Open Source LSU4.9 Wideband Controller

[tmbryhn]

Hi guys!

A friend and I started playing with an analog LSU4.9 controller design bout' three years ago.
I finished it up implementing a micro for doing the pump_i and sensor temp reading, driving the heater FET and also makes a linearized output that corresponds with the Innovate 0-5V output for easy implementation on any aftermarket ECU.

Now I use this design in my homemade ECU and it works really well. I'm very happy with the closed loop controller, keeping the 300Ohm nernst cell stable without oscillations. It also makes the transition from AFR=9.5 to free air in 26ms, which I think is fast enough for tuning you accel enrichments, lol

Anyways, I have driven around with this and a genuine Bosch sensor for about half a year, but suddenly the sensor have started acting up. Don't know if my controller has ruined it, although it seems unlikely. I also have a gentle heater ramp-up from cold iaw. the datasheet.

it seems that the Nernst cell have become unstable. The heater once were 100% stable (PID) in room temp, but this sensor bounces all over the place, and the 450mV also fluctuates as the temp goes up down. Pretty bad oscillation, and the controller can't keep the temp stable at all. It also outputs infinitely lean when this happens (rails the pump current sense amp up to 5V).
Connecting a new sensor, and it's hunky dory

Can anyone chime in on why the sensor has bit the dust, and what the cause may be? I'm happy to provide more details.
Btw. wideband sensors rock!

[toc]

What angle is your sensor installed at?
I wonder if it's finally condensate affected?

I assume you've checked all wiring, as an intermittent wiring issue can cause issues.
As can PCB traces which oxidize over time...

[tmbryhn]

Bench testing the sensor with a different controller in the lab right now. same issue. Unstable. New sensor works fine

I have mounted it slightly face down, as the datasheet describes. It started happening when the weather got slightly colder during the night.
Do you think my ramp ut is to quick? I'm starting at bout' 17DC% for about 6 seconds (condensation phase), then i go straight to 60% and ramping to max 86% in about 5 sec. before the controller goes into regular closed loop mode.

Do you think I can check the nernst output voltage (450mV) with the sensor disconnected? It will probably range from 200-800mV (rich-lean).

[toc]

A quick test of the heating element would be to measure it's resistance.

Not having used widebands before, I can't be 100%, but I'm guessing that the heater is the failure point of these sensors..
Check the resistance of the heater element - expect a few ohms, not a perfect 0, and not open circuit, I'd say 2-10Ohms.

I also know there's a document out there that describes the ideal heater cycle, but truth be told, I think your one is fine - they wanted a gradual heat up over 15 seconds if memory serves (if you are too quick, it'll bust the sensor).

If the heater resistance is gone then that'll be the cause of the erratic readings (Exhaust gas heating the sensor up).

[tmbryhn]

hehe, the heater is obviously in order

The problem seems to be the nernst cell giving erratic readings, making both the pump current amp go bananas, and also the temp feedback.
Have you looked at my schematic?

[toc]

How do you know that the heater is 100% for sure?

I'm just thinking of points of failure, from that it might make better sense.
The nernst cell, I wouldn't think would be the likely failure point, the heater and ceramics would be though...

I did take a squiz at the schematic earlier, it's certainly well laid out.
I also like that you used a PIC.. But the FET being tied to a data line might cause it to turn on / off during programming..?
If not adequately heat sinked, it could cause problems for the FET.

[tmbryhn]

Thanks for the input.

Cool concern, but the programming is not a problem Happens in like two seconds.
The heater measures 3.5 Ohms, so that's good - was the first thing I did. That being said, the heat-up cycle works like it should. It's when it goes into closed loop mode that the controller can't get a steady reading on the nernst cell impedance through the temp sense amp. So when it tries to regulate, it over/under-shoots and the 450mV-output will not stay stable at all, thus making the pump amp go through the roof - see pump amplifier that is used for driving the pump cell with the 450mV nernst voltage reference.

This has worked flawlessly for months in my old VW beetle, hehe. Suddenly the sensor started acting up.

An N-FET driving an average of about 1A in PWM mode does not need heat sinking at all with its exremely low on-state Drain-Source resistance.
But for good practice it's a DPAK with a solid dual-layer-pad with heat transfer vias. I'm using the same footprint for all my DPAKs on the board of course, and I have drivers that source a lot more current than the heater driver.

What I'm really concerned about is how did this happen? Why has the sensor all of a sudden bit the dust? If either my heater strategy, condensation phase or anything else has ruined it, I need to change something since these sensors should have a much longer lifespan.

[tmbryhn]

Here's a picture of the unit.
The Sensor heater N-FET is circled out

[tmbryhn]

After some testing today with my propane torch (yeah, I don't have fancy lab gas lying around), I have made some new discoveries.
If i simulate a very low AFR value with my torch, say 10:1. The AFR drops like it should as a result of the pump amp driving the pump cell in one direction. But suddenly the pump amp gets sorta' latched to free air.
It seems that the 450mV nernst output can't cope or get high enough in response to the pump amplifier, thus resulting in that the output gets "railed" and give a false free air reading.

The Pump amplifier should try to keep both inputs on the pump amp equal (2,95V, that is 2.5V VGND+450mV), but if the Nernst cell does not seem to generate a sufficient 450mV signal in response to the O2 concentration change inside the sensor, consequently the pump amp will try to keep pumping the O2 into the sensor in order to try to stabilize the Nernst voltage to 450mV.

I don't think the diffusion gap is clogged, simply because the sensor responds rapidly if i blow on it, and after I have done that, I can play with it some more - that is until I expose it to a very low AFR (low O2 consentration), and the pump amp. goes high without the nernst cell keeping up again.

Any thoughts on why this sensor have all of a sudden started to play tricks on me after only a couple of thousand kilometers during half a years service?

[Dtm]

Here's a picture of the unit.
The Sensor heater N-FET is circled out

Very nice board! It is your DYI ECU-s? Can you take us more info about this? I think this ECU maked for VAG ECU-s replacing? (I see typical VAG ECU enclosure on a photo).