Wideband Accuracy
Turbo Mini Forums

Yes, they definitely have an impact.

http://www.jbperf.com/

Does anyone know about the PLX controllers health monitoring? Is this a gimmick or something that other manufacturers will copy?

http://www.plxdevices.com/product_info.php...WDBDSMAFR_DM6G3

If they only provide what is mentioned on that page, I don't see what good that is. How do they evaluate the sensor life? What should the response time be and what is actually measured? If they don't say, they're just numbers.

One thing they don't provide that is very useful is the sensor temperature. Might it be that they don't control it that well?

http://www.jbperf.com/

’77 Clubman build thread
http://www.turbominis.co.uk/forums/index.php?p=vt&tid=618189

Siamese 5 port EFI testing
http://www.turbominis.co.uk/forums/index.php?p=vt&tid=611675

Haha, the proverbial elephant in the room.

A few hours research on the net has revealed that we have missed a very important aspect of internal combustion engine science.

A wideband works by measuring the composition of the exhaust gas and calculates the pre-ignition AFR. For a wideband to work correctly it must be placed at a point where "chemical combustion" is complete, otherwise the chemical constituents of the exhaust gas are still dissociated and have not reached chemical equilibrium.

Research in Combustion Kinetics reveals that "chemical combustion" not only takes place in the cylinder, but carries on the exhaust port and pipe.

As stated by the sensor manufacturers, the sensor should be placed more than 0.5m from the valve.

Pre-turbo sampling at the head flange is not going to be very accurate.

PS For reference, LPG would not work as it does not contain any combustion products.

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 ???

Edited by Rod S on 26th Oct, 2014.

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

Steve Clarke also eludes to the issue in this video, after about 1/2 hour or so.

https://www.youtube.com/watch?v=ueqDFWti47g

Presumably, he was quite a way down the exhaust and still picked up variations.

EDIT 35 mins in.

Edited by Paul S on 26th 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."

just considering my current setup for a moment (two sample chambers with connections very close to the head and relatively short sample tubes and a third normal installation approx 0.5m after the turbo).....

This is probably going to sound very simplistic but even if the ones close to the head are reading wrongly, the difference between the two should be correct.
So take the average of the two, subtract it from the reading after the turbo and add that figure to the individual ones pre turbo.

The only caveat is my current sample tubes aren't identical lengths so one will have more of the continued combustion but if it's time dependant (rather than physical distance) that can be compensated for by changing the tube bore to make the two flow rates different.

Hmmm...

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

I think it is a bit more complicated, but not a lot.

Sampling at the edge of the port at the head flange is going to be sampling the flame front at WOT. The relatively cool port/manifold wall creates a gas boundary across which the chemical constituents will vary.

The difference in average cycle gas temperature due to the siamese exhaust port will also cause a variation in the chemical constituents in the inner and outer cylinders as dissociation is temperature dependant.

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."

On sample tubes, the difference in length is relatively insignificant.

There is enough differential pressure under load to drive the air through at critical (sonic) velocity.

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."

On a wideband controller there are many places for inaccuracies to hide.

Bosch Sensors are calibrated at the factory to be 0.01 lambda accurate.

On the wideband controller, dominate errors come from the; the ADC (offset, linear, non linear), Vdd specs (linear error), tolerance of the pump current measurement resistor (linear error). There are other sources depending on the design and the specific hardware.

It is not hard to design a controller that hits 0.01 lambda accuracy from my own experience. So lets say the controller is 0.01 lambda accurate.

If you take the RMS sum of the 0.01 lambda error from the sensor plus 0.01 lambda error from the controller, you will get a ~0.014 lamba error.

Now if the controller had a digital readout or a digital communications channel and you rely on either one of these then I will guess most controllers will yield a max error of 0.014 - 0.02 lambda.

Now if you are interfacing the controller with an ECU like MS then there is an entire set of other sources of error, DAC error of the analog output on the wideband (linear, offset, non linear), ADC specs of the ECU (offset, linear, non linear), ADC input protection circuitry (offset, linear, non linear), the Vdd reference on the ECU (linear).

Then you need to take into account grounding (mostly offset error but I guess you can argue that there is a significant non linear error in there too)

At this point, I figure doubling the prior calculated error is the right ballpark, so the max error will be 0.03- 0.04 lambda.

Generally if you have means to get the AFR/Lambda data digitally then you will be looking at roughly half the error vs using the analog input on your ECU.

These are just ballpark numbers, every controller design is different.

On most of my controllers I put in a calibration signal on the linear output that is 1.66v and 3.33v. Some of my products allow the user to trigger the calibration signal via USB, my other products output the calibration automatically on startup. Now with 2 discrete voltage levels, 1.66v 3.33v, you can calculate the offset and linear error based on the voltage seen by your ECU. So if you calculate the offset and linear compensation correctly and input that into MS, then you will get close to the accuracy of communicating digitally with MS.

