Landtank MAF surprising scangauge results

[SUMOTOY]

Nerdin on MAF

The modifications made to the wheatstone bridge are not implicitly designed to compensate for aging or contamination since both are expressed with some degree of variability. The bridge is definitely optimized to provide the best transient response of the hot wire for the most common operating parameters, but that's about it.

Here is a great whitepaper on the subject from all the way back in 1991:
https://eprints.kfupm.edu.sa/42430/1/42430.pdf
Though it is a bit dated (back when logic hardware was expensive (predating windows 3.1) the principles of Gaussian neural networks are the same.

Been quite a while since I was introduced to Gaussian theory in automotive applications, but IIRC, it had to do with predicting outcomes, ala "Fuzzy Logic" computing. However, it's well documented that the MAF voltage signal in the 80 ecu is not a 'learned' value, it's a tabled value. The subsequent fueling correction of LTFT is a learned and retained value, but MAF is not. It only sets the baseline, all fuel'corrections' are after that value is tabled. The timing values have only temp correction after that MAF tabled baseline. Load has no correction after MAF tabled baseline.

As I understand "Guassian neural network" theory you cite, this would be a rather advanced function used in production vehicle testing of values vs faults. For example, if you have a known good MAF voltage plot of 200,000 events of MAF signal being between 1.25 and 4.455 volts, and the actual values 'learned' are .75 to 4.1, Gaussian theory can predict that the outcome of the events learned is the result of a vacuum leak, not a generic OBDII P0102 MAF code (next gen OBD?). It is also applied in acceleration enrichment (AE) profiles, usually as a tabled function of TPS/MAP. Specifically, if you are driving 'sporty-like' Gaussian modeling can change AE values to give more agressive initial timing or more accurate tip in throttle fueling for the predicted ramp change of TPS/MAP.

Unfortunately, that's not a property of the 80 ecu software of the MAF signal. Nor is it common to the MAF signal in most engine ECU's, because it takes a lot of processing power, certainly not found in the 80 ecu.

With regard to the MAF signal temp and age correction, your comments seem to contradict the manufacturers own documents, which claim the function of the trimmer resistor is to account for age and temp. Since the voltage output value is 97% accurate, it seems your Gaussian plot would just be the same as a 1 standard deviation of normal distribution curve in a non-fuzzy logic software application. That's pretty accurate given that 97% accuracy is throughout the measured mass of air of the actual sensor, regardless of pipe diameter.

Straying far from the 80 MAF application, since you have no voltage plot for the stock or the modded MAF sensor to compare or lay claim of any Gaussian theory. And without learned function ability in the MAF signal, there can only be a voltage vs airflow table, with plots dictated by the clock speed of the ecu processor.

The rest of it is cool reading I suppose.

Cheers

Scott J
'94 FZJ80 Supercharged

 

[Beowulf]

... for people that get this far in future readings (or are currently reading this thread)

Hang in there NaterGator, some of us are still reading. Comprehension is dropping off rapidly but I'm still reading. There seems to be a bit of repetitiveness that causes me to nod off every now and then.

-B-

 

[NaterGator]

Scott, I'm sorry you misunderstand. Unfortunately it is up to you to learn to concepts involved; I've done my time studying the theory and implementation and I can't just impart that to you in a few posts. It requires nitty gritty work and conceptual understanding of the equations, vector calculus, linear algebra, etc.

With regards to MAF drift biasing, reread the IEEE technical discussion quote. It explains that the bias is done on-line (in the software modeling) and primarily utilizes the oxygen sensor feedback.

I can't fathom how you rationalize this:
"without learned function ability in the MAF signal, there can only be a voltage vs airflow table, with plots dictated by the clock speed of the ecu processor."

That's absurd. The plots are completely independent of the sampling rate of the ECU. Sure a nice high sampling rate would provide resistance to aliasing during transients, but that's irrelevant because Toyota would never design a control system that sampled that slowly and there's no reason to discuss sampling rate when the only plot that ostensibly matters (again, according to your beleaguered claims) is the voltage vs airflow curve of the MAF.

 

[SUMOTOY]

Scott, I'm sorry you misunderstand. Unfortunately it is up to you to learn to concepts involved; I've done my time studying the theory and implementation and I can't just impart that to you in a few posts. It requires nitty gritty work and conceptual understanding of the equations, vector calculus, linear algebra, etc.

