Tech Talk:                         On-Board Mixture Analysers & O2 Sensors

A Disscussion from The Hotrod List in 1993 and from various other sources

The following is excerpted from a disscussion on various types of on-board mixture analysers and O2 sensors.
This document is long and  I have published it almost exactly as it was downloaded. The posts have been
reproduced as they were archived without alteration other than to change the layout of some sections for clarity,
and to bold certain text for emphasis. Wherever possible the original participant info and source is included. I
have published this here simply as a technical reference and in no way was involved in this disscussion myself.
Please note that this disscussion dates from 1993 and many (if not all) of the e-mail addresses are going to be out
of date. Lastly, I take no responsibility for the accuracy of the information contained herein, and am sharing this information simply as a courtesy.
Marc Sayer

From hotrod@Dixie.Com Tue Mar 30 11:11:25 1993
Subject: Edelbrock air/fuel ratio monitor
From: (The Hotrod List)


I just bought the Edelbrock air/fuel ratio monitor and hears what I discovered. The sensor is a Bosch PN E971-9F472-AA
(a lot of numbers stamped on it, I think this ones the Bosch part number). I believe this to be a standard three wire
conventual O2 sensor though admittedly I do not have the manufacture's literature. The "little black box" is nothing more
than little and black. Inside there is a surge suppresser, a filter capacitor, one needed and one redundant resistor to
control the LEDs brightness, and a LM3914 Dot/Bar Display Driver. The design is straight out of the application books
with no creativity. Did I mention the lack of input signal conditioning/filtering to remove engine/ignition noise? Given the
speed of the LM3914's comparators, this is a problem and certainly a problem in my implementation.

The electronics clearly runs the sensor in the voltage mode drawing only a 25nA biasing current for the LM3914's internal
buffer. The sensor voltage and air/fuel ratio have the following corresponds:

                      volts      air/fuel     Lambda
                 (open circuit)
                       .250       15.0:1       1.02
                       .375       14.5:1       0.99
                       .500       14.0:1       0.95
                       .625       13.5:1       0.92
                       .750       13.0:1       0.88
                       .875       12.5:1       0.85
                      1.000       12.0:1       0.82

From what I have read in the SAE Transaction pertaining to O2 sensors, this relationship is "bull sh__". :( I can not believe
this mode of operation would let the sensor go as low as .8 lambda let alone have this type of linearity.

I would have at least expected Edelbrock to use a current mode of operation.

Any comments?

My next step is to change Edelbrocks "little black box" into a current mode of operation and call NGK in hopes of finding
an affordable UEGO.

John S. Gwynne

>From hotrod@Dixie.Com Thu Apr  1 12:39:23 1993
Subject: Re: Edelbrock air/fuel ration monitor
From: (The Hotrod List)

Bob Valentine  writes:

>     From the tone, I can tell that the design is not a good one.
>However, I'd like to get away from the DVM hanging under my dash....
>Would using a 3914 (I used those in grade school for really cool audio
>power meters...) suffice with a 1 wire O2 sensor?

From what I've seen so far the electronic portion needs some improvements. The input signal will have to be
conditioned/filtered to gain a steady display. The signal line from the sensor has no loading at all to reduce ignition noise.
In addition, I do not like the voltage generation mode of operation. From some of the generic voltage-vs-current curves (as
a function of lambda), in the SAE transactions, I intend to find a load line that will maximize the sensor's dynamic range
and hope it works well for this sensor too. In the circuit to do this, I will include some filtering. The use of the LM3914 is a
good choice, but I don't see it doing the job alone for a 1 wire or 3 wire sensor.

>[chart showing voltage, air/fuel, lambda deleted]
>So then, is this chart valid?

I don't think so. IMHO I think Edelbrock wanted a product to match MSD's O2 sensor, which has only one tri-color LED to
indicate the mixture condition. Their gimmick was to add more LED's and who would know if these LED's had any real
meaning? They almost flash in a believable fashion.

>I don't have the reference here anymore, but I belive a LM3915
>works on current, rather than voltage.   Nice thing about this series
>is that you can chain several of them together to get a rather precise
>display. (ie., a 20 segment display... hell, just carpet the dash with
>those plug in segments! 8^>)

The LM3914 has a linear display and the LM3915 is logarithmic. Both are a function of input voltage, and both will let you
carpet the dash :) .

