TheRealMRDyno
10+ Year Contributor
- 46
- 5
- Aug 19, 2009
-
Akron,
Ohio
Hi,
I'm the guy that wrote and maintains Mustang Dynamometer's PowerDyne PC software, among other things. (Heap abuse for your pet-peeves with said software here...)
I am also, like many of you, I think, frustrated when trying to estimate the power potential for various modification levels based on posted dyno run data.
And, it appears I need to do >30 posts so I can see if someone has a rear Talon logo piece in good condition for my <20k miles 91 TSI...
So, in an effort to help myself, as well as everyone else, here is a little info on how dynos measure power, and how the way a test is run can influence delivered power. If you read to the end, you will know how to get your dyno operator to show you power/torque values as they will be delivered to the pavement, and which will be consistent from dyno to dyno. Here we go...
To get this out of the way up front, some of us have ourselves to blame for dyno values that vary all over the place. Dyno shop owners need to make money to stay in business, and I have never heard of a customer saying they wanted, or would be happier to see, lower numbers. So, be aware, there are more than a few intentionally mis-calibrated dynos out there, all as a result of pressure from us, the customer. Most of these mis-calibrated dynos are attempting to duplicate Dynojet readings, which are "adjusted" to achieve the expected, but not actually measured, output from some motorcycle tested many years ago.
Now, the dyno stuff...
A dyno measures the power that makes it from the vehicle's engine to the dyno rollers. That power is "sunk" into one of 3 1/2 places:
1) The mechanical inertia of the dyno, which is equivalent to some vehicle weight, given the diameter of the rollers. This is measured via acceleration times the inertia/weight of the dyno. Acceleration is measured to exceptional precision (in the case of Mustang Dynos at least - this requires something on the order of 2 arc-second encoders, or a mapped lower precision encoder) times the inertia/weight of the dyno. Just FYI, for those who have seen a noisy "Accel Torque/Force/Power" trace in the MD software, that is generally due to switching imprecision in the encoder or gear sensor, our noise floor. It can also be caused by driveshaft windup/release in shafted multiple axle designs, which must be addressed via multiple speed sensors and relative weighting.
2) Parasitic losses of the dyno - bearings, belts, gearboxes, and windage. This can not be measured directly during testing, and so must be measured in a parasitic losses measurement routine, and the resulting data stored in a lookup table or fitted curve coefficients. Note that parasitic losses on the "measured" side of any torque sensor must be subtracted from the raw losses of the dyno, so that those losses are not counted twice during testing, since they will be measured directly by said torque sensor during testing.
3) Any active loading element, like an eddy current brake, water brake, or A/C or D/C motor. Loads applied by these controlled elements are measured directly via in-line torque sensors or reaction-arm mounted strain gauges.
3.5) Tire-to-roll interface losses. Despite the 0.5 index here, this is a big deal. Due to increased tire deformation on chassis dyno rollers, compared to on flat pavement, these losses can be pretty significant. This is a bigger deal on small roll designs than on large roll designs, on cradle-roll (two rollers per axle) designs than single-roll designs, and an even bigger deal on multiple-axle designs than on single-axle designs. For performance dynos, multiple-axle pretty much means an "AWD" dyno with (2) axles. The tire:roll interface losses can change significantly from test session to test session if the strapping method/location/tension changes, if the tire pressure changes, if the tire temperature changes, or if the dyno has an adjustable wheelbase and that changes. So, all of these variables need to be fixed (that is, held constant) for all test sessions - alternatively, and much easier to do in practice, these losses can be accounted for as explained below. Note that on multiple-axle dynos with adjustable wheelbases, and cradle-roll type designs, getting the wheelbase set correctly is a big, big deal. If a vehicle is on a cradle-roll type multiple-axle dyno, and the dyno wheelbase does not very (very) nearly match the actual vehicle wheelbase, the tire:roll interface losses, which tend to turn into heat in the tires, can be large enough to cause tire blow-outs.
