Lol, not really. I was hoping to see a discussion on AFR/Lambda from the controller vs. what's being reported by the MPVI. There has to be a point of diminishing returns on the accuracy of the input when compared to the accuracy of the WB output.Originally Posted by gmtech16450yz
Oh. I thought that's kinda the whole point of this? We must be thinking about different aspects of it? Anyway, I don't think it's so much of an accuracy issue as it is a speed and consistency issue. If the data is not EXACTLY as accurate as you would get in a lab environment, but the speed and consistency is good, I'd take that over exact numbers. It's kind of like using a DVOM, I can live with one meter that shows .1 volt higher than another IF it's consistent in that error and can read those voltages quick enough for the task at hand. 12.1 volts or 12.2 volts in a lot of cases doesn't matter as much as if it's a consistant voltage or is it dropping out at times. Speed is most definitely a factor there, if the meter is too slow to see the drop, it's useless even if it's accuracy is +/- .00000001 volt. And accuracy also, if every single car I check with that meter reads 12.1 volts, and I get one that's reading 12.9 volts, I'm not going to be so hung up on whether it's actually 12.1 and 12.9 or if it's actually 12.2 and 13.0. The difference between my baseline 12.1 and the suspect car's 12.9 reading succeeded in letting me know there is a problem. Did that make sense? Is that what you're trying to get more info on? Or am I out in left field without a mitt? lol.
Check out my V8 Sky build video. It's pretty cool!...
https://youtu.be/2q9BuzNRc3Q
https://www.youtube.com/user/gmtech16450yz
I get what you're saying, it's not about the absolute value but the consistency of reporting the difference between expected and actual.
I also understand what the others are saying about lesser WB's reportin larger errors the further away from stoich the gas is.
I was just wondering what the accuracy and acquisition rate of the analog inputs are.
Got it! Ok so I wasn't totally in left field, that's what I thought you meant originally. So yeah, I'm thinking Keith is the only one that's gonna know that answer. You're talking about the ACTUAL analog to digital interface within the mpvi right? So I'm sticking to my original reply then. lol. That's why I was comparing it to an audio D/A. I would think an analog to digital converter that just has to deal with converting a 0-5 volt input should be pretty accurate and fast? So my guess is that the interface within the mpvi is PLENTY fast to handle things like wideband readings during normal, real world engine transients. What that speed actually is, or what the bit depth or sample rates are would be WAY more than I would know! I do know a D/A converter is going to have those traits, like the music I'm listening to right now is being converted at 44.1kHz/64bit, and it looks like that's the info you're after. I don't remember ever seeing any discussion on the actual specs of the mpvi analog to digital converter on here though. Did you happen to look on the HPTuners site to see if they list specs anywhere?
Check out my V8 Sky build video. It's pretty cool!...
https://youtu.be/2q9BuzNRc3Q
https://www.youtube.com/user/gmtech16450yz
I didn't see it anywhere and I too would assume it's relatively quick and accurate. At least more so than the digital to analog signal that's being produced by a WB.
I can see there's a lot of benefit from a CAN-BUS solution by eliminating 2 DA/AD conversions but the devil is in the accuracy of the sensor information interpretation.
Take a look at this screen shot guys. THIS is real world. THIS is why I've tried to say narrowbands are useful. For whatever reason, it really doesn't matter, this guy didn't have his wideband hooked up or being logged. It happens ALL THE TIME, not everyone has a working wideband. I'm guessing like 20% of modded cars do. An even more minute percentage have a $5000 wideband at their disposal. The whole point of the AEM CAN based wideband is that it has a real chance of getting MORE widebands in MORE modded cars. And on top of that, since setting them up is damn near idiot proof, more guys will have usable data.
So we can discuss differences in widebands all day long, which there are, but this is real world. All we have to go by here is the narrowband. How does it look to you guys? Does it really matter that the voltages between a wideband and a narrowband can't be directly linked when the narrowband is reading .028v during this whole 6600rpm WOT event? Holy cr@p it doesn't matter. What can we guess .028v is gonna be reading on a working wideband? 14:1? 15:1? 16:1? It doesn't matter if it's ANY of those. .028v on a narrowband at WOT is blown up engine territory. So are narrowbands valuable? Damn f'ing right they are! I just sent this owner an email telling him to park his car until it gets figured out. That narrowband may have just saved a $13k engine from being scattered.
