Alternative MAF Sensor Implementation
Evolved Member
Joined: May 2006
Posts: 2,399
Likes: 6
From: UK
Bez's speed density patch was really an offshoot of his patch to clip the max MAF signal on lift off, we gather it works for mixmastermatt, but he also had engine problems that were probably unrelated. Bez has done some great disassembly, but some of the first impressions of things he has posted have turned out to need revision/development which he hasn't been involved in since he's been away diamond mining or whatever more exciting things he's doing. More than one person needs to understand how it works so it can be modified/understood/debated/reapplied. Some initial patches from everyone that has been doing them have variables that are orders of magnitude out and just plain don't work as intended until you repeatedly test and re-engineer them.
The IX JDM ECU already has a speed density calculation in there which is used when I unplug the MAF and which I have disassembled, I just need to work out the cleanest way to get the ECU to use this value rather than the MAF value, perhaps then reuse of OEM ECU code would be useful on other ECUs that don't have it. I cannot disagree more with Jack's views on this one, and feel equally polarised against MAF sensors for IMHO equally good reasons and am happy to discuss the technicalities including loads of support from papers and practical experience if required. Even on a stock IX, a MoTeC running speed density runs smoother on throttle transients than the stock ECU. MAF reversion problems disappear. The only issue with the MoTeC is the prolonged cranking required. It was the base for the fastest ever warrantied and emissions legal Evo - the FQ400. The driveability on that was only affected by using a large turbo and limiting the torque for the sake of the drivetrain. It doesn't have speed density related issues at all. I think we can replicate this behaviour on the stock ECU and am determined to try. However, if people really want to pursue alternative MAF sensors I will happily contribute what I can in support of it from what I can glean from disassembly.
The IX JDM ECU already has a speed density calculation in there which is used when I unplug the MAF and which I have disassembled, I just need to work out the cleanest way to get the ECU to use this value rather than the MAF value, perhaps then reuse of OEM ECU code would be useful on other ECUs that don't have it. I cannot disagree more with Jack's views on this one, and feel equally polarised against MAF sensors for IMHO equally good reasons and am happy to discuss the technicalities including loads of support from papers and practical experience if required. Even on a stock IX, a MoTeC running speed density runs smoother on throttle transients than the stock ECU. MAF reversion problems disappear. The only issue with the MoTeC is the prolonged cranking required. It was the base for the fastest ever warrantied and emissions legal Evo - the FQ400. The driveability on that was only affected by using a large turbo and limiting the torque for the sake of the drivetrain. It doesn't have speed density related issues at all. I think we can replicate this behaviour on the stock ECU and am determined to try. However, if people really want to pursue alternative MAF sensors I will happily contribute what I can in support of it from what I can glean from disassembly.
Last edited by jcsbanks; Jan 1, 2009 at 09:55 AM.
I myself would rather use a SD setup for ease of tuning seeing that it's what I cut my teeth on in the Honda realm. I can't say I really see much of a drawback with using map-based SD other than emissions related. Honda successfully used SD until the early 2000's and I suspect the only reason for swapping to a maf based system was from stricter emissions requirements. Heck, even when I tune my car and other Evo's/Subie's I tend to log pressure (when I can and if the vehicle is set up for it) against rpm/load and tune accordingly. The only thing my Evo's maf has caused me is nothing but headaches with part-throttle and idle, while oddly enough my WRX's maf has been pretty consistent with absolutely no issues.
My main concern is with changing filters/intakes/DV and such since I'm not one of those individuals that's hitting the ceiling of the stock maf. My biggest problems are with the drivability issues that I have on a daily basis. A few days ago I decided to fabricate an intake pipe to use other than the stock one that caused me idle issues or the Injen junk that unfortunately is the only intake that fits with their piping kit. I did some logging, made some maf scaling adjustments, and gained some power on the top end. The only problem was during cruise, and it was a big one. Consistently over and over the maf frequency would climb slowly from 150hz to 250hz at ~30% throttle and 2700rpm and the afr would drop into the 9's and choke the engine. It's things like this that really irritate me, and hold back the other DIY guys like me that can't honestly understand why a 6" piece of curved pipe and filter costs almost $200.