To get consistent measurements, here is what you need:
1) Have an accurate value for the dyno's inertia/weight - this is a fixed value, so it does not change (except that some configurable dynos have different inertia/weight values based on configuration).
2) Have accurate values for the dyno's torque sensor calibration. This can vary over time, primarily in terms of an offset drift. The simple solution is to calibrate the dyno every day, and maintain a stable temperature in the dyno cell.
3) Know the dyno's internal parasitic losses, and know the additional tire:roll interface parasitic losses. The simple way to account for the tire:roll interface losses is to run the parasitic losses measurement routine *with the vehicle on the dyno, in neutral*, before each test session. This is where most shops slip up. It is less work to simply measure the dyno's internal parasitic losses once, and always use that data, than to measure the dyno's losses plus the as-strapped-down, at-tire-pressure, etc, tire:roll interface losses. However, it only takes a few minutes to run this test, and it eliminates a fairly large, and fairly variable, variable. Yes, this includes axle bearing losses, differential losses, etc, but those are much smaller than the tire:roll interface losses, so I stand by this practice. This is one way to achieve +/- 1 HP out of 550 HP repeatability across test sessions.
Assuming the dyno operator properly addresses all of the above, there is still one significant variable in terms of getting repeatable, consistent from dyno-to-dyno, values, and that is: the acceleration rate experienced during vehicle testing (bet you thought I was going to say ambient conditions, didn't you? - I give most modern dyno systems credit for having SAE J1349 type corrections built-in, so while they do matter, I think they are pretty well addressed). How much difference can this make? A lot. We did some testing in our demonstration facility with a NASCAR vehicle, making about 550 HP on the dyno. By changing the acceleration rate (by changing loading, or by directly changing the sweep rate), we could move that car's measured output across a 50 HP range.
This is why dyno guys will say things like "you have to test in 4th gear" - they have seen that other gears will generate different power readings. My feeling is that this is generally attributed to different losses in different gears, and while there is some small truth to that, it is not so significant. The acceleration rate, though, is. That is because the engine's output can either a) heat up your drivetrain via frictional losses, or b) accelerate your drivetrain's inertia, or c) get to the ground or dyno rollers (or d) burn up your tires, you are on your own with that - higher speeds help). In the NASCAR example above, all tests were performed in the same gear, so you can be sure the gear losses were not the cause of the 50 HP measurement range. With a high acceleration rate, more of your engine's torque/power is absorbed internally as it accelerates your vehicle's drivetrain, than with a lower acceleration rate. Revving your vehicle in neutral = 100% of engine output going to frictional losses and accelerating drivetrain components. Hence, the pursuit of lighter flywheels, smaller diameter clutches, disconnected balance shafts, etc.
Note that higher load => lower acceleration rate => lower drivetrain losses => higher measured power, in theory and in practice. For those of you concerned about tire slip increasing with more load, it appears that this is not the case, so long as the speed during testing is such that the forces generated are well within the traction capability of the tires.
On a pure-inertia dyno, ala the typical Dynojet, there is no way to control the acceleration rate experienced during testing, it is what it is. This has some advantage in dyno-to-dyno comparisons, since nothing *can* be varied in terms of acceleration rate.
However, on dynos with some active load control element (eddy current brake, water brake, motor, *train* brake pads applied to the rollers), the acceleration rate experienced will be a function of that controlled load (and vehicle output). Since the load or sweep rate is controlled by the operator via software, the acceleration rate has become a variable.
There are two main approaches to controlling the acceleration rate during dyno testing:
1) Actually controlling the sweep rate (= acceleration), by putting the dyno in a ramped speed-control mode, and specifying a speed range and time for the sweep, or by putting the dyno in a direct acceleration control mode. This will indeed maintain the same acceleration rate (within the dyno's limits), regardless of other variables - like if you jacked your boost up by 5 PSI. Note that on the road, that extra 5 PSI would probably increase the acceleration rate somewhat. So, in this example, the measured difference in power due to that 5 PSI on the dyno would be more than you would experience on the street - your vehicle's drivetrain would absorb more torque in the higher acceleration on-street scenario than it would on the dyno, where the acceleration rate, with the additional power, was held at the original acceleration rate.
