Lucifer's Hammer Part 7
Posted: Dec 28, 2006 10:04 PM
Lucifer's Hammer Series Index: Part 1 • Part 2 • Part 3 • Part 4 • Part 5 • Part 6 • Part 7 • Part 8 • Part 9 • Part 10 • Part 11 • Part 12 • Part 13 • Part 14 • Photos
"Yo, Chicken Lips---gets me a beer. The Wordy One has written the sequel to War and Peace . . ."
This next lecture 8) is going to get into some math regarding engine capabilities and some constraints on what's feasible on the S38. I'm including all this detail because (a) there are a lot of misconceptions out there about what kind of power our motors can generate and (b) early on a lot of questions got asked about component selection.
For openers read the following paragraph.
OK A lot of youse gize have, or have read Corky Bell's "Maximum Boost." This is a fine introduction to turbocharging, though it has become a bit dated. There is another book which, IMHO, anyone thinking about forced induction absolutely must read. This is A. Graham Bell's "Forced Induction Performance Tuning." The two Bells are not related, BTW. "FIPT" is available thru Amazon, costs about $35. Gentlemen, this is the best $35 you can possibly spend.
It is not a cookbook how-to manual, but it isn't a graduate engineering text either. It will save you any number of blind alleys and wasted effort. SPEND THE FRICKING MONEY THEN READ THE SUMBITCH. It has a very good section regarding engine building in general, applicable to NA motors as well as FI. As such, it provides some excellent guidance. SPEND THE FRICKING MONEY THEN READ THE SUMBITCH.
Next step. Now for some websites
1. http://www.not2fast.com/wryday This is a very, very good turbo calculation model. It will allow you to do a lot of what-ifing. It is linked to a number of compressor maps for applicable Garrett turbos, and as such, answers a lot of questions. The model doesn't take into account variations in exhaust system design, or head flow capabilities, or what happens with using standalone ECUs or changing octane in fuel, but it is a most useful tool.
2. http://www.injector.com This is Marren Fuel Injection's website and has a number of useful formulas regarding injector sizing. They also sell a very nice injector rail setup for the M30.
3. http://www.RCEng.com This is RC Engineering's website. This outfit is another source for injectors, and also has a very good technical page regarding injector sizing.
4. http://www.turbobygarrett.com Garrett's site. Has an extensive list of the GT-series turbos and their Garrett p/n's. Also has compressor maps.
5. http://www.araoengineering.com For anyone who wonders why a 4-valve head will beat a 2-valve, this is required reading. Arao builds 4-valve heads for US V8 drag motors. Definitely on the Required Reading list.
A recap:
I originally wanted to refresh my S38B35 to restore it's original power.
Because I live at 6200 feet altitude, I am automatically down 20% on power. I wanted to get sea level performance. At altitude.
I wanted a road burner/endurance motor, not a drag strip lord.
"There is no replacement for displacement." Parnelli Jones. Can I stroke this thing?
Go big cams and high compression = having to deal with the environazis.
Don't get greedy and FI will solve your problems. Yaaah. Riiiiight.
Displacement
I went back to my original notes in my "Build Journal" where I've kept track of all the activity related to this project. This had summaries from my conversations with Rowe, Fahey, Korman, EB, etc. EB had said the crank was going to be 94 mm; in fact it is 95. Bore was going to be 94 mm. In fact it is 94.36. Working from EB's numbers, the original displacement was going to be 3914 cc. Actual is 3986. The differences are small, but they will add up as you will see.
Sidebar here.
Early on, there had been some discussion about going out to a 95 (94.95) mm bore. FWIW, the maximum oversize piston on the S38B38 is 94.955 mm; this leaves zero room for any future rebore or refresh. Going out to 95 mm was rejected because if one does that, the amount of material between the siamesed bores becomes extremely thin--around 5 mm. This has the very real potential to allow the castings to crack. Not Real Smart on an FI motor where cylinder pressures are going to be bumped up in any case.
M12 4-banger blocks were used in Formula 1 back in the 80's when BMW was doing turbo Big-Time. HUGE intercylinder wall thicknesses. The factory decreased bore diameters, partially in the interests of reliability. Word to the wise . . . These blocks were M10 production castings with remarkably few changes, BTW.
If someone insists on going to 95 mm, Paul thinks the answer is to hog everything out and fit steel bore sleeves machined accordingly--wetsleeving the block. If you do this, there is some strength and reliability gained, but the cost would be horrendous. In any case, there is nothing you can do about the cylinder's center-to-center distance, which is IIRC,100 mm in both the M30 and the S38 blocks. About all you can say about a 95B x 94S is it puts you at 3998 cc's. So I guess you're closer to 4 liters for bragging rights.
RPM
Based on stroke length, my max rpm is based on the formula 600,000/stroke length in mm. So using the stroke length EB told me would be used, 600,000/94 = 6383 rpm. Using what's there, 600,000/95 = 6315 rpm as a redline. For comparison, a stock S38B35 has an 84 mm stroke. 600,000/84 = 7143 rpm. BMW put a 6900 rpm, redline on the motor, so gave themselves a bit of a cushion.
I know, I know. I have a 95 mm stroke and 6500 rpm programmed into the TEC-3r as a fuel cutoff limit. Paul says the 200 rpm difference isn't a problem, given the internal components used and the balancing that's been done, but I do need to make the adjustments next time I get on the dyno. Total combined rotating mass is balanced within .2 gram. This includes crank, flywheel, clutch, harmonic balancer, rods, pistons, etc. The whole enchilada. There is probably more imbalance due to oil films on the moving pieces. The flow rates we're getting out of the head will support upwards of 7000 rpm, but the amount of time I'll be spending at 6K-plus is going to be pretty limited in any case. Endurance motor, remember?
Volumetric efficiency.
On high-performance 2-valve motors, this number can reach into the mid-90% range with mucho port work, etc. Paul sez NA small-block Chevys with 14-degree valve angles can go upwards of 110%, but this is on billet aluminum drag heads with trick I&E tracts. He's done it, I might add. Theoretical NA Ve limits go into the low 120% range; FI can move than number significantly, from what I understand. The HUGE Ve numbers we have seen come out of the not2fast turbo model were derived by taking outputs from the dyno pulls--fuel consumption, air mass flows, pressure drops across the IC, RPM, measured horsepower and torque--then plugging these measured values into the model. Ve was left as the independent variable. Kowa Bunga. 139% Ve. Holy Bat, Shitman.
I've written earlier about some of the port work that was done on the head. Paul and I think that the efficient flow in and out of the combustion chamber contributes to the 139% Ve. Unfortunately we don't have any flow bench numbers from an untouched S38B35 head; I would dearly love to obtain them to establish a baseline flow value for comparison purposes.
Pressure ratio
Pr = (boost delivered to the combustion chamber + atmospheric pressure)/atmospheric pressure. "Atmo" is the ambient air pressure where the reading is taken. Sea level = 14.7 psi. Denver, actually my house, is 6200 ft ASL. Atmospheric pressure drop is a loss of 3.3% per 1000 ft above sea level. 3.3% x 6.2 = 20.46% less pressure than sea level. 14.7 x (100-.2046) = 14.7 x .795 = 11.68 psi. Use 11.7 psi as we proceed. The effects of air temperature and relative humidity can be measured, but are generally very minor for our purposes. Used to set up the compensating factors on a dyno, but for the moment we'll ignore them.
