The DISI-MZR fuel system (Warning: Science Heavy)

Aiight, those of you that know me know the drill. No TLDRs for you this time; you can get scienced or you can get the fuck out.

Starting out with some simple stuff most people will know:
1. Stock HPFP (high pressure fuel pump) internals need replacing after a certain mod/power level.
2. Stock injectors are good for about 400 WHP, provided you have upgraded HPFP internals to feed them.
3. Stock ITFP (in tank fuel pump) is good for probably about the same 400 WHP; not sure anyone knows for sure but feel free to add if you do.

...And that's about it. That's what I'd wager most people know about the stock fuel system, primarily because until today, that's pretty much all I knew about it too. But now, at least, I have a better idea as to why these 3 (though really I'll only be going in depth on the first two) commonly held ideas ring true for us when they don't always for other platforms with similar technology (like the ST).

Most of you are probably also familiar with what it looks like when fuel runs out; if you are or if you aren't, I'll inform you and/or refresh your memory with these two snippets of log (provided by another, not from me):

View attachment 177


View attachment 178

The common thread between these two logs (and likely any other logs anyone wants to donate to this thread showing running out of fuel on stock HPFP internals), is that pressure falls off around 3800-4k RPM, peaks in badness around 4k-4500 and then comes back up thereafter, returning to normal shortly after 5k RPM.

The most important question I can think of regarding these logs isn't how to fix this issue, since we already have a very easy grasp on that, but rather, why does pressure return? One might think that it's because the stock turbo can't flow enough air, and that is essentially correct; but if you look closely at both of the logs, airflow is increasing with RPM and yet fuel still comes back. In both logs, it's even enriching the AFRs as it does so....So what gives?

Well, my epiphany came with a "well, duh" moment after I realized the math behind it. This came about because I wanted to figure out how much fuel an upgraded HPFP can actually supply (and the answer to that is ...amusing... to say the least).

We (should) all know that injectors are tested at a linear rate and pressure; this is evident in how they are labeled; 1000 cc per minute, 75 pounds per hour, etc; it's simply x flow over y time. Well, this static flow/time scenario doesn't actually apply to the HPFP, specifically because it's what's known as a positive displacement pump.

What is that? Well, I'll put it like this: Let's say you have a 5 gallon bucket of water, full to the brim, and want to remove water from it. Naturally, you step in it with one foot because fuck it, why not right? As you do so, the water that your foot and leg displaces flows out, and when you remove your foot from the bucket (hopefully without getting that shit stuck on your foot like a retard and splashing about), the water level drops. It's what makes hundreds of tons of huge ass boat float; they displace more than they weigh. Kind of like when you're in the pool trying to hold a basketball under the water and it comes back up and knocks your fucking nose back into your skull....But I digress.

Where was I? Oh, right. Positive displacement pumps. You see, much like the car engine, the HPFP has a bore (the face of the HPFP internals) and a stroke (the lift provided by the camshaft lobes), and this gives it a measurable displacement. Because it's connected to the cam, it's easy to figure out how many cycles per second it does and how that relates to engine RPM, which is really important here for the following reason:

While the injectors have the same flow potential (their flow rating) at idle and at 8,000 RPM (for example), the HPFP does not; the difference between idle flow and 8,000 RPM flow basically fucktuples (that's a scientific term, BTW).

Make sense? No? Then I'll add some spreadsheets. Bitches love spreadsheets.

View attachment 182

I'll preface this by saying my lobe measurement is probably off by at least a little bit, and that it doesn't matter that much for reasons that will become clear soon (I promise). Also, this spreadsheet doesn't factor in things like spray window reduction from timing, pressure loss due to HPFP lobe placement, etc.

You can see I've got the bore, stroke, camshaft lobes and the injector size populated (which is cut in half because it's not possible nor safe to run a DI injector at 100% true duty cycle like you can a PI injector, due to it injecting directly into the combustion chamber). These values all represent the stock HPFP internals which don't look like they are too much smaller than the Autotechs, but keep in mind the area of a circle grows drastically with just a small change in diameter; thus, the Autotechs have about 52% more bore area than stock, which is a substantial gain.