2) Applying a road-load simulation during testing. In this case, the load applied to the vehicle would be a road-load curve, plus some additional load proportional to acceleration times the test weight entered for the vehicle, generally:
ForceApplied = (A*1) + (B*Speed) + (C*Speed*Speed) + (TestWeight * Accel)
(That is not the exact math used, but it is clearer than the actual math, which basically changes to "... + (TestWeight * AccelRateThatWouldHappenOnTheStreet).)
In this case, the acceleration rate experienced in dyno testing will be higher with the additional 5 PSI, just at it will be on the street - which means your drivetrain will be eating up more torque/power, and the difference that the 5 PSI makes in reality is less than a controlled sweep rate type test comparison would show.
So: use the vehicle simulation loading method.
And that is most of the dyno stuff.
There are also some vehicle related variables that need to be held constant in order to get the most repeatable data, beyond what can be rolled up into the dyno's parasitic losses data. These all have to do with temperature:
a) Engine temperature
b) Transmission and axle temperatures
c) Intake manifold temperature
(etc etc etc)
z) Intercooler cooling air flow - this is a big, big deal for boosted vehicles, and is difficult to control so that it matches the on-road cooling air flow the vehicle will experience. If anyone is bored, and has some money to go with their time, I think the general approach to pursue would be to 1) measure either a) the air speed behind the intercooler at various speeds on the road, or b) the pressure drop across the intercooler at various speeds on the road, and then 2) setup a closed loop controller with a (large, powerful, high-speed) fan as its actuator, trying to make the same speed or pressure conditions exist on the dyno based on speed. Mustang dynos have an analog output proportional to speed that could be used for this.
Hope this helps someone...
TheRealMRDyno (and yes, I made those, too)
I'm the guy that wrote and maintains Mustang Dynamometer's PowerDyne PC software, among other things. (Heap abuse for your pet-peeves with said software here...)
I am also, like many of you, I think, frustrated when trying to estimate the power potential for various modification levels based on posted dyno run data.
And, it appears I need to do >30 posts so I can see if someone has a rear Talon logo piece in good condition for my <20k miles 91 TSI...
So, in an effort to help myself, as well as everyone else, here is a little info on how dynos measure power, and how the way a test is run can influence delivered power. If you read to the end, you will know how to get your dyno operator to show you power/torque values as they will be delivered to the pavement, and which will be consistent from dyno to dyno. Here we go...
To get this out of the way up front, some of us have ourselves to blame for dyno values that vary all over the place. Dyno shop owners need to make money to stay in business, and I have never heard of a customer saying they wanted, or would be happier to see, lower numbers. So, be aware, there are more than a few intentionally mis-calibrated dynos out there, all as a result of pressure from us, the customer. Most of these mis-calibrated dynos are attempting to duplicate Dynojet readings, which are "adjusted" to achieve the expected, but not actually measured, output from some motorcycle tested many years ago.
Now, the dyno stuff...
A dyno measures the power that makes it from the vehicle's engine to the dyno rollers. That power is "sunk" into one of 3 1/2 places:
1) The mechanical inertia of the dyno, which is equivalent to some vehicle weight, given the diameter of the rollers. This is measured via acceleration times the inertia/weight of the dyno. Acceleration is measured to exceptional precision (in the case of Mustang Dynos at least - this requires something on the order of 2 arc-second encoders, or a mapped lower precision encoder) times the inertia/weight of the dyno. Just FYI, for those who have seen a noisy "Accel Torque/Force/Power" trace in the MD software, that is generally due to switching imprecision in the encoder or gear sensor, our noise floor. It can also be caused by driveshaft windup/release in shafted multiple axle designs, which must be addressed via multiple speed sensors and relative weighting.
