alpha-N (TPS) vs MAP load source

This is going to be a long post so I'll put the question right at the top: am I better off using manifold pressure (MAP) or throttle position (alpha-N) as the engine load source for tuning a rotary with ITBs? The internet seems to conclude TPS is better but if MAP is properly setup is there any benefit to switching?

Continue for more details...

The engine is a naturally aspirated 13B rotary with full electronic fuel and ignition control. The engine is street port ported, which I'd say is about as aggressive as a moderate cam in a piston engine since it happily idles at 1000 RPM. The biggest change over a stock engine is I run an IDA style throttle body and a cast aluminum intake manifold designed for IDA carburetors.

When I first got the car running I used TPS. It generally ran poorly but it was probably more the tune itself than the load source. About a month later, at the recommendation of a friend, I switched to MAP. The vacuum signal is collected in a small block, maybe 0.6 cu. in. in volume, from each rotor and sent to the MAP sensor. Special attention was paid to placing the signal sources as far downstream of the throttle plates as possible but just upstream of the primary/secondary port split and that the distance from source to sensor was the same for each rotor and as short as possible.

The car runs well enough for a track toy. No issues at low/medium up through full throttle. Very light and transient throttle conditions could be improved. I do have to rely on "zero throttle" tables for fuel and ignition to get the car to idle smoothly and not stall when coming to a stop. This is because the engine has low and unstable vacuum at idle and low RPM. At zero throttle max vacuum is -13 psi above 3000 RPM but below that it steadily drops to -6.5 psi at my 1000 RPM idle. This trend holds true for all throttle positions:

That is TPS as a function of manifold pressure and engine speed. I used MegaLogViewer to pull data from dozens of logs taken with my current setup over the past two years. I tried to collect data from points that were frequently hit. That's why there's missing data. The big takeaway though, as expected, is MAP changes significantly with small changes in throttle position. Especially at lower engine speeds. Using 4000 RPM as an example nearly the entire MAP range is used over the course of just 25% throttle change. Since I'm upgrading ECUs I'm wondering if I should take this time to switch to TPS-based load source. Would there be any benefit or is it one of those six-of-one, half-dozen-of-another things? Based on my chart it would appear MAP-based load tuning provides more fidelity at light throttle conditions whereas TPS is better at higher throttle conditions. But if the maps were accurate in both cases would there be any difference in how the car behaves? Are there certain parameters that are always better suited to MAP or TPS based loads?

It's clear that TPS and MAP are not 1:1 so converting from one to the other won't be as simple as changing the load source option in the ECU software. To maintain the same degree of fidelity I'd have to go to a non-linear scale on TPS. Probably something like 0, 2.5, 5, 7.5, 10, 15, 20, 25, 30, 35, 40, 45, 50, 60, 70, etc. and even then there'd be lots of moving of cells. Since this will be quite an effort I want to make sure there will be improvements in how the car drives.

TPS is probably easiest.  I have been running speed density for literally ten years on a bridge port with an ITB like setup, but it gets its vacuum from channels in the throttle body that connect the ports.  I also have a carb cleaner straw as a restrictor in the MAP line, a throwback from my first install on a GSL-SE engine with a MS1/1.01 setup.  It did make throttle response less crisp as there is just enough signal lag that it was noticeable and I had to bondo over that in the tune with acceleration enrichment, which is not "right" but when dealt with properly the drivability is great.

Definitely will want to take advantage of acceleration enrichment scaling with an ITB arrangement.  I believe I have mine set up so 1200rpm is 40% or so and it goes to 100% at 5000 or so.

A MegaSquirt system has an ITB mode that allows you to blend S-D and Alpha-N algorithms.

A vanishing MAP signal like you have is generally the issue with MAP and ITBs leaving alpha-n the best option, especially on a track car as it gives much better resolution near WOT which is all you really car about.  Yes the chosen TPS points will be very non-linear but that doesn't really matter.  The bigger issue I always have is living with only 16 load points (well I refuse to do it so it's not really a problem :) ).... but its workable for track use.  

I've never had a lot of luck with blending schemes, but others swear by it so clearly results my vary.

