Using O2 sensors to tune a running engine

[victorymotorsports]

I'm considering putting an O2 sensor in my exhaust system to aid in tuning. However, I have zero experience using O2 sensors and do not know their limitations and tendencies.

The current set-up is:

4.3L V-6 Even fire Chevy
VP C-10 unleaded racing fuel
4 bbl Q-Jet carburetor
6 into 2 stepped racing headers
Drag racing

My goal is to use the O2 sensor to help shape the fuel curve while testing on the drag strip (can’t afford dyno time yet). First shot at tuning will be to read raw voltage off the sensor, then possibly the raw signal into a data acquisition system.

The research I've done shows two main types of sensors are used, narrow and wide band sensors. The narrow is generally a 2 or 3 wire sensor that shows relative lean/rich conditions and are generally very non-linear. Rich is above 450 mV and lean is below 450 mV. You can use them to tune, but true AFR accuracy between two different "identical" sensors may show different readings for the same AFR, so when a sensor takes a dump it may be somewhat difficult to reestablish a baseline. Plus, leaded fuel poisons the sensor. Wide band O2 sensors are accurate and generally linear, yet much more expensive.

My first thought is to use a basic narrow, heated sensor and hook it up to a digital voltmeter to read in the car. I am concerned with the response time and/or the readings the O2 will give, mostly a very jumpy signal reading directly from the sensor.

What are you thoughts and recommendations?

 

[enginead]

One problem you may find with a switching sensor is the resolution (particularly on the rich side) is relatively poor. At lambda 1 it will read ~0.45V but lambda 0.7 (close to the rich limit of combustion) it only reads ~0.9V and as you mention the response isn't linear. In terms of precision, they tend to be pretty repeatable so I wouldn't worry too much about that (there are literally millions fitted every year to new cars). If you want accuracy though it's much better to go for a wide band sensor, although more expensive they are much better.

If you do go the cheaper way, I have done testing with the sensor wired directly to 12V for the heater (most sensors are heated) and the signal feed to an ADC (analogue to digital converter) with no significant problems, if the signal is noisy (from HT interference for example) you can shield the signal cable. Good luck.

 

[ShaunT]

Forgive me for getting away from the O2 subject here.
If you need tuning help on the Quadrajet side, I have tons of experience including an NHRA SuperStock national title.

Shaun TiedeULTRADYNE Arl,TX(stiede@ev1.net)

 

[victorymotorsports]

enginead,

Would a digital volt meter be the same as an analogue to digital converter? Also, what is "HT" interference?

Shaun,

What can you share about tuning the Q-Jet? Idealy, I think a constant A/F ratio throughout the operating range should be best for best performance. If this is true, can a Q-Jet accomplish this and how would this be done?

Thanks!

 

[ShaunT]

Carbs will never be constant like an injector. You can tune a carb to be what I'd consider excellent for a carb, but when the motor goes through a gear chang at WOT, a carb has a giant hole in it. It momentarily goes ultra lean. Smart injectors will not do that.

Shaun TiedeULTRADYNE Arl,TX(stiede@ev1.net)

 

[patdaly]

Victory, start here:

http://www.diy-efi.org/

Good info on WB O2s

 

[enginead]

Victorymotorsports,

The volt meter is essentially the same as an ADC but you can record or view an ADC input on a computer and thus look at it later and also sample at a higher rate than a volt meter may update at. The HT reference is High Tension as in 10,000V through the HT leads between the coils and spark plugs. This electrical noise is a common issue in current production vehicles and a lot of effort (and legislation) is required to ensure EMC (Electro Magnetic Compatability) requirements are met.

 

[Andy330hp]

Standard O2 sensors can be installed in millions of vehicles because they will all switch from high to low voltage at 14.7:1. No OEM, that I know of, try to do what you are. All they do is high=rich and low=lean. Sometimes the code will make it look as though they are, as most will mention 0.450 or 0.500 mV as if it truly corresponds to 14.7, but its just a threshold for switching. What more is that the curve gets moved around be the temperature of the exhaust. So, you you could make a change such as adding fuel which could lower exhaust temps and thus cause the O2 output voltage to decrease, which would suggest a LEANER mixture. Feel free to try it, you will probably get it to run pretty well, but it will not be optimized. I dont think you'll really be able to "shape" your curve as you desire.

Andy

 

[ShaunT]

Consider individually probing each and every exhaust port seperately with EGT probes and a digital set up.

Shaun TiedeULTRADYNE Arl,TX(stiede@ev1.net)

 

[enginead]

Andy,

There is a linearisation curve for switching sensors (see Bosch Automotive Handbook) which does describe lambda as a function of voltage output. It's how post catalyst sensors are used as fuelling trim. In terms of the effect of temperature on the shape of the curve I'm pretty sure that it's also a 'switch', ie. above a certain temperature (~300°C) they work, below they don't. I've certainly never experienced the characteristics of the sensor changing with temperature once they are warm enough to operate.