2) Parasitic losses of the dyno - bearings, belts, gearboxes, and windage. This can not be measured directly during testing, and so must be measured in a parasitic losses measurement routine, and the resulting data stored in a lookup table or fitted curve coefficients. Note that parasitic losses on the "measured" side of any torque sensor must be subtracted from the raw losses of the dyno, so that those losses are not counted twice during testing, since they will be measured directly by said torque sensor during testing.
3) Any active loading element, like an eddy current brake, water brake, or A/C or D/C motor. Loads applied by these controlled elements are measured directly via in-line torque sensors or reaction-arm mounted strain gauges.
3.5) Tire-to-roll interface losses. Despite the 0.5 index here, this is a big deal. Due to increased tire deformation on chassis dyno rollers, compared to on flat pavement, these losses can be pretty significant. This is a bigger deal on small roll designs than on large roll designs, on cradle-roll (two rollers per axle) designs than single-roll designs, and an even bigger deal on multiple-axle designs than on single-axle designs. For performance dynos, multiple-axle pretty much means an "AWD" dyno with (2) axles. The tire:roll interface losses can change significantly from test session to test session if the strapping method/location/tension changes, if the tire pressure changes, if the tire temperature changes, or if the dyno has an adjustable wheelbase and that changes. So, all of these variables need to be fixed (that is, held constant) for all test sessions - alternatively, and much easier to do in practice, these losses can be accounted for as explained below. Note that on multiple-axle dynos with adjustable wheelbases, and cradle-roll type designs, getting the wheelbase set correctly is a big, big deal. If a vehicle is on a cradle-roll type multiple-axle dyno, and the dyno wheelbase does not very (very) nearly match the actual vehicle wheelbase, the tire:roll interface losses, which tend to turn into heat in the tires, can be large enough to cause tire blow-outs.
To get consistent measurements, here is what you need:
1) Have an accurate value for the dyno's inertia/weight - this is a fixed value, so it does not change (except that some configurable dynos have different inertia/weight values based on configuration).
2) Have accurate values for the dyno's torque sensor calibration. This can vary over time, primarily in terms of an offset drift. The simple solution is to calibrate the dyno every day, and maintain a stable temperature in the dyno cell.
3) Know the dyno's internal parasitic losses, and know the additional tire:roll interface parasitic losses. The simple way to account for the tire:roll interface losses is to run the parasitic losses measurement routine *with the vehicle on the dyno, in neutral*, before each test session. This is where most shops slip up. It is less work to simply measure the dyno's internal parasitic losses once, and always use that data, than to measure the dyno's losses plus the as-strapped-down, at-tire-pressure, etc, tire:roll interface losses. However, it only takes a few minutes to run this test, and it eliminates a fairly large, and fairly variable, variable. Yes, this includes axle bearing losses, differential losses, etc, but those are much smaller than the tire:roll interface losses, so I stand by this practice. This is one way to achieve +/- 1 HP out of 550 HP repeatability across test sessions.
Assuming the dyno operator properly addresses all of the above, there is still one significant variable in terms of getting repeatable, consistent from dyno-to-dyno, values, and that is: the acceleration rate experienced during vehicle testing (bet you thought I was going to say ambient conditions, didn't you? - I give most modern dyno systems credit for having SAE J1349 type corrections built-in, so while they do matter, I think they are pretty well addressed). How much difference can this make? A lot. We did some testing in our demonstration facility with a NASCAR vehicle, making about 550 HP on the dyno. By changing the acceleration rate (by changing loading, or by directly changing the sweep rate), we could move that car's measured output across a 50 HP range.
