NGK AFX vs. AEM 30-0300 CAN Comparison

[Z06-Hammer]

Interesting......

noise.png

[MeentSS02]

Interesting......

noise.png

Where did this information come from?

I trust the AFX more than any AEM. Properly calibrated and set you get very precise and consistent results.
I know this is not a wideband comparison, but I won't trust any AEM wideband to tune a forced induction engine, ever. I'd better use a Innovate instead of an AEM.
The AEM is good 30-0300 is great for NA tunes, stock or mildly modified cars only. Well, that's my opinion

Best Regards.

I don't have a dog in this fight, but thought I'd chime in anyway. I've only ever used an Innovate LC-1 on my old ride ('02 LS1 Camaro) and the AEM 30-0300 on this one (Viper)...I'm not a professional tuner, and would never claim to be. The CAN support is what pushed me towards the AEM unit - otherwise, I was going to go with the Ballenger AFR500v2. I must say that I've been happy with my choice thus far. I can't verify whether or not it is accurate since I don't really have anything to test against, but it has been extremely consistent. I was able to get my WOT fueling within 1-2% after the first logged runs, and after a couple more sessions it's down to a couple tenths of a percent. If it's telling me the wrong thing, it's at least doing so every single time.

I personally hope that other manufacturers step up to the plate and offer something similar - the ability to put the wideband data on the CAN and pull it off as a PID is incredibly convenient. One less thing to worry about calibrating is a win in my book.

[matty b]

Signal to noise ratio has nothing to do with accuracy, it just means that the signal its putting out is not corrupted or heavily filtered. Thats not a bad thing, but it only counts to the fidelity of the signal.

[dr.mike]

Signal to noise ratio has nothing to do with accuracy...

Not strictly true. In a case where the AFR value is constant ( like a stationary engine), you can just average as many sample, as needed, together, to get a value. In an automotive application, where the mixture values change rapidly, most error in AFR measurement comes from latency. And latency is GREATLY increased when it is necessary to average several samples, over time, to get a usable signal.

On the test bench, with calibrated test gasses, the AEM 30-0300 and NGK Powerdex read within 1% of each other. But, the 30-0300 responds to changes in AFR much faster. On the order of 10X faster. So, it is much more accurate in keeping the AFR data in synch with the RPM and manifold pressure signals, which change rapidly under acceleration, braking, shifting, etc. This makes sure that the AFR data winds up in the right fuel map cell, when logging.

A second source of error is the sensor's exhaust gas pressure sensitivity. At idle, where exhaust back-pressure is near zero, and AFR is near 14.7, there is no appreciable difference. But, as you approach maximum gas pressure, at WOT, and move away from 14.7 AFR, the AFR readings are affected.

The NTK sensors are about 4X as sensitive to pressure changes, as compared to the Bosch sensors.

As an example; at an exhaust gas pressure of ~6psi, an uncompensated NTK sensor will read 17.0AFR, when the actual value is 16.5AFR ( +0.5 AFR ) and 10.3AFR when the actual value is 11.0AFR ( -0.7AFR )

By comparison, the Bosch sensor will read 16.6AFR ( +0.1AFR) and 10.8AFR ( -0.2AFR ) under the same conditions.

So, at high exhaust gas flow / back-pressure, I would trust the AEM, more.

The sources of error in wideband measurements are, in order of descending magnitude:

1. Ground voltage offsets ( when using analog signals )
2. Response time/Latency.
3. Exhaust gas pressure effects.
4. After that, you get into the +/- 1% ranges of the differences between the steady-state responses of various units.

So for most reliable measurement you want:
a. Digital data transfer. Serial or, better, CANbus, for no offset errors.
b. The fastest possible response time. To avoid latency error.
c. either a pressure compensated system or VERY low back-pressure, or, at least, a Bosch sensor.

[gmtech16450yz]

Not strictly true. In a case where the AFR value is constant ( like a stationary engine), you can just average as many sample, as needed, together, to get a value. In an automotive application, where the mixture values change rapidly, most error in AFR measurement comes from latency. And latency is GREATLY increased when it is necessary to average several samples, over time, to get a usable signal.