>If you can find a UEGO, let me know.   Last I heard they were
>made of "Unobtanium", and way out of bounds on price.

The data sheet I have from HORIBA on their UEGO sensor for the MEXA-110 says it's made of zirconia and ceramic :) .
But at a price of $900, it is unobtainable. If I can find the time, I'm still going to look for an affordable one.

[I mentioned awhile back the MOTES A/F analyzer that competes with the MEXA but costs less than Horiba's sensor.
They told me they were using, I believe, the Bosch UEGO sensor that is much cheapter than the one Horiba uses.
Someone else noted that one model of Honda lean burn engine uses a UEGO. Might want to check the archives for that
info.  They're at JGD]

John S. Gwynne

>From hotrod@Dixie.Com Mon Jul 26 12:46:06 1993
Subject: Re: More about Holleys
From: (The Hotrod List)

> O2 Sensor setup, etc <

I drilled a hole in the cast iron exhaust manifold right before the outlet flange, and reamed it out until an 18mm anti-fouling
spark plug adaptor was a hammer fit. I ground down the adaptor until its inner end was flush with the interior wall of the
manifold, and then welded it in place.

The O2 sensor I used was a freebie 1-wire type. A 3-wire is better for reasons I'll go into in a bit, but costs more. I picked
up ground locally using a valve cover bolt through a loop soldered in the end of the ground wire, and connected the O2
sensor to its mate, solder and shrink-wrap. I used some automotive "trailer wire" (the 4-conductor ribbon used for hooking
up auxiliary lights etc.) which seems to be taking the heat and grease reasonably well. The reason for taking a local
(engine block) ground is to minimize any error induced by body/block return currents and any voltage offset that might
develop from loads like A/C, lights if you used the dash as ground.

I am using a spare analog VOM to read O2 volts; it seems to be working well enough, even though I've been told that the
O2 sensor needs a very high load impedance. This meter has about 100Kohms (50K/volt, 2.5V scale) and I think it's about
on the edge of affecting the sensor output accuracy; if I switch to the next lower scale I get a slightly lower value reading
indicating Zload sensitivity to me.

The reading ranges from 0 to 1V (approx), with 0 indicating lean and 1V indicating rich. 0.5 is supposed to be
stoichemetric. When warm I seem to be pegged rich, or close to it, except for a "hole" during throttle transitions.

The one-wire sensor depends on exhaust heat to activate it; the element has to be pretty hot to develop any voltage and
low temperatures result in a reduction or absence of output. I find that there's not enough heat at idle to keep the sensor
fired up. If I idle for more than a few seconds the output voltage starts to decay. The other thing is that it takes a while to
start reading. For a tailpipe test rig I made, a 3-wire was a necessity since by that point there's not enough heat ever.

For a permanent installation I think the best solution would be to use a small panel meter with a single-supply op amp
as a unity gain buffer and a calibration resistor to get 1V full scale. A shunt resistor from amplifier output to ground
might help the amp in the low end. I tried cobbling up a bar graph display using an old LM3914 I had in the parts-is-parts
tin but it seems to have developed an attitude. For some reason Radio Shack has stopped selling these although they
still sell the bar graph displays.

From hotrod@Dixie.Com Mon Aug 16 21:12:16 1993
Subject: 02 sensor help!
From: (The Hotrod List)

> O2 Sensor setup, etc <

>The O2 sensor I used was a freebie 1-wire type. A 3-wire is better for
>reasons I'll go into in a bit, but costs more.

We are interested in putting an 02 sensor in a Datsun 240Z SCCA ITS class road race car. The Datsun has a six into 
two header (where three cylinders being fed by one carburetor) into a dual exhaust system. We are interested in 
monitoring the air / fuel ratio of each carb (3 cylinders) while the car is at speed to determine efficiency at high RPM.

We understand K & N makes an air/fuel monitor that retails for $139.00. Can we inexpensively duplicate this device or 
do you have a better, less expensive alternative.