Starting at the top, the first red box reads "Max Flow RPM;" this is the RPM I calculated that the injector would no longer be able to keep up with the HPFP flow wise; obviously, this doesn't quite add up with the logs posted earlier as pressure was still falling up to 4600 to 4800 RPM; there are any number of reasons for this including but not limited to there being no pressure generated by the HPFP during injection events, or possible float/hangs in the internals that cause it to not cycle properly. More on this at a later time, if you all want to discuss that; I have multiple pumps I can tear down and show the goods, so to speak.

Moving on, the first black line separates the section for overall HPFP flow based on bore, stroke, and RPM (cycles per minute). The gray row has RPM, the purplish row has the flow data in CCs (note this is overall flow and not per injector flow).

Below that, however, is injector flow information. The gray line is RPM as above, but the purple one is different. This is the overall HPFP flow output divided by the number of injectors (with appropriate considerations for cycles, RPM, injection events, etc). Below that, the green line shows HPFP flow (per injector) vs the listed injector flow number in yellow at the top. Negative numbers here mean that the injector is capable of flowing more fuel than the HPFP can supply, positive numbers mean the reverse. Notice anything yet? That's right, the number goes positive (even with stock HPFP internals) after 4,000 RPM. It is likely that my measurements are a tad off and this should actually go positive at/around 4500 according to the previous logs, but again, this doesn't take into account lots of variables.

Probably the most amusing thing to note here is that even with stock HPFP internals, the math suggests that a car spinning 8,000 RPM would benefit from aftermarket injectors that are DOUBLE the stock flow rate. That would be roughly somewhere near 400 horsepower.

You're probably asking yourself "So wait, this motherfucker is saying you don't need HPFP internals to go big turbo!?" Well, yes and no. Chances are, on pump gas and without a tune/turbo setup to properly match that kind of configuration, you're going to need a lot of JBWeld to fix the hole in your block from running out of fuel at low to mid RPMs. For my build, however (which is a destroke high RPM build, just look for "DISI-MZResponse" in the Genwon build diaries section), it *might* be *potentially* feasible to run stock HPFP internals for lower power tunes. It would be hilarious, at least, and I might even try it briefly if for no other reason than to satisfy my own curiosity once my motor is built.

Back on topic. The nasty flow math gets better with Autotechs, obviously. I'll let you peruse this one on your own and compare to the stock one above:

View attachment 189


+50% flow at 4,000 RPM when stock internals were just breaking even; a listed full 2600 cc worth of injection room at 8,000 RPM. That is power waiting to happen.

But wait, there's more to this than just flow rates and pressures. Timing is pretty important too, as is the type of fuel you run. I'll actually need people to chime in with their logged IDCs (injector duty cycles) to further contribute to this idea, but I'd wager that those running smaller mixes of E85 can get away with higher IDC numbers than people running full E85 before the engine starts to stutter or misfire; this has a lot to do with spray window, timing, and the chemical properties of the fuel. For example, I've heard of people running in excess of 130% IDC without issue, while my car wasn't able to go past 105% without misfiring.

Some explanation for those that might not know:
In a port injected car, the fuel generally sprays on top of the intake valve and will sit there until the valve opens and allows the fuel to enter the cylinder along with a bunch of air (under high IDC values, anyway; for emissions and economy, most modern cars will start to spray just as the valve(s) start to open).

On a direct injected car, however, we can spray fuel during both the intake stroke (when air is drawn in through an open intake valve) and during the compression stroke (when all the valves are closed and the piston is on its way back up to top dead center for firing and the power stroke). Spraying fuel during the spark event can cause it not to fire at all or to fire weakly/misfire as the flame front is blown out (AKA spark blowout). This is also why DI injectors may have a huge listed rating, but can only operate at half that value.