Another item, often overlooked, is the "depression" or pressure loss going into the compressor due to the resistance from the air filter. This has to come off the "atmo" in the denominator. Using a large K&N cone filter, I probably have a depression of about .5 psi. So Pr = (15.0 +11.7)/(11.7-.5) = 2.38 Pr. We stongly suspect EB used values which he would see at sea level, i.e., the textbook numbers. (14.7 +14.7)/14.7 = 2.00 Pr. FWIW, the original plan called for 1 bar of boost, 14.7 psi.
We think EB, not knowing what kind of CFM numbers the head would flow, probably used very conservative Ve's. Probably on the order of 90%. Plugging in these kind of values makes the not2fast model generate plausible outputs pointing towards using a GT-35R.
Run all these through the CFM equation.
(3.914 x 6383 x 90 x 2.00)/5660
(4,496,951)/5660 =794.51 CFM. x .07 air density conversion to pounds per cubic foot = 55.62 pounds air mass/minute delivery.
This works out fine on paper for a GT-35R, but on our dyno runs,when the boost got over 8 psi, the air mass value is way into choke, given the way my motor flows air. This is demonstrated by the dropoff in Manifold Air Pressure (MAP) as the RPM went towards redline.
For comparison purposes, the values on my motor as built:
(3.986 x 6315 x 139 x 2.383)/5660
(8,327,265)/5660 = 1471.2 CFM x .07 air density = 102.99 pounds air mass/minute.This puts us right in the middle of GT-45R territory. go look athttp://www.turbobygarrett.comthen look at the map for the GT-45.
Stupefying. For a reality check, I ran these factors into the not2fast model. This yielded a more plausible 86.50 # flow requirement. This works out OK for a GT-42 as far as how far into "choke" is concerned, but gets very close to "surge" at low boost/loading/airflow levels. This implies poor response at best when coming off idle or at low loading levels.
So what to do about getting some good numbers for Ve? According to Paul, normal practice is as follows:
1. Obtain exact cc volumes for each cylinder--combustion chamber, cylinder proper, piston dome or dish, head gasket thickness, intake and exhaust tracts.
2. Perform a flowbench test on each cylinder. Steps 1 and 2 are done on the head as it comes off the motor--no modifications whatsoever, including cleaning. This provides real-world baseline values.
3. Do the work on the head--port and tract reworking, removal of irregularities, surface finishing, multi-angle grinding, boring pot throttle bodies, ceramic coating, etc.
4. Repeat the flow bench measurements along the way after each modification, being sure the work is staying uniform hole-to-hole, and to see what changes in flow are coming from each change. Keep track of the values in writing.
5. Repeat the flowbench test. Compare to the original values. NOW you have something to go on to evaluate the head's efficiency.
With these new flow and volume numbers, there is a basis for estimating Ve and thus air mass flow capabilities. This, in turn, will give you the elements to do the math for the turbo sizing.
If you have done the homework, it will become readily apparent that a GT-35R begins to run out of breath on a 4-liter S38. Keep in mind my motor is 3986 cc. The stock M30 has 3430 cc, or about 14% less displacement for openers. Additionally, the S38's 4 valves make a significant difference in the ability to move air. Some of this is due to port design--see Rich Euro M5's fantastic pictures looking through the ports across the combustion chambers--but a great deal is due to the head's huge valve curtain area.
Valve curtain area: Valve diameter x pi x valve lift x number of valves.
In my motor, with 38.5 mm intakes, 38.5 x 3.1416 x 11 mm lift x 2 valves = 26.61 square mm.
For comparison, use Duke's 47 mm intakes on the Hartge H6 head: 47 mm x 3.1416 x 11 mm lift x 1 intake valve = 16.24 square cm. This translates into a 63.85 % curtain area advantage in the 4-valve head. This isn't dissing anyone; it's just a fact, girls.
Think of curtain area as the size of the doors at the end of a hallway going into an auditorium. If you are emptying or filling the auditorium, having more doors or wider ones makes it much easier to move people in or out. The same is true with the flow of air. Reread arao Engineerring's web page http://www.araoengineerring.com.
Intake valve area is a critical element in how much power the motor will produce, as it directly affects airflow volumes moving through the cylinders. That airflow, in turn, affects the volume of fuel burned and thus the power generated. Keeping in mind that tract design with consistent shaping and few sharp bends as possible is also a huge factor. The idea here is to maintain airflow velocity.
So using the not2fast model and a GT-35R compressor map, I find I can move around 55 # of air at my desired 15 psi of boost. Each pound of air will yield around 10 to 10.5 flywheel horsepower, depending on the amount of fuel available. An efficient turboed motor will require somewhere between .5 and .55 pounds of fuel per hour per horsepower. This is known as Brake Specific Fuel Consumption, or BSFC. NA motors tend to run somewhat lower BSFC values; supercharged somewhat higher. Plugging in those kind of values, we can see that a GT-35R can deliver somewhere around 577 flywheel horsepower at 15 psi boost. Push the boost number upwards and one drops off the compressor's "efficiency" islands and the compressor begins to go into a stall or "choke" condition where the air simply heats up and doesn't get to the intakes as efficiently. So just cramming more boost pressure isn't necessarily the way to more power. The answer is to move more air. This is why a larger compressor at lower pressure can yield more power than cranking the daylights out of a small unit. Yes, there are issues regarding turbo lag with larger compressors, but I think the concept should be apparent.
Once again, the solution is to move air to and through the combustion chambers, so anything we can do on the intake side has to be complemented on the exhausts. The more of the spent charge that can be removed, the more (relatively) cool intake mass can be drawn or forced in. The degree of charge interchange, called volumetric effiiency or Ve, measures this relationship. a good NA motor can get close to 95%+; a strong FI motor, with optimized tract configurations and judicious cam selection can get well above 100%, IF the exhaust system can flow the spent charge. This is the reasoning behind a tuned header exhaust layout. Ve is difficult to measure in a model such as not2fast, but it can be inferred from the engine's operating parameters. Plug in values obtained during dyno pulls and from a datalogging wideband sensor, and Ve can be estimated. We did this with Lucifer's Hammer and came up with some stupefying Ve values. But then, Paul did everything he could to make this sucker flow . . . .
Where this left us was looking for a turbo which could flow on the order of around 80 to 85 # at 15 psi. The models pointed us toward a GT-40 or-42. This past year, Garrett made the compressor maps for the GT-45 available. In theory, a GT-45 would be optimal, but two factors come into play. First, the physical size of these units meant we ran into space constraints. Second, these large compressors have real lag considerations. The latter isn't a problem if the engine is built strictly for the drags and you're well up on RPM when the lights go green, but not so wonderful if one is pulling from off-idle and running in real-world stop-and-go traffic.
