Page 3 of 3

Re: MAP but not SD as intermediate solution?

Posted: Mon Sep 01, 2014 10:54 pm
by Nobody
What optimum sampling method is, will depend on many variables even intake manifold design. Most current intakes have runners tuned for ~3rd order resonance and large plenum volumes. Large plenum volumes dampen pressure variations. Not sure why older Japanese engines has such wimpy plenum volumes?

Averaging is not the ideal solution because your available sample size for a window will vary with RPM. On a very well optimized engine you will see over 100% VE in small RPM window. The resonance will actually create above atmospheric pressure for a short period.

Re: MAP but not SD as intermediate solution?

Posted: Tue Sep 02, 2014 12:23 am
by kb1gtt
Another issue with averaging is the signal delay. The 1mS response is an mechanical equivalent of an RC filter. The signal delay will vary with for a variety of frequency components in the signal. AKA 1Hz will have one signal delay, while 100Hz will have another delay. To get a real reading you really need to do some de-warping then followed by a Kalman filter. However that's all complicated stuff, and for an early version the rough windowed then averaged filter will provide a functional starting point. It's not optimal or perfect, but it's functional and will work for many applications.

Re: MAP but not SD as intermediate solution?

Posted: Tue Sep 02, 2014 1:00 am
by Nobody
Keep in mind that some modern engines have pseudo Atkinson cycle, such as Mazda Skyactiv in which they claim 14:1 CR. In most cases they are bleeding off compression ratio by closing inlet valve late, so some air is returned back to intake manifold, effectively creating 2+ higher pressure pulses (reverse flow from late valve closing and standing waves - which can combine). Similar can happen with large cams with tons of overlap.

Then you have BMW with variable valve lift, to minimize throttling losses… Different rules again.

Simple averaging maybe OK for low tech lightly loaded engines…

Re: MAP but not SD as intermediate solution?

Posted: Tue Sep 02, 2014 2:10 am
by kb1gtt
OEM's often have those extra cavities to muffle the pulses. This quiet's the motor as well it dampens the signal seen by the up stream sensor, removing much of the pulse. A MAP that is really close to the intake is not so fortunate. These pressure pulses are one of many reasons why the min is often used instead of the max, as the valve overlap will generally create an excess in pressure, not a min pressure. There are certainly other better algos. For now lets get something that's flexible and upgrade-able in the future. Low effort and functional. Then we can add options later one down the road.

Re: MAP but not SD as intermediate solution?

Posted: Tue Sep 02, 2014 3:11 am
by Nobody
kb1gtt wrote:OEM's often have those extra cavities to muffle the pulses. This quiet's the motor as well it dampens the signal seen by the up stream sensor, removing much of the pulse. A MAP that is really close to the intake is not so fortunate. These pressure pulses are one of many reasons why the min is often used instead of the max, as the valve overlap will generally create an excess in pressure, not a min pressure. There are certainly other better algos. For now lets get something that's flexible and upgrade-able in the future. Low effort and functional. Then we can add options later one down the road.
As mentioned elsewhere I don’t play with Japanese cars, jetskis and motorcycles yes (Yamaha - FZR and R1). Anything built in the last 15-20 years that I play with does indeed have close coupled MAP sensor.

Google “MAP Sensor” and select images, 90% will support above comment.

Now even Bosch uses Freescale sensors in their MAP housing. Most good ECUs will run with lost cam signal, because they can secondary reference MAP sensor to determine intake stroke.

Now as far as secondary resonator on intake tract, this is done 90% of the time to keep MAF stable at low demand, nothing to do with MAP. Dampening pulses inside intake manifold is not desirable from performance standpoint. Oddly wasn’t it the presumption earlier in this thread that MAF was linear and not a curve? Wasn’t SD a magic table? Couldn’t complete an equation without example on how to covert KG to grams?

