Wideband Accuracy
Turbo Mini Forums

Following on from Graham T's thread "1312cc Ti", where there was a frank exchange of views about the accuracy of the 0-5v output from wideband gauges. Opinions are obviously polarised, but there does not seem to be anything other than anecdotal evidence mainly from people on the MS Forums reporting differences in the AFR reported by the gauge and that displayed in logs or Tuner Studio. I don't know if other tuner forums are reporting similar issues.

The differences in reported values can be explained in most cases:

Ground offsets - the gauge/controller is earthed to the car bodywork rather than a common earth point.

Incorrect scaling - TS has not been set up to correctly convert the -05v signal to AFR or Lambda.

For the rest, the error is attributed to poor conversion of the digital data derived from the Wideband controller into a 0-5v signal and then back to a digital value in the ECU, known as DAC - ADC conversion. Then there are potential errors due to signal noise and voltage source.

I'm happy with the ADC end of the conversion, provided it is set up correctly. The MS ECU uses the same architecture to derive variables for IAT, CLT & MAP. They are far more critical than the wideband value as they are used to derive the amount of fuel, the wideband being a monitoring instrument in the majority of cases. If electrical noise is an issue, then you would expect to see it with the IAT and CLT values, after all they are derived from a variable voltage from wires strewn across the engine bay. The MS MAP sensor is not subjected to noise as it is mounted on the main board of the ECU, however it relies on accurate ADC conversion.

So, my thinking is that the main problem is with the DAC conversion of the wideband controller which may also be vulnerable to variations in power supply or electrical noise.

I am at a point in my latest builds where I can lay my hands on brand new AEM and Innovate wideband systems. So I propose to carry out a bench test using variable supply voltages running my MS3/X on a JimStim, a laptop or 2 running Tuner Studio/Logworks/Hyperterminal to display both the derived AFR from the 0-5v analogue signal or the serial digital values. The sensors will be installed in a fabricated sample chamber.

To test the 5v or lean value, I will run the sensor in air. To test the rich value, I plan on using unburnt hydro-carbon in the form of LPG. I'm not sure what result we will get from LPG but as long as the results are consistent, then that's not an issue, or is it?

Fire away if you think that I'm chasing my tail.

Saul Bellow - "A great deal of intelligence can be invested in ignorance when the need for illusion is deep."
Stephen Hawking - "The greatest enemy of knowledge is not ignorance, it is the illusion of knowledge."

interesting, thoughts,

I havent read the other thread completly, but i always had the VEMS WB guage and a display from the MS on the 7 port, so was able to see if the figures stacked up, normally within about 0.1 AFR, which I put down to calibration of the analog output, also I never really saw any sharp spikes that I thought were noise ( after I didtched the LC1 anyway...)

If accuracy was a concern I'd look at CAN busing the Wideband signal into the MS, i seem to recal on the MS3 you can stick in a bunch of WB's via CAN, i think its doable on the MS2E code too but havent looked.

Yes, interesting thoughts.

My first point would be that, assuming we are talking about a Megasquirt based system (MS2 or MS3), there are only two makes of wideband controllers that can communicate digitally with the MS, and then only through one of Jean's IOx variants unless you are going to build your own secondary microcontroller (eg, Phil, arduino?) and write your own CAN code.

So, keeping it simple it, Innovate or 14point7

Others can talk digitally to a PC/laptop via their own serial protocol but not to an MS/IOx.

So unless you use a controller that can talk to an MS digitally and analogue, you won't be making the comparison on a common platform.

I have seen the difference on my (now unused) TechEdge units but I don't have it recorded for the reasons above. The TechEdges have digitally driven LED displays, or a digital serial link to their own software on a PC/laptop. Once their software became incompatible with TS, I had to put their 0-5V output into the MS setup via IOX analogue inputs because of lack of ports on the MS2 (my previous BluRay/DVD analogy) and there was always a 0.1 - 0.2 difference. I'm convinced I had no scaling/calibration issues on the analogue and my grounding regimes are as good as the manual, if not better.

Now that I have swapped over to 14point7 SLC-OEMs I too can do what you propose - I have enough spare inputs on my IOx-OEM to display the analogue data as well as the digital.

Watch this space.