In reply to mke :

Rotaries have the benefit of having twice as many intake pulses per port.  Between the two intake ports, vacuum never "vanishes" as each port gets vacuum roughly 75% of the time.  You do get a bouncy signal with the lack of plenum volume.

As far as load points, I think I could have done well with a 4x4 map.  The VE curve is almost flat once over about 2000rpm and rotaries are pretty insensitive to fuel mixture up to the point where it starts misfiring.  One of the things that annoyed me about going to MS2/Extra is that I could not keep using a 8x8 map, the smallest was 12x12.  Good enough really is good enough sometimes  I especially liked the one OE computer (Chrysler SBEC?) that didn't have VE maps, it had a VE algorithm.  No lookup table, just mathematics.

Pete. (l33t FS) said:

In reply to mke :

As far as load points, I think I could have done well with a 4x4 map.  

right click, delete (or add) column...oh wait, you don't have an enginelab ecu 

In reply to Pete. (l33t FS) :

FWIW, pretty much all OEM's use an algorithm of some type to estimate the air in the cylinder.  Some are really complex, where there are multiple tables to provide inputs to that equation, though.  Which makes the air calculation almost impossible to calibrate in the vehicle anymore.  I miss the good old days of the MAF sensors.

In reply to alfadriver :

The way it was described to me, there WERE no lookup tables per se, just some math that described a section of a curve in load and a section of a curve in RPM and they were added (or multiplied?) together.  I suspect this is more than sufficient for a sedate engine like a 2.5l OHC meeting 80s emissions and performance standards.  I also suspect that they tailored the ENGINE's capabilities to match what the computer could do.

SDS EFI is similar except you actually get a load "table" and an RPM "table", no math, just numbers. 

I was curious, because I never compared throttle position to manifold pressure before, so I loaded up my latest datalog and set up the histogram.

Just for S&Gs I looked at what injector pulswidths were, too.

In reply to Pete. (l33t FS) :

Back in the simple world, that worked.  Now, though....  

It's hard to say what the specific equation is- as I can hardly figure out the one we used.  But I know it's basically 

Ax^2+Bx+C.  

And we have tables that cover how each A, B, and C changes over the operating range of the engine.  It's not exactly that simple, but there is a slope (B), offset (C), and quad (A) term.

Having seen the GM system via HPTuners, it's similar in concept, even using the same basic formula, but the physics model is slightly different.

Things vary based on engine speed, temps, intake and exhaust cam position, EGR... it can go on to the engineer's dreams of what matters.  There are even some terms that end up being circular.

Most people think the WB sensor provides enough feedback to run it live- but it really can't be good enough, fast enough.  It's good for a/f control and making the car drive well with great emissions, but it's not really fast enough to direct feed back to the core estimate at a really high rate.

Geez, just thinking about that last part- we have a handful of long term corrections to the a/f estimate, which is a small fraction of all of the inputs- so there's no way to have a correction for everything.

Pete. (l33t FS) said:

I was curious, because I never compared throttle position to manifold pressure before, so I loaded up my latest datalog and set up the histogram.

I'm trying to understand what I' looking at....MAP on the Y axis and TPS in the cells? Color the amount of times it occured?  ....but the numbers look odd because obviously you can have 100% throttle at any RPM so it has me confused?

In reply to mke :

That had me thinking a lot too.  It can't be a direct correlation of TPS to MAP for the reasons you state.  As best as I can figure, this was from an 8 megabyte datalog from a long street drive, so it is all sedate driving, so correlation is more accidental than anything.

What would make more sense, if one were looking for a MAP-TPS correlation, would be to have TPS on the Y axis and MAP in the cells.  Like this.

I had to extend the Y axis down to zero.  My normal histogram setup is a copy of my VE table settings so that I can just copy cells over, there's no sense having cells go below 20kpa if the engine never goes anywhere near 20kpa

I am reasonably certain that the vacuum numbers are accurate, as the engine IS barely running when it is down at 800rpm.  Or 1200 for that matter.  I think it is only in the 600rpm cell (a concession to needing to invent more cells) in the period between cranking and running.