I’m not here to argue, simply stating more to it than that, on a 20+ year old grocery getter, yes its simple enough. Now switch to something performance oriented and it changes in a heartbeat… Also comments saying WOT is easy or forgiving, its razor thin if after max HP, especially with those stuck with 91 octane, with accuracy and repeatability being critical. Lots of corrections required and timing is usually crutch.

Install proposed ECU on a performance application with a person who will race their own shadow, then report back.

BTW - have you ever really tuned an ECU? I'm talking blow the tires off (spin) at 60+ MPH?

Re: MAP but not SD as intermediate solution?

Posted: Tue Sep 02, 2014 9:55 am
by kb1gtt
I think we are saying nearly the same thing. We agree there are better algo's for predicting how much O2 there is in the cyl. The difference is that I'm encouraging a get'er done approach, with enhancement options down the road.

I understand that OEM MAF is designed to have a "linear" region. A curve showing signal vs flow will have distortion at the low end and high end of the flow curve. As well the "linear" region is often skewed slightly. The linear region typically means you can put on a tolerance of say +/- 1%, then you can draw a line and the line can guarantee you are with in 1% at any time. You can get a slightly more accurate signal if you don't use the "linear" equation in the "linear" region. However getting the flow curve can be a bit hard, as many MFG's only publish the mV per CFM number and the min and max flow range. If you have calibrated equipment, you can characterize you intake and do a kind of de-warping of the signal to get a valid signal a the limits, but the tolerances are going to start to get ugly, so I wouldn't rely on a de-warped signal long term as the tolerance of the de-warped sections can go blah quickly. For OEM's if you can't get a good signal at the high end and low end, they put on the MAP, and generally only use the MAP for high end, then make the MAF small enough to be good for the low end. Then they blend as required to use the most accurate sensor at the optimal time. Performance people often don't care much about idle, or simply make it idle at 2kRPM or something like that. So it's common they don't use the MAF for low end stuff. After all who needs that weight, intake restriction, and you probably don't have calibrated test fixtures to take better measurements either. So keep it simple is the approach for most people.

My goals are science based, not racing based. So no I have not pushed the limits of performance. I generally find the safe tolerances and stay on my side of that tolerances. Also I run AWD, so you can't blow the tires off at 60MPH even if you wanted to. Rally folks like to make the car stick to the road. If you can blow the tires off at 60MPH, I'd recommend fixing your power deliver system :)

Re: MAP but not SD as intermediate solution?

Posted: Tue Sep 02, 2014 8:05 pm
by E4ODnut
I have to agree with the approach to keep it simple at first, add complexity only if required.

On my engines, the system I use is very basic. MAP is sampled once every 8 milliseconds. There is no attempt to synchronize it with any particular crankshaft position. There is no averaging of readings. The 460 has no plenum to speak of, the tubing from the manifold to the MAP sensor is ~ 8ft long and I have no restrictions in the tubing. The readings are quite smooth with good response time. On the 300 the situation is similar except this engine has a good size plenum. The 351s, also have a plenum and about the same length of tubing, but I had to put a 0.030" restrictor in the tubing to smooth out the readings. Response time is still more than adequate for this application. On all the engines when I compare my MAP readings with a good quality vacuum gauge and do the conversion, they correlate very closely.

My point is that even with this comparatively primitive method of sampling MAP it can work just fine for relatively low performance multi cylinder engines. If you have a single or twin cylinder engine then this method probably wouldn't work at all and some sort of synchronized sampling will be required. If synchronized sampling will work for a single or twin, it would probably work better than what I am doing for a multi cylinder engine.

Re: MAP but not SD as intermediate solution?

Posted: Tue Sep 02, 2014 9:20 pm
by Nobody
kb1gtt wrote:I think we are saying nearly the same thing. We agree there are better algo's for predicting how much O2 there is in the cyl. The difference is that I'm encouraging a get'er done approach, with enhancement options down the road.