Anywho, let's see what the timing and fueling relationship looks like:
View attachment 191

Here's what we've got going on here:
1. RPM across the top in grey (first value can be changed so it's yellow in my sheet)
2. MS Per Degree is how long it takes in milliseconds per degree of crankshaft rotation
3. The listed DI injection spray Window, which is essentially 180 degrees of crank rotation in milliseconds (and why it gets smaller as RPM goes up; another drawback that port injection doesn't have)
4. Ignition timing, which is always in degrees of crankshaft rotation
5. Timing MS which is how much time in milliseconds the spark event takes off the spray window (because you don't want to spray fuel during/after the spark, that's bad, mmmkay?)
6. The Effective DI Window which is the maximum injection spray window in milliseconds
7. And finally the % Window Lost shows you how much spray window you lost to timing in a number that's more relate-able.

What's interesting to note about this table is that the same timing values affect the fueling window by the same amount, regardless of RPM. This is because fueling window and timing are both based off the same core number: RPM and more specifically, MS Per Degree.

Going back to how timing affects fueling, you can see that 14 degrees timing reduces the fueling window by 3.89%, while 22 degrees reduces it 6.1%. Why is this relevant? Well, most pump gas cars will run a maximum timing value at or around 14 degrees, while most corn mix cars will run max timing at or around 22 degrees; this all depends on their redline, however, and going further up in RPM winds up needing more and more timing to keep the power band flat. There is a reason for that, but I'm not really prepared to go into that at this time.

SO, the difference in timing alone costs a car going from pump gas to a corn mix about 2.21% fueling headroom, assuming both cars only go up to 100% IDCs.

And on that note, I'll end this thread and give everyone interested time to ask questions which I can do my best to answer and add to this post at a later time.

 

I like the premise, but would also like to see more logs corroborate the 4-4.5K RPM "deficit". http://revisionsrus.com/logs/128 might also be useful to you. This log has my car hitting 2.5 load at ~4.4K RPM, and it hasn't been corrected for the hidden open loop trims. You can see it lean out from 11.8 to 12.7 AFR and recoup 8% IDC, and basically stays there until I lift at 6.7K. Normally it would approach ~115% IDC in a bit warmer weather. I don't think going lean at mid-range RPM is a particularly good strategy, but I suspect that does open up the prospect of tuning someone without HPFP internals on a K04.

 

A stretched out pump gas with no aux tune. Stressing out the fueling system for you. I can throw it on the dyno as early as tomorrow if need be. I also attached the actual log.

Also I hit some pretty decently high loads around the 4k area and seem to maintain hpfp pressure pretty well. I also have some logs here @Enki somewhere that I used stock internals on a GTX2867 lol. Don't ask. They were merely at spring pressure though which was 14psi I believe.

 

I mostly wanted to look at cars with stock HPFP internals to show this, since out flowing the injectors with Autotechs (or the like) is pretty easy.
One thing I did on my car, was run a corn mix on stock HPFP internals and a tune; remove VVT, you drop airflow in the danger range and can make the same power on a lot less boost.

 

A couple things:

1. PI cars (specifically boosted ones) can still cram shitloads of fuel into the cylinder with air to match, even at high RPMs. This has a lot to do with cam profiles. A PI car can just leave the injector open (@100 IDC, which isn't the greatest idea), but DI cars actually have to time it properly or it can hurt parts; spraying before the exhaust valves are fully closed can raise EGTs or even melt a catalyst if you're running it hard, for instance.

2. Higher rod:stroke ratio actually increases piston dwell at TDC and reduces dwell @ BDC, with shorter r:s ratios being the opposite (which is why most vehicles run a r:s ratio in the 1.5-1.6 range, as most cars aren't meant for racing). This hurts VE/torque at low to mid RPMs since there's less time before the piston starts back up the cylinder (and thus reduced cylinder filling/VE).

3. I'll be going as high compression as I possibly can on my build, which will also reduce required timing before MBT. Higher compression also increases thermal energy harnessed by the engine (efficiency), but requires lots of octane to work properly that way (because compressing anything makes it hotter and thus more prone to detonation in the case of air/fuel mixes).