On the test bench, with calibrated test gasses, the AEM 30-0300 and NGK Powerdex read within 1% of each other. But, the 30-0300 responds to changes in AFR much faster. On the order of 10X faster. So, it is much more accurate in keeping the AFR data in synch with the RPM and manifold pressure signals, which change rapidly under acceleration, braking, shifting, etc. This makes sure that the AFR data winds up in the right fuel map cell, when logging.

A second source of error is the sensor's exhaust gas pressure sensitivity. At idle, where exhaust back-pressure is near zero, and AFR is near 14.7, there is no appreciable difference. But, as you approach maximum gas pressure, at WOT, and move away from 14.7 AFR, the AFR readings are affected.

The NTK sensors are about 4X as sensitive to pressure changes, as compared to the Bosch sensors.

As an example; at an exhaust gas pressure of ~6psi, an uncompensated NTK sensor will read 17.0AFR, when the actual value is 16.5AFR ( +0.5 AFR ) and 10.3AFR when the actual value is 11.0AFR ( -0.7AFR )

By comparison, the Bosch sensor will read 16.6AFR ( +0.1AFR) and 10.8AFR ( -0.2AFR ) under the same conditions.

So, at high exhaust gas flow / back-pressure, I would trust the AEM, more.

The sources of error in wideband measurements are, in order of descending magnitude:

1. Ground voltage offsets ( when using analog signals )
2. Response time/Latency.
3. Exhaust gas pressure effects.
4. After that, you get into the +/- 1% ranges of the differences between the steady-state responses of various units.

So for most reliable measurement you want:
a. Digital data transfer. Serial or, better, CANbus, for no offset errors.
b. The fastest possible response time. To avoid latency error.
c. either a pressure compensated system or VERY low back-pressure, or, at least, a Bosch sensor.

OMG Mike I just got the email saying someone posted in this thread and when I saw it was you I HAD to go look!!! THANK YOU THANK YOU THANK YOU!!! I really felt like I was banging my head against the wall here. It's kind of hard to stay in the fight trying to spread the CORRECT information when a few closed minded people end up calling you all kinds of names and saying how ridiculously stupid you are. I had given up here because I figured it was a losing battle, but I felt confident that sooner or later what I, and now you were trying to say would start to sink in. Bottom line for ME is the AEM 30-0300's, going into HPTuners through the CAN network, has been the most accurate and reliable wideband I've ever seen. I'm no expert on the actual technical side of widebands like you are, but I do know how to tune with one. And tuning with the CAN based AEM's have been stupid easy!

Thanks again Mike! I know what I had posted here had nothing to do with why you jumped in, I'm sure you just happened to see some erroneous information and wanted to give your input. I'm so happy though because it was basically what I was trying to get across! Hopefully it will mean a lot more coming from someone like yourself that's expertise and occupation depends on knowing this stuff.

[gmtech16450yz]

Jesus Christ. You just don't get it.

It appears that Dr. Mike doesn't get it either steck. Maybe you, lsxpwrdz, matty b and montecarlodrag should educate him like you tried to educate me. I'd be interested in watching a technical discussion between you guys and him.

[jsllc]

I own them both.

Had dr. mike build 2 units for me. Had the same issues. My old tune using Ballenger AFR500 was dramatically different. I am lucky though as a friend works for DynoJet and they have a facility just a few miles away. Ran some tests with the car using a lab quality unit. The results showed that the AEM was the more accurate of the two. The issues that caused the deviation were 2. Response time/Latency. 3. Exhaust gas pressure effects. Mostly 2. In a nutshell both were very close with static content. The latency makes the logs inaccurate.

thanks again dr. mike for the units. they work great.

[dr.mike]

There is a lot of "religion" that makes its way into the whole wideband discussion. A lot of it has to do with the whole NTK vs. Bosch sensor issue.

Both sensors have their advantages and disadvantages. From a lot of testing both, on the bench and in road/dyno testing, the results are pretty clear.

Pluses for the NTK sensors.

1. NTK sensors are more durable, under high and changing EGT conditions.
Bosch sensors can suffer damage if operated outside of their intended temperature range.

2. NTK sensors have less sensitivity to EGT error in their readings.
Bosch sensors must maintain very good temperature control to maintain accuracy.

3. NTK sensors seem to be manufactured to tighter tolerances.
Bosch sensors use their LASER trimmed calibration resistor to "null out" variations.