Also, does the 02 sensor develop its own voltage or is it a variable resistance device? Any explanation (reference 
material, etc.) on how this operates would be greatly appreciated.

Roger Hensley
Patriot Motorsports
1099 N. 16th St.
Otsego, Mi 49078

[Cyberdyne makes an in-dash A/F indicator that lists for $29 in Summit's catalog.  Don't pay a dime more.  Many of 
these things are labeled with wide range A/F ratios.  Since the standard O2 sensor covers a narrow range around 
stoich, this labeling is fraudulent.  As long as standard production O2 sensors are used (UEGOs sensors by themselves 
cost several hundred bux), all the displays must work the same.  If you want an analog readout, a simple analog meter 
with an op-amp buffer does the trick. The voltage range from the sensor is 0-1 volt at high impedance so gain isn't 
necessary; only buffering.

For references, see the archives plus SAE's "Sensors and Actuators" publication from the last few years.  More info 
than you can stand. Remember though, lead-free gas only.  JGD]

From hotrod@Dixie.Com Mon Oct 11 23:39:00 1993 Subject: Re: custom O2 gague (& others) From: (The Hotrod List) In message <> , you write: >these questions deal with the air/fuel mix sensor (like cyberdyne's in the summit catalog): >where does the sensor need to be located, ie real close to the engine (reporting on just one cylinder (and hoping it is '>typical')), or can it be put down on the collector, so it would report on how the cylinders (as a whole) are doing? How  >does the placement affect the accuracy/linearity (or what else?) Sensor location is important for the non-heated EGO sensor. Too close to the engine and it will become too hot and  shorten the life expectancy. Too far way and the sensor may not stay warm enough to work at idle.  Location is not a  problem for the thick film heated sensors (TFHEGO), and placement in the collector is fine.
>does the sensor come with a full spec sheet (ie what the voltage or resistance is for different a/f ratios, or can one be 

>which leads me to the thought that brought on the questions: i want to make a small 'gage' for a/f ratios (i don't like the 
>look of the cyberdyne unit), with 5 lights or led's on it:
>1) red    -> way too lean
>2) yellow -> fuel-economy cruise (slightly lean)
>3) green  -> stoich
>4) yellow -> acceleration (slightly rich)
>5) red    -> way too rich

>As far as I know (which may not be much :) ), all of these gauges for under $150 that claim to cover such a wide A/F 
>range (specifically on the rich side) ARE bullshit...  I bought the Edelbrock air/fuel monitor (see post end of March/93 
>in the archives -- and traced out the circuit. It uses a TFHEGO in the voltage generation mode which 
>can really only indicate 2-states (rich/lean). Needless to say that I was not happy and you should not waste your 
>money on that one....Before you buy one of these, run down to that auto parts store and get a generic 3-wire sensor.
To build a wide-range Air/Fuel meter, you need something like a UEGO sensor (again see the archives) that would use 
a oxygen cramming method or oxygen bias method (see SAE journals). This can cover the range .6<lambda<1.6. 
I checked into buying one of these from Horiba, but they wanted $900. I now know Ceramic Sensor Co. makes this for 
them, but I don't know how much it would be to buy direct (I would guess $500).
> another question, this on air temperature reading: i have been in a
> jeep with an out- door thermometer (i assume it's for the fuel
> injection system), and was wondering how difficult it would be to
> create one?  is there such a thing as a reasonably linear thermocouple
> (or a thermocouple & amp pair that produce a linear result)?  i'm
> hoping to do better than  'go to k-mart & rip open a digital
> thermometer', but if that's really the best way to go....

