NO wideband with flash!!!!!
Jan 1, 2007, 10:57 AM
#16
Evolved Member
iTrader: (58)
Join Date: Dec 2003
Location: Princeton, NJ
Posts: 3,120
Likes: 0
Received 0 Likes
on
0 Posts
i would get a wideband. this way if you add more powertrain parts, you can make small adjustments. or if you decide you want to run more racegas at certain situations/events, you can tune yourself a map based on the vendors map but slightly leaner and timing.
personally this is what i am doing.
personally this is what i am doing.
Jan 1, 2007, 11:05 AM
#17
Evolving Member
Join Date: Dec 2006
Posts: 276
Likes: 0
Received 0 Likes
on
0 Posts
Also, note that accuracy varied wildly accross the board in a few states in a couple of test cases. If you plot these results and do a statistical deviation, you will see that, regardless of ranking, all the units accept the benchmark unit (which was accepted as Gospel) are dramatically outside their manufacturer's specs.
I'm sorry, I realize that you may like the promotional implications, but you cannot accept the rankings if you do not accept the raw data. I say the results are bunk, with a relatively new sensor in good condition, Zeitronix should be perfectly suitable for a meth injection failsafe, but if you want to argue that it is off by .6 AFR when a I need it most, well, that is another story!
I think that there are two fundemental flaws in the test, both of which probably stem from a lack of understanding of instrumentation in general and lambda measurements in particular.
First, the benchmark was wholly arbitrary. Like you, the 'testers' appear to have been impressed with price ("$1200 MOTEK"). However, no effort appears to have been made to validate the accuracy or reponse time of the NTK unit with calibrated gasses. For an explanation of why this is necessary, I'd suggest Bowling and Grippo's excellent articles on their open efforts to develop a precision wideband controller.
Second, all the units were monitored at the same time. There are two implications to this, first, the sensors were at different locations in the exhaust stream. Second, no effort was made to profile the response times of the various controllers - and response times vary dramatically.
Among the inexperienced, or people who have only been exposed to slow, current based, wideband measurements, there is sometimes a perception that AFR (or more precisely lambda) is a very smooth and steady thing that can be measure in the exhaust. This is absolutely false. Nozzles are imperfect, fuel balance between cyls is never perfect, and inducution systems always have some pressure waves. Cycle to cycle there is constant fluctuation.
Again, if you are only used to sluggish and not super precise measurements, these constant fluctuations can come as a surprise. Since the article writers probably weren't even aware that fluctuations exist, it is natural that it did not occur to them that the exact AFR at different points in the exhaust system are not identical at any given instant in time.
Regarding time, it almost certainly did not occur to them that the units themselves were shifted various amounts in time in their readings, depending on their exact measurement implementation and the specific sensor used. In other words, it never occured to them that the readings on an LC-1 were typically 5 to 50 mS old while most of the current based measurements where 150 to 400 mS old.
So you have comparisons of different points on a fluctuating stream with instruments of varying lag times on a car that was probably not being held in a particularly precise, steady state - all against a benchmark that was wholly untested...
Still, I don't think methodology alone explains these outlier results (we participate in published shootouts and testing all the time, so we've got a pretty good idea what to expect from various units). I think that the testers were probably also making additional mistakes. For example, if any of the units, including the NTK sensor, were overheated or mounted in an improper orientation, the readings could be significantly off. As far as I know, only we - and other units licensing our measurement principle, can detect non-catastrophic sensor overheating. Current based units simply report erroneous readings.
Or, perhaps the testers were attempting to use tail pipe clamps, which have many gotchas of their own. Maybe a bung had a small air leak - or multiple bungs were place too close together (the sensors switch a couple of amps of a relatively high impedence load - and can adversely effect each other in close proximity)...
Seriously, if it had been a simple case of our unit testing poorly, that would be one thing. I'd still be thinking defective sensor or possible operator error (calibration?), but I wouldn't be highly suspicious of the overall methodology and results. But when all the units vary wildly and are grossly out of spec? I'm sorry, odds are just incredibly high that the testers didn't know what they were doing.
-jjf
Last edited by CharlesJ; Jan 2, 2007 at 01:17 PM.
Jan 1, 2007, 11:21 AM
#18
Evolving Member
Join Date: Dec 2006
Posts: 276
Likes: 0
Received 0 Likes
on
0 Posts
What does change is the speed of measurements. Lead literally forms a type of crystal coating on the sensor, so our pulses get longer and longer (you can literally see it on a current probe and a scope on the feedback). For optimum tuning, speed does matter, since delayed results can mislead you when it comes to charting fuel and timing maps. So I always recommend keeping a 'tuning' sensor and a day to day sensor. (Edit: It should be noted that Klaus, our guru and inventor of the measurement principle, poo-poo's this. His stance is that, even our 'slow' readings are fast by industry standards - but two sensors is how I approach it.)
