Page 2 of 3 FirstFirst 123 LastLast
Results 21 to 40 of 52

Thread: Ngk Afx wideband?

  1. #21
    Advanced Tuner
    Join Date
    Dec 2005
    Location
    Posts
    599
    What you are calling "overshoot" on the AEM is a known result of the pressure wave on switching gasses. i.e. it matches the output of a high-speed pressure sensor. It happens because the pump cell in the sensor has to work against the gas pressure until it equalizes with the test chamber.

    It is also visible on an LC-1, at full speed ; which does not use a PID at all and, has no concept of "overshoot". So far, only the LC-1 and the new AEM systems have a fast enough response to see it. There should be a similar event at the low-to-high transition. Unless your high test gas is off the scale ( e.g. air ) in which case, it will just clip and look squared off.


    Here is an LC-1 in "Instant mode" tracking the pressure wave.
    LC-1.JPG

    This is a 30-0300 on the same rig. With more resolution ( and less noise ) than the LC-1
    30-0300.JPG

  2. #22
    Advanced Tuner
    Join Date
    Dec 2005
    Location
    Posts
    599
    Can you try the same tests while triggering off of the solenoid voltage? That way the time before the signal gets into the display range is equalized.
    Nice clean signals on the v2

  3. #23
    Tuner in Training bmotorsports's Avatar
    Join Date
    Mar 2008
    Location
    Richmond, VA
    Posts
    34
    Quote Originally Posted by dr.mike View Post
    What you are calling "overshoot" on the AEM is a known result of the pressure wave on switching gasses. i.e. it matches the output of a high-speed pressure sensor. It happens because the pump cell in the sensor has to work against the gas pressure until it equalizes with the test chamber.
    I'm not sure why the words PID loop came out of my fingers but yes, you are right. I was trying to simplify the mechanics. That being said, the AFR500v2 represents the gas mixture more accurately in this case, disregarding the electromechanical effect on the sensor.

    Interesting graphs for LC-1 and AEM X-Series unit. How old is that LC-1? That noise is terrible though I know they were working on it through the product life.

    Quote Originally Posted by dr.mike View Post
    Can you try the same tests while triggering off of the solenoid voltage? That way the time before the signal gets into the display range is equalized.
    Nice clean signals on the v2
    You know your day isn't going well when you try to edit your long post with many images on your phone and it gets deleted instead. It is manual activation so I would need to set up a trigger signal. It is possible to do, there are a lot of other improvements and metrics I could add to the testing. It all takes time. If I ended up creating a lab grade testing environment to benchmark various systems and their response speed, I'd be in deep on time, effort, & cost. I looked at doing it before the AFR500v2 and calculated cost around $20k+ with all costs and time properly accounted for. Now to recover my post...
    Last edited by bmotorsports; 03-08-2016 at 07:42 PM.

  4. #24
    Tuner in Training bmotorsports's Avatar
    Join Date
    Mar 2008
    Location
    Richmond, VA
    Posts
    34
    Here are my results from sweep testing the AEM X-Series, AFR500v1, AFR500v2, AFRM GEN1, AFRM GEN2. This test is not a laboratory controlled test and I did this quickly just to get some results. You can call me impartial because I sell all three systems or not because one is branded. I can only say that I didn't favor any system or alter my testing in any way for one unit vs another. The AFR500v1 & v2 can use many different sensors, I only selected a couple for brevity and simplicity.

    This is only a sweep test for speed! This is just ONE criteria for a good wideband system and is only meant to specifically address the point of response speed brought up in this thread. I chose the scope parameters to make my life easier.

    The AEM X-Series with the Bosch LSU 4.9 is FAST! There is overshoot and undershoot as the wideband goes rich but doesn't stay near max rich:



    The AFR500v1 with the NTK Calibration Grade Sensor is slower but suffers no overshoot, hitting its min and max targets properly:



    The AFR500v2 with the NTK Calibration Grade Sensor is FAST! As the AFR500 has no overshoot/undershoot, it reaches an accurate, stabilized value faster than the AEM!