This is why dyno guys will say things like "you have to test in 4th gear" - they have seen that other gears will generate different power readings. My feeling is that this is generally attributed to different losses in different gears, and while there is some small truth to that, it is not so significant. The acceleration rate, though, is. That is because the engine's output can either a) heat up your drivetrain via frictional losses, or b) accelerate your drivetrain's inertia, or c) get to the ground or dyno rollers (or d) burn up your tires, you are on your own with that - higher speeds help). In the NASCAR example above, all tests were performed in the same gear, so you can be sure the gear losses were not the cause of the 50 HP measurement range. With a high acceleration rate, more of your engine's torque/power is absorbed internally as it accelerates your vehicle's drivetrain, than with a lower acceleration rate. Revving your vehicle in neutral = 100% of engine output going to frictional losses and accelerating drivetrain components. Hence, the pursuit of lighter flywheels, smaller diameter clutches, disconnected balance shafts, etc.
Note that higher load => lower acceleration rate => lower drivetrain losses => higher measured power, in theory and in practice. For those of you concerned about tire slip increasing with more load, it appears that this is not the case, so long as the speed during testing is such that the forces generated are well within the traction capability of the tires.
On a pure-inertia dyno, ala the typical Dynojet, there is no way to control the acceleration rate experienced during testing, it is what it is. This has some advantage in dyno-to-dyno comparisons, since nothing *can* be varied in terms of acceleration rate.
However, on dynos with some active load control element (eddy current brake, water brake, motor, *train* brake pads applied to the rollers), the acceleration rate experienced will be a function of that controlled load (and vehicle output). Since the load or sweep rate is controlled by the operator via software, the acceleration rate has become a variable.
There are two main approaches to controlling the acceleration rate during dyno testing:
1) Actually controlling the sweep rate (= acceleration), by putting the dyno in a ramped speed-control mode, and specifying a speed range and time for the sweep, or by putting the dyno in a direct acceleration control mode. This will indeed maintain the same acceleration rate (within the dyno's limits), regardless of other variables - like if you jacked your boost up by 5 PSI. Note that on the road, that extra 5 PSI would probably increase the acceleration rate somewhat. So, in this example, the measured difference in power due to that 5 PSI on the dyno would be more than you would experience on the street - your vehicle's drivetrain would absorb more torque in the higher acceleration on-street scenario than it would on the dyno, where the acceleration rate, with the additional power, was held at the original acceleration rate.
2) Applying a road-load simulation during testing. In this case, the load applied to the vehicle would be a road-load curve, plus some additional load proportional to acceleration times the test weight entered for the vehicle, generally:
ForceApplied = (A*1) + (B*Speed) + (C*Speed*Speed) + (TestWeight * Accel)
(That is not the exact math used, but it is clearer than the actual math, which basically changes to "... + (TestWeight * AccelRateThatWouldHappenOnTheStreet).)
In this case, the acceleration rate experienced in dyno testing will be higher with the additional 5 PSI, just at it will be on the street - which means your drivetrain will be eating up more torque/power, and the difference that the 5 PSI makes in reality is less than a controlled sweep rate type test comparison would show.
So: use the vehicle simulation loading method.
And that is most of the dyno stuff.
There are also some vehicle related variables that need to be held constant in order to get the most repeatable data, beyond what can be rolled up into the dyno's parasitic losses data. These all have to do with temperature:
a) Engine temperature
b) Transmission and axle temperatures
c) Intake manifold temperature
(etc etc etc)
z) Intercooler cooling air flow - this is a big, big deal for boosted vehicles, and is difficult to control so that it matches the on-road cooling air flow the vehicle will experience. If anyone is bored, and has some money to go with their time, I think the general approach to pursue would be to 1) measure either a) the air speed behind the intercooler at various speeds on the road, or b) the pressure drop across the intercooler at various speeds on the road, and then 2) setup a closed loop controller with a (large, powerful, high-speed) fan as its actuator, trying to make the same speed or pressure conditions exist on the dyno based on speed. Mustang dynos have an analog output proportional to speed that could be used for this.
Hope this helps someone...
TheRealMRDyno (and yes, I made those, too)
thanks
