O2 sensor response times

Wed Apr 19 06:36:47 GMT 2000

On Tue, 18 Apr 2000 20:42:21 -0500, Tom Sharpe <twsharpe at mtco.com>
wrote:

>Garfield Willis wrote:
>>
>>All the commercial AFR meters go down into the 10sAFR; it's an area people want
>>to at least get a glimpse at, even if you're only there for a brief moment.

>Gar, could you possibly post a 'scope reading from a running engine to sort of
>'see' the response time? Can we see the individual pulses? How do we compute an
>average AFR?  Tom

Hmm, there be a couple topics possibly combined here. When I said you'd
only be in the 10sAFR briefly, I meant in terms of several/many cycles,
but not so short a time as individual cylinder pulses. Events of the
duration of throttle/AE transients is what I had in view in the comment
above, not so short as the duration of an individual firing.

But I've also mentioned previously that there have been some mixed
signals about the actual response time of these O2 sensors, indeed even
from the NTK guys themselves. Here's a snippette from a previous post:

>>One of the NTK WBO2 pump type sensors is generally capable of picking up a
>>misfire on a V-8 at 3000 rpm---so pretty quick for these!
>
>This is an interesting question.
>
>As Dr. Pelican and Greg have mentioned (in the original NTK article,
>they give these same numbers for misfire detection on an I-4), there are
>some "varying claims". If you add in the high-end meter mfg's specs for
>response time (I remember seeing one claim 0.05secs!; ECM Inc says their
>NTK-based box'$ response time is "<150ms", while curiously NTK doesn't
>spec the response time of their own blue box system/interface), so
>you've got all kinds of all-over-the-place specs. Even in the NTK
>article, there seem to be potentially conflicting numbers; in one place,
>they show the response time for AFR deltas/excursions of around +-4AFR,
>and show these response times in the several hundreds of milliseconds!
>Then, in discussing the ability to detect misfires (seeing the lean
>excursion that occurs from one cyl's mix not burning), they show blips
>in the sensor's outputs that are as narrow as say <100ms, and conclude
>from that "...by using the lean signal [the lean excursion blips I
>mentioned prior] with the ign timing signal, the misfired cylinder also
>can be distinquished". How this all corresponds to picking out misfires
>at 3-4kRpm I'm still unclear on. The numbers just don't seem to match
>up. Add to this, the diagram they (NTK) give for their misfire test
>setup shows all four exhausts collected and then the AFR being measured
>*downstream* just prior to the cat. Thus, there is definitely some gas
>mixing. Remember that mass transport (exh gas movement) occurs at a much
>slower speed than pressure wave propagation, so just because the exhaust
>puffs indeed have a distinct and sharp delineation in pressure, gotta
>consider that in a manner akin to waves propagating in the ocean, the
>pressure phenom may be very distinct along with considerable
>mixing/diffusion of the mass of the media taking place as it transports
>downstream. As an extreme example of this phenom, consider the flow at
>the end of the tailpipe, where even tho the pressure pulses from each
>cylinder are still clearly distinct, the individual mixture packets from
>each cylinder are by then pretty thoroughly mixed/averaged.
>
>Having given all those caveats, what seems to be universally accepted is
>that these sensors are indeed noticeably faster than say the older HEGOs
>like the LSM-11s (Frank Parker showed me an example where his NTK box
>caught a funny rich blip/excursion on his Typhoon when you quickly
>released the throt, but IIRC, his LSM-based equipment misses that
>excursion...correct me if I err here FP). And what's probably most
>important, they DO seem to be fast enough to catch any possible mixture
>excursions that present EFI controllers can produce. Hardly a "spec",
>but reassuring nevertheless. We just haven't had time to set up an
>apparatus to do our own response time testing, but I can tell you that
>EGOR does absolutely NO smoothing or averaging of the sensor output, so
>whatever the bandwidth of the sensor itself is, you're getting/seeing
>all of it at the output of our electrics. I have no way of knowing if
>this is true for other mfgs.
>
>If I had to guess from the mongrel mix of specs, the response time is
>probably somewhere between 100-200ms. Just a swag, tho.
>
>Gar

OK, if you're still with me after that major deja, there's one more
facet of this to consider. Even in the case of the NTK's waveforms
showing misfire, the blips in the AFR are such short spikes that they
don't appreciably affect the overall/average AFR. So given the numbers
above, you could definitely see misfire AFR spikes superimposed on the
average gas ratios, but it's not like you have a sudden major blast of
lean mixture for one cylinder's time, because that puff of exhaust is
having to mix with the gas already in the exhaust system. And as close
as you can get the O2 sensor, it's never RIGHT at the mouth of the
dragon. Remember, we're talking about the kind of misfire artificially
induced, where the IGN is shut off for one firing randomly, a certain
percentage of the time (usually a very small percentage; NTK used 1.5%
misfires in their study). If the engine stumbles from some major screwup
like a sticky injector or faulty carb, then of course THAT will affect
the AFR readings noticeably, but that will also not appear as an AFR
spike, but a much longer (several cycles) event.

So here's the long answer: if you equip your setup with a sync on the
IGN pulse for #1 cyl, and look at the O2 sensors in a
"sensor-per-cylinder" setup where you've got the sensors close and
sequestered to each blast furnace, (you'll be able to afford doing this
for the very first time, with EGOR modules :), you SHOULD be able to see
the individual AFR "signal"s, but from measurements we've taken, it
doesn't look like the NTK's are much faster than 100msec at the very
outside. So don't plan on picking out individual events real clearly at
6Krpm is all I'm saying. Heh.

Certainly don't worry about having to "integrate" the AFR signal in
order to "compute the average AFR", or worse, sample it to find the
"instantaneous AFR"or find the "effective AFR". This was an early
concern when we all had the impression that these sensors were
lightening fast. They ARE fast, but not that fast. Reading the NTK
papers after seeing response times on real engines opened my eyes a bit
to the promotion those guys put on for their sensors. You need to look
carefully at the time scales on some of their waveforms. I'll just say
that like those guys who propose to dig out the AFR from the ionization
signal, the AFR guys seem to want to dig out misfire detection from the
AFR signal. To paraphrase Harry Calahan, "a sensor's just gotta know
it's limitations". Ionization technology is great for detecting misfire
and knock; wide-band O2 technology is great for determining AFR. Change
roles and you have the proverbial French policeman and English cook,
instead of vice versa.

As far as scope pictures showing response times and the like, none of
that will be out until EGOR's coming out party, which is just around the
corner BTW :). Take a look at the waveforms in the NTK paper (SAE
#920234) if you want to puzzle over the conflicting stories until then.
Best I can offer at the moment.

We'll have some interesting results on O2 sensing on a twin-turbo rotary
('95 RX7); one of our choice testbeds, since it gives some really nice
rich excursions. We also intend to publish findings on the question of
sensitivity of these type of O2 sensors to exhaust pressure, another
area nobody seems to want to factually discuss, but for reasons which
should be obvious, we want to hold off on all that also, until we
announce our wee 'product line'. :)  Testing/calibrating with sample
gases has also raised some interesting questions on the relative
accuracy of NTK's own blue-boxes and their AFR curves, in the very rich
region (<12AFR). The long cloak of commercial secrecy surrounding these
sensors appears to have hidden a few interesting stories betwixt the
curtain folds.

Gar

----------------------------------------------------------------------------
To unsubscribe from diy_efi, send "unsubscribe diy_efi" (without the quotes)
in the body of a message (not the subject) to majordomo at lists.diy-efi.org