With regards to MAF drift biasing, reread the IEEE technical discussion quote. It explains that the bias is done on-line (in the software modeling) and primarily utilizes the oxygen sensor feedback.

I can't fathom how you rationalize this:
"without learned function ability in the MAF signal, there can only be a voltage vs airflow table, with plots dictated by the clock speed of the ecu processor."

That's absurd. The plots are completely independent of the sampling rate of the ECU. Sure a nice high sampling rate would provide resistance to aliasing during transients, but that's irrelevant because Toyota would never design a control system that sampled that slowly and there's no reason to discuss sampling rate when the only plot that ostensibly matters (again, according to your beleaguered claims) is the voltage vs airflow curve of the MAF.

With all due respect, I see smoke and mirrors, and not much to back it up. In the IEEE technical quote you cite, it's possible to use fuzzy logic to the MAF signal to account for wear and age. I.E. (pun intended), this "Gaussian neural network" within the software would use other signals to verify the 'actual' tabled value of the MAF - before it's tabled. If we look at the above graphs for MAF 3 example, at 4000rpm steady state cruising (say 22% actual load), the engine should have a MAF voltage of 3.9393v, but if the actual voltage is 3.8212 for enough events, the GNN could modify the MAF signal (slope). Your Gaussian theory to MAF signal then, could predict that after 30,000 events at that learned value (all other values normal), the change in MAF signal is valid, and would then be software modified ('hidden layer' modifier = age, wear, temp, mods...) to read 3.8212 at 4000rpm steady state load cruise at 22% actual load (now Actual Load = MAF Calculated Load).

That said, once the MAF signal is tabled at that value, the processing follows the same as the 80 ecu. This fuzzy logic modeling you reference can account for minor MAF wear, minor engine tolerance, minor engine modification, and age. That's NOT HAPPENNING in the 80 ecu.

With regard to sampling rate vs processing speed: That's exactly how MAF 'resolution' is defined in software. The 80 software uses less MAF sampling of the voltage slope than newer high speed processor ecu's. The slope of the MAF doesn't change, only the points on it used by the software changes. NG, this is how MAF work. Why both Christo and I claim earlier in this thread that using a 'better' MAF sensor is only as good as the software/processor it's attached to.

I've used fuzzy-logic computers in automotive applications. I believe I understand how they work, and the relationship between Gaussian distribution vs normal distribution. When I see "Gaussian" combined with "neural network", you define a fuzzy logic controller circuit. The 80 ecu has none of this function in the MAF signal to the software tables.

NG, if you think this IEEE quote applies to the MAF in the 80 ecu, please document where that is found in the 80 ecu/MAF documentation. Every document I've seen regarding the 80 MAF signal, is it's just a voltage value for baseline Load, timing and fuel. There is no 'hidden modifier' in the MAF voltage as proposed in the IEEE quote you reference.

Straying this far from how a 80 ecu inputs MAF voltage is the absurd part. I grew up with a father who was Assistant Dean of Engineering and Applied Sciences at Yale for years. And was taught early on, attempts at complicating basic concepts, doesn't change the outcome.

Thanks for your input

Scott J
94 FZJ80 Supercharged

 

[Ebag333]

With all due respect, I see smoke and mirrors, and not much to back it up.

Were you able to actually type that with a straight face?

NG, if you think this IEEE quote applies to the MAF in the 80 ecu, please document where that is found in the 80 ecu/MAF documentation.

Why doest NatorGator have to prove that it's in the 80's documentation, while when others ask you for proof for your statements you can quote from documentation from non-FJ-80 (or even Toyota) vehicles?

 

[SUMOTOY]

Were you able to actually type that with a straight face?

Why doest NatorGator have to prove that it's in the 80's documentation, while when others ask you for proof for your statements you can quote from documentation from non-FJ-80 (or even Toyota) vehicles?

Are you claiming that the toyota references regarding MAF operation Christo used do not apply to the 80? How exactly do they differ? I'm positive that NG has the IEEE reference wrong, as the quote applies to how a fuzzy logic MAF signal 'hidden modifier' works. Not *post* software table modification, but in pre-software voltage tabling. If you want to claim that adaptive fuzzy logic is there, then the voltage slope of the stock and modified MAF's should be the same, and no 'other' modification (i.e, removing the FPR reference) should be required.

I presented my understanding of how these systems work in great detail. I am interested how I may have misinterpreted these systems, including fuzzy-logic applications (whether they are used or not).