To return to constraint #1 for a minute. There is very little space on the exhaust side of the S38 motor between the block and the engine bay sheet metal and engine mounting rails. What space there is got taken up when we used the equal-length header arrangement. It was this problem that led to the "remote mount" turbo location we have used. Recently 666 Fabrication has shown a top mount GT-40 put on an M30 motor; this was the result of some very nice tubular header fabrication. A top mount configuration was discussed with EB back in late 2003, but for whatever reason was rejected out of hand on his end. If one was willing to do a fair amount of sheet metal reshaping on the hood, I think a top mounted GT-40/-42/-45 might well be a viable possibility.
Back to turbo sizing. While the models point towards a given turbo comprsssor size, the models also seem to assume a more or less constant loading level. Think of a static emergency power diesel generator, or an offshore racing inboard. Hugh MacInnes, in his book, "Turbocharging," addresses this set of circumstances. MacInnes is of the opinion that the "theoretical" optimal turbo sizing is around 10% too large for real world road-going applications. He contends that the amount of time spent at full boost is so small that trying for the huge turbo gives away far too much in terms of driveability. If one accepts that premise, sizing down from a 84 or 86 # requirement to say a 72# delivery, very little is lost. The turbo will still push the larger number up to a point, but at a much lower compressor efficiency. Not a problem as long as the time spent at the higher flow number is kept brief to minimize the heat buildup in the intake charge.
If you plot a 72# airflow against a 2.38 Pr on a GT-35R map, it's apparent that this operating point is well into choke. Additionally, we plotted different boost levels (Pr) against airmass delivery over a range of rpm levels. What came out of this is that for the S38 4-liter, if the turbo used is a GT-35R, you're limited to about 10 psi of boost at my altitude. Pr = 1.94 (10.0 + 11.7)/(11.7-.5). Nothing necessarily wrong with that, but it doesn't use a great deal of the motor's capabilities.
So why was a GT-35R originally recommended? We suspect several things.
One, by staying on the small side, the airflow capacity would limit potential power output and reduce stresses on the motor, and with that, the possibility of the thing breaking. Smaller, in this case, would also mean more rapid spooling and quicker response from a lower rpm level. In other words, less lag.
Two, Uncertainties related to the Ve. If the head wasn't flow-benched, the builder would have only guesswork to estimate gas flows. Knowing that a good NA motor will get up into the 95% range, using a similar value would reduce the calculated airflow volume. Pressure ratios probably assumed a sea level environment, thus yielding a lower Pr. Max rpm assumptions are unknown, but a longer stroke points toward a lower value. Again, reducing the redline means less stress on the motor, and less exposure to having to replace something under warranty.
Three, EB may have chosen to use what was readily available, and what would physically fit in the space in the engine bay.
In short, we simply don't know.
Having experienced first-hand the limits on the GT-35 based on the dyno pulls in late 2005, the physical sizing and practical flow levels pointed us towards using a T66 with a ball bearing rotating assembly to minimize spool-up time. Go back to LH Part 1 for a summary of the outputs. Once we determined whut this bad boy could do, common sense cut the boost levels down. Using the 2-channel boost controller, "normal" boost is capped at 7.5 psi. The turbo and the motor love it. The "ya dumb sh*t" level is 15 psi. Plans call for reduciung that down to about 12 psi this coming spring. 15 psi works beautifully, but the compressor moves quickly into choke, way off the island to the right, and delivered boost drops off due to charge heating. Besides at ~12 psi I can break the tires loose all the way up the ladder; 15 psi scares the living snot out of me.
Having ranted for a while about the turbo, I want to shift over to the fuel issue.
The stock S38 injectors flow, I believe, 24# per hour. I can't swear to this, but I think Bosch and BMW cap the duty cycle at 80%. So .80 x 24 = 19.2 #/ hour flow. Assuming a BSFC of around .4 to .45 on the efficient but thirsty S38, we get 19.2/.45 = 42.67 horsepower per cylinder. Multiply by 6 and we get 256 flywheel hp. Which is where BMW rated the US-spec S38B35.
OK for the moment on the injectors. If we take 19.2#/hour on the injectors, x 6 injectors we get 115.2#/hour combined flow requirement. Assume that fuel weighs 7.4#/gallon, by dividing we get a max flow rate of 15.57 gallons per hour. 1 gallon = 3.785 liters. So 15.57 gallons = 58.9 liters/hour delivery. I think the stock BMW main pump will push around 140 liters per hour at 3 bar (43.5 psi) in the lines.
Have patience, you'll see where this is going.
Now let's say the turbo is calling for 84# of air, fully cranked up. It implies something around 840 flywheel horesepower. We know that my motor will show a BSFC of between .49 and .52, running 11.5 A/F ratios when at full boost. 840 flywheel x .52 = 436.8#/hour fuel demand. Oof-da! 438.6/6cylinders = 72.8#/.hour fuel flow requirement per injector. Granted this is only for the short periods when at full (15 psi) boost, but under those circumstances, the last thing I want is going 'way lean.
So it is pretty apparent that a stock BMW fuel pump ain't-a-gonna-get 'er done.
This leads to my conversations with Aeromotive. They build pumps that will keep a AA Fueler happy. Website: http://www.aeromotiveinc.com. I call these guys and outline my needs. Their answer: use a pair of their A-1000 pumps. The A-1000 will deliver 600#/hour at 43.5 psi and 13.5 volts into the pump. Pump delivery will decrease proportionately as line pressure goes up. One can assume line pressure will increase at least 1-for-1 as boost pressure rises. So if I am runnning 15psi, I can expect my fuel line pressure at the injectors to be 43.5 + 15, or 58.5 psi. Call it 60 psi for good measure. at 60 psi, Aeromotive tells me the pump will flow approximately 435#/hour.
Now here's where it begins to get interesting. Aeromotive strongly suggests the pumps not see more than 67% of their rated capacity if the demand is continuous. 100%, fine for short bursts, e.g., on the dragstrip, but the strain is far more severe in a continuous situation, say driving halfway across the US to 5erFest, even though the feed to the fuel rail is lower. So assuming a worst-case situation, my 436 pound demand is at the deliverable capacity on a single pump. Not Good.
Back to injector sizing. Both Paul and Justin Pierce, the Electromotive expert at Mile High Performance, insist that you NEVER want to go over 75% of duty cycle. You may get away with 85% for short bursts, but you run the risk of the injector "locking" and fuel delivery goes in the toilet forthwith.
Not A Good Thing for the injector either. On the other hand, putting in the Injector From Hell, say a Ford Motorcraft 160# item has its problems. At low (read normal street use) demands the duty cycle may be down at a couple of percent or so and running very narrow pulse widths. The injector won't properly cycle and the fuel dribbles out of the orofice rather than atomizing. The result is a very balky motor. So we want to keep the injectors running somewhere between say 10% and 75% duty cycle. So in my case, we are looking at something close to 73 or 74 pounds delivery at a 75% duty cycle. 74/.75 = 98.67 pounds at a nominal rated 100% capacity. But we have the Injector From Hell problem showing up nearly all the time. Face it, we damn near never use full boost. Think about the way you drive day-to-day.
