using narrowband sensors for lpg on disco

[pibby]

Running oxygen sensors on a duel fuel 1998 discovery 14cux. Have installed leds in cabin to show oxygen sensor output. The gas system is the injected type with its own cpu which is a slave to the 14cux and from which I alter the fuel mapping for the lpg.

Lambda of 1.0 for petrol is 14.7:1 AFR, lpg is around 15.5:1 AFR. The titania sensor(s) on the rover is fed a supply voltage of 1v and its resistance varies depending on the amount of unburnt oxygen in the exhaust gases. The narrowband sensors aim for a switchpoint of 0.45v to achieve lambda of 1. I have read in a number of places they are accurate between AFR of 15:1 to 12:1.

Using the leds in the cabin I can see how the 14cux manages the petrol fuelling and I replicate this fuelling pattern to do the lpg mapping. However, what I find is this consistently gives suboptimal fuel consumption so I find myself fuelling until the leds are about to light up which gives a better exhaust note (it sounds rich otherwise), approximately the same power but much better economy.

So I’ll get to my question – is the switching point voltage for lpg and petrol the same to achieve a lambda of 1 and if it is different what is the stoichemetric switching voltage on narrowband oxygen sensors for lpg or how is it achieved using a narrowband sensor? I have spoken to the bosch automotive technical department on this and also the ntk (ngk) technical department on this. Unfortunately they completely contradicted each other with respect : if the sensor voltage changes based on the % of unburnt oxygen in the exhaust stream will the voltage be the same for lpg and petrol to achieve a lambda of 1. I have gone to TAFE automotive department to try and get an answer and I have gone cross eyed trying to the find the detail on the web. I have spoken to a couple of people in the trade and still can’t get an answer. If I had a wideband controller I’d plug it in and get my answer straightway, but I don’t and the cheapest I can find is around $300 out of the states.

My next step was to ring holden or ford to find out what they do on their factory gas vehicles which I am told have narrowband sensors. But surely someone should know this stuff. If it is the same voltage for lpg and petrol I am curious to understand why as I’ve read for automotive purposes a rich mixture on one end of the spectrum is 0% oxygen whilst on the other end a lean mixture is around 3-4% oxygen with atmospheric oxygen around 21% so with different AFR for each fuel it follows that the voltage would be different.

The lpg mapping software (it is an elko system) that I use is a single dimensional tool ie it looks at the petrol injector opening time and you can increase or decrease it. the problem I am finding is the petrol ecu 14cux calculates the injector opening time on a matrix including tps/rpm/temperatures etc. so for a 5ms opening time you might be under heavy load at 4000 rpm or light load at 1200rpm however when this is mapped for lpg you can only map a single gas injector opening time for each petrol opening time ie there is no option to factor in other variables. So what you end up with is a compromise.

I hope this all makes sense. Anyone able to help??

[bee utey]

AFAIK the lambda sensors will sense lambda the same voltage with both fuels. They are not designed to operate exactly at lambda as the ecu's using them use the output to float across this 0.45V value for the health of the catalytic converters, when fitted. So they are a poor tool for accurate A/F ratios. They are also very temperature dependent off lambda.

My mapping software for gas injection (not Elko) has a separate USB key which allows full setting of the lpg fuel map. Something you might ask your supplier if there is upgraded software available.

I would say to make full use of Landy oxy sensors you could interface them to your ECU. This has been done and documented on this forum I believe.

[PhilipA]

Using the leds in the cabin I can see how the 14cux manages the petrol fuelling and I replicate this fuelling pattern to do the lpg mapping. However, what I find is this consistently gives suboptimal fuel consumption so I find myself fuelling until the leds are about to light up which gives a better exhaust note (it sounds rich otherwise), approximately the same power but much better economy.

I am a bit reluctant to comment on this, as I do not know the answer.

BUT you do not mention whether you actually have the O2 sensors connected to the 14CUX. I assume not because in my case anyway , there is no stability to the mixture when in closed loop . It constantly cycles up and down to average (I assume) lambda . This is the 14CUX reacting to the O2 input, hence "closed loop."
BTW I measured this variation with a separate Zirconia sensor mated to a Jaycar LED indicator. The reason I say "I assume" Lambda is that I did not calibrate the LED pot on a known lambda source, but the behaviour of the mixture is quite different open vs closed loop, and it is pretty correct.

If you have a steady reading then it is not closed loop. In closed loop my assumption is that the injector on time is constantly changing significantly at one throttle/load setting, so it would not be much help in deciding a single map for the gas.
I am just recalling some info I read quite a while ago that Lambda for alcohol is different to lambda for petrol and is in the 15+:1 range, so it is entirely possible that Lambda for gas is also different to petrol. If this is so , then it follows that you would have to lean off to get gas lambda.