Check out my V8 Sky build video. It's pretty cool!...
https://youtu.be/2q9BuzNRc3Q
https://www.youtube.com/user/gmtech16450yz
It was never argued or said as far as I know that narrow bands are useless or shouldn't be used - not by one single person on this posting to my knowledge and readings... "I" also never said that this wideband was useless... "I" just stated that you should never rely on a narrow bands voltage output to dial in a wideband.... ALSO anyone out there with the above log - pray your narrow band is shorted internally or something - that air fuel ratio is in the 16's or higher and during boost at that - Yikes......... How is that thing not miss firing?
Last edited by GHuggins; 05-18-2016 at 10:02 PM.
2010 Vette Stock Bottom LS3 - LS2 APS Twin Turbo Kit, Trick Flow Heads and Custom Cam - 12psi - 714rwhp and 820rwtq / 100hp Nitrous Shot starting at 3000 rpms - 948rwhp and 1044rwtq still on 93
2011 Vette Cam Only Internal Mod in stock LS3 -- YSI @ 18psi - 811rwhp on 93 / 926rwhp on E60 & 1008rwhp with a 50 shot of nitrous all through a 6L80
~Greg Huggins~
Remote Tuning Available at gh[email protected]
Mobile Tuning Available for North Georgia and WNC
A NB can't tell you when your WB is right. But, it can , often, tell you when the WB is wrong.
Thats a great amount of info, but latency is different than S:N and is different than accuracy.
First would you agree that you could have low latency and be inaccurate?
Second, would you agree that a very low S:N could also be inaccurate?
Im not brand biased as I own neither and actually dont plan on owning either, but you didn't address S:N is your response, in fact, you responded to everything BUT S:N.
You pointed out what can cause inaccuracy versus a plot that relies on time and that is totally valid, yet says nothing to what you quoted me about and that is how clean or not a signal is. Im glad to have an in depth, tech based discussion here and you will not find me getting all giggly if you half way agree with me. Im here to learn as much as I am to teach, but lets make sure we talk about the same things when discussing or correcting someone.
Last edited by matty b; 05-18-2016 at 11:44 PM.
I got from all this that the new Bosch 4.9 sensor reacts much faster to change in actual afr in the exhaust. I don't know how useful any faster than the NTK sensor is going to be. Logging the ECM/NTK combo I get around 0.1-0.2 of a second reaction to actual fueling in the exhaust on my logger. Most of the time its 0.1 on ECM's that log at a faster rate. Who knows that this actual delay is from the sensor, processing, or the actual distance of the sensor from the cylinders. There are alot of variables I don't see discussed on these topics. I question the legitimacy of the pressure vs accuracy testing. This was a big jump for the NTK sensors prior to the Bosch 4.9 sensors coming out. The ability of them to handle heat and pressure and resist leaded fuel. Just because ECM makes the NGK/Ballenger doesn't make it gold but I'm inclined to believe a product from them which makes calibration equipment along with those lab grade lambda meters that are the standard that have been bench-marked to be leaps and bounds more accurate than any others on the market. So to claim a new AEM (which has a terrible track record for accurate and quality equipment) is better than or equal to the NTK/NGK/ECM/Ballenger is a bold statement that will obviously have to be backed up with facts other than some users randomly claiming so.
The newest version of the ECM wideband supplied to Ballenger Motorsports supports the new Bosch 4.9 sensor so at that point the only thing that could be claimed to be better is the serial/CAN communication aspect. I for one use a dedicated power source for my wideband which kills any noise on the analog circuit. For me I don't see any benefit for the can/serial communication over the current analog signal. Sure it'd be nice to have that option but when I already have less than 0.001 lambda fluctuation from my analog input there is no gain in "accuracy" there for the serial communication. Those that use them in car sure I can see it helping a ton. And those that are after that 4th decimal point accuracy will be chasing your tail with a single wideband anyway. Little do you know that there are extremely larger fluctuations cylinder to cylinder that are just getting baked into an average anyhow before it hits the element of that single sensor sitting in the exhaust. Have any of you played with sensor placement in the exhaust? What did you see? The variables just stack up the deeper you dig.
On the narrowband topic, I am a firm believer in watching those if they are still in the car. You can pickup misfires and all other sorts of problems by watching them that even widebands don't see due to processing. The narrowbands are direct connection to an input where as a wideband has to be processed through a controller then that signal transmitted either via analog or digital communication. So in that sense the narrowbands generally are more "responsive" than just about any wideband on the market however they are highly inaccurate when trying to use a certain mV reading to equate to a lambda reading.