4. NTK Sensors are easier to control, electronically.
Bosch sensors require more advanced PID controls to maintain proper temperature.

Pluses for the Bosch sensors.

1. The Bosch sensors have much lower latency in their measurements.
The NTK sensors are about 10X slower to respond, depending on the controller.

2. The Bosch sensor has a much lower exhaust gas pressure sensitivity error in its measurement.
The NTK sensors are about 4X more sensitive to exhaust gas pressure error.

3. Bosch publishes their calibration/trim resistor data. So controllers can use the sensors "out of the box"
NTK sensors, generally, require a free-air calibration to normalize he sensor output, in aftermarket applications.


On balance, I go with the Bosch sensors because of the latency and pressure sensitivity issues.
The temperature sensitivity advantage of the NTK sensor can be overcome with proper sensor temperature regulation.
Pressure sensitivity can not be compensated for without the use of an exhaust gas pressure sensor and additional electronics.

Also, "free-air" calibration is a bit of a mixed bag. It needs to be done very carefully, under controlled conditions, to get accurate results.

[schpenxel]

Facts in a discussion? Blasphemy.

I was hoping you'd chime in dr.mike. You're one of very few who actually has facts to add to this debate.

[10_SS]

Facts in a discussion? Blasphemy.

I was hoping you'd chime in dr.mike. You're one of very few who actually has facts to add to this debate.

where's the like button.

[gmtech16450yz]

Facts in a discussion? Blasphemy.

I was hoping you'd chime in dr.mike. You're one of very few who actually has facts to add to this debate.

Why do you think I jumped in? I didn't see a discussion, all I saw was 3 or 4 word smart ass comments based on nothing but emotion and brand bias. I tried to make it a discussion with actual data, but it just turned into a lynch mob trying to make the guy saying the earth is round look like an idiot. The idiot knows better than to enter a battle of wits with a bunch of unarmed men. lol. That's why I bowed out for awhile.

Update, the earth IS round. And no, you can't just universally say "Brand X is the best". That's an ignorant statement. Brand X might have some excellent products, but that doesn't mean the Brand Y, even if they have made crappy stuff in the past, can't come up with a better product. Anyone that knows electronics can tell you that A LOT of the time it's simply who has the LATEST product because technology is always advancing. Newer isn't always better, but in things like this it often is. CAN based widebands are kinda new. Analog wideband use WILL be known as the caveman approach once all the other brands are up to speed. That's a fact.

[schpenxel]

What data did you post? Your posts weren't much better than the other guys. You never made a succinct point or had any proof other than "it's faster" and telling me the points I was making weren't valid because they weren't under the same conditions, then tried to make points of your own that weren't any better.

We all know it's faster, no one is arguing that.

Mike's the only one in this thread that has truly tested both widebands under different conditions to prove which setup is better under different conditions.

And yes, brand bias is stupid. I wasn't a fan of AEM widebands either, but this one is a whole different ballgame.

[LSChev]

Since this is an HPTuners forum I am curious, is there a way to intergrate the CAN output from the AEM into the HPTuners scanner?

Thanks in advance.

[schpenxel]

Yes--that is the biggest positive of it

http://www.hptuners.com/forum/showth...d-beta-testers

[LSChev]

Awesome!

Now this makes more sense, I have some reading to do.

Cheers,

[gmtech16450yz]

schpenxel
What data did you post?

This was just one of my examples of data that I posted. This was on page one. Dr. Mike, does what I said here in this post make any sense to you? Seriously. I thought this was a pretty darn good example of what you can do with the data from the AEM 30-0300 wideband through CAN. I've personally never been able to achieve fuel control like this. No, this isn't a direct comparison between the AFX because I don't have one, but to ME it is clear "data" that shows not just fast but accurate wideband data. Am I full of cr@p in thinking this? Seriously, let me know.