This post reckons lambda for gas is 15.5:1. If so there is your answer. So you need a wide band sensor.

Air/fuel meter gauges on LPG cars [Archive] - calaisturbo.com.au
http://newsgroups.derkeiler.com/Arch.../msg00075.html
Regards Philip A

[pibby]

Bee utey,

I was hoping you’d see this post given your experience in this area. I do have the oxygen sensors running off the 14cux. From what I have read the narrowband are very accurate at attaining lambda 1.0, more so than wideband. However the further you get from lambda 1.0 the signal goes from less accurate to meaningless. I’ll attach a picture from a bosch pamphlet for the sake of those who aren’t familiar with how a narrow band works (which included me until a few months ago) which shows why it is so accurate at lambda 1.0.

I still can’t understand how lpg and petrol can have the same voltage at lambda 1.0 on narrowband, if you are using a wideband then 14.7:1 and 15.5:1 will be at different voltages.

The software for the lpg is what they use on all of their installs. This is why I am trying to understand how to get the 14cux to control the fuel level in the first instance and then I just fine tune it with the lpg software. My feeling is that it is setting the fuel at 14.7:1 given the resultant fuel consumption, exhaust note (it sounds rich) and the extra blackness on the exhaust and I was expecting to see a much lighter colour on the spark plugs. If it walks like a duck, quacks like a duck, looks like a duck….well it’s probably running rich. I have also wired in a trim pot and diode to induce an increase in voltage as seen by the 14cux from each of the sensors individually. I haven’t finished testing this yet but I was thinking that if I introduce 300 or 400 mV then it will trick the 14cux into thinking it is running rich and effectively offset the fuel curve. I have tested it so far on one sensor and I noticed the appropriate change in the signal but I wasn’t too clear on whether it actually changed the injector opening times on that bank of the engine. I will know better once I wire it across both banks of the motor as I have found that the 14cux doesn’t do a good job of balancing each bank on gas and I have to use the lpg software to balance it up.

Is the software you use able to control an injected gas system including individual injector banks or is it for the mixer type where lpg is introduced into the air stream?

[bee utey]

Bee utey,

I was hoping you’d see this post given your experience in this area. I do have the oxygen sensors running off the 14cux. From what I have read the narrowband are very accurate at attaining lambda 1.0, more so than wideband. However the further you get from lambda 1.0 the signal goes from less accurate to meaningless. I’ll attach a picture from a bosch pamphlet for the sake of those who aren’t familiar with how a narrow band works (which included me until a few months ago) which shows why it is so accurate at lambda 1.0.

I still can’t understand how lpg and petrol can have the same voltage at lambda 1.0 on narrowband, if you are using a wideband then 14.7:1 and 15.5:1 will be at different voltages.

The software for the lpg is what they use on all of their installs. This is why I am trying to understand how to get the 14cux to control the fuel level in the first instance and then I just fine tune it with the lpg software. My feeling is that it is setting the fuel at 14.7:1 given the resultant fuel consumption, exhaust note (it sounds rich) and the extra blackness on the exhaust and I was expecting to see a much lighter colour on the spark plugs. If it walks like a duck, quacks like a duck, looks like a duck….well it’s probably running rich. I have also wired in a trim pot and diode to induce an increase in voltage as seen by the 14cux from each of the sensors individually. I haven’t finished testing this yet but I was thinking that if I introduce 300 or 400 mV then it will trick the 14cux into thinking it is running rich and effectively offset the fuel curve. I have tested it so far on one sensor and I noticed the appropriate change in the signal but I wasn’t too clear on whether it actually changed the injector opening times on that bank of the engine. I will know better once I wire it across both banks of the motor as I have found that the 14cux doesn’t do a good job of balancing each bank on gas and I have to use the lpg software to balance it up.

Is the software you use able to control an injected gas system including individual injector banks or is it for the mixer type where lpg is introduced into the air stream?

I am having trouble here. You are using Rover oxy sensors coupled to the 14cux? The ones used on the P38A are 5 volts lean 0 volts rich. The narrow band sensors used by Holden and Ford generate their own voltage, 0V lean, 0.9V rich. They run between 0.2V and 0.8V approx. while operating in closed loop. The injector on-time varies a small amount to achieve this, a lot less than the voltage swing suggests. If your 14cux is running on sensors it will adjust fuelling on closed loop regardless of which fuel you are on. The LPG computer has only conversion figures from petrol to gas, not absolute figures independent of petrol. Adjusting the conversion map will affect LPG fuelling on rapid changes and closed loop operation.
Without knowing much about other sensor types, I assume they are measuring oxygen relative to atmospheric, so A/F is dependent on the chemical composition of the fuel.
The richness you smell is more likely to be tracer gas levels, this will depend on the composition of LPG from the refinery. I hve not achieved a totally clean exhaust on any engine controlled by narrow band sensor as there will always be rich moments during open loop operation. Looking at plug colour would be useful only after a very long run on closed loop to allow cold start deposits to fully burn off.
My injection software allows 2 bank operation, not that I have used that feature on Rovers yet. I will be getting a P38 in for a 12 month check-up tune shortly and will have a look.
The mixer type systems I use have only a simple processor without accessible software so they don't have much tunability.
How is your ignition advance set?