Thermocouples are inherently nonlinear (at least over a large temperature range). Check out the National Semiconductor 
data acquisition / linear devices data book.  They make some absolute linear temperature sensor (LM34A as an example) 
that cover -50 to +300 deg. F with an accuracy of +-2 deg. (10mV/deg F).  Other temperature ranges are listed.
Posted by: John S Gwynne <emory!!jsg>

From hotrod@Dixie.Com Tue Jul  5 13:15:15 1994 Subject: Honda LAF sensor From: (The Hotrod List) >The wide-range O2 sensor used in some of the Hondas is called an LAF sensor in England, UEGO over here.  There's  >a chart in the October 1991 Car Design & Technology which shows the response curves of a conventional Lambda O2  >sensor and the Honda LAF sensor The LAF's response isn't linear, as depicted on the chart.  It starts at about -5v at 10:1 A/F, moving up sharply to about  -1v at stoich, then progressing slowly up to maybe +2v at 25:1.
From: (James W. Swonger) Subject: Re: Determining AFR from O2 Voltage readout >A non-heated sensor will behave as described if driving a low impedance load. This is due to the low exhaust temperature at >idle, which reduces the current available from the cell. There will be an error induced, which is less into a higher impedance >DVM than a resistive input such as a VOM (I have used both and observed this directly). The bargraph indicator may have an >input resistance low enough to make this noticeable. On my van, I could see the sensor output voltage indication drift >downward at about 100mV/sec after dropping onto idle. Eventually it would hit the floor, just "go out". Revving it for a few >seconds would light it back up, and cruise was enough to keep it hot. With a very high input impedance meter, the indication >remains steady even at idle. You can substitute a heated sensor to get around the problem of inadequate exhaust temperature; three wire Bosch sensors can be gotten cheap at the AutoZone or whatever, if the future high school graduate behind the counter can be made to  understand the concept of reading a parts list. I used a heated generic sensor to make a slip-on tailpipe sniffer.   You could  also try using header wrap to maintain exhaust temperature up to the point where the sensor is located.
From: <>
Steve, I saw your post on and thought I'd lend you my  .02.  Somone on the Corvete mialing list had a similar 
type of  question, so I thought I'd just copy the note over to you too.  If you have any questions, EMAIL me.  For some 
reason I can't post to

Standard O2 sensors basically measure rich or lean operation.  Their response is very nonlinear and can not be 
used to measure precise A/F.In computer controlled cars, the computer operates on a limit cycle, switching between rich 
and lean in order to average out to a particular A/F (usually stoich.).  Sensors to measure actual A/F and not just rich/lean do 
exist, they are called Universal Exhaust Gas Oxygen sensors (UEGO) or Linear EGO sensors.  Essentially a UEGO sensor 
is two EGO sensors physically linked.  One half of the sensor is exposed to exhaust gas, the other to O2.  To make a long, 
complex story short,  just as O2 causes a potential difference across the walls of the O2 sensor, pumping current into the O2 
sensor can cause O2 ions to flow through the sensor.  By measuring how much current must be pumped into the sensor in 
order to maintain a constant voltage across the O2 detecting cell (450 mV), the A/F can be determined. These devices are 
common place in dyno/engine calibration labs (at least here at Motorola and at Ford).  The interface circuit for the sensor is 
extremely complex, as the sensor must be maintained in a closed loop and the currents measured are very small (into the 
tens of uA's).  Besides having to control the O2 detecting and pumping cells, the heater must also be controlled and 
maintained at 10V RMS, battery voltage will burn it out.  However, battery voltage (actually 12V RMS)is used to quickly warm 
up the sensor, after which the heater voltage is reduced.  Thus, another circuit is used to detect when the sensor is warm.  
If current is pumped into a cold sensor, it will "blacken" and be destroyed.  Operating temperature of the sensor is determined 
by measuring its electrical impedance.

I worked on such a circuit for Ford a few years back with the intention of putting it into production vehicles.  It turned out that
it was not cost effective for the benefits (at least that was their  reason for killing the project).  Instrumentation sensors like this
exist, along with their digital readout, but they cost about ?8K! NTK  (a division of NDK Spark Plug) is a very popular one.

I think that it would be possible to make a crude A/F meter using a standard ZrO2 EGO sensor and measuring the time that is 
spent lean (<450mV) vs. rich (> 450mV).  Note that nearly all (if not all) EGO sensors are actually HEGO sensors, H for heated.  
With these standard O2 sensors, battery to the heating element is sufficient.