Last edited by LSxpwrdZ; 05-19-2016 at 12:05 AM.
James Short - [email protected]
Located in Central Kentucky
ShorTuning
2020 Camaro 2SS | BTR 230 | GPI CNC Heads | MSD Intake | Rotofab | 2" LT's | Flex Fuel | 638rwhp / 540rwtq
2002 Camaro | LSX 427 | CID LS7's | Twin GT5088's | Haltech Nexus R5 | RPM TH400
Accuracy is what's left, after the error contributions of the various error sources are summed. So, having any one source of error low does not guarantee accuracy. In theory, one could have zero noise ( for example ) and, still have poor accuracy, due to other sources.
Noise, latency, analog offsets, etc. are all error contributors.
Noise and latency are often linked because the filtering that is used to remove the noise expresses as increased latency. i.e. the time that it takes to gather the samples needed for filtering/averaging, or the response time of the filter element(s), is added to the raw latency.
You get the greatest increases in accuracy by attacking the error source with the greatest contribution.
Like I said before , in practice, the sources of error in wideband systems, in descending order of magnitude, are :
1. Analog offset errors ( especially when using the A/C or EGR methods )
2. Latency ( including latency added by noise filtering )
3. Exhaust gas pressure effects ( +/- 4%/psi for NTK +/- 1%/psi for Bosch )
4. Sensor/controller DC accuracy. ( indicated lambda vs. calibration gas )
The AEM 30-0333 is getting a lot of attention because it completely eliminates analog offset errors by bypassing that stage, entirely.
That is... if the Gauge shows Lambda 1.009 HPTuners shows Lambda 1.009 because it is transferred digitally, with CRC checks and all.
It is also has , far and away, the the least total latency measured to date.
As for exhaust gas pressure error, it uses the Bosch sensor, which helps a bit, but does not have a pressure sensor for compensation.
The bench accuracy ( calibration gasses ) is as good as any other system tested. Most are pretty good, in this respect.
I'm probably comparing a little more superior wideband that I classify as the NGK. The ECM AFM1600L (modified NGK) deals with the latency and filtering as compared to the NGK. From the guys at ECM they said the AFM1600L is basically an AFM1000 in an NGK box. I've not used an NGK in a little while now however I still stand that a properly setup analog signal will easily get the job done IF you know how to get that signal calibrated. We have been pushing for a serial option with the ECM widebands to obviously do what AEM and others have and eliminate the analog noise to take out one of those errors from the equation.
James Short - [email protected]
Located in Central Kentucky
2020 Camaro 2SS | BTR 230 | GPI CNC Heads | MSD Intake | Rotofab | 2" LT's | Flex Fuel | 638rwhp / 540rwtq
2002 Camaro | LSX 427 | CID LS7's | Twin GT5088's | Haltech Nexus R5 | RPM TH400
Little do some know. Using an AEM 30-0310 the following was recorded on an LS-2 with a broken header bolt. Slight exhaust leak, sucking air on scavenge.Little do you know that there are extremely larger fluctuations cylinder to cylinder that are just getting baked into an average anyhow...
exhaust gasket leak.png
Because the AEM wideband has an actual latency of around 10ms or less, it was able to track the single cylinder lean events ( ~40ms ). With the oscillations synchronized to RPM/2.
Latency makes a huge difference. if your t63 response time is around 100ms; then, after 100ms, your reading is still 37% off. No need to go looking for that 4th digit.
A lot happens in an engine in 100ms.
The ECM/NGK controller has a pretty clean output, noise-wise. IIRC it also has an analog ground line for devices with differential inputs. This would help null offset errors in devices with differential inputs. The MPVI does not have the differential type inputs. So, it is subject to the offset errorsYou may be using some system that does have the differential inputs. So, that may be helping out some ( quite a bit, actually ).
I have been designing analog front-end systems for 30+ years. And I can promise you that, in an automotive environment, there are always issues with 0-5v analog signals. It's a wonder that they never went to 4-20ma current loop ( industrial style ) communication.