I understand that OEM MAF is designed to have a "linear" region. A curve showing signal vs flow will have distortion at the low end and high end of the flow curve. As well the "linear" region is often skewed slightly. The linear region typically means you can put on a tolerance of say +/- 1%, then you can draw a line and the line can guarantee you are with in 1% at any time. You can get a slightly more accurate signal if you don't use the "linear" equation in the "linear" region. However getting the flow curve can be a bit hard, as many MFG's only publish the mV per CFM number and the min and max flow range. If you have calibrated equipment, you can characterize you intake and do a kind of de-warping of the signal to get a valid signal a the limits, but the tolerances are going to start to get ugly, so I wouldn't rely on a de-warped signal long term as the tolerance of the de-warped sections can go blah quickly. For OEM's if you can't get a good signal at the high end and low end, they put on the MAP, and generally only use the MAP for high end, then make the MAF small enough to be good for the low end. Then they blend as required to use the most accurate sensor at the optimal time. Performance people often don't care much about idle, or simply make it idle at 2kRPM or something like that. So it's common they don't use the MAF for low end stuff. After all who needs that weight, intake restriction, and you probably don't have calibrated test fixtures to take better measurements either. So keep it simple is the approach for most people.

My goals are science based, not racing based. So no I have not pushed the limits of performance. I generally find the safe tolerances and stay on my side of that tolerances. Also I run AWD, so you can't blow the tires off at 60MPH even if you wanted to. Rally folks like to make the car stick to the road. If you can blow the tires off at 60MPH, I'd recommend fixing your power deliver system :)
I’m not here to debate your definition of linear…

Exhibit 1
Image
Old Porsche and BMW type

Image
Another Porsche but better MAF

Exhibit 2
Image
LS1 1997-2004
LS3 2008-2013
LS3 is Hitachi MAF is used by many car manufacturers.

It’s a curve as I have always stated and you are still stuck thinking linear slope. Go back a few pages and reread what you said.

I guess my 650+ HP engine that idles at 750 RPM doesn’t count? It doesn’t buck, surge and generally almost like stock in terms of daily driving. My previous 900+ HP car (sold) now that did idle at ~900 RPM but can certainly be easily driven. After you get over a certain HP there isn’t a magic "street" tire that will plant power until you run out of torque multiplication with transmission (rear wheel drive, in my case).

As far as AWD, yup it’s nice, but then that’s what my Tahoe is for.

Are you serious about the science thing? Your assumptions on engine management would contradict that LOL.

BTW late model Mustangs are MAF all the way through idle to WOT.

Re: MAP but not SD as intermediate solution?

Posted: Tue Sep 02, 2014 9:27 pm
by Nobody
E4ODnut wrote:I have to agree with the approach to keep it simple at first, add complexity only if required.

On my engines, the system I use is very basic. MAP is sampled once every 8 milliseconds. There is no attempt to synchronize it with any particular crankshaft position. There is no averaging of readings. The 460 has no plenum to speak of, the tubing from the manifold to the MAP sensor is ~ 8ft long and I have no restrictions in the tubing. The readings are quite smooth with good response time. On the 300 the situation is similar except this engine has a good size plenum. The 351s, also have a plenum and about the same length of tubing, but I had to put a 0.030" restrictor in the tubing to smooth out the readings. Response time is still more than adequate for this application. On all the engines when I compare my MAP readings with a good quality vacuum gauge and do the conversion, they correlate very closely.

My point is that even with this comparatively primitive method of sampling MAP it can work just fine for relatively low performance multi cylinder engines. If you have a single or twin cylinder engine then this method probably wouldn't work at all and some sort of synchronized sampling will be required. If synchronized sampling will work for a single or twin, it would probably work better than what I am doing for a multi cylinder engine.
Be careful of long vacuum reference lines and restrictors - they can introduce tip in knock. Or on fast throttle transients MAP is playing catchup.

Re: MAP but not SD as intermediate solution?

Posted: Tue Sep 02, 2014 10:08 pm
by E4ODnut
Nobody wrote:Be careful of long vacuum reference lines and restrictors - they can introduce tip in knock. Or on fast throttle transients MAP is playing catchup.
Possibly. All I can tell you is that what I've got works well for me.