So to keep the injectors in their sweet spot, Justin comes up with a staged dual injector soulution. Injector set 1 lives in the stock location at the top of the individual throttle bodies. These are upsized to 45#. 45 x 75% duty cycle = 33.75#/hour delivery. 33.75/.52 BSFC = 64.9 flywheel hp per cylinder, or about 389 hp. Entirely feasible on 7.5 psi of boost. But what happens when I go to 15 psi?
Not2fast's model tells me I need to be running 985 cc/minute (93.8#/hour) injectors at 75% to meet the demand. Marren Fuel Injection's model tells me I need 97.76#/ hour. As we saw above, the max expected draw is going to be a bit under 73#. So 72.8/97.76 = 74.5% duty cycle. So a second set of injectors appears to be the solution.
If the primary injectors, rated at 45# are giving me 33.75# at 75%, the second set needs to be capable of delivering 39.05#/hour at 75%. (72.8 -33.75 = 39.05 ) 39.05/.75 = 52.1#/ hour. If we run 74# injectors as the second set, we're staying around 52.7% of duty cycle. Should be fine, especially if we lower the "crossover" point on the primary set to around 67-70%. The larger capacity secondary injectors also give us a cushion on fuel flow if, for whatever reason demand has been underestimated.
FWIW, when we did the dyno pulls with the GT-35R, we were running a single set of 74# injectors. Balky at low loads, these went to around 90% duty cycle when the GT-35R was cranked up. Dyno runs ceased forthwith.
But how to avoid stressing the pump? The answer is to run two pumps in parallel, one feeding each fuel rail.
Pump 1 sees 45# x .75 x 6 injectors = 202.53/hour flow. This assumes we'ld be running 7.5 psi boost, and not calling on the secondary injectors. 43.5 psi baseline fuel pressure + 7.5 from boost = 51 psi in the lines. 43.5/51 = 85.3% of baseline pump capacity of 600#/hour = 511.8# adjusted capacity. Multiply 511.8 by the 67% continuous use factor and we get 342.9. 202.5/342.9 = ~59.0% continuous load factor. No sweat for the pump.
Secondaries. 74# x 6 injectors = 333#/hour theoretical max delivery requirement. At full boost (15 psi), fuel line pressure = 43.5 +15 = 58.5 psi. 43.5/58.5 = 74.4% x rated nominal pump capacity of 600#/ hour, or 446#. Multiply 446 by 67% gives 289.9. Thus the theoretical delivery requirement of 333# exceeds the sustainable capacity by about 11%, but well within the "burst" capacity of 446#. Going back to the calculated secondary feed requirement of 39.05#, we have 39.05 x 6 injectors = 195.3# flow requirement. 195.3/298.9 = 65.3% We should be just fine.
This setup means several things for the fuel delivery system.
1. Each pump is run through a separate relay-protected controller which manages pump speed based on fuel demand--RPM and load; fuel pressure is monitored via the fuel pressure regulator.
2. Each fuel rail is fed separately, so the system is redundant. If one pump were to fail, the other can supply the motor.
3. The second set of injectors is run by a separate set of drivers in the TEC-3r. Since the system is set up as full sequential, this required building a second driver circuit, and programming the secondary injector bank to open up only under defined loads and RPM. This avoids overfueling at low MAP and when coming off low throttle settings. This was a serious effort on the part of Electromotive, and took some rather original work by Justin Pierce. My hat's off to these guys.
One minor quirk: when the second injector bank opens up, there is a momentary "overenrichment" until the load increases. Not noticeable when driving, but it shows up on the dyno printouts as a hiccup in the A/F ratio.
4. Getting the second injector bank into place was not easy. Justin ended up drilling the bottom sides of the individual throttle bodies and welding in a set of injector bungs, angled slightly into the TB throats, and 180 degrees opposite the primaries. Accessing the injectors, if needed, means pulling the plenum. Wiring was a whole buncha fun.
5. The two fuel pumps live above the rear subframe and feed the two 5/8" ID copper fuel lines going into the two fuel rails. The lines each ahave a 100-micron inline filter before the pump, and a 10-micron filter after the pump. Fuel pickup is from a pair of oversize "duckfeet" in the bottom of the former in-tank pump carrier, then to the pump inlet lines. The in-tank transfer pump was eliminated, and fuel backflow is controlled by a backcheck valve in the pickup lines. The amount of drainage back through the lines is minimal; when the ignition is turned on, there is a very noticeable whine as the pumps are charged and the fuel system comes up to operating pressure.
The fuel regulator is externally adjustable, but not from the driver's seat. It is remote-mounted in the right front of the engine bay, above the ABS controller. The return fuel line back to the tank is a 3/4" ID copper tube. The capacity is dictated by the sizing of the two 5/8" delivery lines, and assures there is no "pooling" of fuel in the injector rails when fuel demand is low. This arrangement helps to dissipate any acquired heat in the fuel picked up via the pumps or being in close proximity to the heat-shielded turbine which lives below the fuel rails. The use of copper in the lines also serves to act as a radiator. Delivery back to the tank via the 3/4" line enters the front of the tank and is terminated away from the fuel pickup point, thus allowing the fuel in the tank to function as a heat sink before it is sent to the pumps.
While this sounds like overkill, it means that the pumps are not being stressed which keeps the heat they add to the fuel stream to a minimum and extends their life expectancy.
For comparison purposes on the fuel lines, the stock lines are 8 mm ID. These have a cross-sectional area of 50.26 square mm.The 5/8" (15.875 mm) lines have an area of 197.93 square mm each. The total cross-sectional area of 395.87 square mm is 7.87 times the stock figure. That ought to deal with flow requirements . . .
For you guys with the M30s who are thinking FI:
IMHO the stock E28 fuel pumps aren't up to the task. Getting to where I had confidence in the adequacy of the fuel delivery has been a real PITA. I think you are going to have to get some pretty exact measurements on the fuel pump delivery rates, injector sizing and duty cycles before a turboed M30 sees much serious road use. IMHO simply assuming the stock pump and lines are up to the task is pretty risky.
Stock pump in the M5 is, I think, 140 liters/hour. The ETK sez it is the same p/n in a 535i, but I'm not 100% certain on this.
Fuel weight = 7.4#/gallon, or about 1.96#/liter. Assume the pump's sustainable max loading is 67% of 140 (see Aeromotive's advice above). 140 x 67% = ~94 liters/ hour. 94 x 1.96 # = 184#/hour. Assume a set of six 71# injectors at 75% duty cycle = 320#/hour, considerably above the stock pumps capabilities.
A Walbro 255liter/hour unit can probably do it (255lph x 67% = 170.9 liters/hour. 170.9 x 1.96 = 335#/hour sustainable delivery), but I have real reservations about the stock 8mm fuel lines sizing. I am very skeptical that the stock lines can meet the injector draw at full demand. This translates into fuel starvation and a lean condition under full boost. Not Pretty.
I am here to tell you that getting all this figured out and installed was a cast-iron bitch.
Go to graphite's website, http://www.doggunracing.com/mye28/LucifersHammer as there are a buncha pictures on this.
Chapter 8 will pick up with my visit to Paul's shop and the next episodes of parts chasing and the assembly adventures.
There. Everybody got that?