[pibby]

thanks for the replies philipA and bee utey. i think you can see my problem. the way i read it, one of your responses suggests a narrowband will achieve lambda whatever fuel while the other response tends to suggest the narrowband is suitable for achieving lambda 1.0 for petrol and any other fuels which have a stoich point of 14.7:1.

just to clarify, i am using the titania sensors with the 14cux and as you mention bee utey they are different to the zirconia sensors which need to reference the atmosphere and generate their own voltage (the titania are fed 1V and their resistance changes based on the unburnt oxygen in the exhaust which sends the voltage once it has been subject to the resistance back to the 14cux). they are definitely switching between 0 and 1 volt. i have read of earlier titania sensors which switched between 0 and 5 volts. have i simply got the wrong voltage sensors? i shall have a look at the rave cd.

if the narrowband are not the best tool for lpg (i also map the fuel via the lpg software but this is after the 14cux has derived its mapping and is a one dimensional tool) i was trying to work out how to get the mapping done via a wideband sensor. that's why i was asking about your software as the elko software can only adjust on petrol injector opening times so maybe i should look at some other software to control the lpg mapping and takeover the lpg hardware which is able to use a wideband or has a matrix which includes tps and rpm.

with regard to my timing, that's as good as gold. i've ditched the dissy and put in edis (well the dissy is still there underneath the coils minus cap) so she's perfectly matched to lpg (i've also got a switchable map in for petrol which is automatically activated whenever the lpg is not running). this is using megasquirt. to use a wideband with megasquirt you still need to buy the controller, not just the sensor. from what i've read it will work straight out with a narrowband and it has an option to set the switching voltage you want to aim for eg petrol is approx 0.45v. this is where i was coming from when asking about the switching voltage for lpg on narrowband.

at this stage i really don't want to run fuel on megasquirt, i have read some of philipA posts and believe much like he does the 14cux does a pretty good job on starting and warming up. there's a lot of work to get it running like the 14cux with little upside.

i had a look at those links as well philipA. intuitively it makes sense that voltage would be different for different fuels. i would have thought that bosch and NGK would have this information documented. maybe the factory gas cars from ford/holden have wideband sensors fitted???

[PhilipA]

i had a look at those links as well philipA. intuitively it makes sense that voltage would be different for different fuels. i would have thought that bosch and NGK would have this information documented. maybe the factory gas cars from ford/holden have wideband sensors fitted???

I haven't really got my mind around it because I have not HAD to think about it, but intuitively the ECU in the OEM petrol/gas cars could have completely different maps for petrol and gas which take into account the differences in output of the o2 sensors.
Intuitively they could be narrow band as they would show Lambda output at 15.5:1 on gas. You will note one of the posters said that a wide band sensor has a different program for gas. This could be incorporated with narrow band in the ECU, one would think.

I guess even E85 could have narrow band if the ECU has appropriate maps.
Without seeing the sensors I have no idea.
Just looking at ebay the normal Bosch 4 wire zirconias are listed up to VE but only for V8 rear and not for V6 at all. So my guess is that both have wideband even with petrol only. Not completely authoritive but indicative.

For Rear - Left & Right Only. Different sensors required for Front Left and Right

That is for V8. The rears are after the catalysts usually just to confirm they are not dead.

Regards Philip A

[jazzaD1]

that may help you a bit

a narrow band sensor is really only useful for keeping the air:fuel ratio at a stoichiometric mixture, that is a lambda value of 1

from wikipedia:
In theory a stoichiometric mixture has just enough air to completely burn the available fuel. In practice this is never quite achieved, due primarily to the very short time available in an internal combustion engine for each combustion cycle. Most of the combustion process completes in approximately 4-5 milliseconds at an engine speed of 6000 rpm. This is the time that elapses from when the spark is fired until the burning of the fuel air mix is essentially complete after some 80 degrees of crankshaft rotation.
Catalytic converters are designed to work best when the exhaust gases passing through them show nearly perfect combustion has taken place.
A stoichiometric mixture unfortunately burns very hot and can damage engine components if the engine is placed under high load at this fuel air mixture. Due to the high temperatures at this mixture, detonation of the fuel air mix shortly after maximum cylinder pressure is possible under high load (referred to as knocking or pinking). Detonation can cause serious engine damage as the uncontrolled burning of the fuel air mix can create very high pressures in the cylinder. As a consequence stoichiometric mixtures are only used under light load conditions. For acceleration and high load conditions, a richer mixture (lower air-fuel ratio) is used to produce cooler combustion products and thereby prevent detonation and overheating of the cylinder head.