I'm actually expecting my MBT (with 2.2:1 r:s ratio and 13:1 compression) to be somewhere around 6-10 degrees @ 6k RPM or thereabouts on full E85. Both of these should expand my overall spray window considerably, and my HP per airflow (and thus MPG) should improve as well, since upping the compression ratio makes better use of the fuel from a thermal efficiency standpoint, and having a longer r:s ratio does the same for cylinder pressures with less overall timing (better torque at lower boost and airflow levels).

 

I totally agree that injection timing is critical for DI, I was just saying that we might not be as constrained as originally thought.

After thinking about it yes, TDC dwell time would be increased which would make the engine more knock prone but the rest is still true with there being more dwell time at BDC and allow for a longer injection window.

I think that a high rod/stroke ratio with high compression are good compliments for each other and that you're going to get good results from it. The longer dwell time at TDC would allow less chance for abnormal combustion and make cycle to cycle combustion events more consistent which is very good. My gut tells me you should be able to run 10 degrees at 6k but we'll see when you start tuning. I'm still curious about the other end of the spectrum with lowering the rod ratio and what that might bring to the table for us. The biggest downside I see to it is increased wear due to the higher rod angles but that can be mitigated with an offset gudgeon pin. Another disadvantage would be that the engine would be a bit more knock prone.

 

MAJOR UPDATE!

OK, some of you might be aware I was working on "something special" for the past couple of days, and I'm happy to report that I've got it mostly figured out, barring a couple things. Special thanks to @Poon_Flavored_Tang for the logs he provided that helped me figure at least this section out, and to the others that tried, but ultimately could not provide what I was looking for.

First Revelation:
Everyone (myself included, originally) thinks that logged IDC is a bullshit number because the math never quite adds up. How can you add timing and have the IDC not change? That doesn't make sense.
For now, I'm going to say that this is false, but more research is needed (and could be answered by one log). Let me further state that it's probably false until certain conditions are met, and then it becomes true; more on that later.

Second Revelation:
Injector Duty Cycle is unaffected by VVT.
This is data based on the logs I've looked at, at any rate, but that may change if someone somehow exceeds 100% IDC while VVT is in full on pork flow mode. I doubt anyone wants to test this for me though, so I'll just leave it at that.

Third Revelation:
Injector duty cycle is always based on 360 degree injection window. This is, of course, barring the results from the logged information that I still need to examine. But wait, how is this possible? You can't spray fuel anywhere outside of intake and compression strokes!
Incorrect.

Fourth Revelation:
Stock camshaft timing plays a big role here (specifically the exhaust), and makes me fearful regarding upgraded cams possibly hurting the engine due to how the fueling window is actually calculated.


How it all fits together:
First thing's first; let's look at the cam profiles:
View attachment 1102
Most of you, if you're reading this, should know what valve overlap is. If not, it's where both intake and exhaust valves are open at the same time. It's likely the Mazda engineers chose a profile with 16 degrees of separation (that's -16 degrees overlap) for a couple reasons; one of those being emissions, as camshaft profiles with even a little overlap tend to allow raw fuel into the exhaust on port injected cars. Port injected cars, however, don't have a fueling window limit like DI cars do; this is why we may have 1700 CC per minute injectors, but can only use about 850ccs (~400 WHP @ 100% IDC) worth of fueling through them. The second reason Mazda might do this is to either simplify injection timing logic or to preserve potential power production up to a point. I don't think they anticipated this platform running any sort of alcohol fuel when it was designed, so I'm going to go for the former rather than the latter.

Moving on, let's look at the math:
View attachment 1103

The spreadsheet diarrhea above is about 3 solid days worth of banging my head against the keyboard. On the far left side, the yellow cells represent actual logged values, and are entered into the formula which calculates everything out in stages. With a listed RPM of 6388, each degree of crankshaft rotation is exactly 0.0261 milliseconds; not a lot of time to be sure. Just below it is the base DI window as calculated for one complete rotation of the crank at that RPM in milliseconds. Still not much time. Below that is ignition timing, which is in degrees of crankshaft rotation before top dead center.