My main concern is with changing filters/intakes/DV and such since I'm not one of those individuals that's hitting the ceiling of the stock maf. My biggest problems are with the drivability issues that I have on a daily basis. A few days ago I decided to fabricate an intake pipe to use other than the stock one that caused me idle issues or the Injen junk that unfortunately is the only intake that fits with their piping kit. I did some logging, made some maf scaling adjustments, and gained some power on the top end. The only problem was during cruise, and it was a big one. Consistently over and over the maf frequency would climb slowly from 150hz to 250hz at ~30% throttle and 2700rpm and the afr would drop into the 9's and choke the engine. It's things like this that really irritate me, and hold back the other DIY guys like me that can't honestly understand why a 6" piece of curved pipe and filter costs almost $200.
Evolved Member
Joined: Jan 2006
Posts: 953
Likes: 2
From: Oxfordshire
For those of you looking to gain a little more insight to SD, this is interesting if a little math heavy.
Speed density .doc download
Speed density .doc download
Karmann sensors definitely have issues, their vibration and turbulence sensitive, but are regarded as more accurate than other types of flow based sensors. I don't intend to argue SD over MAF as you tend to see SD on race cars, and MAF on production cars.. production vehicles get years of R&D to determine what offers the best driveability, where Race car R&D doesn't really take drivability into account, and I'm sure if it did, it probably uses MAF systems of some sort.
I have always had problems with the stock EVO MAF, tried everything to get better drivability out of it when I started making other modifications, and I agree it causes headaches. When I switched over to a Ford Cobra Maf and converter electronics, all of those drivability issues went away, and I gained performance. Don't get me wrong, it has its issues and tradeoffs, and takes time to learn to work with. But it does offer the best set of compromises.
Unfortunately the FQ300 Motec being Emission Legal, does not make it OBD Legal in the USA, no matter how clean a new car runs, if it doesn't have an OEM ECU, or at least support OBD-II Transparently, it will not be legal. Its unfortunate there is no exception process for people who really do have clean running cars.
Speed Density has its place, as does a carburetor, and MAF based systems. But so does Throttle Body, Batch Fire, and Sequential Fire EFI systems.. Again, we are talking about a production vehicle where its just as important to be able to drive the car comfortably, as it is for good performance. There's a reason why years of R&D goes into this stuff, there's also a reason why technology shifts away from one to another.
Speed Density is a simpler solution, it is based on calculations and minimal sensor data, that can never possibly accommodate all conditions seen in daily driving. It also "glosses" over things which makes some things seem smoother, when in reality they aren't, their just hidden until the conditions change. This might be fine for some people, But for me this is unacceptable.
I have a speed density TBI EFI system on my Malibu, It runs a wideband and is always running closed loop, under cruise and load. Every time the weather changes, the drivability changes too. It works well, way better than my carburetor did, but if you were to log the AFR and EGT data, it varies as conditions change and the calculations don't. Fully closed loop systems certainly help, but its just not the same.
Ultimately, I agree with other statements, if people WANT a speed density patch that works awesome and is easy to use, its an option to me, I might ultimately use it, and I'd certainly support the R&D to help out. It certainly would make finding service parts no more difficult than finding OE parts. But I would need to see a level engineering that makes for a good and reliable system in most conditions.
I have always had problems with the stock EVO MAF, tried everything to get better drivability out of it when I started making other modifications, and I agree it causes headaches. When I switched over to a Ford Cobra Maf and converter electronics, all of those drivability issues went away, and I gained performance. Don't get me wrong, it has its issues and tradeoffs, and takes time to learn to work with. But it does offer the best set of compromises.
Unfortunately the FQ300 Motec being Emission Legal, does not make it OBD Legal in the USA, no matter how clean a new car runs, if it doesn't have an OEM ECU, or at least support OBD-II Transparently, it will not be legal. Its unfortunate there is no exception process for people who really do have clean running cars.
Speed Density has its place, as does a carburetor, and MAF based systems. But so does Throttle Body, Batch Fire, and Sequential Fire EFI systems.. Again, we are talking about a production vehicle where its just as important to be able to drive the car comfortably, as it is for good performance. There's a reason why years of R&D goes into this stuff, there's also a reason why technology shifts away from one to another.