    The AFR500v2 with the Bosch LSU 4.9 is FAST! You can see the AFR500v2 is presenting more reasonable values than the AEM overshoot / undershoot when going rich. The time to reach a stabilized, accurate value is similar with an edge going to the AFR500v2.



    The NTK AFRM GEN1 & GEN2 showed similar performance on the analog output speed. There is a 100ms pause between value changes on the analog output which is not ideal. I had hoped that NTK would address this in the GEN2 but I see it is the same. That said, the ZFAS-U2 sensor has a very small tip which may have affected testing. I still like this unit for a host of other reasons but it is clearly not in the ballpark of the AFR500v2 or AEM X-Series for transient response times.



    I did not expect the AFR500v2 to outperform the AEM X-Series on response time. I'm happy to see that we have this new crop of amazing products available to the consumer. This is better than it has ever been at every level of tuning & performance.

    Pick any of the three below as they are all solid options:
    AFR500v2
    AEM X-Series
    NTK AFRM GEN2

  5. #25
    Advanced Tuner
    Join Date
    Dec 2005
    Location
    Posts
    599
    I hate when that happens ... grrrr.

    The reason I ask about the trigger point is because, in your pics, the AEM gauge shows a much steeper ( almost brick-wall vertical ) slew rate, entering the display range. So, it likely spent less time between the gas change and the 1st deflection on the analog signal. In the 2 pics I posted, the purple line is the solenoid switching. There is about 12ms from solenoid to the 1st deflection for both the AEM and LC-1.

    In my pic, the "dip" is lambda 0.865 with a gas of 0.885. ( it varies with the gas pressure ) So, it is possible that the V2 has less error ( < 0.02 lambda ) over that period, as long as it is not still showing the higher gas value, at that time. Maybe I should get one to play with

    Looking again, the AFR500v2 seems to be going rail-to-rail, all the way from 5v to 0v? So, it couldn't overshoot, if it wanted to ? What are the gas values ?
    i.e. the stabilized value is the 0v rail (ground).

    Rough eyeballing it from the AEM trace, it looks like the rich gas is about lambda 0.7? Is 0.7 lambda = 0.0v on the output of the AFR500v2 ?
    Last edited by dr.mike; 03-08-2016 at 08:33 PM.

  6. #26
    Advanced Tuner
    Join Date
    Dec 2005
    Location
    Posts
    599
    Just to be clear, I am not picking on the AFR500's ( Powerdex ) they are among the best made widebands available. And, I have tested and taken apart just about all of them.

  7. #27
    Tuner in Training bmotorsports's Avatar
    Join Date
    Mar 2008
    Location
    Richmond, VA
    Posts
    34
    Quote Originally Posted by dr.mike View Post
    Just to be clear, I am not picking on the AFR500's ( Powerdex ) they are among the best made widebands available. And, I have tested and taken apart just about all of them.
    I thought about it and went through my notes while everything was still fresh. My final conclusion is that both the AFR500v2 and AEM X-Series are amazing widebands for speed. The AFR500v2 is amazing for a lot of other reasons and the AEM X-Series has a strong feature set.

    Quote Originally Posted by dr.mike View Post
    The reason I ask about the trigger point is because, in your pics, the AEM gauge shows a much steeper ( almost brick-wall vertical ) slew rate, entering the display range. So, it likely spent less time between the gas change and the 1st deflection on the analog signal. In the 2 pics I posted, the purple line is the solenoid switching. There is about 12ms from solenoid to the 1st deflection for both the AEM and LC-1.

    In my pic, the "dip" is lambda 0.865 with a gas of 0.885. ( it varies with the gas pressure ) So, it is possible that the V2 has less error ( < 0.02 lambda ) over that period, as long as it is not still showing the higher gas value, at that time. Maybe I should get one to play with

    Looking again, the AFR500v2 seems to be going rail-to-rail, all the way from 5v to 0v? So, it couldn't overshoot, if it wanted to ? What are the gas values ?
    i.e. the stabilized value is the 0v rail (ground).