With current based measurement, depending on the exact design, the speed of measurements remains relatively constant (current systems are generally relatively slow to begin with), but the accuracy of the readings will suffer over time. Like speed, accuracy matters, so perhaps those users would benefit from the same approach (two sensors).
-jjf
Last edited by jfitzpat; Jan 1, 2007 at 11:25 AM.
Jan 1, 2007, 01:29 PM
#19
Evolving Member
iTrader: (6)
Join Date: Feb 2004
Location: Maryland
Posts: 442
Likes: 0
Received 0 Likes
on
0 Posts
But it also tested out of Zeitronix's spec by a factor of 6 (off by .6 AFR, instead of the .1 AFR spec'ed) at peak pressure and RPM. If you convert this to lambda, it would appear to be too innacurate for an aggressive meth injection tune.
Also, note that accuracy varied wildly accross the board in a few states in a couple of test cases. If you plot these results and do a statistical deviation, you will see that, regardless of ranking, all the units accept the benchmark unit (which was accepted as Gospel) are dramatically outside their manufacturer's specs.
I'm sorry, I realize that you may like the promotional implications, but you cannot accept the rankings if you do not accept the raw data. I say the results are bunk, with a relatively new sensor in good condition, Zeitronix should be perfectly suitable for a meth injection failsafe, but if you want to argue that it is off by .6 AFR when a I need it most, well, that is another story!
I think that there are two fundemental flaws in the test, both of which probably stem from a lack of understanding of instrumentation in general and lambda measurements in particular.
First, the benchmark was wholly arbitrary. Like you, the 'testers' appear to have been impressed with price ("$1200 MOTEK"). However, no effort appears to have been made to validate the accuracy or reponse time of the NTK unit with calibrated gasses. For an explanation of why this is necessary, I'd suggest Bowling and Grippo's excellent articles on their open efforts to develop a precision wideband controller.
Second, all the units were monitored at the same time. There are two implications to this, first, the sensors were at different locations in the exhaust stream. Second, no effort was made to profile the response times of the various controllers - and response times vary dramatically.
Among the inexperienced, or people who have only been exposed to slow, current based, wideband measurements, there is sometimes a perception that AFR (or more precisely lambda) is a very smooth and steady thing that can be measure in the exhaust. This is absolutely false. Nozzles are imperfect, fuel balance between cyls is never perfect, and inducution systems always have some pressure waves. Cycle to cycle there is constant fluctuation.
Again, if you are only used to sluggish and not super precise measurements, these constant fluctuations can come as a surprise. Since the article writers probably weren't even aware that fluctuations exist, it is natural that it did not occur to them that the exact AFR at different points in the exhaust system are not identical at any given instant in time.
Regarding time, it almost certainly did not occur to them that the units themselves were shifted various amounts in time in their readings, depending on their exact measurement implementation and the specific sensor used. In other words, it never occured to them that the readings on an LC-1 were typically 5 to 50 mS old while most of the current based measurements where 150 to 400 mS old.
So you have comparisons of different points on a fluctuating stream with instruments of varying lag times on a car that was probably not being held in a particularly precise, steady state - all against a benchmark that was wholly untested...
Still, I don't think methodology alone explains these outlier results (we participate in published shootouts and testing all the time, so we've got a pretty good idea what to expect from various units). I think that the testers were probably also making additional mistakes. For example, if any of the units, including the NTK sensor, were overheated or mounted in an improper orientation, the readings could be significantly off. As far as I know, only we - and other units licensing our measurement principle, can detect non-catastrophic sensor overheating. Current based units simply report erroneous readings.
Or, perhaps the testers were attempting to use tail pipe clamps, which have many gotchas of their own. Maybe a bung had a small air leak - or multiple bungs were place too close together (the sensors switch a couple of amps of a relatively high impedence load - and can adversely effect each other in close proximity)...
Seriously, if it had been a simple case of our unit testing poorly, that would be one thing. I'd still be thinking defective sensor or possible operator error (calibration?), but I wouldn't be highly suspicious of the overall methodology and results. But when all the units vary wildly and are grossly out of spec? I'm sorry, odds are just incredibly high that the testers didn't know what they were doing.
