Introduction

Software such as that available on PGM-Fi.org and from Hondata make use of the Honda ECU's oxygen sensor input to allow real time display and data logging of wideband oxygen sensors. This involves feeding the analogue output of the wideband sensor's controller in through the stock (narrowband) O2 input. The known relationship of voltage to AFR is then entered into the software package you are using to allow it to log a real world value.

The wideband set-up I purchased a while ago from Innovate Motorsports features two independent programmable outputs with a maximum 5v range of operation. This allows you to set each output to supply a voltage of 0-5v in almost any linear relationship to AFR in the controller's operating range, typically 10:1-20:1 AFR. The most logical way to configure the unit would be 0v for 10:1 and 5v for 20:1, but there is a limitation with the stock ECU that prevents us from doing this.

This modification should not be performed unless you have a specific need to read from 0-5v in your O2 sensor input.

 

The Problem

The stock Honda O2 sensor input will only read to a maximum of around 3.8v, this does not present any problem when the stock narrowband O2 sensor is used as hovers about the 1v area in operation. While you could just configure your WB controller to output from 0-3.8v from 10:1 to 20:1 AFR, it would be a lot more accurate if you could read to 5 volts. Another problem is that many WB O2 controllers (like the AEM, TechEdge etc) do not allow you to alter the wideband signal output - this means any AFR above the 3.8v threshold will not be read! Various people in the past have used op-amps to rescale the 0-5v output of these wideband controllers to meet the restraints of the ECU, and while it gets the job done it would be nice to have everything done properly.
 

The Solution

To get around this limitation I started tracing the input stage to for the O2 sensor input, as most other analogue inputs on the ECU will read to 5 volts. A block diagram of the O2 sensor input stage of the USDM P28 ECU is shown below - I have checked many other OBD-1 ECU's and they all operate in a similar fashion.

 

 

Three main components are shown, an Op-Amp (NJM2904), Multiplexer (Mux) (MC14051) and the ubiquitous 66207, or more correctly one of the 66207's analogue to digital converter (ADC) inputs. The diagram is simplified and only shows what we are concerned with.

Op-Amp - as its name suggests is an amplifier, and allows the incoming signal to be processed, depending on its configuration.

The analogue Mux is best thought of as a multi-pole switch with many inputs and one output (which is fed to the ADC) as the switching action does not affect the signal on its way through. The Mux has pins that allow the ECU's processor to select which input it wants to read at any one time, and the ECU cycles through all of the inputs very fast - this allows it to take a snapshot of many different inputs, with only one ADC input. This saves money, as what would be one input channel now becomes many. The time the CPU takes to scan all of these inputs and get back to the start does not worry us here, as the signals it is reading do not change fast enough to be a concern - the CPU "switches" between inputs very fast compared to the speed of change of the input signal, so for all intents and purposes its as good as a direct ADC input.

The ADC is converts the analogue voltage of 0-5v in this case into a numerical representation of that voltage that the CPU can understand, and store in its memory. Notice I say 0-5v - this is because the ADC is actually configured to read from 0-5v although the O2 input is processed so to limit the maximum voltage the ADC sees to 3.8v.

If you want to read more about Op-Amps, Mux's and ADC's have a search on Google.

The most simple way to find out what the problem is would be to apply 5v onto the O2 input, and watch for it becoming altered along the route. That done it was discovered that the Op-Amp stage was limiting the voltage seen by the ECU. This being the case I bypassed the op-amp by lifting a resistor and applied 5v directly to the Mux, and without much surprise we now see the ECU reading to 5v (or as close as it will ever get) - great stuff!

That done, another problem reared its head. I decided to set the WB up in the car along with the ECU I was testing with side by side in the garage to check operation. I set my WB to read 0-5v over a 10-20:1 AFR range, the engine was started and was getting an approx 14.7:1 reading on the Innovate's LCD, and the same on the PC after some calibration. However switching to the graph mode, showed some anomalies - most notably the huge spikes on the graph? I stopped the engine.....and with the AFR reading a steady value on the Innovate WB's LCD - the spikes were still appearing - how could that be possible? Checking with a fast response voltmeter revealed the same story which, proved it was something relating to the wideband controller.

The graph above shows the AFR ratio spiking almost one whole unit, even though the engine was turned off at the time!

Next, I powered off the wideband controller took a break and came back. When I turned the controller back on again I noticed after the warm up period the AFR kept to a steady straight line.....for around 20 seconds it started to oscillate again. Interesting.......it was fine until just after the initial warm up period.

I decided to get the scope out and take a look at everything that could be a source for this noise, earth loops and the like. One of the first sources I thought the noise could be from, was an inadequate earth on for the wideband unit itself - my WB gets its supply directly from an unused accessory terminal on the fuse box, and a sturdy ground point direct to the body so I didn't expect to find anything wrong. I also checked the ground for the analogue output vs. the ECU's O2 input - again no problems there. However while monitoring the current drawn by the wideband, I noticed the fluctuations in AFR were exactly matched to surges of current drawn by the WB - now this was worth a look! The WB controller turns on and off the O2 sensor's heater automatically after the initial warm-up to maintain a steady temperature - so this was the source of the fluctuations.

I checked the supply to the WB controller - no voltage fluctuation on either rail. I internally checked the WB controller's DAC (Digital to Analogue Converter) also no fluctuation of the supply rails, so I assume its a problem with the WB controller's input stage from the O2 sensor, which the scope appeared to point towards.

So, to get round the problem I decided smoothing the O2 input of the ECU would be the easiest approach, I tried various values of Capacitor/Resistor combination until I found something that dampened the oscillations, but would not add lag to the changing input so that the recorded AFR would be skewed in relation to the actual AFR the WB controller was seeing. I chose to quickly modify the input stage as follows:

 

 

I added a switch to swap in and out the stock Op-Amp input stage, and a 10uf capacitor which dampens the oscillations perfectly almost to the point they are undetectable yet still gives a very crisp response to changing input levels.

So, when you want to hook up a wideband you can flick the switch to use the modified input stage and datalog as normal, or switch back to the stock input stage with ease. Of course - the 10uf capacitor can be omitted if you have a wideband with a noise free output!

Testing with the modified input stage produced excellent results with recorded and actual AFR's being spot on.