    Rough eyeballing it from the AEM trace, it looks like the rich gas is about lambda 0.7? Is 0.7 lambda = 0.0v on the output of the AFR500v2 ?
    The overshoot / undershoot / stabilization is what took time for the AEM. I did some more testing today. I ran into a few oddities. When testing different power supplies, the AEM X-Series analog didn't work properly on a 10Ah battery. I also found the analog stopped responding before the gauge stopped moving lean (the gauge reads higher than 20:1). I will have to look at the battery issue later.

    Regarding the overshoot/undershoot on the AEM I was able to validate this was due to excessive flow rate. I was able to reduce the flow rate and get an excellent response graph. The AEM is crazy fast and is right there on speed with the AFR500v2 with the NTK sensor. That said, at the flow rates I was testing yesterday (not measured but consistent for the days testing), the AFR500v2 did not exhibit the overshoot/undershoot problem.




    I'll reiterate that this testing is only for response speed! This is just ONE criteria for a good wideband.

    Both the AFR500v2 and the AEM X-Series are excellent systems. We sell and can wholeheartedly recommend both. The AFR500v2 has an edge in sensor flexibility & reliability with extreme accuracy, precision, and speed in all conditions. The AEM X-Series has shown it is blazingly fast with an excellent feature set.
    Last edited by bmotorsports; 03-09-2016 at 05:58 PM.

  8. #28
    Advanced Tuner
    Join Date
    Dec 2005
    Location
    Posts
    599
    1. You are right about the analog output vs. display. The analog output was made to match the old AEM 4100/4110 gauges.
    The range is 0.5v = Lambda 0.58 to 4.5v = Lambda 1.22. 1.22 is about 18.0AFR, which was the max for the old gauges.
    The new 30-0300 display and serial output go to 20.0AFR. And, the CANBus data goes to about lambda 6.6 ( ~100AFR ).
    But, also gives Oxygen % data up to 20.9% ( free-air ). The CAN data also goes down to about lambda 0.55 ( 8.0AFR ). Or maybe 0.50.... I forget.

    The extra 0.5v at the top and bottom of the range are there to give headspace to deal with ground offsets.

    It looks like you are running into the 0.5v "soft rail" limit on your new 30-0300 trace. This would be a lambda gas of about ~0.58. The analog output will not go below this, unless signalling an error, or during warmup. So, any "overshoot" to about lambda 0.56, would be hidden by clipping.

    2. On the overshoot vs. gas flow. It is true that reducing the pressure will reduce the "overshoot". But, it should still be visible.

    It's really difficult setting up this kind of test because the amount of pressure needed to simulate fast transitions that happen in the high speed/volume exhaust stream, using a little solenoid and hose, cause those pressure waves.

  9. #29
    SeƱor Tuner MeentSS02's Avatar
    Join Date
    Jun 2005
    Location
    Dayton, OH
    Posts
    1,132
    Quote Originally Posted by bmotorsports View Post
    Regarding the overshoot/undershoot on the AEM I was able to validate this was due to excessive flow rate. I was able to reduce the flow rate and get an excellent response graph. The AEM is crazy fast and is right there on speed with the AFR500v2 with the NTK sensor. That said, at the flow rates I was testing yesterday (not measured but consistent for the days testing), the AFR500v2 did not exhibit the overshoot/undershoot problem.
    How does the V2 fare with the NTK production grade sensor since it appears that you used the calibration grade variant for the test?
    2008 Viper - now with HPToona - 1/4 Mile Shenanigans Here
    11.02 @ 130

  10. #30
    Advanced Tuner
    Join Date
    Dec 2005
    Location
    Posts
    599
    That said, at the flow rates I was testing yesterday (not measured but consistent for the days testing), the AFR500v2 did not exhibit the overshoot/undershoot problem.
    That was my original point. it COULDN'T overshoot because the rich gas value was 0v. Unless the analog output can generate negative voltages ? 0v just means lambda 0.62 or lower.