Re. your test proposal, I don't think air vs propane is a valid test - they are just extreme ends of the scale.
An engine running lambda 1 (14.7 AFR) would be more valid,

Schrödinger's cat - so which one am I ???

Edited by Paul S on 25th Oct, 2014.

Saul Bellow - "A great deal of intelligence can be invested in ignorance when the need for illusion is deep."
Stephen Hawking - "The greatest enemy of knowledge is not ignorance, it is the illusion of knowledge."

Schrödinger's cat - so which one am I ???

I installed an AEM controller / display on the dashboard and used the 0-5V ouptut to talk to Megasquirt.

With Tunerstudio open it was clear that both displays were perfectly in sync, but there was a big offset at each end of the scale, maybe 1-2 AFR. I just played with calibration in TS until it was closer to 0.1 AFR difference. After several weeks of tuning, I never noticed the 2 readings drift further apart than this. I'd calibrate any other 0-5V signal (TPS etc) so it only seems normal to do the same with the AEM.

EDIT:-

My CAN bus turned up this morning so will be having a play with that shortly. I need 3 inputs in total, but I'll probably keep the original sensor plugged directly in to the MS.

Edited by PhilR on 25th Oct, 2014.

Good point.

If you calibrate the analogue 0-5V to match the digital output then I guess obviously it will read the same.

My comparison was based on using the standard ratio supplied by the manufacturer (ie, 0.5V =x, 4.5V =y) that is already in the tables and then looking at the difference.

Doing your own calibration will obviously fix the difference.

But how many people would do that ?

Again, simple or complex

Schrödinger's cat - so which one am I ???

I had the gauge on the dash to compare them side by side, so it was obvious from the start. With no gauge, most people would make a huge assumption that the figure their ECU sees is correct - not good.

I looked up a few calibrations and tried several, but annoyingly I couldn't get any as close as they should be. In the end it was just quicker to trial and error the calibration points until it looked right.

Edited by PhilR on 25th Oct, 2014.

Lifes too short to worry about 2%.

Scrap that idea.

Saul Bellow - "A great deal of intelligence can be invested in ignorance when the need for illusion is deep."
Stephen Hawking - "The greatest enemy of knowledge is not ignorance, it is the illusion of knowledge."

If you KNOW you only have 2% and are happy with it, that's fine be me.

It was you who posed this new question.

I would say I'm at the best end of the scale.

Others who just accept the factory analogue scaling and don't have all the other good points you mentioned earlier (ie, ground offsets) may not be in such a good position.

But it was your question.

I've just given the best answer I can at this moment in time.

Schrödinger's cat - so which one am I ???

The way I was rebuked yesterday for suggesting that you could just connect the 0-5v directly to the ECU gave me and possibly others the impression that the level of inaccuracy was far greater, say 5-10%, in which case it would be worth doing something.

An tolerance of +/- 2% will have absolutely no impact on the way you set the car up or its performance.

Saul Bellow - "A great deal of intelligence can be invested in ignorance when the need for illusion is deep."
Stephen Hawking - "The greatest enemy of knowledge is not ignorance, it is the illusion of knowledge."

At the risk of repeating myself I said I have seen +/- 2%

And that was not logged for the reasons I have explained.

Others have seen more.

But it's the usual case of no actual recorded data to plonk up on a graph.

I can put some up in a few days (failed injector O ring at the moment) but I doubt mine will be more than 2% out because my grounding/wiring is good.

I'l bet I could make it 5-10% by moving a few ground wires but why should I ???

As Graham has already mentioned, the digital data needs no filtering, it is stable, clean, and very fast.

If people have the hardware to do it that way, why decry us because we do ???

You keep going on about making EFi simple, well yes, it can be simple but don't keep knocking people who want to make it a bit more accurate at minimal cost.

Enough said unless you want real facts.

Schrödinger's cat - so which one am I ???

The way you were rebuked?