i use an innovate LC-1 wideband sensor on my megasquirt install, and if you want to know exactly what the lambda value, or AFR is, it is the only way to go

but considering you arent using megasquirt to control fuel, is it really necessary? i mean, you are right, there is only so much you can do with the stock mapping

here's a question: do you have the oxygen sensor wired to your lpg ecu also? as i know in the software you can specify what type and voltage of sensor you have, then the lpg system becomes less 'one-dimensional'

of course i am yet to try this with my elko system, as i am still finalising my petrol map in megasquirt

[bee utey]

The Zavoli injection system can read any type of oxy sensor and display the values on screen, but they aren't used to correct the fuel map. The most recent version uses OBD2 input to correct fuelling but that wouldn't be available in your case.
The LPG ECU does not independently set gas injection time, it always is up to the petrol ECU to adjust fuelling time to correct the sensor voltage on closed loop. Then the LPG follows.
A minor difference in mapping will make no difference in fuelling on closed loop. Only when the petrol system is in open loop are the LPG injection times adjustable by the software to vary fuelling.
So if you are running wideband sensors you can definitely offset the voltages to affect the LPG tuning, switch the offset in and out as required.

[pibby]

Ok, I think it’s looking like the narrowband being run through the 14cux is not the best way to put an lpg map together. I have used the led kits from jaycar to tune from visually in conjunction with using the gas software where I can see the actual sensor voltage numerically for each bank. I think it’s been a fairly good map I was able to put together but the weakness has been in the gas tuning software. Prior to putting edis on the car I had 3 runs of 23, 24 and 25 litres per 100km on lpg around town, the motor is a 4.6 litre. From memory the rave cd had 22.8 l/100km around town for the 4.6 litre which had the land rover coil packs and ignition mapping. I am hoping to get a little better than this with the edis, there is a noticeable difference with more ‘poke’. Town being Hobart which is not flat and using the lpg software I can see the immediate increase in injector opening time (petrol or lpg) as soon as you hit an incline. It’s almost enough to make you want to move to somewhere flat.

Thanks for looking up those sensors philipA, I know zip about commodores and falcons.

jazzaD1 you will find the sensor in the software is purely for visual information, the software does nothing with it unfortunately, if it did I would have my answer. The graph you included is part of what my problem is. From my understanding lambda is just a ratio of two afr eg petrol stoich is 14.7:1 so if you were running at 13.0:1 then you would be 13.0/14.7=0.88 lambda. Lpg stoich 15.5:1, so if you were running 14.7:1 then you are running 14.7/15.5=0.95 lambda which is the reason why I feel I am running rich as the narrowband sensor switches around 0.45V which is the stoich point for petrol. My tests have shown around a 10-15% difference in fuel economy between a map based on 14.7:1 and my map which is probably around 15.5:1. Unfortunately your graph doesn’t talk about the voltage for the narrowband sensor for each of these AFR. You’re right, the wideband would handle this no probs but I wasn’t that keen to run fuel through megasquirt. edis I’m happy with as it has proven itself over many installations and if either edis or megasquirt ignition fail then as you know it will default to 10 btdc in limp home mode (I have mine around 15 degrees btdc physically and use the ignition maps to take this offset into account. 10 btdc is a bit undercooked for lpg). I will eventually carry around the dissy cap too as it’s only three bolts and the edis is off then reconnect wires back to the oem coil and you’ve got the dissy back.

So after this convoluted route I believe these are the options for a self controlling vapour injection lpg fuel map on the disco able to withstand variables such as the timing being changed etc :

1/find gas software which can control the existing hardware and which has a mapping tool which can use the feedback from either a wideband or narrowband sensor to do the fuel allocation by (sounds like what bee utey uses might be able to do this)
2/use megasquirt which has the ability to use narrowband input and let you select the switching voltage or use a wideband sensor but still have to buy the controller
3/finish off my testing where I am introducing voltage into the signal to the 14cux from the sensors to trick it into believing it is running rich and find the point at which the mixture is 15.5:1. trial and error here!

I hope I’m missing an obvious way of doing this which some one knows as I’m not keen on reinventing the wheel.