I would like to start a discussion that is of great interest to those here on PPB: The dyno. Thru the years of discussion and analysis on forums like these we have all grown our collective view of what the dyno means as a tool.
In the beginning everyone compared their dyno results without delay. Shortly after that, people realized that the same car on one dyno somehow had different results on other dynos. Some time later, people figured that all dynos of brand X at least read the same, so we could assign some time of scaling factor to move from one brand to another. (recall the Vishnu dyno spreadsheet). Then at yet another date, we discovered that even the same brand dyno doesn’t always read the same. The next thing you know, the same dyno would read the same car differently on different days. Oh hell, it must just be futile.
No, not really futile just difficult and challenging. There are two significant areas in relation to the use and comparison of dynamometer results that deserves some further discussion. Let us start with the ever persistent ‘correction factor’.
The notion of the use of some kind of ‘correction factor’ is based on original work by many different standards bodies to inject some amount of reproducibility in power measurements. In the US, SAE is certainly the most well known, but among the rest of the world, standards from ISO and DIN have equal penetration into the engineering mindset. Lest we Americans think not the world revolves around us.
The concept of having a correction factor is derived from both the changes in engine operation due to atmospheric differences, but also testing procedures, steady state and dynamic efficiencies, and of course repeatability. Most people referring to “SAE Corrections” in the US are talking about SAE1349, which is the current and most updated and recognized industry standard. Like all standards, it is VERY easy to talk about with an air of authority without actually TALKING about it. I took the time to read the standard and much to my surprise it was VERY easy reading. This document is nothing like the usual standards body work in both obfuscation and terminology. I highly recommend it for any professional tuner.
While SAE prevents publication of the document itself, US Copyright fair use provisions do allow quotation and analysis of the document. There are several interesting factors used in the SAE correction, but the “Power Correction Factor” is what most people here would be interested in. Let is take a quick look at the following text:
This section describes the corrections used for non standard atmospheric conditions. There are three critical things to get from this text.
(1) Input conditions to the engine MUST be the same as the calibration is configured for. This practically means you must use an inlet air temperature sensor that measures the inlet conditions, not just the room temperature. (of course air inlet in not intended to be the actual vehicle air inlet, but the effective vehicle air inlet in normal operation. This allow changes the use of fans directed at the air inlet path, if such a path would not see this additional flow in normal operation. SUMMARY: Most dyno’s measure the room temperature, which while better then nothing, is still not a very precise method.
(2) Turbocharged cars need not apply. The use of the standard SAE air density corrections for turbocharged applications is plain wrong. As stated in the standard, without any need for additional clarity, “For example, boosted engines with absolute pressure controls shall not be corrected for ambient barometric pressure”. It need not be any clearer. You will notice by the way that all of the PDXTuning posted data is non corrected. Further in the document, the use of intercoolers is discussed, and there are changes to the air temperature corrections based on the efficiency of these intercoolers.
(3) Using a correction factor of more then 6% total (3% air and 3% fuel) is considered non standard. This is VERY important. Consider that every line of text in this standard has undergone significant review and discussion. This section is very clear for a reason. To help understand this issue, I’ll make the case for it below.
Of the three above, the first two are pretty obvious. While denser air in a normally aspirated car has a predicable change based on the models used in SAE1349, the variable conditions in a boost controlled turbocharged or supercharged engine make such corrections very difficult to model. The rapid slope of the efficiency plots of a turbo charger make changes in inlet pressure (and thus changes in the turbocharger pressure ratio) very non-linear. A small change on some pressure points will make a much greater difference then the same change at a different part of the curve. Since the boost control system will adjust the output pressure, the actual air pressure delivered to the motor will be more dependant on that system then the differences in air density (SIGNIFICANTLY)
The third factor, the limit to 3% air correction is critical. The notion that an engine made 300whp at 6000ft in Colorado, but would make 400whp at sea level is absurd WITHOUT even the simplest analysis. The basis of the correction factor assumes several very important things. It assume the fuel and ignition system, and volumetric efficiency, the exhaust flow limits, and the knock thresholds of the motor scale in the same linear fashion. It also assumes the engineer performing the tests is proficient. It assumes the test parameters are known, and the engineer can intelligently ascertain that at the corrected level, the engine would accurately perform as the scaling indicates. Translation: This means you have actually tested the engine at the other location, and there is plenty of sensor, fuel flow, knock limit, and well as all of the other mechanical and electrical limits that are not railed. To make this more simple, consider the following:
“I made 300whp on my 550cc injector based WRX at 6000 ft. My fuel injectors were at 100% duty cycle, and the boost was 18 psi.. My dyno said the correction factor was 1.22%, so I really made 366whp. Of course when I went down to sea level I ran the same boost with the denser air, my injectors were again at 100%, my AFR was 13:1 on pump gas and the car blew up. Why didn’t anyone tell me injectors can’t run at 122%!!!’
The same story above can be said for so many different parts of the car. If you hit the MAF limit at 6000 ft, do you think somehow at sea level the MAF would magically have more headroom for all this additional magic power?
If you see someone using a correction factor of more then perhaps 6% for temperature corrections (a 2x jump over the official standard view), you should discount at a rate of 2:1 just for the stupidity penalty.
Last but not least, consider section 5.6 of SAE 1439:
Yes, that does say “These correction formulas are not intended for altitude de-rating”.
As I mentioned at the beginning, “There are two significant areas in relation to the use and comparison of dynamometer results that deserves some further discussion”. The second area is what interests me the most. This is the area that I am very much looking for contribution and discussion from other parties. As there are several real SAE engineers (unlike myself, who is only an engineer pretending to play an SAE engineer on TV) present on this forum, as well as countless others who have experiences to contribute.