"Yo, Chicken Lips---gets me a beer. The Wordy One has written the sequel to War and Peace . . ."
This next lecture 8) is going to get into some math regarding engine capabilities and some constraints on what's feasible on the S38. I'm including all this detail because (a) there are a lot of misconceptions out there about what kind of power our motors can generate and (b) early on a lot of questions got asked about component selection.
For openers read the following paragraph.
OK A lot of youse gize have, or have read Corky Bell's "Maximum Boost." This is a fine introduction to turbocharging, though it has become a bit dated. There is another book which, IMHO, anyone thinking about forced induction absolutely must read. This is A. Graham Bell's "Forced Induction Performance Tuning." The two Bells are not related, BTW. "FIPT" is available thru Amazon, costs about $35. Gentlemen, this is the best $35 you can possibly spend.
It is not a cookbook how-to manual, but it isn't a graduate engineering text either. It will save you any number of blind alleys and wasted effort. SPEND THE FRICKING MONEY THEN READ THE SUMBITCH. It has a very good section regarding engine building in general, applicable to NA motors as well as FI. As such, it provides some excellent guidance. SPEND THE FRICKING MONEY THEN READ THE SUMBITCH.
Next step. Now for some websites
1. http://www.not2fast.com/wryday This is a very, very good turbo calculation model. It will allow you to do a lot of what-ifing. It is linked to a number of compressor maps for applicable Garrett turbos, and as such, answers a lot of questions. The model doesn't take into account variations in exhaust system design, or head flow capabilities, or what happens with using standalone ECUs or changing octane in fuel, but it is a most useful tool.
2. http://www.injector.com This is Marren Fuel Injection's website and has a number of useful formulas regarding injector sizing. They also sell a very nice injector rail setup for the M30.
3. http://www.RCEng.com This is RC Engineering's website. This outfit is another source for injectors, and also has a very good technical page regarding injector sizing.
4. http://www.turbobygarrett.com Garrett's site. Has an extensive list of the GT-series turbos and their Garrett p/n's. Also has compressor maps.
5. http://www.araoengineering.com For anyone who wonders why a 4-valve head will beat a 2-valve, this is required reading. Arao builds 4-valve heads for US V8 drag motors. Definitely on the Required Reading list.
A recap:
I originally wanted to refresh my S38B35 to restore it's original power.
Because I live at 6200 feet altitude, I am automatically down 20% on power. I wanted to get sea level performance. At altitude.
I wanted a road burner/endurance motor, not a drag strip lord.
"There is no replacement for displacement." Parnelli Jones. Can I stroke this thing?
Go big cams and high compression = having to deal with the environazis.
Don't get greedy and FI will solve your problems. Yaaah. Riiiiight.
Displacement
I went back to my original notes in my "Build Journal" where I've kept track of all the activity related to this project. This had summaries from my conversations with Rowe, Fahey, Korman, EB, etc. EB had said the crank was going to be 94 mm; in fact it is 95. Bore was going to be 94 mm. In fact it is 94.36. Working from EB's numbers, the original displacement was going to be 3914 cc. Actual is 3986. The differences are small, but they will add up as you will see.
Sidebar here.
Early on, there had been some discussion about going out to a 95 (94.95) mm bore. FWIW, the maximum oversize piston on the S38B38 is 94.955 mm; this leaves zero room for any future rebore or refresh. Going out to 95 mm was rejected because if one does that, the amount of material between the siamesed bores becomes extremely thin--around 5 mm. This has the very real potential to allow the castings to crack. Not Real Smart on an FI motor where cylinder pressures are going to be bumped up in any case.
M12 4-banger blocks were used in Formula 1 back in the 80's when BMW was doing turbo Big-Time. HUGE intercylinder wall thicknesses. The factory decreased bore diameters, partially in the interests of reliability. Word to the wise . . . These blocks were M10 production castings with remarkably few changes, BTW.
If someone insists on going to 95 mm, Paul thinks the answer is to hog everything out and fit steel bore sleeves machined accordingly--wetsleeving the block. If you do this, there is some strength and reliability gained, but the cost would be horrendous. In any case, there is nothing you can do about the cylinder's center-to-center distance, which is IIRC,100 mm in both the M30 and the S38 blocks. About all you can say about a 95B x 94S is it puts you at 3998 cc's. So I guess you're closer to 4 liters for bragging rights.
RPM
Based on stroke length, my max rpm is based on the formula 600,000/stroke length in mm. So using the stroke length EB told me would be used, 600,000/94 = 6383 rpm. Using what's there, 600,000/95 = 6315 rpm as a redline. For comparison, a stock S38B35 has an 84 mm stroke. 600,000/84 = 7143 rpm. BMW put a 6900 rpm, redline on the motor, so gave themselves a bit of a cushion.
I know, I know. I have a 95 mm stroke and 6500 rpm programmed into the TEC-3r as a fuel cutoff limit. Paul says the 200 rpm difference isn't a problem, given the internal components used and the balancing that's been done, but I do need to make the adjustments next time I get on the dyno. Total combined rotating mass is balanced within .2 gram. This includes crank, flywheel, clutch, harmonic balancer, rods, pistons, etc. The whole enchilada. There is probably more imbalance due to oil films on the moving pieces. The flow rates we're getting out of the head will support upwards of 7000 rpm, but the amount of time I'll be spending at 6K-plus is going to be pretty limited in any case. Endurance motor, remember?
Volumetric efficiency.
On high-performance 2-valve motors, this number can reach into the mid-90% range with mucho port work, etc. Paul sez NA small-block Chevys with 14-degree valve angles can go upwards of 110%, but this is on billet aluminum drag heads with trick I&E tracts. He's done it, I might add. Theoretical NA Ve limits go into the low 120% range; FI can move than number significantly, from what I understand. The HUGE Ve numbers we have seen come out of the not2fast turbo model were derived by taking outputs from the dyno pulls--fuel consumption, air mass flows, pressure drops across the IC, RPM, measured horsepower and torque--then plugging these measured values into the model. Ve was left as the independent variable. Kowa Bunga. 139% Ve. Holy Bat, Shitman.
I've written earlier about some of the port work that was done on the head. Paul and I think that the efficient flow in and out of the combustion chamber contributes to the 139% Ve. Unfortunately we don't have any flow bench numbers from an untouched S38B35 head; I would dearly love to obtain them to establish a baseline flow value for comparison purposes.
Pressure ratio
Pr = (boost delivered to the combustion chamber + atmospheric pressure)/atmospheric pressure. "Atmo" is the ambient air pressure where the reading is taken. Sea level = 14.7 psi. Denver, actually my house, is 6200 ft ASL. Atmospheric pressure drop is a loss of 3.3% per 1000 ft above sea level. 3.3% x 6.2 = 20.46% less pressure than sea level. 14.7 x (100-.2046) = 14.7 x .795 = 11.68 psi. Use 11.7 psi as we proceed. The effects of air temperature and relative humidity can be measured, but are generally very minor for our purposes. Used to set up the compensating factors on a dyno, but for the moment we'll ignore them.