Another point of argument in this thread was my premise that narrowbands can be a "sanity check" for your widebands. Again, Dr. Mike, does that hold water? Can you use the data from a narrowband sensor to compare to a wideband and get useful information from the comparisons? OBVIOUSLY it's not a direct correlation like some tried to say I said. I never said that. Of course you can't compare them voltage for voltage. I believe what you CAN do is compare transient swings and speeds of those swings, and also relative voltages, realizing that the ranges and curves for the two different sensors will NOT totally line up. I've seen wideband sensors that in logs are obviously reacting WAY slower to mixture changes than the factory narrowbands. Factory narrowbands are pretty fast. A lot of aftermarket widebands I've seen logged are nowhere near as fast. When I see that, I actually depend on the narrowband readings MORE than the wideband's slower reacting readings. If the narrowband sees something that the wideband totally misses, which data should you use? I'd say the narrowband. And that's talking about actually looking at the traces. If all you're depending on is histogram data, the wideband will be totally missing those cell hits because they never saw them. Again, I could be totally full of cr@p. Please let me and the others know the FACTS and TRUTH as far as your experience knows those to be. Thanks.

Here's an example of what happens when you're tuning with accurate wideband data. This is a 2 minute screenshot. Anyone that's compared widebands to narrowbands knows that anything off of stoich and the narrowbands are either going to go full rich or full lean. It's very hard to get a narrowband to actually read a steady mid-range value, they inherently want to fall towards rich or lean, that's how they work.

What this is showing you is that the AEM 30-0300 in this log is pretty much locked at 1.00 lambda, or stoich obviously. The mixtures ARE actually so close to stoich, and staying there consistently that the narrowbands will actually stay somewhat steady in their midrange. (I realize .450 is the actual center point, but expecting the narrowband to sit EXACTLY in it's center is asking a little too much.) This is telling me that the wideband reading 1.00 lambda HAS to be pretty darn close to actual stoich otherwise the narrowbands would hang on the rich or lean ends. (This engine is running in open loop so there is no O2/mixture switching. That's a different situation and of course the narrowband won't sit in the midrange if switching was on.) For you guys that are trying to learn something from this, switch off closed loop and O2 switching and see how your wideband compares to your narrowbands. Most likely even if your wideband is reading stoich, your narrowbands are going to be hanging around on either the rich or lean ends, not perfectly in the middle like this log shows. Dave, maybe you can show me an AFX wideband log that shows this same correlation? Again, if the wideband is even slightly off, the narrowbands WILL NOT sit steady like this. Getting narrowbands to sit steady at the crossover point is like balancing a beach ball on a seals nose. This screenshot shows a beach ball balanced on a seals nose for a full 2 minutes.

Schpenxel- if you want a sanity check for your wideband vs. narrowband, do the open loop test like this. When your wideband is reading stoich, your narrowbands will either be full rich or full lean. At least that will give you an idea of which way your wideband is biased. If you get lucky and your narrowbands hover near the midpoint like this screenshot, then your wideband is pretty darn accurate, at least in relation to your narrowbands. (Which are usually pretty darn accurate when working properly.)

[dr.mike]

Narrowband can give useful information. But, you have to avoid the temptation to think that you can relate a given NB voltage to a specific AFR/Lambda value, except for the midpoint at 14.7.
The voltages produced by a NB sensor are dependent on temperature and exhaust gas pressure, as well as air-fuel ratio.

Narrowband sensors are, generally, faster than widebands because they are "single-stage" direct reading sensors. While widebands are 2-stage sensors that work in a control loop, to produce readings. That's why there is no such thing as a "narrowband controller". Narrowband sensors can react to changes in air-fuel ratio in 1ms to 2ms. Which is much faster than any wideband. Some narrowband sensors have smaller gas port holes and, can be a bit slower. But, still, generally, better than 10ms.

While you can't convert the narrowband voltage directly to an AFR measurement, you CAN correlate changes in the narrowband voltage to changes in actual AFR. i.e. if the NB voltage jumps from 800mv to 900mv, you know that the mixture got richer, at that point. But, you need to interpret the wideband to see where it started and where it ended up. If you see an event on the NB sensor, there should be a matching event on the WB. The exception being right at 14.7, where the NB voltage swings wildly, over a very small AFR range.

So, the narrowbands are a good sanity check for the widebands. Especially if the NB and WB sensors are located close together, If you see a quick negative spike (lean) in the NB trace, there SHOULD be a corresponding positive spike (lean) on the WB trace, within just a few milliseconds. etc. If the NB is reading lean, the WB had better not be reading on the rich side. If the WB crosses 14.7, the NB should switch. If not, there is a problem.

If you don't have a fast wideband, divining the NB traces is the only clue that you get for gear-sift events, recovery from fuel cutt-off, et al. that occur over just a few milliseconds.