Let us look at the effects of measured power output based on the type and usage model of the dyno itself. Consider first the simpliest case, the dynojet. The dynojet dyno uses a simple inertial method of measuring power output, and uses the engine rpm to down calculate the torque produced. It does not directly measure torque output, and does not allow for variability of load to the car. When using the dynojet, we are measuring the transient power output by sweeping the car thru a range of engine speeds. This is not a measurement of steady state output.
The SAE standard does have something to say about this:
The most interesting part of the above paragraph is the following: “ The method for determining test conditions used for rating engines from light duty vehicles is to obtain and record time synchronized data on all engine control parameters from an engine installed in a vehicle during a transient maneuver and then duplicate these control settings during steady state operation on a dynometer.” This leads us to a very important conclusion. The rate of acceleration on the dyno must match the rate of acceleration on the road. This means the rate must include effects of aerodynamic drag, wheel drag, angular inertia drag, as well as the effects these loads have on the power produced. Not only must the rate (not just in total, but also in shape of delivery) be the same, but the vehicle operating parameters must match the real transient maneuver. That means the same boost, timing, fueling, etc.
It is instructive to look at how the three most common dyno’s handle the above requirements:
Dynojet: Load is a static load based on the weight of the rollers. (actually the angular inertia). If your car happens to weight the same, your closer the if it doesn’t.
Dynapak: The Dynapak has a dynamic load cell that can provide any possible load curve. However as the software is currently written, it appears to offer only a static sweep rate. The static rate allows you to make a RPM range (say 2000-7000 rpm) to take 12 seconds, in a linear slope. This does have the benefit of allow the operator to make the dyno load more or less, but however it is very flawed in simulating real world conditions. If you car produces 500whp vs 200whp, the same time sweep means that the higher hp car is seeing much more load. In the turbocharged application, this can results is abnormally high initial torque readings as the dyno applies reverse torque to match to specified sweep rate. To properly use this dyno, one would have the test the vehicle on the road, and make the sweep time the same as the road test vets.
Mustang: The Mustang, in theory, has the best possible simulation. It does both a static load (weight of the rollers), a dynamic weight (based on the vehicle weight, as set by the user), and a second dynamic weight based on the aerodynamic drag (specified as hp@50mph). This allows the dyno to provide a load that is similar to what would be seen on the street. However, in my experience, this method works only as well as the user has loaded the right calibration files. These calibration files can make s HUGE difference in shape and size of the curves produced. This is evident by comparing results from the Mustang dyno here, the Gruupe-S, FIS, and many others.
Not only does the load (how fast the pull goes) change the engine loading and time for turbine spool, but it also effects the amount of energy load in the flywheel to wheel conversion.
This conversion from flywheel to wheel is often of the most interesting debate. Before looking in that aspect of dyno tuning, let us first look at the effect of sweep rate on energy used to accelerate the drivetrain components.
As we all know from high school physics, it is easy to model the effects of drum accelerating using only a few simple calculations. I’ll save the space for now and only show the abbreviated approach. The goal is to determine the amount of power used to accelerate the drivetrain and wheels(called wheels below), given a fixed inertial weight dyno drum (no dynamic load).
Twa = Torque accelerating the wheels
Tds = Torque applied to accelerate the drum
Ti = Torque input to the system (from the engine)
Fds = Force at the drum-wheel surface(drum)
Fws = Force at the drum-wheel surface (wheel)
Rw = radius of the wheel
Rd = radius of the drum
Fds=Fws (unless the wheels break loose)
Tds = Ti – Twa
Twa = Iw*Aw
Fds = (Ad*Id)/Rd
Since Fds = Fws
(Ad*Id)/Rd = (Ti – IwAw)/Rw
Simplification yields what we are looking for:
Twa = Ti * ( 1 / ( [Rw^2*Id]/[Rd^2*Iw] ))
In less mathematic form, this suggests the loss to accelerate the wheels is based on the ratio of the inertial load of the wheels vs the load of the drum. To put some quick calculations in place: A 3200lb equivalent drum, used on a car with 4 32lb wheels, results in the loss of torque used to accelerate the wheels at about 2%. Keep in mind this is using a simple filled cylinder approximation for angular inertia for the wheels (the wheels are not a solid cylinder of course). The wheel loss might actually be a bit higher, and of course the transmission and shaft load is something non zero as well.
None the less, it is clear that if you change the load weight from 3200 lbs to 1600 lbs, you get 4% loss. This difference in loss is more then the typical SAE correction used. This figure is probably under rated, given the real weight and inertial mass of the components.
Translation: The rate of acceleration on the dyno (controlled or uncontrolled) will directly effect the measured power. On a dyno like the dynojet the load is consistent so the same car with the same wheels and driveline should experience the same comparative load. This comparative loss is not ties to the power produced. In this way, the dynojet model does produce results that can be more consistent across the same kind of dyno. (given the same weight drums, which is not true of course).
On a dyno like the dynapak, the above equations are useless because the torque load delivered to the engine is not a static value. It varies based on the power output of the car, which makes even run to run comparisons very difficult.
This does not make the results on the Dynapak useless, or any less useful for tuning. It does however make the results VERY VERY difficult to compare to results from other kinds of dynos. It unfortunately also makes comparisons of different cars on the same dyno also a bit difficult. The comparisons can still be made, but differences in the 2-5% range should not be considered significant.