Help understanding sampling rate and filtering

I am new to all of this, learning as I go along - and grateful to everyone for their help. I state this to explain that much if not all is over my head. But I am learning.

The ECU log in my car samples everything at the same rate. I have a choice of 100, 50, 33, 25, 20, 10 and 1 Hz. The majority of my work, I do it at 50Hz. This works well on some sensors - but not so well on others.

I have been chasing some gremlins on my coolant sensor. I see some unexplainable spikes(really depressions) in the readings. On occasion, they are enough to create issues with the tune. Some have indicated that negative spikes are typical of a wiring problem. I have replaced the sensor multiple times, redone the wiring, tried different input channels - can't find it.

I've learned I can minimize the problem by turning the filtering on the coolant sensor way up. Yet I also read here that no filtering is really better. I can't get to the source of the issue, it seems this approach is only masking it.

Since I can't set the sampling rate by specific items, can anyone suggest a work-around? Or maybe some pointers on how to learn more about this so as to figure it out?

If I can't set my desired sample rate (for fluid temps, usually 1Hz), a very reasonable alternative it is to filter to desired rate or slower. Think of it this way, filtering helps you see the trend of the data without being confused by the noise. Now if the ECU is acting upon this data at a faster rate, then it does help to occasionally log the data at the same rate it is used. Particularly if you are seeing spikes in some table that uses the sensor input (like Coolant Temperature compensation).

If it's not wiring and not the sensors -- then you just might have found a bug in the ECU firmware. Best you can do is report this to the manufacturer, provide as much information and logs as you can, and hope they can fix it.

I know you're using the PE3 ECU, and I have heard they have some issues with the ECU firmware development / updating, so it might be something you just need to live with, or if too many similar issues arise, decide to change ECUs. This is usually a whole lot easier than it seems - often you just remove the wires one at a time from the old ECU connector and put a new terminal on and insert into the correct place in the new ECU connector.

Paul that's an interesting issue. I don't think I've ever actually never run into this without it being a sensor or wiring issue which got resolved.

I have dealt with some sensors which use old packard style connectors that have intermittent connectivity while nothing is broken or incorrectly installed per say, so I wonder if that's the sort of thing you're running into. I've replaced those with other sensors that have connectors which maintain more reliable pin contact to avoid the issue.

All that said, since water temperature doesn't change quickly, I would agree that filtering the value relatively heavily would be possible without valid data loss, but I only expect that to help you out if your errant values are present for a very brief duration.

If you log the vehicle at 100Hz to get a good look at the issue, how long do the errant values last?

Mike (and David)

Appreciate your comments and thoughts. It's great to have someone to help me talk my way thru this. So far, in chasing this gremlin, I've changed the thermostat, changed coolant sensors (4 different ones), changed sensor plugs (3 different ones), rewired to bypass a firewall bulkhead plug, just in case it has bad connectors. I haven't changed the ECU or the plug that mates to it. Starting to run out of options.

The sensor is a typical GM #15326386. The connector is a Delphi(Packard) shrouded GT150 2 way. I've tried both original Delphi parts and also re-pops with no difference in results.

When it acts up, it lasts somewhere between 1 to 30 seconds. I also see 30ms spikes where the reported coolant temp drops 10-30 degrees then right back to normal. It was pointed out that a typical drop in a value is more indicative of a wiring or connection break, which I have been trying to find, but without success so far.

The short ones at higher rpm really are no big deal. The issue is when you get a long one while at idle. I am seeing 60-degree temperature swing when the thing acts up. That's enough to throw off my fuel compensations and some Idle Air settings. Again at higher rpm, no big deal, but at an idle, it gets messy. And you can imagine what it does to my confidence level in wanting to drive the car 200 miles away from home.

I am talking with the Mfg. and trying some workarounds before we jump to a new plug and ECU. ECU is an easy change, but the plug requires changing 34 wires - which I really prefer to avoid. PE Ltd. has been great to work with, it is just frustrating since I can't definitively find what's causing it.

Again, thanks for your help - I'm not giving up, I'll lick this one yet. Your guys experience and comments has been very helpful in troubleshooting

Paul

It appears to be a 2 wire, resistance type sender?

From what you're describing, I was wondering if there may be something on a related circuit that's affecting the wiring potentials?

If another device, that operates on an intermitent basis, has a ground, or supply, that has a common point with the thermostat switches on/off that can easily fool the ECU into 'thinking' the resistance, and hence temperature, has changed. While a high draw device, like a cooling fan, may be most obvious, it could be something quite minor if the wire gauge is on the small side, and the comparitive devices are significantly different.

if you haven't done so, I'd go back over the wiring* and make sure the thermostat was independent of anything else.

*perhaps you could post a sketch, or schematic, for it - preferably as a jpeg, or similar?

Apologies in advance, I am not trying to be "funny":

Have you also replaced the female pins within the connector and the pin of the signal wire on the ECU-side when you made the two external wires?

Gord - way ahead of you. It is a two wire device. And it is 'isolated' from the other circuitry as it connects to the sensor ground circuit of the ECU, it does not directly connect to the vehicle ground. The circuits that also connect to the sensor ground are oil pressure, fuel pressure, TPS, MAP, Barometer, Speed sensor, and Crank & Cam sensors.