    Same for the 0.5v reading on the AEM. it just means lambda 0.58 or lower.

    That's why it is better to make these tests with either CO2 ( lambda 1.00 ) or BAR97 H gas ( lambda 0.885 ) That way, you stay in the usable range, and, away from the clipping rails.

    Same goes for the rich-to-lean transitions. Straight air gives lambda infinity. So it always clips to the maximum reading ( 5v, 4.5v, etc. ). A mixture of 3% O2 in N2 will give a lambda value of 1.18; which should give an actual measurement reading after the transition. Instead of just pushing to the clipping rail voltage.

    Short version: analog out voltages at the clipping rails ( 0v, 5v, etc. ) are ambiguous. And, cant be used to demonstrate the measurement characteristics of the devices.

    Like I said, these tests are really hard to do right

  11. #31
    Tuner in Training bmotorsports's Avatar
    Join Date
    Mar 2008
    Location
    Richmond, VA
    Posts
    34
    Quote Originally Posted by bmotorsports View Post
    Here are my results from sweep testing the AEM X-Series, AFR500v1, AFR500v2, AFRM GEN1, AFRM GEN2. This test is not a laboratory controlled test and I did this quickly just to get some results.
    Quote Originally Posted by bmotorsports View Post
    It is possible to do, there are a lot of other improvements and metrics I could add to the testing. It all takes time. If I ended up creating a lab grade testing environment to benchmark various systems and their response speed, I'd be in deep on time, effort, & cost. I looked at doing it before the AFR500v2 and calculated cost around $20k+ with all costs and time properly accounted for.
    Quote Originally Posted by dr.mike View Post
    Like I said, these tests are really hard to do right
    We completely agree on this point. It is hard, takes a lot of time and a lot of money. This is why I chose to announce the VERY limited scope of the testing.

    Quote Originally Posted by dr.mike View Post
    That was my original point. it COULDN'T overshoot because the rich gas value was 0v. Unless the analog output can generate negative voltages ? 0v just means lambda 0.62 or lower.

    Same for the 0.5v reading on the AEM. it just means lambda 0.58 or lower.
    The overshoot/undershoot was happening at ~0.73, 1.4v lambda on the AEM X-Series which is in range on the AFR500v2. With the VERY limited scope of testing, this would take more testing to flush out and examine. I re-tested the AEM X-Series the next day and found it had excellent performance at the reduced flow rate. Thank you for the on-point questions that led me to further thinking & testing to the benefit of consumers who find better information as a result.

    Quote Originally Posted by dr.mike View Post
    That's why it is better to make these tests with either CO2 ( lambda 1.00 ) or BAR97 H gas ( lambda 0.885 ) That way, you stay in the usable range, and, away from the clipping rails.

    Same goes for the rich-to-lean transitions. Straight air gives lambda infinity. So it always clips to the maximum reading ( 5v, 4.5v, etc. ). A mixture of 3% O2 in N2 will give a lambda value of 1.18; which should give an actual measurement reading after the transition. Instead of just pushing to the clipping rail voltage.

    Short version: analog out voltages at the clipping rails ( 0v, 5v, etc. ) are ambiguous. And, cant be used to demonstrate the measurement characteristics of the devices.
    Yes. You could use a lot of different test mixes, etc. We could take over the thread and talk about testing specifics for a very long time. I wanted to perform a VERY limited scope test for analog output rail to rail response on the meters I have on hand. I've already spent a lot more time on images & posts than testing . In this very limited scope, I think I've managed to achieve what I set out to do, which is address the question of speed & response. The short version is that the AEM X-Series & AFR500v2 excel. The NTK AFRM has a 100ms stepped response that is not ideal.

    I can wholeheartedly recommend the Ballenger Motorsports AFR500v2 as well as the AEM X-Series for response speed based on my testing.