If you get lucky and your narrowbands hover near the midpoint like this screenshot, then your wideband is pretty darn accurate,

This is true and not true. That the WB and NB agree on what 14.7 ( lambda 1.000 ) is, is nice. But, 14.7 for a wideband is really "zero" for the sensor. You cant say that a WB is accurate because it reads 14.7 any more than you can say that a volt-meter is accurate because it reads 0.00v with the test-probes shorted together. It is a pretty crummy WB ( or volt-meter ) that can't show its zero value at "zero". To test accuracy you need a calibrated test gas, or for the volt-meter, a precision voltage reference, to compare with. i.e. if the voltmeter reads 5.000v with a precision 5v reference voltage, THEN it is accurate If a wideband reads Lambda 0.883 with the matching calibration gas, THEN it is accurate. But reading 14.7 ( Lambda 1.00 ) in 100% CO2 ( perfect 1.00 gas ) is, actually, not that impressive.


TLDNR

NB sensors are better at telling you WHEN the AFR changed than WHAT the AFR is.

[gmtech16450yz]

OMG I am CRACKING UP at something you said Mike! I even had to Google search it because I had no idea what it meant! This... TLDNR. hahahaha. After all these years of my posts usually being WAY too long, I've NEVER had anyone say that! How did I not know that? Hella funny. And yes, you taught me something that had absolutely nothing to do with cars or sensors! THANKS! lol.

Also, thank you so much for taking the time to post that. I read every word! Everything you said made perfect sense and pretty much mimics, agrees with or adds to what I've learned and understood about narrowbands and widebands. It's interesting, we come from very different areas of knowledge and careers but I feel that our different backgrounds give things like this a VERY wide view of how everything works. I have that "down in the trenches" side and you have the technical/Engineer side. I have actually gone to hundreds of GM Training Center classes that deal with things like this, but I'm no Engineer and I've never worked in a lab of any kind. I would imagine you've done your fair share of work under a hood too, but obviously the technical side of what we're talking about is where you shine. What an awesome combination! I'm very happy that you're here contributing! And you even bold important words like I put them in caps! You clearly realize the importance that ONE single word can make to a statement. haha.

So since you admitted that my novel was too long to read, can you please re-read the part about the narrowband and the wideband reading relatively steady for 2 minutes? I totally agree and understand what you said, but I'm not sure you realized that was in a car. That's a trace from HPTuners of my running car. THAT'S why I was impressed. Of course doing something like that in a lab with calibration gasses wouldn't be that impressive, but in a running car to be able to get the fueling that steady and accurate that the narrowband and wideband stayed that consistent is something I don't think I've ever seen in the real world/ running car situation. And yes, I also totally agree and understand that those values themselves (lol) mean very little and can't be directly compared. It's the relationship and the fact that the narrowband was hovering in the middle range and not on the lean or rich ends that was impressive. To me at least!

Again, Thank you SO much for taking the time to contribute to the discussion!!!

[wetcoast]

Can someone please point me in the direction of a discussion that addresses the accuracy of the analog to digital conversion done by the MPVI device?

[gmtech16450yz]

Can someone please point me in the direction of a discussion that addresses the accuracy of the analog to digital conversion done by the MPVI device?

Hmmm. Good question. Keith would probably be the one to answer that. My thoughts are it's not really so much a problem with whatever A/D converter is in the MPVI, it's more about what voltage offsets you plug in? It should be a simple conversion after that? I can't imagine it's anywhere near as complex as say an audio device converting an analog signal to digital or a digital file to an analog output. That's pretty complex and can absolutely be done "wrong".

My problem is with aftermarket widebands there are lists or presets of what offsets to use for each certain wideband. That's where I see the errors, I don't believe you can just plug in a "generic" set of numbers and have the scaling be exact for the particular vehicle YOUR wideband is wired into. I know my PLX and analog AEM widebands ended up being quite a ways off from what PLX and AEM listed as what to use. And now having the AEM through CAN, I could do a direct comparison and see that in my particular vehicle the "generic" offset values WERE way off of what the digital/CAN setup was showing. The values I had come up with on my own for my own application were very close to what the new CAN wideband setup shows, but certainly not as consistent as the vehicle's voltage and loads change.

Did that help? lol.