Now, you would normally think that ignition timing would detract from the fueling window, because spraying fuel after the spark event could be seriously bad juju as far as cylinder temps and power production goes, so you'd want to remove the timing from the IDC calculation. Makes sense, right? Well, if you do that, your numbers will never, ever line up. In the middle of the screenshot above, the IDC column at the bottom shows 102.98%; this is the row we will be working with (the one separate from the ones above it).

Keeping that number in mind, manually calculating the IDC works out as follows (logged IPW is listed in a yellow field above):
Injector Duty Cycle = Injector Pulse Width / (crankshaft degrees * Milliseconds Per Degree) * 100%

Now with numbers:
IDC = 9.67 / (343 * 0.0261) * 100 -=== note that 343 degrees comes from 17 degrees timing minus 360 degrees rotation) ===-
IDC = 9.67 / (8.9523) * 100
IDC = 1.08016.... * 100
IDC = 108.016%

As you can see, it doesn't add up to the logged value.
However, if you use the same exact formula above, but leave the crankshaft rotation at 360 degrees, it comes out to 102.91%. I know it's not exact, but that can be explained by the response time of the logging system pulling down an accurate RPM number. With the rounding I'm using, the RPM window for accurate math comes out to 23 RPM or in this case between 6374 6397 RPM; a super loose variance of 14 RPM below actual logged values.

So how did I figure this all out? Good question. I'll repost the above screenshot to minimize scrolling for the explanation:
View attachment 1103

Now on the right hand side, there's some obvious columns like IDC (Injector Duty Cycle), IPW (Injector Pulse Width) and IgnTim (Ignition Timing), which are all values pulled right out of the log. On the left hand side, are some random values floating to the side of the big block of code we were talking about earlier; this is how I found out what was going on. In that third column, next to the "MS Per Degree" column, is a formula I was randomly throwing numbers into multiplied by the MS Per Degree value; I was doing this to get the value next to "Eff. DI Window" value up high enough to make the value next to "Logged IPW," which is actually the spreadsheet calculated IDC, to match the logged IDC values. This is the value I placed in the "Adjust" column. It took a while to realize that the adjustment was actually really close to the logged timing values, so that's when I added the timing column for comparison.

The "Closed" column has the timing information from each dataset I manually entered, which turned out to be the exact same value (in some cases, dead on balls accurate in some) as the "Timing MS" value on the left (light purple color). This represents how far before the intake valve opens (time wise) that the injector starts firing.

The "DegA" column represents a conversion from milliseconds back into degrees of crankshaft rotation, and uses both the "Adjust" and "Eff. DI Window" (dark green on the left) values combined and mathed based on the "MS Per Degree" value. This is when I noticed the trend that sent me down this rabbit hole.

Finally, the "DegU" column is the IPW expressed in degrees of crankshaft rotation, and matches the IDC perfectly (when divided by 360).


The Great Epiphany & TLDR:
What does this all mean?

1. Our stock cam timings are part of the fueling equation, and changing them can have unexpected results on fueling window; I suspect that this hasn't been much of an issue because people running aftermarket cams are also running aux fuel to make use of them in most cases.

2. Cars running high timing are likely to encounter a lower IDC threshold before running lean and/or misfiring. This is especially true since the piston is likely to still be encroaching on top dead center when the spark event goes off.

3. Cars running high timing are probably more likely to spray fuel into a cylinder with an open exhaust valve.

4. The picture is not yet complete; I need a car with specific configuration (including tune) to log data for me so that I can figure out which one of the following possibilities is true:
a. The fueling logic is static and as listed above; this would mean ultra high timing is worse than high IDCs on most non aux fuel vehicles
b. The fueling logic adjusts the fueling window (and thus IDC calculation) based on timing up to 16 degrees of camshaft deadspace; this would mean that high timing and high IDCs are equivalent in their "badness" as far as most cars are concerned.