  12. #32
    Tuner in Training bmotorsports's Avatar
    Join Date
    Mar 2008
    Location
    Richmond, VA
    Posts
    34
    Quote Originally Posted by MeentSS02 View Post
    How does the V2 fare with the NTK production grade sensor since it appears that you used the calibration grade variant for the test?
    We used the NTK Calibration Grade sensor and the Bosch LSU 4.9 in the prior tests.

    As for the NTK Production Grade sensor, the results are very good! The AFR500v2 is MUCH faster than the AFR500v1. The NTK Calibration grade shows an edge as the mixture gets very rich and we recommend the Calibration grade sensor for very rich mixtures and methanol applications. The NTK sensor options are less susceptible to lead & exotic fuel fouling than the LSU sensors and are well known for their reliability, longevity, and accuracy.



  13. #33
    Advanced Tuner
    Join Date
    Dec 2005
    Location
    Posts
    599
    The overshoot/undershoot was happening at ~0.73, 1.4v lambda on the AEM X-Series which is in range on the AFR500v2.
    I guess I am probably just not saying it right. Yes, 1.4v is about 0.73 lambda on the AEM output. What is confusing me is that the same 0.73 lambda gas should read about 1.14v ( in range ) on the AFR500V2 ( from the manual, "Lambda = 0.62 + Vout x 0.096" ). But, your traces show the AFR500V2 vout as 0.0v ( where the "1->" marker is ) Which would be out of range. Unless you have a different scaling set up ?

    It could just be me. I have probably looked at more of these transition traces than any three people on the planet. So, everything just kinda jumps out at me.

    I'll grab an AFR500V2 w/ the calibration grade sensor to test, as soon as they are available. The Lab grade sensor is too rich for my blood.
    Maybe I can make an OBDII adapter for it

  14. #34
    Moderator
    Join Date
    Mar 2014
    Location
    Raleigh, NC
    Posts
    6,347
    I vote this best thread ever

  15. #35
    Tuner in Training bmotorsports's Avatar
    Join Date
    Mar 2008
    Location
    Richmond, VA
    Posts
    34
    Quote Originally Posted by dr.mike View Post
    I guess I am probably just not saying it right. Yes, 1.4v is about 0.73 lambda on the AEM output. What is confusing me is that the same 0.73 lambda gas should read about 1.14v ( in range ) on the AFR500V2 ( from the manual, "Lambda = 0.62 + Vout x 0.096" ). But, your traces show the AFR500V2 vout as 0.0v ( where the "1->" marker is ) Which would be out of range. Unless you have a different scaling set up ?

    It could just be me. I have probably looked at more of these transition traces than any three people on the planet. So, everything just kinda jumps out at me.

    I'll grab an AFR500V2 w/ the calibration grade sensor to test, as soon as they are available. The Lab grade sensor is too rich for my blood.
    Maybe I can make an OBDII adapter for it
    Scaling is the same. You are seeing the same issue I described on initial testing. The X-Series should have shot down to 0.5v on the rail and held there. AFR500v2 didn't have the issue, AFRM didn't, reduced flow rate resolved on second test I did with AEM X-Series which was beautiful. Could be testing error or a number of other factors on this limited scope, non laboratory test.

    Just a few more weeks for the AFR500v2 units, expecting to ship 2nd week of April 2016. An OBDII adapter would be a nice addition!

    Quote Originally Posted by schpenxel View Post
    I vote this best thread ever

  16. #36
    Advanced Tuner
    Join Date
    Dec 2005
    Location
    Posts
    599
    The X-Series should have shot down to 0.5v on the rail and held there. AFR500v2 didn't have the issue,
    If the AEM gauge "should have shot down to 0.5v on the rail and held there.", then, the test gas must be richer than 0.58 lambda. Which is 0.5v on the AEM.

    0.58 lambda is "out of range" for the AFR500; as lambda 0.62 is at 0.0v.

