of situation is to suspect a triggering problem. But the "cold start fail" log he posted contains a bunch of resets, which to me says voltage problem.

Of course knowing Murphy's teachings, once the voltage problem is sorted out there will probably be a triggering problem!

[dave]

yeah i dunno what's up w/ the logs. and I can't get megatune to connect so tunerstudio is all we get.

So I decided to man up and brave the cold. Gonna post what I have so far while I warm up.
This is what I got:

First off, I DO have a seperate wire for electronics, it's not sharing with the starter.

(During this entire test the car cranked just fine but no spark)
Voltage at the battery while cranking:

We start at a hair over 12V, plunge to 9V then it rises up over 10V.

different timescales:


On this slower scale you can clearly see the starter hitting compression loads. Stays up at 11V..

OK. Moving on.. This is voltage under the hood at the junction where the front-back cable runs meet the main harness. This is on the electronics cable, not the starter cable.


Aside from an additional blip when I hit the key it seems about the same. I suspect that blip and the increased fuzz is because I have the probe grounded to the nearest chassis point I could find. There is not a dedicated ground, just chassis, so that's the best we get. Starter noise will be there.

OK. Now we are inside the car. I have a terminal block mounted directly on top of the MS unit for the wideband to interface in, so easy access for +12, gnd, and AFR.
This is on that terminal block.



So here it looks like it dips to 8V and rides up to only 10V. A little lower than the others but still seems to me like it should be fine. Engine still turning fine.

Now. I did figure out a way to watch Voltage from TS. I set up the history gauge and took a screenshot.


This MS voltage looks like read a full .5V lower even while just sitting there. This may be calibration issue. I'll get a DMM out next time and see if it reads closer to the scope or to MS.
The 8V dip Seems smoothed out, the cranking V is around 9.5V here. If there is a problem here then it's happening on too fast of a scale to see here.

With the linear regulator Vf of .3, there should be PLENTY of juice to keep the 5V strong. And the circuit that's acting up runs completely off the 5V. If I'm reading things correctly. Maybe the crank sensor puts out lower signal in the cold? But then why would jump-starting the car work? I'm a bit confused about this.

It's getting too late tonight to get into pulling the cover on the MS, I'll likely need to solder on a bunch of probe hook points for this next part to see what's going on with this op-amp circuit. And have schematics printed, etc. plus it'll be in the 30's tomorrow, that's TWICE AS WARM.

[DaveEFI]

I use a bare single core loop at both Tach Select and Tsel on the component side. Convenient place to hook up a scope to see what the VR input is doing. You need to see approx 5 volt pulses at Tsel for MS to sync.