Supposing you have a controller that does not output calibration voltages, the best thing you can do is ground the unit properly. The sensor heater is usually pulsed current in the khz range, with peak current ~3-5 amps. That is a significant amount of current, and I have seen it cause alot of headaches in many installations. On all my controllers, I separate the electronics and the heater grounds into 2 wires, you ground the electronics ground where your ECU and grounded, and the heater ground you ground it to the chassis. The heater ground can be a sh*tty ground and it does not matter as the sensor heater is just a big resistor so you can have a noisey or crappy ground and it will generally be fine. Some controllers have a single wire which combines the heater and the electronics ground, that is a really poor way to save costs.

There are many other considerations like pressure compensation, altitude compensation, temperature compensation, etc... The lowest hanging fruit is have dedicated lines for the heater ground and the electronics ground, after that it is output calibration voltages. Most controllers do not even implement those basic things, so getting worked up about other potential causes of error is moot.

Interesting stuff Alan.

I remember seeing between 0.1 and 0.2 error (between digital and analogue) on my previous TechEdges but there was (and still is) no way to log it as their digital output is only compatible with their own digitally driven displays or their software (not sure it's actually theirs but what they recommend) but not with MS/TS.
As far as I can tell from their rather unclear schematics the ground on the analogue output connector block is the same ground as the main power supply.
I was very concious of this when I first considered switching over to the SLC-OEMs and made sure they were separate on my own carrier PCB (I included the option of analogue in case I ever wanted it) and I've seen you still separate them even down to the SLC Free.
The Innovates I've played with claim to have them separated but I'm not so sure. (I've seen a difference with them too but they were so unreliable it was never clear if it was a real difference or just them not working again...)
It will be interesting to see the difference in a real scenario. Later this week I'll do a comparison with one of your SLC-OEMs and an Innovate (if it will actually work since I dumped it back in its box).
I can't do a proper comparison with the TechEdges (only visual) for the reasons above.

One further thought on the number of people (anecdotal?) who mention these errors on various forums - every manufacturer seems to have a different calibration "curve" (mostly straight lines before any one gets pedantic) for their AFR range vs voltage. Some use 0-5V others 0.5 - 4.5, AFRs 9-19, 10-20, no end of variations.
Sometimes I wonder if people have actually selected the right option for their EMS ECU or even realised they had to.
Hopefully the people who have used digital understand this but I do wonder.

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

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 was still thinking in the realms of 4mm ID.

If I was doing it, I would not change the sample tube diameter or length, but just introduce an orifice into your existing tubes to slow things down. Easier then to test the impact of different size orifices and velocities.

The orifice could be formed simply by welding the end of the tube and drilling, starting at 0.5mm, then going up in increments to measure the impact on the AFR readings.

Edited by Paul S on 27th 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."

Changing the tube size is easy.

The fittings for 6mm Od or 4mm OD are all 1/8" male threaded (either NPT, BSP or BSPT but they are all so close at 1/8" I won't be a perfectionist)

That gives ID options of 5,4,3 or 2mm.

I think changing the tube ID should be more effective than an orifice and just as easy in my case although I can't start at 0.5mm with tube.

This could be interesting.

I'm more and more thinking my after turbo higher reading is because of this, not all the other things I previously assumed.

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

I agree that it doesn't give much information. At best you could be able to check if you have an offset and correct for a different reference voltage by having the full lean value and the full rich value. It would be similar to what Alan is doing but with 0V and 5V. But using the extreme values is more problematic.

And using a calibration gas would be ideal but that would be a very expensive solution. Alan does use such a gas but he also builds and designs controllers (and test competitors).

http://www.jbperf.com/

Scott Clark in the link only uses the wideband data it as a guide combined with the results on the dyno to work on as he states "what the engine wants"

I guess your approach to tuning the car will have an effect on how accurate you need the wideband to be? on the road you wont get the HP to torque curves

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

Tech edge uses a differential output, it is superior to splitting the grounds as a differential output will also take into account the current used by the wideband controller itself.

Splitting the ground as I do will negate the effect of the 3-5 amps due to the sensor heater, it does not negate the offset of the typical 100ma used by the wideband controller itself to drive everything else.

Offset voltage due to 100ma is insignificant IMHO. I am comfortable defending my position of splitting grounds rather than using a differential output.

1275 N/A Sprite, 998 T2 Turbo Mayfair
1275 EFi Turbo

^^^^
I agree with both of you but Paul has cast some doubt on whether the inner and outer ones are reading the same offset as each other if the sample point is close to the head. We have both measured very different temperatures between inner and outer at that point in the past.
Later this week I'll re-connect my thermocouples and get some readings under boost.

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

1275 N/A Sprite, 998 T2 Turbo Mayfair
1275 EFi Turbo

Didnt JohnK do some work with EGT's on each exhaust pipe when they were working on the EFI kit that they sell?

I seem to remember some footage of digital meters connected to the sensors with a 5 port on the dyno? It might help with some comparison.

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