-jjf
Also, note that accuracy varied wildly accross the board in a few states in a couple of test cases. If you plot these results and do a statistical deviation, you will see that, regardless of ranking, all the units accept the benchmark unit (which was accepted as Gospel) are dramatically outside their manufacturer's specs.
I'm sorry, I realize that you may like the promotional implications, but you cannot accept the rankings if you do not accept the raw data. I say the results are bunk, with a relatively new sensor in good condition, Zeitronix should be perfectly suitable for a meth injection failsafe, but if you want to argue that it is off by .6 AFR when a I need it most, well, that is another story!
I think that there are two fundemental flaws in the test, both of which probably stem from a lack of understanding of instrumentation in general and lambda measurements in particular.
First, the benchmark was wholly arbitrary. Like you, the 'testers' appear to have been impressed with price ("$1200 MOTEK"). However, no effort appears to have been made to validate the accuracy or reponse time of the NTK unit with calibrated gasses. For an explanation of why this is necessary, I'd suggest Bowling and Grippo's excellent articles on their open efforts to develop a precision wideband controller.
Second, all the units were monitored at the same time. There are two implications to this, first, the sensors were at different locations in the exhaust stream. Second, no effort was made to profile the response times of the various controllers - and response times vary dramatically.
Among the inexperienced, or people who have only been exposed to slow, current based, wideband measurements, there is sometimes a perception that AFR (or more precisely lambda) is a very smooth and steady thing that can be measure in the exhaust. This is absolutely false. Nozzles are imperfect, fuel balance between cyls is never perfect, and inducution systems always have some pressure waves. Cycle to cycle there is constant fluctuation.
Again, if you are only used to sluggish and not super precise measurements, these constant fluctuations can come as a surprise. Since the article writers probably weren't even aware that fluctuations exist, it is natural that it did not occur to them that the exact AFR at different points in the exhaust system are not identical at any given instant in time.
Regarding time, it almost certainly did not occur to them that the units themselves were shifted various amounts in time in their readings, depending on their exact measurement implementation and the specific sensor used. In other words, it never occured to them that the readings on an LC-1 were typically 5 to 50 mS old while most of the current based measurements where 150 to 400 mS old.
So you have comparisons of different points on a fluctuating stream with instruments of varying lag times on a car that was probably not being held in a particularly precise, steady state - all against a benchmark that was wholly untested...
Still, I don't think methodology alone explains these outlier results (we participate in published shootouts and testing all the time, so we've got a pretty good idea what to expect from various units). I think that the testers were probably also making additional mistakes. For example, if any of the units, including the NTK sensor, were overheated or mounted in an improper orientation, the readings could be significantly off. As far as I know, only we - and other units licensing our measurement principle, can detect non-catastrophic sensor overheating. Current based units simply report erroneous readings.
Or, perhaps the testers were attempting to use tail pipe clamps, which have many gotchas of their own. Maybe a bung had a small air leak - or multiple bungs were place too close together (the sensors switch a couple of amps of a relatively high impedence load - and can adversely effect each other in close proximity)...
Seriously, if it had been a simple case of our unit testing poorly, that would be one thing. I'd still be thinking defective sensor or possible operator error (calibration?), but I wouldn't be highly suspicious of the overall methodology and results. But when all the units vary wildly and are grossly out of spec? I'm sorry, odds are just incredibly high that the testers didn't know what they were doing.
-jjf
Thanks for the post.
Last edited by CharlesJ; Jan 2, 2007 at 01:19 PM.
Jan 2, 2007, 11:17 AM
#20
Evolving Member
Join Date: Dec 2006
Posts: 276
Likes: 0
Received 0 Likes
on
0 Posts
I made reference to some articles earlier. An outstanding explanation of how a modern O2 sensor works, as well as an explanation of the current measurement principle, can be found here. Don't let the pictures daunt you, the material is extremely well presented.
I'm going to have to wrap up my participation here. It is very difficult for me to talk about technicalities without relating it to our specific products. This appears to put me squarely into forbidden territory. Rather than make more work and grief for the moderators (they not only have to edit/delete my posts, but any references to them), I'll lay low.
I hope my participation to this point is perceived by most folks here in the spirit I truly intended it.
Best Regards,
-jjf
I'm going to have to wrap up my participation here. It is very difficult for me to talk about technicalities without relating it to our specific products. This appears to put me squarely into forbidden territory. Rather than make more work and grief for the moderators (they not only have to edit/delete my posts, but any references to them), I'll lay low.
I hope my participation to this point is perceived by most folks here in the spirit I truly intended it.
Best Regards,
-jjf
Thread
Thread Starter
Forum
Replies
Last Post