Speed Density is a simpler solution, it is based on calculations and minimal sensor data, that can never possibly accommodate all conditions seen in daily driving. It also "glosses" over things which makes some things seem smoother, when in reality they aren't, their just hidden until the conditions change. This might be fine for some people, But for me this is unacceptable.
I have a speed density TBI EFI system on my Malibu, It runs a wideband and is always running closed loop, under cruise and load. Every time the weather changes, the drivability changes too. It works well, way better than my carburetor did, but if you were to log the AFR and EGT data, it varies as conditions change and the calculations don't. Fully closed loop systems certainly help, but its just not the same.
Ultimately, I agree with other statements, if people WANT a speed density patch that works awesome and is easy to use, its an option to me, I might ultimately use it, and I'd certainly support the R&D to help out. It certainly would make finding service parts no more difficult than finding OE parts. But I would need to see a level engineering that makes for a good and reliable system in most conditions.
lol
Karmann sensors definitely have issues, their vibration and turbulence sensitive, but are regarded as more accurate than other types of flow based sensors. I don't intend to argue SD over MAF as you tend to see SD on race cars, and MAF on production cars.. production vehicles get years of R&D to determine what offers the best driveability, where Race car R&D doesn't really take drivability into account, and I'm sure if it did, it probably uses MAF systems of some sort.
MAP is a non invasive way to tune the engine. It can be VERY accurate and account for just about any condition you can think of. The down fall is setup. It takes a lot more tuning to get it perfect, hence the reason most professional racing teams use telemetry and real time tuning systems. Even after hours of tuning in a controlled lab, it still requires more tweaking to get it perfect at every single track under every different condition. But a MAF would be no different and would just pose as one additional “restriction” in the intake track.
I am almost certain that F1 and other NA racing circles use alpha-n almost exclusively, not speed density. These applications tend to be built and rebuilt to exacting details, and aren't concerned about changes that take place over long-term use. Alpha-n systems with basic correction capability are even in use for modern street vehicles, such as BMW's CSL (much to my surprise).
In a turbo car, speed density has the benefit of eliminating the MAF, but the tradeoff is the inability to compensate for changes over the long haul. I'm not entirely disappointed by that, but I'm not going to switch over unless I know it works without a hitch, and I don't feel like I need to drive around with a MAF in my trunk as a contingency measure.
In a turbo car, speed density has the benefit of eliminating the MAF, but the tradeoff is the inability to compensate for changes over the long haul. I'm not entirely disappointed by that, but I'm not going to switch over unless I know it works without a hitch, and I don't feel like I need to drive around with a MAF in my trunk as a contingency measure.
I am almost certain that F1 and other NA racing circles use alpha-n almost exclusively, not speed density. These applications tend to be built and rebuilt to exacting details, and aren't concerned about changes that take place over long-term use. Alpha-n systems with basic correction capability are even in use for modern street vehicles, such as BMW's CSL (much to my surprise).
In a turbo car, speed density has the benefit of eliminating the MAF, but the tradeoff is the inability to compensate for changes over the long haul. I'm not entirely disappointed by that, but I'm not going to switch over unless I know it works without a hitch, and I don't feel like I need to drive around with a MAF in my trunk as a contingency measure.
In a turbo car, speed density has the benefit of eliminating the MAF, but the tradeoff is the inability to compensate for changes over the long haul. I'm not entirely disappointed by that, but I'm not going to switch over unless I know it works without a hitch, and I don't feel like I need to drive around with a MAF in my trunk as a contingency measure.
None - I've never held a negative opinion about SD. All I am saying is that I am reluctant to give up a fully functional MAF system for a buggy SD conversion (where the factory ECU is concerned).
A couple of our ECU gurus have whittled out an SD conversion, but it has not yet demonstrated the capability to function in a suitable trouble-free fashion.
A couple of our ECU gurus have whittled out an SD conversion, but it has not yet demonstrated the capability to function in a suitable trouble-free fashion.