Feel free to correct my misunderstanding of these systems if you are up to it. I claim MAF EFI of a 90's vintage 80 series toyota is pretty straight forward and really basic.

YOMV, but the voltage slopes would be required to present 'exceptions' to Mr. T's explanations to operation of the MAF sensor.

Scott J
94 FZJ80 Supercharged

 

[Ebag333]

Are you claiming that the toyota references regarding MAF operation Christo used do not apply to the 80? How exactly do they differ?

It's not surprising that except for the first sentance nothing in your post has anything to do with what I asked.


You said:

NG, if you think this IEEE quote applies to the MAF in the 80 ecu, please document where that is found in the 80 ecu/MAF documentation.

Yet here's what you offer for proof. Please note that I've limited this to just 3 quotes, if I went back and referenced every single post you made about a vehicle other than an 80 somehow proving something about the 80's MAF, this post would be rediculously long.

This trend of yours has already been pointed out before.

I already cite site the 4in Lingenfelter MAF scaling in this regard back in post 444. I'm sure you just didn't take the time to review the content of the cite-site (pun intended) I referenced.

http://www.lingenfelter.com/sites/l...files/LPE100mmMAFsensorflowcurvesv1.0_000.xls

I have measured voltage slopes on the exact same MAF sensor insert used for a late model v6 N/A motor and a twin turbo version of a late model v6 motor (identical-interchangeable part number).

Many vendors, like Lingenfelter for instance, will give you the MAF slopes for their modified sensors - others give the MAF transfer function to remap the MAF software tables
http://www.lingenfelter.com/sites/l...files/LPE100mmMAFsensorflowcurvesv1.0_000.xls

Just sayin' - carry on!

If you are going to demand that someone back up their claims with only Toyota documentation, then you should as well. That's the difference between you and Christo, virtually everything he stated was directly from an 80. When you quote Toyota documentation, it's the exception, not the rule.

And that's exactly why Christo's input was useful, while yours is not.

 

[Walking Eagle]

I reject your reality and insert my own.

He won't see it...

 

[semlin]

it is well over 100 posts since sumotoy revived this thread. he has contributed nothing in that period other than extremely tenuous speculation, and he has shown that he is completely immune to logic and totally unable to digest relevant information.

i think we are accomplishing nothing here debating him further. he is either a fool or a troll. either way, i suggest everyone put him on ignore as i have done and we move on.

 

[SUMOTOY]

It's not surprising that except for the first sentance nothing in your post has anything to do with what I asked.

And that's exactly why Christo's input was useful, while yours is not.

I have referenced many of Christo's docs he used, including the MAF reference documentation. I've read speculation on how a 80 ecu is different, including a guy mis-stating it has some sort of fuzzy logic MAF signal adaptation. Even a vendor of a modified MAF meter claiming 'better' without slopes, proper theory or operation understanding.

MAF based EFI operation is simple, it's well documented, and the 80 isn't exempt from how it works. In lieu of LT actually posting up voltage slopes, we can look at several examples of slopes, Output slopes vs (effective) diameter, how they work, and how they are tabled in an ECU. Toyota not publishing their software programming doesn't mean MAF operation has changed in any way. It's a 0-5v signal that gets tabled into software, just like every other toyota 0-5v MAF signal that gets tabled in the software. You can look at *current* toyota MAF signal references (or any autoshop101 toyota references), and not find an exception to the concepts I presented. Again, do you claim the 80 is different than I presented? How EXACTLY?

Saying MAF operation is different in the 80 MAF ecu because there is no direct 80 documentation to claim it's the same, seems a bit backwards to me. The vendor of a modified MAF Sensor should be able to clearly address the similarities and differences, and document what they are. Er, that's a MAF voltage slope function first?

Carry on

Scott J
94 FZJ80 Supercharged

 

[landtank]

will someone just lock this thread already. Any value it might have had is gone.

I started a dedicated thread for Sumotoy to carry on in so he doesn't destroy an archive of good info.

But he refuses to do that for whatever reason and insists on continuing this destructive nature of his.

 

[landtank]

Oh Ya, I AM NOT A VENDOR! I'm a guy who made something that others liked and asked for, PERIOD.

I'm not about to be bullied by Slee or Sumotoy in that respect. Who declared either of you to have that right was wrong.

 

[firetruck41]

The vendor of a modified MAF Sensor should be able to clearly address the similarities and differences, and document what they are. Er, that's a MAF voltage slope function first?