Hi Steve.  You asked: >If an AFR sensor is basically two EGO sensors (one on the inside of the exhaust manifold, and one on the outside), and  >the two compare readings, I don't see why it would be so hard (so expensive) to make a device to do this.  Probably the  >circuitry in determining the AFR from the comparison of inside/outside voltages I guess. They are not two separate sensors.  As a simplification, you can think of them as two separate sensors, but they are  physically linked so that the O2 ions can flow from one side to the other.  The O2 reference is created by applying a very  small bias current to the sensor.  This creates an O2 reference in part of the cell.  It is not exposed to outside air, however. Mark Stavropoulos Motorola AIEG Northbrook, IL
>From  Sat Nov 12 20:49:30 EST 1994 In article <>, John Avery  wrote: >The November 1994 issue of SEMA News (Specialty Equipment Marketing >Association) shows the folowing "tuning aid" as they call it: >Accell...introduction of a new air/fuel ratio meter for all automotive >and light-truck applications. The new product is a tremendous tuning aid >for enthusiasts with both carbureted and fuel-injected engines. The >meter allows the driver to the monitor the air/fuel ratio during actual >operating conditions and is driven by a fast response oxygen sensor that >is bolted into the exhaust system. Close calibration of fuel curves >becomes quite easy, allowing precise fuel metering at all engine >speeds." The address is: >Accel >8700 Brookpark Rd. >Cleveland, OH 44129 >(216) 398-8300 >I have nothing to do with this company. I just thought it was relevant >to this thread. Hope it helps somebody. It definitely is.  Before I started this thread, I knew very little about O2 sensors, and was hoping that I could use one to  determine exact AFR in my engine.   Since then, I have received much information from everyone (thanks everyone!) and  have learned quite a bit about O2 sensors.  Still would like to know more though. I'm going to repost with all the info I have received in a few days, maybe somebody with some more disk space would like  to keep all the info as FAQ material? Anyways, from the descriptions of how O2 sensors work, and operate in a closed loop system, it appears obvious to me  now that all a regular O2 sensor can do is sense rich or lean conditions with reference to the stoich mixture of  14.7:1   In fact, that is all it NEEDS to do to keep a closed loop system working well, since in closed loop, the computer  wants the mixture to stay as close to stoich as possible (the rich and lean voltage indications of the O2 sensor just keep  bouncing signals back to the computer, which in turn keep the mixture very close to 14.7:1) From the many descriptions, and opinions I have received, there is NO steady relationship (linear, cubic,  quadratic, etc.) between the O2 sensor voltage and the AFR.   Therefor, from many others conclusions, I have  come to the conclusion that no products which offer an AFR from a regular O2 sensor can even be the least bit  accurate in determining a numerical AFR. On the other hand, a system like MSD's which only indicates rich/lean is plenty accurate, because that information CAN  be obtained from the O2 sensor. Whether it is worth $150 is up to the owner, seeing as how the same readings can be  obtained by a regular O2 sensor and a good voltmeter. According to the information I have received, only a UEGO  (Universal exhaust gas oxygen) sensor can determine accurate  AFR readings.   I won't go into how one of these gems works right now, it will be in the repost. Also, according to the  information I have received, these sensors cost a LOT of money, and the interface circuitry is extremely complex and  expensive. So, that's pretty much where it is left with me, any product which works off of a regular O2 sensor and claims to  measure AFR is a scam.  And the products which actually can measure AFR cost a pretty penny. This is NOT to say that the regular O2 sensor can't be used for tuning. Because it can be very helpful for tuning for  emissions, or returning an engine to a previous mixture setting (which you had recorded in O2 sensor volts). So in conclusion, I would like to know how Accel's unit operates.  Does it operate off of a regular O2 sensor, or a UEGO  sensor?  (rhetorical question - unless someone has the answer) Steve
From: (Larry Piekarski) Subject: Re: New Question (was Re: Determining AFR...) > Second Question: > > If the above is true, how would I go about wiring up one of these bar graphs > from radio shack, as to not short anything out, or ruin any of my current > wiring?  (i.e. what resistors, etc. would I need for this readout) > I'm not much of a circuit builder -but am trying to learn. First, the O2 sensor is extremely high impedience, 22 MOhm. If you are building a readout device, make sure you start  with a very high impedience voltage follower. Second, most voltmeters and scopes have only 1 MOhm inputs, so be  cautious of the readings you get.
End Of Discussion