The ECM/NGK controller does a great job for what it has to work with. The NTK sensors, while accurate on the test bench, are VERY slow and are very sensitivity to exhaust gas pressure changes. I have, personally, observed this effect on the test bench, by running up the pressure with calibration gasses. The signal is off by ~4% per psi, when I change the pressure. The Bosch sensors only change by about 1% per psi, under the same conditions.
NTK actually calibrates their sensors at about 1.5psi, just to have the error start somewhere in the middle of the usable range. Figuring that AFR at very low back-pressures are probably around lambda 1.00 anyway. I know there are some ECM/NTK systems that include a pressure sensor for the exhaust gas. ( like the AEM 4-channel units do ). This is a much better solution.
I do not trust a non-pressure-compensated NTK based wideband at full throttle ( high exhaust back-pressure ).
Well thank you for admitting that, almost.
Its neither theory nor the cause of outside sources that you could have a SNR of 0 and still be inaccurate. The controller could output a signal that read 4.50000000000vdc with zero deviation with zero latency (yes impossible) and the signal SHOULD be 3.500000000vdc. If every time the sensor encountered a situation that should output 3.500000000vdc but read 4.50000000000vdc without fail, you would have amazing precision, unworldly latency, zero SNR yet have horrible ACCURACY.
Im sure the AEM and ECM/NGK controllers are more that good enough for 99% of people, but you quoted me replying to someone regarding SNR and why that somehow counts towards accuracy. I hope Im making it abundantly clear that metric does not mean anything regarding latency or accuracy and in the real world, a better SNR on the output side is usually coupled with higher latency. I'd argue with consumer electronics that the better the SNR the higher the latency compared to professional meters. Im not here to argue that though, just what I said and only what I said (and was quoted for saying).
I've trusted mine on plenty of high hp cars without issues... It's all in what you do with the data and not just the wideband data. ECM does offer a pressure compensation wideband however it's out of the price range of most shops. The addition of support for the 4.9 will obviously help in that department.The ECM/NGK controller has a pretty clean output, noise-wise. IIRC it also has an analog ground line for devices with differential inputs. This would help null offset errors in devices with differential inputs. The MPVI does not have the differential type inputs. So, it is subject to the offset errors
I have been designing analog front-end systems for 30+ years. And I can promise you that, in an automotive environment, there are always issues with 0-5v analog signals. It's a wonder that they never went to 4-20ma current loop ( industrial style ) communication.
The ECM/NGK controller does a great job for what it has to work with. The NTK sensors, while accurate on the test bench, are VERY slow and are very sensitivity to exhaust gas pressure changes. I have, personally, observed this effect on the test bench, by running up the pressure with calibration gasses. The signal is off by ~4% per psi, when I change the pressure. The Bosch sensors only change by about 1% per psi, under the same conditions.
NTK actually calibrates their sensors at about 1.5psi, just to have the error start somewhere in the middle of the usable range. Figuring that AFR at very low back-pressures are probably around lambda 1.00 anyway. I know there are some ECM/NTK systems that include a pressure sensor for the exhaust gas. ( like the AEM 4-channel units do ). This is a much better solution.
I do not trust a non-pressure-compensated NTK based wideband at full throttle ( high exhaust back-pressure ).
As far as your logged data of the header leak, most wouldn't be able to see that with HPT either depending on the scan rate. Most of the ECMs at their fastest rate will log at a 20hz/50ms rate which is still more than that oscillation from the exhaust leak was. That is unless they are updating the software to poll the serial input faster than their analog input on the cable itself (which is possible).
James Short - [email protected]
Located in Central Kentucky
2020 Camaro 2SS | BTR 230 | GPI CNC Heads | MSD Intake | Rotofab | 2" LT's | Flex Fuel | 638rwhp / 540rwtq
2002 Camaro | LSX 427 | CID LS7's | Twin GT5088's | Haltech Nexus R5 | RPM TH400
Like I said, accuracy is what's left after ALL of the error contributors are summed. The contributor that you mention would be the "steady state scale/offset error" of the sensor/controller.ts neither theory nor the cause of outside sources that you could have a SNR of 0 and still be inaccurate. The controller could output a signal that read 4.50000000000vdc with zero deviation with zero latency (yes impossible) and the signal SHOULD be 3.500000000vdc. If every time the sensor encountered a situation that should output 3.500000000vdc but read 4.50000000000vdc without fail, you would have amazing precision, unworldly latency, zero SNR yet have horrible ACCURACY.