Just for more S&Gs I also plotted pulsewidth, which, if the injector constants are accurate, should be a decent indicator of airmass per cycle.  After filtering out transients, of course.

Pete. (l33t FS) said:

I am reasonably certain that the vacuum numbers are accurate, as the engine IS barely running when it is down at 800rpm.  Or 1200 for that matter.  I think it is only in the 600rpm cell (a concession to needing to invent more cells) in the period between cranking and running.

hmmmm....normally an idling engine is pulling vacuum, the lack of air is what holds the RPM low.  If it was truly a MAP of 98 at idle that would mean it has all the air it could want and opening the throttle would change nothing, so there would be no increase in RPM.

I think this is all proof that alpha-n is the way to with most ITB setups 

In reply to mke :

Most of the reason I never went alpha-N was a lack of a repeatable TPS.

And yes, throttle response is a mite soggy down there  I keep the idle at 1200 or so. Most people turn the idle speed up to 1500-1800 because of that, but that is a bit rowdy on the street. The idle goes from a rolling crackle to sharp power chords.

I suspect the OP's engine will be happier, plus two 50mm (55mm?) blades have different area vs. angle characteristics vs four 42mm blades.  What is interesting is that once over a certain speed or load, the difference between a street port and a bridge port is very small.

In reply to Pete. (l33t FS) :

Standing out in the shop fighting with my timing chain cover I realized why your graphs look the way they look....histogram+lag.  Its showing info from the log but not in the form of matched sets so for example you hit the throttle, then MAP goes up, then RPM goes up so there is a time shift and it doesn't actually show causation, its more a loose correlation caused by how you happened to be driving.  The classic example is when ice cream sales increase so do crime rates.... but ice cream doesn't cause crime, the relation is warm summer days which make ice cream taste better and staying out late more pleasant but tempers less pleasant creating a lot more opportunity. 

I tried for a couple years to run Speed-Density on my ITBs (not a rotary, just so you know).  I finally have up and ran Alpha-N and never went back. I only use the MAP for constant baro correction now; it's not even hooked up to the motor.

I have very non-linear bins for the throttle position, chosen based on where I spend most of my time in throttle and what I'm trying to tune out.  The top end of the RPM is much less precise and it doesn't seem to matter there. Down low in every day throttle and street driving, the more bins the better your off-idle is going to be.

I won't go back to SD on any ITB engine again.  Very happy with Alpha-N.

Back after a long weekend and short vacation so catching up now. For what it's worth I filter the MegaLogViewer histogram so I'm not catching edge cases. And yeah, more green = more hit points. Maybe I'm thinking about it incorrectly but I thought the values were accurate. You can have 100% TPS at any RPM but it's always going to correspond with atmospheric pressure so the only fields with 100 in them will be the far right on my plot or the very top on Pete's. What was hard for me to wrap my head around was that relationship between TPS and MAP isn't constant for any condition other than 100% throttle. For example, if I go from a steady state cruising condition at like 5% throttle then quickly jump to 25% and hold it MAP will jump to nearly atmospheric then pull more vacuum as engine speed rises. You can see this in my plot using 10% throttle as an example. If the initial throttle movement occurs at 2000 RPM MAP will be about 2 psi below atmospheric. Holding it there will increase vacuum to around -9 psig by 5000 RPM. It'll peak around -10 or -11 psig above 5500 RPM and go straight up from there since it requires more than 10% throttle to hit those points even with no load.

I don't know if the Haltech can enable multiple engine load sources for a single table. That would be a neat option, though. MAP for low load/road driving and TPS for high load/track stuff. It does say that even with alpha-N as the load source a MAP signal is still required so perhaps it's automatically taking that into consideration. Like Pete said the VE table is generally pretty flat except for a jump around the -2 psi mark. That's just because 80% of the MAP columns represent only up to 30% TPS. It looks like I might try to make TPS work. If for anything just to experiment. I'll get the car running on the new ECU with speed density since I know that works 

infernosg said:

It does say that even with alpha-N as the load source a MAP signal is still required so perhaps it's automatically taking that into consideration.

Alpha-N requires a barometric correction, that might be what they are talking about? 