Another item, often overlooked, is the "depression" or pressure loss going into the compressor due to the resistance from the air filter. This has to come off the "atmo" in the denominator. Using a large K&N cone filter, I probably have a depression of about .5 psi. So Pr = (15.0 +11.7)/(11.7-.5) = 2.38 Pr. We stongly suspect EB used values which he would see at sea level, i.e., the textbook numbers. (14.7 +14.7)/14.7 = 2.00 Pr. FWIW, the original plan called for 1 bar of boost, 14.7 psi.
We think EB, not knowing what kind of CFM numbers the head would flow, probably used very conservative Ve's. Probably on the order of 90%. Plugging in these kind of values makes the not2fast model generate plausible outputs pointing towards using a GT-35R.
Run all these through the CFM equation.
(3.914 x 6383 x 90 x 2.00)/5660
(4,496,951)/5660 =794.51 CFM. x .07 air density conversion to pounds per cubic foot = 55.62 pounds air mass/minute delivery.
This works out fine on paper for a GT-35R, but on our dyno runs,when the boost got over 8 psi, the air mass value is way into choke, given the way my motor flows air. This is demonstrated by the dropoff in Manifold Air Pressure (MAP) as the RPM went towards redline.
For comparison purposes, the values on my motor as built:
(3.986 x 6315 x 139 x 2.383)/5660
(8,327,265)/5660 = 1471.2 CFM x .07 air density = 102.99 pounds air mass/minute.This puts us right in the middle of GT-45R territory. go look athttp://www.turbobygarrett.comthen look at the map for the GT-45.
Stupefying. For a reality check, I ran these factors into the not2fast model. This yielded a more plausible 86.50 # flow requirement. This works out OK for a GT-42 as far as how far into "choke" is concerned, but gets very close to "surge" at low boost/loading/airflow levels. This implies poor response at best when coming off idle or at low loading levels.
So what to do about getting some good numbers for Ve? According to Paul, normal practice is as follows:
1. Obtain exact cc volumes for each cylinder--combustion chamber, cylinder proper, piston dome or dish, head gasket thickness, intake and exhaust tracts.
2. Perform a flowbench test on each cylinder. Steps 1 and 2 are done on the head as it comes off the motor--no modifications whatsoever, including cleaning. This provides real-world baseline values.
3. Do the work on the head--port and tract reworking, removal of irregularities, surface finishing, multi-angle grinding, boring pot throttle bodies, ceramic coating, etc.
4. Repeat the flow bench measurements along the way after each modification, being sure the work is staying uniform hole-to-hole, and to see what changes in flow are coming from each change. Keep track of the values in writing.
5. Repeat the flowbench test. Compare to the original values. NOW you have something to go on to evaluate the head's efficiency.
With these new flow and volume numbers, there is a basis for estimating Ve and thus air mass flow capabilities. This, in turn, will give you the elements to do the math for the turbo sizing.
If you have done the homework, it will become readily apparent that a GT-35R begins to run out of breath on a 4-liter S38. Keep in mind my motor is 3986 cc. The stock M30 has 3430 cc, or about 14% less displacement for openers. Additionally, the S38's 4 valves make a significant difference in the ability to move air. Some of this is due to port design--see Rich Euro M5's fantastic pictures looking through the ports across the combustion chambers--but a great deal is due to the head's huge valve curtain area.
Valve curtain area: Valve diameter x pi x valve lift x number of valves.
In my motor, with 38.5 mm intakes, 38.5 x 3.1416 x 11 mm lift x 2 valves = 26.61 square mm.
For comparison, use Duke's 47 mm intakes on the Hartge H6 head: 47 mm x 3.1416 x 11 mm lift x 1 intake valve = 16.24 square cm. This translates into a 63.85 % curtain area advantage in the 4-valve head. This isn't dissing anyone; it's just a fact, girls.
Think of curtain area as the size of the doors at the end of a hallway going into an auditorium. If you are emptying or filling the auditorium, having more doors or wider ones makes it much easier to move people in or out. The same is true with the flow of air. Reread arao Engineerring's web page http://www.araoengineerring.com.
Intake valve area is a critical element in how much power the motor will produce, as it directly affects airflow volumes moving through the cylinders. That airflow, in turn, affects the volume of fuel burned and thus the power generated. Keeping in mind that tract design with consistent shaping and few sharp bends as possible is also a huge factor. The idea here is to maintain airflow velocity.
So using the not2fast model and a GT-35R compressor map, I find I can move around 55 # of air at my desired 15 psi of boost. Each pound of air will yield around 10 to 10.5 flywheel horsepower, depending on the amount of fuel available. An efficient turboed motor will require somewhere between .5 and .55 pounds of fuel per hour per horsepower. This is known as Brake Specific Fuel Consumption, or BSFC. NA motors tend to run somewhat lower BSFC values; supercharged somewhat higher. Plugging in those kind of values, we can see that a GT-35R can deliver somewhere around 577 flywheel horsepower at 15 psi boost. Push the boost number upwards and one drops off the compressor's "efficiency" islands and the compressor begins to go into a stall or "choke" condition where the air simply heats up and doesn't get to the intakes as efficiently. So just cramming more boost pressure isn't necessarily the way to more power. The answer is to move more air. This is why a larger compressor at lower pressure can yield more power than cranking the daylights out of a small unit. Yes, there are issues regarding turbo lag with larger compressors, but I think the concept should be apparent.
Once again, the solution is to move air to and through the combustion chambers, so anything we can do on the intake side has to be complemented on the exhausts. The more of the spent charge that can be removed, the more (relatively) cool intake mass can be drawn or forced in. The degree of charge interchange, called volumetric effiiency or Ve, measures this relationship. a good NA motor can get close to 95%+; a strong FI motor, with optimized tract configurations and judicious cam selection can get well above 100%, IF the exhaust system can flow the spent charge. This is the reasoning behind a tuned header exhaust layout. Ve is difficult to measure in a model such as not2fast, but it can be inferred from the engine's operating parameters. Plug in values obtained during dyno pulls and from a datalogging wideband sensor, and Ve can be estimated. We did this with Lucifer's Hammer and came up with some stupefying Ve values. But then, Paul did everything he could to make this sucker flow . . . .
Where this left us was looking for a turbo which could flow on the order of around 80 to 85 # at 15 psi. The models pointed us toward a GT-40 or-42. This past year, Garrett made the compressor maps for the GT-45 available. In theory, a GT-45 would be optimal, but two factors come into play. First, the physical size of these units meant we ran into space constraints. Second, these large compressors have real lag considerations. The latter isn't a problem if the engine is built strictly for the drags and you're well up on RPM when the lights go green, but not so wonderful if one is pulling from off-idle and running in real-world stop-and-go traffic.
To return to constraint #1 for a minute. There is very little space on the exhaust side of the S38 motor between the block and the engine bay sheet metal and engine mounting rails. What space there is got taken up when we used the equal-length header arrangement. It was this problem that led to the "remote mount" turbo location we have used. Recently 666 Fabrication has shown a top mount GT-40 put on an M30 motor; this was the result of some very nice tubular header fabrication. A top mount configuration was discussed with EB back in late 2003, but for whatever reason was rejected out of hand on his end. If one was willing to do a fair amount of sheet metal reshaping on the hood, I think a top mounted GT-40/-42/-45 might well be a viable possibility.