Ted B, you are probably right on the F1 and I didn't even really think about that as alpha-n is a "simplified" version of speed density (although obviously the maps can be very elaborate). On an N/A car, TPS is often used instead of MAP as it provides a more stable value for calculations. If I recall correctly, didn't GM use alpha-n in the throttle body injected motors?
However, on boosted applications, do they use alpha-n?
The point remains though, drivability is a major concern on race cars.
For what it's worth, almost every very quick car I know running speed density ends up tweaking the fuel map at every race. I'm not saying it's a better option for everybody. I'd prefer it, yet I started this thread on using other MAF sensors, so obviously I see the merits in both systems.
However, on boosted applications, do they use alpha-n?
The point remains though, drivability is a major concern on race cars.
For what it's worth, almost every very quick car I know running speed density ends up tweaking the fuel map at every race. I'm not saying it's a better option for everybody. I'd prefer it, yet I started this thread on using other MAF sensors, so obviously I see the merits in both systems.
For the benefit of the masses . . .
Alpha-N says, "at this TPS and this RPM, IPW = X", and that's the end. A fancy system will give an IAT and BARO compensation adjustment, otherwise, those must be made manually. If you have a NA engine with ITBs and velocity stacks, this is what you want. It requires the least processing power to operate and gives the quickest throttle response. This is a no-go for turbo and most SC apps, but I can see it working with a roots (positive displacement) blower.
Speed Density says, "at this MAP and this RPM, Load should be X, so IPW = Y". Simpler than MAF, yes, but if we make a change that affects VE (Load) at some rpm, there is no open loop compensation. SD doesn't work well for any app with ITBs or open-air plenum.
Mass Airflow says, "at this MAF and this RPM, Load = X, so IPW = Y'. Unlike SD, it has some open loop compensation capability, but in return requires the most processing power, and has the slowest transient response.
It is all a matter of give and take - sacrifice something to gain something else. As to which works best for any specific application is a combination of practicality and personal preference.
Alpha-N says, "at this TPS and this RPM, IPW = X", and that's the end. A fancy system will give an IAT and BARO compensation adjustment, otherwise, those must be made manually. If you have a NA engine with ITBs and velocity stacks, this is what you want. It requires the least processing power to operate and gives the quickest throttle response. This is a no-go for turbo and most SC apps, but I can see it working with a roots (positive displacement) blower.
Speed Density says, "at this MAP and this RPM, Load should be X, so IPW = Y". Simpler than MAF, yes, but if we make a change that affects VE (Load) at some rpm, there is no open loop compensation. SD doesn't work well for any app with ITBs or open-air plenum.
Mass Airflow says, "at this MAF and this RPM, Load = X, so IPW = Y'. Unlike SD, it has some open loop compensation capability, but in return requires the most processing power, and has the slowest transient response.
It is all a matter of give and take - sacrifice something to gain something else. As to which works best for any specific application is a combination of practicality and personal preference.
Newbie
Joined: Oct 2004
Posts: 45
Likes: 0
awdgsx91, thank you for clearing things up.
To me, it seems like we would be at the same end result for this particular application. You intend to do all the work in the MAF, making it a universal product that can be setup for virtually any car. I am suggesting doing the conversion in our ECU directly. Both method has it's benefits and obviously as a vendor, your method would provide a much larger market.
I have a question about your HPX MAF. It sounds like you are allowing the sensor output to be scaled almost however you like? Thus a 3" tube MAF can be scaled to read from say 30lb/min = 5V to 70 lb/min = 5V? Or for our purposes, you could scale so 30lb/min =1600 HZ or 70lb/min = 1600Hz and so forth?
Also, the HPX sensor has an IAT output available, correct?
To me, it seems like we would be at the same end result for this particular application. You intend to do all the work in the MAF, making it a universal product that can be setup for virtually any car. I am suggesting doing the conversion in our ECU directly. Both method has it's benefits and obviously as a vendor, your method would provide a much larger market.
I have a question about your HPX MAF. It sounds like you are allowing the sensor output to be scaled almost however you like? Thus a 3" tube MAF can be scaled to read from say 30lb/min = 5V to 70 lb/min = 5V? Or for our purposes, you could scale so 30lb/min =1600 HZ or 70lb/min = 1600Hz and so forth?