In the case of widebands, in an automotive environment, this is one of the smallest contributors to error. Differences between any two given widebands may be around 1% - 2%, in steady state w/ sufficient filtering. That is, when I put any given wideband on the test bench with calibrated gasses, they are usually pretty close. There are some unfortunate exceptions.
The reason SNR is so often entwined with latency is because of the latency introduced in order to filter it down to whatever is considered a usable level.
And, latency contributions can be very large, in a dynamic system. i.e. 37% at the t63 response time of the wideband. Again, in a stationary engine, this is not an issue. But trying to tune fueling for a power-shift, it is critical. Most tuning is somewhere in between.
And, what you CAN do with the data, given its limitations. The LSU4.9 sensor update to the ECM unit should help it, across the board, in real-world applications.It's all in what you do with the data ...
The 3.x release of HPTuners supports a scanning rate of 100Hz/10ms which should be enough to see this. This is, probably too fast for serial data, though. Especially in a Windows PC USB/Serial system, with unknown buffering latency. Windows is not remotely real-time.As far as your logged data of the header leak, most wouldn't be able to see that with HPT either depending on the scan rate. Most of the ECMs at their fastest rate will log at a 20hz/50ms rate which is still more than that oscillation from the exhaust leak was. That is unless they are updating the software to poll the serial input faster than their analog input on the cable itself (which is possible).
And there in itself is another potential problemAnd, what you CAN do with the data, given its limitations. The LSU4.9 sensor update to the ECM unit should help it, across the board, in real-world applications.
The 3.x release of HPTuners supports a scanning rate of 100Hz/10ms which should be enough to see this. This is, probably too fast for serial data, though. Especially in a Windows PC USB/Serial system, with unknown buffering latency. Windows is not remotely real-time.. With the lag of windows and/or the software it makes you wonder how the readings will lineup in the datalog. I'm sure like everything else we have discussed it will be minimal. My biggest question is the actual accuracy of the AEM unit to process the reading from the sensor. We all know they all use the same sensors whether it be Bosch or NTK, it's how the controller controls and interprets the signal and re-generates that signal either into an analog or serial/digital signal that really matters. Maybe AEM has hit a homerun with this new unit
I just went back and looked at some logs and you are right. It's polling the analog input very quickly through HPTuners... I'm seeing roughly 12.5ms refresh on one of my scans of the analog input.
James Short - [email protected]
Located in Central Kentucky
2020 Camaro 2SS | BTR 230 | GPI CNC Heads | MSD Intake | Rotofab | 2" LT's | Flex Fuel | 638rwhp / 540rwtq
2002 Camaro | LSX 427 | CID LS7's | Twin GT5088's | Haltech Nexus R5 | RPM TH400
Oh, that one is easy. It is easily +/- 1% with the several test gasses that I used on the test bench. Slightly more accurate than the ECM/NGK unit that I tested ( Powerdex ).My biggest question is the actual accuracy of the AEM unit to process the reading from the sensor.
BUT, the test was performed at atmospheric pressure ( 0psig ). And, I , now, know that the Powerdex was expecting the test gas at ~1.5psi ( 0.1bar ). So, they may be equal, overall. HOWEVER, IF YOU DO A FREE-AIR CAL AT ATMOSPHERIC PRESSURE, YOU GET THE SAME ORIGINAL ERROR!!!
This is one of those cases of chasing the 4th digit.
As for time-skewing the data due to Windows timing. When using CAN data the MPVI, all of the data is collected IN the MPVI, then sent, together, over USB. So, it is properly correlated, at least to itself.
The problem is with USB/Serial ports. They take a separate path from the MVPI data. And that data has an unknown time skew, as compared to the MPVI data. This is PROBABLY not a big issue because serial data-rates are generally pretty low. ( in the 10Hz / 100ms range ). So, 20ms-50ms of skew isn't "that bad".
But, it is still an issue when looking at what I call "cell velocity". "Cell velocity" is the rate at which the RPM and MAP readings move through VE table cells. i.e. During dynamic events, the VE cell may change 30 times in a second ( 33ms per cell change ). So, 50ms of data skew means that the wideband data could be 1 or 2 cells off, from where it SHOULD be applied. Another 100ms of latency makes it even worse.
This kind of skew is a big part of what makes tuning an art. Knowing how to read between the lines.
If the wideband data always wound up in the right VE cell, anybody could tune
Reply With Quote