Carry on

Scott J
94 FZJ80 Supercharged

Since you are not comfortable with purchasing from a vendor who has not documented his modified MAF, and he probably (and understandably) is not going to sell you one, then we have reached an amicable conclusion. If Landtank promises not to sell you a MAF, will you let this thread die? Now is your chance to make your own MAF mod and create a thread about developing it.

I will now prepare for an argument on a point that I never made, in which you will conclusively prove I am wrong.

 

[Walking Eagle]

will someone just lock this thread already. Any value it might have had is gone.

I started a dedicated thread for Sumotoy to carry on in so he doesn't destroy an archive of good info.

But he refuses to do that for whatever reason and insists on continuing this destructive nature of his.

Started going through suspension travel thread from three and a half years ago, and your advice to me at that point was "just let it go." And in three and a half years, Sumo's not become anymore logical, or honest, or any less dangerous.

 

[sleeoffroad]

Oh Ya, I AM NOT A VENDOR! I'm a guy who made something that others liked and asked for, PERIOD.

If you want, please explain the difference in another thread. You use this site to sell stuff. How is that not a vendor? You build people's truck for profit, you sell parts and solicit input on making more parts (4Runner MAF). I believe you do all those for a profit and not charity, so how is that not a vendor?

I'm not about to be bullied by Slee or Sumotoy in that respect. Who declared either of you to have that right was wrong.

I have said my piece and have not posted anything in this last round that Scott started. Please do not put me and him in the same boat.

 

[landtank]

You use this site to sell stuff. How is that not a vendor? You build people's truck for profit, you sell parts and solicit input on making more parts (4Runner MAF). I believe you do all those for a profit and not charity, so how is that not a vendor?
.

you're right, I don't run a charity. On the other hand I'm not the only one that will make some of their designs available for others to buy if they like either.

So when you decide to round up everyone and chastise them for that then feel free to include me. Otherwise singling me out seems a little personal.

 

[SUMOTOY]

Oh Ya, I AM NOT A VENDOR! I'm a guy who made something that others liked and asked for, PERIOD.

I'm not about to be bullied by Slee or Sumotoy in that respect. Who declared either of you to have that right was wrong.

You are a vendor by definition. All caps doesn't change the definition. Bullied? Explanations of how MAF EFI systems works, really has nothing to do with you. If your 'vendor mod' falls into the category of explaining how a larger MAF works, specifically how it can't work, you are welcome to explain how the 80 is different from my definition of MAF based EFI systems in general, or specific to the 80.

I think that requires voltage slopes of the stock and modded MAF to even start presenting a different position, which without having done/doing so as a vendor, likely creates a feeling of 'bullied'.

At this point, I almost wish 80 EFI MAF applicaton was different.

Scott J
94 FZJ80 Supercharged

 

[Ebag333]

you're right, I don't run a charity. On the other hand I'm not the only one that will make some of their designs available for others to buy if they like either.

So when you decide to round up everyone and chastise them for that then feel free to include me. Otherwise singling me out seems a little personal.

As someone who got chastised a bit for even considering offering a product, I don't think you're being singled out here.

I'll admit, I was a bit upset at first over the chastisement from a few of the people, but looking back at it I do appreciate it and understand that it's not personal.

Slee offered up a valid concern regarding the FPR line, just like he offered up a valid concern regarding product insurance and an X-Link. I don't think he's trying to beat either of us up, but to offer constructive criticism and valid concerns.

Just two cents from someone who got "chastised" over a (potential) product.

 

[sleeoffroad]

you're right, I don't run a charity. On the other hand I'm not the only one that will make some of their designs available for others to buy if they like either.

So when you decide to round up everyone and chastise them for that then feel free to include me. Otherwise singling me out seems a little personal.

I never said you can not do it or chastised you for doing it. This site is a great avenue for people to do just that.

I just have a little difficulty in understanding how define vendor vs not vendor and the responsibilities that come with that in terms of products being sold on this site. Is there a $ limit? A requirement for a business license?

The $'s from the customers are the same. They buy a product with their $'s irrespective of the origin of the products.

I can claim the same, I started making sliders in my garage and people liked it and started buying it. Got a little bigger than that, but so what I have x number of products, you have y. So where is the distinction?

 

[landtank]

the difference in my eyes is that it's a hobby to me not a business. One design fuels the next. Honestly, I'd prefer to be a behind the scenes wholesaler. Which since I can only deliver products in batches it's pretty much what I am.