Back to turbo sizing. While the models point towards a given turbo comprsssor size, the models also seem to assume a more or less constant loading level. Think of a static emergency power diesel generator, or an offshore racing inboard. Hugh MacInnes, in his book, "Turbocharging," addresses this set of circumstances. MacInnes is of the opinion that the "theoretical" optimal turbo sizing is around 10% too large for real world road-going applications. He contends that the amount of time spent at full boost is so small that trying for the huge turbo gives away far too much in terms of driveability. If one accepts that premise, sizing down from a 84 or 86 # requirement to say a 72# delivery, very little is lost. The turbo will still push the larger number up to a point, but at a much lower compressor efficiency. Not a problem as long as the time spent at the higher flow number is kept brief to minimize the heat buildup in the intake charge.
If you plot a 72# airflow against a 2.38 Pr on a GT-35R map, it's apparent that this operating point is well into choke. Additionally, we plotted different boost levels (Pr) against airmass delivery over a range of rpm levels. What came out of this is that for the S38 4-liter, if the turbo used is a GT-35R, you're limited to about 10 psi of boost at my altitude. Pr = 1.94 (10.0 + 11.7)/(11.7-.5). Nothing necessarily wrong with that, but it doesn't use a great deal of the motor's capabilities.
So why was a GT-35R originally recommended? We suspect several things.
One, by staying on the small side, the airflow capacity would limit potential power output and reduce stresses on the motor, and with that, the possibility of the thing breaking. Smaller, in this case, would also mean more rapid spooling and quicker response from a lower rpm level. In other words, less lag.
Two, Uncertainties related to the Ve. If the head wasn't flow-benched, the builder would have only guesswork to estimate gas flows. Knowing that a good NA motor will get up into the 95% range, using a similar value would reduce the calculated airflow volume. Pressure ratios probably assumed a sea level environment, thus yielding a lower Pr. Max rpm assumptions are unknown, but a longer stroke points toward a lower value. Again, reducing the redline means less stress on the motor, and less exposure to having to replace something under warranty.
Three, EB may have chosen to use what was readily available, and what would physically fit in the space in the engine bay.
In short, we simply don't know.
Having experienced first-hand the limits on the GT-35 based on the dyno pulls in late 2005, the physical sizing and practical flow levels pointed us towards using a T66 with a ball bearing rotating assembly to minimize spool-up time. Go back to LH Part 1 for a summary of the outputs. Once we determined whut this bad boy could do, common sense cut the boost levels down. Using the 2-channel boost controller, "normal" boost is capped at 7.5 psi. The turbo and the motor love it. The "ya dumb sh*t" level is 15 psi. Plans call for reduciung that down to about 12 psi this coming spring. 15 psi works beautifully, but the compressor moves quickly into choke, way off the island to the right, and delivered boost drops off due to charge heating. Besides at ~12 psi I can break the tires loose all the way up the ladder; 15 psi scares the living snot out of me.
Having ranted for a while about the turbo, I want to shift over to the fuel issue.
The stock S38 injectors flow, I believe, 24# per hour. I can't swear to this, but I think Bosch and BMW cap the duty cycle at 80%. So .80 x 24 = 19.2 #/ hour flow. Assuming a BSFC of around .4 to .45 on the efficient but thirsty S38, we get 19.2/.45 = 42.67 horsepower per cylinder. Multiply by 6 and we get 256 flywheel hp. Which is where BMW rated the US-spec S38B35.
OK for the moment on the injectors. If we take 19.2#/hour on the injectors, x 6 injectors we get 115.2#/hour combined flow requirement. Assume that fuel weighs 7.4#/gallon, by dividing we get a max flow rate of 15.57 gallons per hour. 1 gallon = 3.785 liters. So 15.57 gallons = 58.9 liters/hour delivery. I think the stock BMW main pump will push around 140 liters per hour at 3 bar (43.5 psi) in the lines.
Have patience, you'll see where this is going.
Now let's say the turbo is calling for 84# of air, fully cranked up. It implies something around 840 flywheel horesepower. We know that my motor will show a BSFC of between .49 and .52, running 11.5 A/F ratios when at full boost. 840 flywheel x .52 = 436.8#/hour fuel demand. Oof-da!
438.6/6cylinders = 72.8#/.hour fuel flow requirement per injector. Granted this is only for the short periods when at full (15 psi) boost, but under those circumstances, the last thing I want is going 'way lean.So it is pretty apparent that a stock BMW fuel pump ain't-a-gonna-get 'er done.
This leads to my conversations with Aeromotive. They build pumps that will keep a AA Fueler happy. Website: http://www.aeromotiveinc.com. I call these guys and outline my needs. Their answer: use a pair of their A-1000 pumps. The A-1000 will deliver 600#/hour at 43.5 psi and 13.5 volts into the pump. Pump delivery will decrease proportionately as line pressure goes up. One can assume line pressure will increase at least 1-for-1 as boost pressure rises. So if I am runnning 15psi, I can expect my fuel line pressure at the injectors to be 43.5 + 15, or 58.5 psi. Call it 60 psi for good measure. at 60 psi, Aeromotive tells me the pump will flow approximately 435#/hour.
Now here's where it begins to get interesting. Aeromotive strongly suggests the pumps not see more than 67% of their rated capacity if the demand is continuous. 100%, fine for short bursts, e.g., on the dragstrip, but the strain is far more severe in a continuous situation, say driving halfway across the US to 5erFest, even though the feed to the fuel rail is lower. So assuming a worst-case situation, my 436 pound demand is at the deliverable capacity on a single pump. Not Good.
Back to injector sizing. Both Paul and Justin Pierce, the Electromotive expert at Mile High Performance, insist that you NEVER want to go over 75% of duty cycle. You may get away with 85% for short bursts, but you run the risk of the injector "locking" and fuel delivery goes in the toilet forthwith.
Not A Good Thing for the injector either. On the other hand, putting in the Injector From Hell, say a Ford Motorcraft 160# item has its problems. At low (read normal street use) demands the duty cycle may be down at a couple of percent or so and running very narrow pulse widths. The injector won't properly cycle and the fuel dribbles out of the orofice rather than atomizing. The result is a very balky motor. So we want to keep the injectors running somewhere between say 10% and 75% duty cycle. So in my case, we are looking at something close to 73 or 74 pounds delivery at a 75% duty cycle. 74/.75 = 98.67 pounds at a nominal rated 100% capacity. But we have the Injector From Hell problem showing up nearly all the time. Face it, we damn near never use full boost. Think about the way you drive day-to-day.So to keep the injectors in their sweet spot, Justin comes up with a staged dual injector soulution. Injector set 1 lives in the stock location at the top of the individual throttle bodies. These are upsized to 45#. 45 x 75% duty cycle = 33.75#/hour delivery. 33.75/.52 BSFC = 64.9 flywheel hp per cylinder, or about 389 hp. Entirely feasible on 7.5 psi of boost. But what happens when I go to 15 psi?