Edit 2:
5. It is still unclear if an increase in IDC will go only forwards (after the spark event), before (eat up more cam "dead" space) or both.
End Edit

6. This affords us a unique opportunity; the opportunity to change fueling logic for the tuning devices that will support doing so, and if upgraded injectors come out, this will ultimately be a requirement for cars that are trying to push the envelope without aux fueling, as the bigger the injector the more loss there is when spraying during an open exhaust valve and/or after the spark event


Theories:
1. The cause of the lean event on most cars is not a result of high IDCs, but rather, the injector being open when cylinder pressures are high for a higher fraction of the injection cycle. This is less running out of fuel than it is running out of window.

2. A car running high timing is more likely to experience an IDC overrun misfire (spark blowout; aggressive leaning) than it is to run lean gradually. This would be because cylinder pressures will be significantly higher with the higher timing as the spark event happens while the compression stroke is still taking place, and an open injector is likely to have a higher backpressure into the rail in those conditions which would drastically decrease flow into the cylinder vs cars running lower timing.

3. Aftermarket cams are likely to cause the engine to run leaner than reported, as some raw fuel will find it's way into the exhaust, thus skewing AFRs. Thus, they are more likely to blow up or hurt otherwise stock engines that are not running a aux fuel; better yet, run aux fuel for cooling only and don't tune for it; let the engine run rich if you're going to run cams for safety.


Footnote:
If anyone is willing to fill in the last bit of info I need to understand this whole thing completely, here is what I need specifically:
1. Any gen vehicle
2. Must be big turbo
3. Must be running a corn mix
4. Must NOT have aux fuel installed or enabled (stock injectors should be the only thing supplying fuel)
5. Ignition timing must be at/over 18 degrees
6. Injector duty cycle must be at/over 105% (if not misfiring/leaning out; if it is, this information is still helpful)
7. The following information *MUST* be present in the log (at a minimum; more information is OK) to be of any use:
a. RPM
b. IDC
c. AFR
d. Ignition Timing
e. IPW
f. Though not required in the log itself, it would be helpful to know what your fueling target is.

EDIT:
Next on the agenda:
Trying to figure out maximum safe IDC before having issues like leaning out or misfires
"Fueling Cheat Sheets" where applicable based on the above result

Questions? Comments? Concerns?

 

Review of a couple new logs suggests that fueling logic follows the static method originally proposed in Epiphany 4; thus, going over 16 degrees timing probably isn't the best idea if your exhaust is free flowing at higher RPMs, and it would be better to add boost instead to an unidentified degree.

 

So @rfinkle2 challenged my thought processes behind the prior big update post (in a constructive way) and this is the result of further discussion and looking over some logs he provided for analysis:

1. VVT does not affect IDCs even a little; logged vs calculated output was between .01% and .001% accuracy, and this was done with a car running 10 degrees of VVT @ 6063 RPM with 99.19% IDC, 15 degrees timing and 9.82 IPW.

2. HPFP pressure variance is unlikely to have much effect on fuel targeting; as stated previously by Steve over at VersaTune, fuel flow increases with the square of pressure. I'm sure this isn't quite right, but the way I picture this statement is as follows:

Flow = 1 Pressure = 1
Flow = 2 Pressure = 2
Flow = 3 Pressure = 4
Flow = 4 Pressure = 8
Flow = 5 Pressure = 16
Flow = 6 Pressure = 256

Another analogy is that the difference between 1800 and 2000 PSI on our cars is probably equivalent to the difference in flow between 40 and 44.5 psi on a PI car; not that much.

Maybe someone can chime in with a better example; further questions, comments, and concerns are always welcome.

 

Need to post a correction; valve overlap is actually -38 degrees, not -16.

Don't really think it changes the math any, though.

 

Knowledgebased! Discuss this further: http://mazdaspeeds.org/index.php?threads/the-disi-mzr-fuel-system-warning-science-heavy.648/