 

[Andy330hp]

My point is this: You cannot take that picture from the Bosch book, enlarge it, scale it off, and say "740mV=13.8:1" or whatever. You will at best be able to say "740 mV will put me somewhere between 14.3 and 13.5, give or take". Again, this is as I have heard, but I don't think that curve is well established and reliable. Only the switching point is.

As for the temperature effect, Heywood shows that changing exhaust temps from 900 to 1300F (a change which could easily occur from low to high load) will change the output a full 100mV. If you are only tuning WOT, then maybe the swing will not be as much, but I don't think you'll be able to maintain one exhaust temp throughout the range of your engine map

I do admit ignorance on the use of post-cat oxygen sensors, but I thought they were used primarily as indicators of catalyst efficiency. The response of a secondary sensor would be more gradual for a couple reasons:
1. The cat stores and releases oxygen in a cycle much slower than the engine's transition from rich to lean, creating a broader waveform
2. More constant temperature and pressure

Nonetheless, I still don't think eaxct voltages are measured for fueling changes. I think the ecu is still just looking to make sure that it is going rich-lean-rich, and not staying lean longer than it stays rich or vice-versa.

Andy

 

[franzh]

One additional item:
The PCM is looking for constant switching, but also the duty times at each switch. Ideally, we would like to see 50% duty times, indicating a stoich mixture. Duty times over 50% (about .450 V) and the PCM recognizes a richer mixture and begins to modify the fuel mapping tables to take fuel away (negative fuel trim) and vice versa. The sensor manufacturers purposely setup the output of the sensor to be non-linear during the middle of the output to indicate a change in fuel mixture for this reason.

Using an averaging voltage feature with a 2 or 3 second refresh on a digital volt meter with a switching sensor will yield some interesting results. You can also use the averaging feature on duty cycle. True, linear wide range sensors give a pretty accurate picture of what is happening, especially if you intend to run in high Lambda regions where a switching sensor is relatively inaccurate.

I have tuned many hundreds of engines at stoich with the $25.00 Bosch sensor and the DVOM. Since most catalysts are formulated for stoich mixtures, and comparing pre and post cat sensors is a function of both OBD-II and many tuners, it makes good sense to use a switching sensor.

When using a non-electronic controlled carburetor, switching is relatively non-existent, whereas an electronic carburetor or fuel injection constantly switches.

Franz

 

[enginead]

Hi Andy,

My point is that you can scale the linearisation curve of the switching sensor (it is done in ME7 as the whole closed loop fuelling control including rear trim is done in the lambda domain). The problem is, because of the relatively flat response either side of the switch point the resolution of the sensor is quite poor, ie mV/AFR is a small number, therefore to accurately describe the lambda you need a very fine quantisation of the signal (at least 12bit).

It's all a bit academic really as the engine in question is carb fed so the accuracy and repeatability are going to be far outside the levels of detail we're discussing. Interesting debate though

 

[Andy330hp]

I'll agree with that!

Andy

 

[FourJawChuck]

What is wrong with a type K thermocouple? I have a two channel calibrator that I have been using for at least 10 years now that does exactly what you are asking, O2 sensors are for factory cars. Forget tuning by A/F ratio and stoichiometric mixtures and start learning to do it by exhaust temp, you don't see O2 sensors in aircraft for a good reason...they are too slow and not linear in their response. It's not like you need cats on your race car do you?

EGT is the only way to go.

 

[enginead]

Is that the jingle of spurs I hear.....

AFR is very important (particularly in a racing application)as the max power occurs rich of stoichiometric (lambda 0.9) and not at a fixed exhaust temperature (which is very much a factor of spark timing).

 

[tstom]

I'm going with enginehead on the egt's we dyno a lot of engines and use egt's in every cylinder and we find them to be of very little use for fuel curve tuning I think as enginehead has said they are affected by too many other things We have seen engines with 1000 deg F egt burn pistons because everything was burned before it got to the probe and we have seen 1500deg F temps be dead rich because the timing was retarded

I do often wonder why they are used in aircraft Could it be they are useful in a constant speed application ?

Victory M.S. a carbureted drag car fuel curve does not nescessarily want to be completey flat

Tom

 

[techguru]

There is a slightly easier way, pick up an CO sensor. they make them just for tuning cars, less than 130$.

 

[franzh]

One other issue when using EGT for tuning engines:
Retarded ignition timing, low compression, out of phase cam timing, can all lead to a high exhaust temp, even with the final engine out fuel mixture at or very near stoich.
Using EGT to tune an engine is fine, as long as you have an exact temp where that specific engine is at stoich, and a scale to show the AF/EGT for that specific engine.
Using a CO meter is fine, but it does not accurately reflect a lean condition. I still like a real world O2 sensor, linear UEGO if possible, but a switching sensor at minimum. As far as not reacting quickly enough? A modern freshly seasoned sensor can show individual firing impulses on an O-scope. That's pretty quick. The last Type K I used took about 2 seconds to reflect a change of 100 deg F.
Franz