If you look at Phil's post above, he does mention 1-2 AFR. That is huge. I don't think it is usual for other controllers but seems quite common with AEM controller. Of course, that's apples and oranges if you need to change the controller to use the digital CAN AFR.

http://www.jbperf.com/

As for simplicity, there is a controller being tested that will only require a CAN connection and it will automatically detect if it is connected to an MS2 or MS3 and send the data to the correct location. Settings on the MS2/MS3 will be minimal and the same as they would be for a 0-5V connection.

http://www.jbperf.com/

I have a blitz ecu display in the skyline along with an aem in narrowband mode and at certain points i was getting the blitz saying 0.85lambda and the aem 14afr
This was usually as im falling into a cruise the lambda catches up to the afr and then it settles into a 15.5 afr cruise the ecu reports a lambda just over 1 ( as set by me)
As i had a second aem i swapped them over ( gauge and sensor but not wires)
The second one ( oldest one off the mini ) doesnt do this
The first one is now in 0-5v mode and working fine on the mini :) lol

Yes i moved to the darkside

Instagram @jdm_brett

To confirm, the 1-2 AFR difference that I saw was due to calibration. Figures around 14.7 were spot on, then either side of stoich the AFR got further out: +2 AFR at one extreme and -2 AFR at the other end; Not something that I'd put down to a connection problem.

There are many brands on the market, but no good reason to expect them to conform to the same 0-5V scale. If you use the 5V output, the need to calibrate is very likely. With a digital output, the need to calibrate goes away, but (at least for my install) that's where the advantage ends. I won't worry about inaccuracy any more than I would for any other 5V sensor, MAP , TPS etc.

One issue with wideband controllers is that they are not ratiometric devices like all the other standard sensors used with the ECU. And since they have their own 5V reference, if the manufacturing tolerances are too large (like they seem to be on many AEM units) you end up with a wrong calibration. Having to create your own calibration curve instead of using the manufacturer standard curve does not make sense and is not something everyone will be comfortable with or even able to do.

Another issue is that some controllers only have a single ground for both the signal and the heater. The heater circuit can be quite noisy which can lead to a noisy signal. Also, you can't connect this ground to the ECU since it is too noisy and that may lead to a ground offset.

These things are why I don't like the AEM and I'm not fond of the latest Innovate unit. The first is that the analog output seems to require a custom calibration in a lot of installations and the second only has one ground wire which seems to be problematic for many.

But I don't have first hand data for either so I can't say how inaccurate they are nor how easy/hard it is to correct to an acceptable range.

However, I think it is better for users and potential users to know that these issues exist and that simply plugging the analog signal in the ECU might not be as straightforward as it is for other sensors.

Edited by jbelanger on 25th Oct, 2014.

http://www.jbperf.com/

Edited by jbelanger on 25th Oct, 2014.

http://www.jbperf.com/

Edited by PhilR on 25th Oct, 2014.

I don't know enough about the way the sensor works to say exactly what goes on with the butane but I do know that it is a bit more complex than that. The sensor actually "breathes" O2 from its wiring loom and the cell pump current to come to a Lambda 1.0 mix in the cell is what the controller uses to compute the Lambda value at the sensor.

And I know that the controllers are now using linear calibrations but I used the curve term as a generic term as there were some controllers with a non-linear output (and there may still be some).

I also don't know how AEM does it but you're not the first and only one to report huge mismatches between the official AEM calibration and what is needed to get the same as what their own display gives. It has been assumed that the display is correct so they must have an appalling lack of quality control and/or a very bad output circuit design to get this that bad. Or their gauge design is flawed but then they have a bigger problem and so do their customers.

Edited by jbelanger on 25th Oct, 2014.

http://www.jbperf.com/

It might be of interest to those who want to see more about how the sensors work: http://www.megamanual.com/PWC/LSU4.htm

So both CO and H2 are involved in addition to O2.

http://www.jbperf.com/

Surely the sensors themselves have a tolerance that would need factoring in?

I run a supercharger and I don't care the TB is on the wrong side.
VEMS + 12 PSI + Liquid Intercooler = Small Bore FUN!

There is a laser etched resistor in the connector. That should be used by the controller to adjust to the sensor. Some controllers do not use it. I know the 14point7 controllers do make use of it. Others claim that free air calibration is sufficient.

http://www.jbperf.com/

So would the same idea follow that some of the tolerance could also come from the controller hardware and code then?

I run a supercharger and I don't care the TB is on the wrong side.
VEMS + 12 PSI + Liquid Intercooler = Small Bore FUN!