I have verified there are no other connections to the sensor ground circuit to be sure there isn't a ground loop somewhere.

DynoDom: On the sensor side, I have replaced the female pins twice. On the signal side, that pin is part of the ECU connector and is not easily interchanged. That is the one part of the circuit I am currently working on replacing.

Paul that's a far more severe issue than I was initially picturing. That style connector has been nothing but trouble in my experience, but that doesn't mean it's a problem in your situation so I have a few questions.

When there's a 60 deg change does that value drop to an error default value which would indicate a loss of connection?

When this occurs, could you key off, then on, unplug those other sensors and see if the coolant reading goes back to normal? If this has no impact, that narrows things down a bit.

Paul,

If it hasn't been done, you could also monitor voltage across both wires of the sensor, when backprobing at the ECU / the sensor. If you can log the voltage in the ECU as well, you could check if both values match when the temp value drops. As it seems to be a long "spike", you could use a multimeter, or if you have access to an oscilloscope, it is usually quicker to detect voltage spikes.

Mike

A little confused about the connector comment. GM has used this same piece on millions of cars.

This is a trace of the 30sec-60 degree excursion I referred to earlier. When doing this, I was working on some Idle Air Control settings. You can imagine what the radical temperature change did to all the fuel & idle compensations based on temperature.

The occurrence is very random and typically, never this bad. When it does happen, it doesn't throw any error codes. The ecu watches for out of range conditions and will default to a given temperature, but this error still looks normal to the ecu - the input is still between -20 and 250F.

I will have to watch for the condition to occur again while I am at idle and a condition in which it is safe to re-key to see what happens. I haven't had that opportunity. This particular occurence was one of the worst and it was about 30 seconds. All of the others are typically 3-10 seconds and it reverts back to what it should be. Again, at 65mph, 2200 rpm cruise, the 5-second excursion is no big deal. But at idle, and more importantly, my confidence level, it's a big issue.

Frank: Pretty difficult to probe the connectors or put on an oscilloscope since the issue is very random. As an example, on a recent run of 22 minutes, it happened once for 13 seconds, 14 minutes into the run. I was at speed so the symptoms were barely noticeable.

Paul,

For a bit of comic relief, I had this issue on my stock except tune 2001 Ford truck which comes with that sensor. Sometimes they work, but I don't use them. I don't want to assume that's your issue though, just present it as a possibility.

Back to solutions, without having the car on the dyno, your ability to test certain things while the issue presents itself are a bit limited, but...do any other sensors change in their readings while this is occurring? I know you said your sensor ground wire is discreet, but they're often joined within the ECU, so an issue with another sensor, wire, can still impact others.

You could still try unplugging the other non critical sensors to see if anything changes i.e. IAT, oil and fuel pressure, speed, baro. You may have to enter reasonable failsafe values for the vehicle to run well during the test.

On the chance what you're seeing is "real", where is the sensor located in the system? Any chance you're seeing air pockets? How brief is the one downward spike between the two long drops? That brief dip my do more to give us confidence this is electrical rather than mechanical.

Mike

None of the other sensors show any issues. In fact, I compare the IAT sensor which is essentially the same thing with a different housing and it is rock solid.

Air pockets is an interesting thought. Would an air pocket on that style sensor at 180 degrees cause it to drop 60+ degrees? I don't know the physics/characteristics of them.

Detailing the incident shown above, the overall length was 33 seconds. Within that span, there were 5 excursions

1st 8 secs long, 50 degree drop

2nd 320ms, 15 degree drop

3rd 200ms, 43 degree drop

4th 340ms, 48 degree drop

5th 14 secs, 62 degree drop

No type of pattern in the spacing or position within the span.

My next step is moving the sensor to a different channel in the ECU as a way to try and isolate any internal ECU issues. Could possibly a capacitor in the circuit be acting up, losing a charge intermittently, and the ECU responding accordingly? Just thinking out loud

Last step after that is to change out the ECU. Just trying to pinpoint the issue.

Yes an air pocket can cause a massive shift in the reading.

Using a different ECU channel is a great test, well done.

This morning, I tested it using a different channel as input for the temperature sensor. At the same time, I turned off the filtering, so it would run in the worst scenario possible. Two 45 minute runs, and the temperature was rock solid. I could trace where the thermostat opened and closed and the temperature stayed constant for these tests at 185.2 +/- 1.8 degrees.

I am leaning towards blaming a defective capacitor in the default cooling channel as the culprit. It is the only thing I can think of that would cause a shift like I was experiencing. Heat might cause a change in resistance but it would correct itself a few seconds later and I know the circuit board wasn't changing temperature that fast. I believe that typically, a semiconductor or discrete chip either works or it doesn't. The only thing I can point to is a change in capacitance that would throw the circuit off. Unfortunately, I don't have access to the wiring diagrams to dig into.

The fact it's run for 90 minutes without throwing the error tells me I am probably onto something. I couldn't go more than 10 minutes on the other channel.

I believe the next step is replacing the ECU

Love it. Great work Paul.

ECUs can often be repaired rather than replaced. I'd give contacting your dealer a chance to see if they can set a repair up. There may be a cost associated, but it's usually far less than a new ECU.