Also, the HPX sensor has an IAT output available, correct?
Example: 3" tube area 7.065 3.5" tube area 9.616 9.616/7.065=1.361 1.361x100 lb/min = 136 lb/min for a HPX installed in a 3.5" tube.. and yes they have an IAT.
Newbie
Joined: Oct 2004
Posts: 45
Likes: 0
Question:
Since a MAF's transfer function must be calibrated with respect to a MAF housing of fixed volume to be accurate, wouldn't you need to know in advance the intended tubing diameter and pre-program accordingly? If we match the transfer function of an HPX to a Mitsu MAF when that HPX is situated in a 2.5" tube, that will equate to a very different air mass if we switch that HPX to a 4" tube.
Any sensor is potentially a bad fit if data critical to its accuracy is ignored. Without IAT, any hot wire or hot film MAF is incomplete: PV=nRT
'A little programming' is required for the SD mod, but that also adds another layer of calculations and apparenty doesn't work very well either. If another $150-200 saves weeks (if not months) worth of programming and tuning hassles, I feel it's worth it.
For Mellon, myself, and others who will surely follow, the factory MAF is a limiting factor that clamps a ceiling on how much power we can obtain from our setups. For us, resolving this is a genuine need, the only alternative being a standalone. We couldn't care less if the upgrade is a draw through, blow through, or whatever, so long as it doesn't amount to a cheap, unreliable band-aid. A few hundred bucks is a drop in the bucket of what it takes to get to the point whereby this becomes a need. If this was just an exercise in inconveniencing myself in pursuit of 'ricer hawtness', I wouldn't waste my time.
Since a MAF's transfer function must be calibrated with respect to a MAF housing of fixed volume to be accurate, wouldn't you need to know in advance the intended tubing diameter and pre-program accordingly? If we match the transfer function of an HPX to a Mitsu MAF when that HPX is situated in a 2.5" tube, that will equate to a very different air mass if we switch that HPX to a 4" tube.
Any sensor is potentially a bad fit if data critical to its accuracy is ignored. Without IAT, any hot wire or hot film MAF is incomplete: PV=nRT
'A little programming' is required for the SD mod, but that also adds another layer of calculations and apparenty doesn't work very well either. If another $150-200 saves weeks (if not months) worth of programming and tuning hassles, I feel it's worth it.
For Mellon, myself, and others who will surely follow, the factory MAF is a limiting factor that clamps a ceiling on how much power we can obtain from our setups. For us, resolving this is a genuine need, the only alternative being a standalone. We couldn't care less if the upgrade is a draw through, blow through, or whatever, so long as it doesn't amount to a cheap, unreliable band-aid. A few hundred bucks is a drop in the bucket of what it takes to get to the point whereby this becomes a need. If this was just an exercise in inconveniencing myself in pursuit of 'ricer hawtness', I wouldn't waste my time.
It's very simple, You simply input the area or size of the housing you will be using into the software when you set up the maf.
Newbie
Joined: Oct 2004
Posts: 45
Likes: 0
Sorry, unfortunately I can't see the pictures posted due to filters at my work.
Is the PMAS sensor a voltage or freqeuncy output? The Ford sensors are voltage, so a Ford replacement sensor hooking directly to an ECU looking for freqeuncy is throwing me off a bit.
The transfer function from what I have found simply looks to be a look-up table of voltage vs. airflow.
Is the PMAS sensor a voltage or freqeuncy output? The Ford sensors are voltage, so a Ford replacement sensor hooking directly to an ECU looking for freqeuncy is throwing me off a bit.
The transfer function from what I have found simply looks to be a look-up table of voltage vs. airflow.
We have two HPX sensors.. one is V the other has a hz output..
If I remember correctly the EVO pcm strategy uses a flow/hz scalar and a compensation table opposed to a transfer funtion table like most other pcm's, but I'm sure you guys have figured out a way to go around that.
Last edited by awdgsx91; Jan 8, 2009 at 07:53 AM.
Is the transfer function in the HPX completely programable? For example, most hot wire MAFs seem to be nearly exponential on the airflow-output curve. Can this sensor be programed to produce a linear output with respect to airflow (y=mx+b)?