Which Haltech do you have?...I'm happy to take a quick look through the manual.

Alpha-n doesn't require baro, its just a really good idea. 
 

if idle map is ~70 kpa or higher I will go right to alpha-n. On street cars I may use blended tables (alpha-n * map lookup) or itb mode if its on ms. Race cars, alpha-n and send it. 

In reply to mke :

Haltech Elite 1500 ECU using their ESP software. The excerpt from their Help file regarding using TPS as the primary load source is:

When selected with VE as the Tuning Method, a MAP sensor or a Barometric Pressure Sensor is also required. The MAP sensor is used as a secondary load sensor for vacuum and/or boost. A Barometric Sensor is used for barometric or altitude correction. (See Fuel MAP Correction table)

The definition of the Fuel MAP Correction table is given as:

The percentage correction table for changes in Manifold Pressure. Up to 3D mapping is available with this table. When the Tuning Method is set to VE and when the load source is TPS, the Manifold Pressure is automatically corrected for, as part of the VE fuel calculation. In this case, the table should initially be set with all values at 0% (default). The table is then used as a correction for any error in the automatic calculation. When the Tuning Method is set to Injection Time, there is no VE calculation auto correction, and all values need to reflect a raw correction with changing Manifold Pressure.

For more info I'm actually using an external 1 bar MAP sensor instead of the Haltech internal unit. I did this to keep the length of vacuum hose to a minimum. So in theory, if I switched to alpha-N as the load source I could take advantage of both the MAP and barometric correction tables.

Paul_VR6 (Forum Supporter) said:

Alpha-n doesn't require baro, its just a really good idea. 

I guess it depends on the ECU. Being a good idea I'm pretty sure some ECUs require the input. and I think he was saying the haltech required a MAP, which I was just guess was really Baro....but I haven't done a haltech for probably 10 years  so I'd need to look though the manual.  I remember the motecs I did  the manual said to hook it up, but I'm not sure what would have happened if the signal was missing.  The way I wrote the code (model) in my ECU a baro input it is required as I did not include any setup item to disable the input....but thinking about it now the ECU has 5V pullups on all the AN inputs so I guess if there is no baro sensor attached then its read as  the 106kPA sensor limit I guess.  

infernosg said:

 So in theory, if I switched to alpha-N as the load source I could take advantage of both the MAP and barometric correction tables.

That is a nice setup if you can get a decent MAP reading , which it sounds like you have.  This is the MS blended idea but without table switching.....I think its much easier to tune and more flexible,

It looks like it's doing an air density calculation like a baro correction, but using the manifold density which effectively is adding resolution to the TPS.  I set a haltech up like that years ago when there was no correction table so numbers were effectively all set to 0 for me, but it worked really well even before they made it a real feature.

Edit: Since you are using an external MAP it will probably let you set the internal to Baro...and I suspect will use  MAP/BARO or similar for the correction which is probably better yet.  

The real PITA here is it's all very ecu dependent and some don't include good documentation. 

In reply to mke :

That's the plan: get the car running on the new ECU using what is essentially the same old speed density tune, then mess with converting to alpha-N as time permits. It looks like I'm going to have to adjust the base Haltech barometric correction table, though. I don't understand where they got these numbers from at all:

Should be 0% at 14.7 psia assuming SL. Also, I don't know where on earth one can drive where atmospheric pressure is 7 psia... Since I'm at around 200 ft my baseline barometric pressure is closer to 14.6 psia. I don't expect to ever be driving above 5000 ft or below SL so my total range should only be 12.2-14.7 psia. Going solely off air density that's a range of like -13.5 to +0.6% fuel correction if I doing the math correctly. Won't know for sure until I get the new ECU setup and find out what the onboard sensor thinks local atmospheric pressure is.

In reply to infernosg :

Haltech is always been a bit odd by todays standards but in general the way its been is they expect you to set corrections to 0% when you tune, then correct from there.  So whatever the barometer says when you tune, you set that point in the table to 0, then use the ideal gas law to fill in the rest as a baseline that should never need to change.  So, yes, your table will be different from theirs.