    If you are using a gas that is SO rich that it forces both units past the rails ( like lambda 0.50 ) then you are essentially treating these units like narrowbands. As both the lean and rich gasses are outside of the usable ranges. You can't really deduce anything about speed or accuracy from that. Because, who knows how long it would take either unit to settle on a value of 0.50 lambda, if they COULD display it. That all happens "off-screen" Same goes for how long it took for the 1st reaction to show up, after the gas change, moving off of the high-side rail. ( 5v or 4.5v )

    With 0.50 gas, how long is takes to get to 0.58 or 0.62 does not tell you much. as a LOT of the time is taken, settling that last few 0.001's of lambda. And, that is hidden, if you are off the scale, "beyond the rails".

  17. #37
    Moderator
    Join Date
    Mar 2014
    Location
    Raleigh, NC
    Posts
    6,347
    Ok, what the heck do you two do for a living?

  18. #38
    Advanced Tuner
    Join Date
    Dec 2005
    Location
    Posts
    599
    Mixed-signal instrumentation engineer.

    In the last year, I characterized the functionality of several widebands, before they were tested by 360 Product testing, for AEM.

    These are the response time results that 360 degree guys got. My results were similar except:

    1. My numbers were all about 10ms faster, across the board because 360deg included the solenoid reaction and gas transport times.
    2. My LC-1 results were about 6ms faster than theirs. On top of the 10ms difference.
    3. The noise level on the PLX and Motec units contributed to their outputs crossing the t63 boundaries before their actual signal did.
    i.e. once you average together enough samples to get the noise down to a usable level, they are much slower than shown.

    ( From AEM's website )
    AEM_X-Series_Wideband_Response_Time_Comparison_Chart.jpg

  19. #39
    Tuner in Training bmotorsports's Avatar
    Join Date
    Mar 2008
    Location
    Richmond, VA
    Posts
    34
    Quote Originally Posted by dr.mike View Post
    If the AEM gauge "should have shot down to 0.5v on the rail and held there.", then, the test gas must be richer than 0.58 lambda. Which is 0.5v on the AEM.

    0.58 lambda is "out of range" for the AFR500; as lambda 0.62 is at 0.0v.

    If you are using a gas that is SO rich that it forces both units past the rails ( like lambda 0.50 ) then you are essentially treating these units like narrowbands. As both the lean and rich gasses are outside of the usable ranges. You can't really deduce anything about speed or accuracy from that. Because, who knows how long it would take either unit to settle on a value of 0.50 lambda, if they COULD display it. That all happens "off-screen" Same goes for how long it took for the 1st reaction to show up, after the gas change, moving off of the high-side rail. ( 5v or 4.5v )

    With 0.50 gas, how long is takes to get to 0.58 or 0.62 does not tell you much. as a LOT of the time is taken, settling that last few 0.001's of lambda. And, that is hidden, if you are off the scale, "beyond the rails".
    The gas is richer than the scale on all systems tested. This was not a test of accuracy, only speed as I've mentioned. I've announced the VERY limited scope with all results. You seem to keep wanting it to be more than it is. It isn't. We've covered that a proper lab test takes a lot of time, effort, money, and is difficult.

    This is just a sweep test with a rich gas done VERY quickly on a bench, not in a lab setting. The test achieved the specific goal of showing how times vary between meters & sensors at maximum switching speed. Testing was done, goals were achieved, results showed variation. I re-tested the AEM X-Series and it performed well, per posted results. Any proper lab test on on the AFR500v2 will have to wait for production hardware due out shortly.

    I think we've had a great discussion but I want to keep the thread on topic. We could derail this thread for days talking about my 2 hour test that was intentionally limited and how it is not an expensive, time consuming, laboratory test item by item.

    Quote Originally Posted by schpenxel View Post
    Ok, what the heck do you two do for a living?
    Engineers working in automotive controls

  20. #40
    Advanced Tuner
    Join Date
    Dec 2005
    Location
    Posts
    599
    Now I see what you did Yes. You will get those kind of traces, if you use a super-rich gas that is way off the scale.

    That's why I kept asking what the gas was. Now, it makes sense