Not2fast's model tells me I need to be running 985 cc/minute (93.8#/hour) injectors at 75% to meet the demand. Marren Fuel Injection's model tells me I need 97.76#/ hour. As we saw above, the max expected draw is going to be a bit under 73#. So 72.8/97.76 = 74.5% duty cycle. So a second set of injectors appears to be the solution.
If the primary injectors, rated at 45# are giving me 33.75# at 75%, the second set needs to be capable of delivering 39.05#/hour at 75%. (72.8 -33.75 = 39.05 ) 39.05/.75 = 52.1#/ hour. If we run 74# injectors as the second set, we're staying around 52.7% of duty cycle. Should be fine, especially if we lower the "crossover" point on the primary set to around 67-70%. The larger capacity secondary injectors also give us a cushion on fuel flow if, for whatever reason demand has been underestimated.
FWIW, when we did the dyno pulls with the GT-35R, we were running a single set of 74# injectors. Balky at low loads, these went to around 90% duty cycle when the GT-35R was cranked up. Dyno runs ceased forthwith.
But how to avoid stressing the pump? The answer is to run two pumps in parallel, one feeding each fuel rail.
Pump 1 sees 45# x .75 x 6 injectors = 202.53/hour flow. This assumes we'ld be running 7.5 psi boost, and not calling on the secondary injectors. 43.5 psi baseline fuel pressure + 7.5 from boost = 51 psi in the lines. 43.5/51 = 85.3% of baseline pump capacity of 600#/hour = 511.8# adjusted capacity. Multiply 511.8 by the 67% continuous use factor and we get 342.9. 202.5/342.9 = ~59.0% continuous load factor. No sweat for the pump.
Secondaries. 74# x 6 injectors = 333#/hour theoretical max delivery requirement. At full boost (15 psi), fuel line pressure = 43.5 +15 = 58.5 psi. 43.5/58.5 = 74.4% x rated nominal pump capacity of 600#/ hour, or 446#. Multiply 446 by 67% gives 289.9. Thus the theoretical delivery requirement of 333# exceeds the sustainable capacity by about 11%, but well within the "burst" capacity of 446#. Going back to the calculated secondary feed requirement of 39.05#, we have 39.05 x 6 injectors = 195.3# flow requirement. 195.3/298.9 = 65.3% We should be just fine.
This setup means several things for the fuel delivery system.
1. Each pump is run through a separate relay-protected controller which manages pump speed based on fuel demand--RPM and load; fuel pressure is monitored via the fuel pressure regulator.
2. Each fuel rail is fed separately, so the system is redundant. If one pump were to fail, the other can supply the motor.
3. The second set of injectors is run by a separate set of drivers in the TEC-3r. Since the system is set up as full sequential, this required building a second driver circuit, and programming the secondary injector bank to open up only under defined loads and RPM. This avoids overfueling at low MAP and when coming off low throttle settings. This was a serious effort on the part of Electromotive, and took some rather original work by Justin Pierce. My hat's off to these guys.
One minor quirk: when the second injector bank opens up, there is a momentary "overenrichment" until the load increases. Not noticeable when driving, but it shows up on the dyno printouts as a hiccup in the A/F ratio.
4. Getting the second injector bank into place was not easy. Justin ended up drilling the bottom sides of the individual throttle bodies and welding in a set of injector bungs, angled slightly into the TB throats, and 180 degrees opposite the primaries. Accessing the injectors, if needed, means pulling the plenum. Wiring was a whole buncha fun.
5. The two fuel pumps live above the rear subframe and feed the two 5/8" ID copper fuel lines going into the two fuel rails. The lines each ahave a 100-micron inline filter before the pump, and a 10-micron filter after the pump. Fuel pickup is from a pair of oversize "duckfeet" in the bottom of the former in-tank pump carrier, then to the pump inlet lines. The in-tank transfer pump was eliminated, and fuel backflow is controlled by a backcheck valve in the pickup lines. The amount of drainage back through the lines is minimal; when the ignition is turned on, there is a very noticeable whine as the pumps are charged and the fuel system comes up to operating pressure.
The fuel regulator is externally adjustable, but not from the driver's seat. It is remote-mounted in the right front of the engine bay, above the ABS controller. The return fuel line back to the tank is a 3/4" ID copper tube. The capacity is dictated by the sizing of the two 5/8" delivery lines, and assures there is no "pooling" of fuel in the injector rails when fuel demand is low. This arrangement helps to dissipate any acquired heat in the fuel picked up via the pumps or being in close proximity to the heat-shielded turbine which lives below the fuel rails. The use of copper in the lines also serves to act as a radiator. Delivery back to the tank via the 3/4" line enters the front of the tank and is terminated away from the fuel pickup point, thus allowing the fuel in the tank to function as a heat sink before it is sent to the pumps.
While this sounds like overkill, it means that the pumps are not being stressed which keeps the heat they add to the fuel stream to a minimum and extends their life expectancy.
For comparison purposes on the fuel lines, the stock lines are 8 mm ID. These have a cross-sectional area of 50.26 square mm.The 5/8" (15.875 mm) lines have an area of 197.93 square mm each. The total cross-sectional area of 395.87 square mm is 7.87 times the stock figure. That ought to deal with flow requirements . . .
For you guys with the M30s who are thinking FI:
IMHO the stock E28 fuel pumps aren't up to the task. Getting to where I had confidence in the adequacy of the fuel delivery has been a real PITA. I think you are going to have to get some pretty exact measurements on the fuel pump delivery rates, injector sizing and duty cycles before a turboed M30 sees much serious road use. IMHO simply assuming the stock pump and lines are up to the task is pretty risky.
Stock pump in the M5 is, I think, 140 liters/hour. The ETK sez it is the same p/n in a 535i, but I'm not 100% certain on this.
Fuel weight = 7.4#/gallon, or about 1.96#/liter. Assume the pump's sustainable max loading is 67% of 140 (see Aeromotive's advice above). 140 x 67% = ~94 liters/ hour. 94 x 1.96 # = 184#/hour. Assume a set of six 71# injectors at 75% duty cycle = 320#/hour, considerably above the stock pumps capabilities.
A Walbro 255liter/hour unit can probably do it (255lph x 67% = 170.9 liters/hour. 170.9 x 1.96 = 335#/hour sustainable delivery), but I have real reservations about the stock 8mm fuel lines sizing. I am very skeptical that the stock lines can meet the injector draw at full demand. This translates into fuel starvation and a lean condition under full boost. Not Pretty.
I am here to tell you that getting all this figured out and installed was a cast-iron bitch.
Go to graphite's website, http://www.doggunracing.com/mye28/LucifersHammer as there are a buncha pictures on this.
Chapter 8 will pick up with my visit to Paul's shop and the next episodes of parts chasing and the assembly adventures.
There. Everybody got that?
This, for the moment, is 15 psi. OK, the dyno showed 14.95. Frankly, it is a tad much. The car without me in it, but with a half tank of gas weighed in at 3360 #. The dyno showed 707 hp at the rollers. This translates to about 4.75 # per horsepower. This spring the game plan is to dial Stage 2 back to about 12 psi. This is being done for reasons of economy. I got a bit tired of having to supply Depends for the passengers . . . . 
