Okay, I've been mulling this over for a bit, as it seems people come up with different numbers all the time from their dyno runs. Of course you have the argument of exactly how much driveline loss there is with any given car, but that's not what I'm after in this post. I'm only interested in the *correct* way to do the math, provided you know the percentage loss.

Here's my take on it:
Let's say in one example you have a car with a 20% power loss from engine to rear wheels. (No particular car, just an example.) :) Since the power is being generated at the engine, and then being *lost* on its way to the wheels, I would assume the formula would be like this:
(engine power) * (.80) = (wheel power)
So, if you do a dyno run and find the wheel power, you would divide the wheel power by .80.

Does this sound about right?

X=wheel hp, 80/100 x X/?. Which is just about the same way your doing it but without the decimal point.

If you approximate your driveline losses as a constant percentage, then you are assuming that the torque needed to overcome the driveline friction is constant with RPM.

Which is an OK approximation. If you had a dyno plot of the engine on a stand, you could then take the torque line and move every point on it down 20% (or whatever you think your driveline losses are). Same for the HP line (as HP is just torque*RPM/5252) to get what the net car would be.

Conversely, you could take a wheel-dyno plot and move the lines up by the right fraction. Divide by 0.8 in this example.

But, there is more in going from the wheel dyno to the shaft. You also have to scale for whatever gear you were in in the dyno run, and the final gear ratio. If you want to calculate shaft HP, you need to take out the "lever arm" of these gear ratios. Most reputable wheel dyno run plot already take this into account, however.

Also in real mechanical systems, friction isn't constant with velocity of parts. This is especially true if there are any viscous coupling type parts. Mecifully, the audis still have the old Torsen gizmo that is all gears. So the approximation ought to be close. But if you have an auto or tip car, then I'm pretty sure your torque converter has a non linear drive line loss effect. This would mean needing to move your torque line around by a different fraction depending on RPM at that point on the curve, and then recomputing the HP curve.

Of course since we all can't agree on what the basic driveline loss is for our cars, there's no point in trying to get the non-linear terms right. I've heard anyplace from 15% to 45% driveline losses for the quattro. Sometimes I think people like to believe there are large losses because then they can multiply/divide and conclude they have a nice big engine output due to their mods.

I'd love to see a stock car, tuned to basic specs, with my exact drivetrain on a dyno. Then we could do your math to get the right scale factor for the drive-line. All I ever get to see are non-stock cars, and guesses about hwo to go from dyno to shaft.

I completely agree. The losses are probably non-linear through different speeds, however you are right that there's just no use arguing over that when people can't even agree on a linear value. :)

However, a while back BOOSTD posted a dyno run he did over at APR on their 4-wheel dyno. They did a stock, unmodified A4 to get a baseline, which is rated by Audi to put out about 150 hp. It was getting about 118 at the wheels on it's best run. So, that gives you about a 22% loss.

I don't have the link to the post handy, but I'm sure it can be found on the search engine relatively easily. Found it last night in just a few minutes.

...because driveline loss is NOT a percentage - it is a fixed torque figure for each RPM point.

This has been bantered about on the E30SIG site numerous times. The bottom line is that they have dynoed both stock and modified engines first on a dyno stand (engine only) and then on a driveline dyno. The result: the tranny and rear end consume the same torque (equateable to hp in conversion) at a given RPM regardless of the engine.

Logic further dictates that unless you have made modifications to the drivetrain itself, it will take the same torque to get it to spin 1000 rpm - regardless of power source (electric motor, stock engine, modified engine, guy on a stationary bike, etc.).

Sorry,
Mike O.

Take the torque curve and move the line down by a fixed value (offset) as opposed to a percentage (gain), to go from shaft HP/torque to wheel.

That kind of makes sense. If you believe that the loss in the driveline is static friction and not dynamic viscous loading.

But some of it has to by dynamic loading. The moment of intertia of the spinning components dictates that if I go from 0-6000 RPM in 5 seconds, it would take more torque to do the same thing in 4 seconds. Here the driveline loss is not a constant with RPM, because it has a rate-of-change-in-RPM term.

Do the math. Four (quattro) 36 lb wheel/tires spun from zero to 60 MPH in equivalent wheel RPM takes 28 ftlbs of torque. Which would be seen as driveline loss. If you took 12 seconds to do it, you would only take half that to this loss.

If you slowly climb through the RPM on your dyno the above effect would not be seen I think.

The loss has a constant term for static friction, an RPM dependant term for viscous friction and a delta-RPM term for rotational components.

People say percentage. Or constant. It's just an appoximation. The dynamic system in the time domain is:

TsubA = C + C(omega) + C(omega-dot)


- Z

Comparing rearwheel dyno data to flywheel dyno data is a real can of worms. Losses vary not only due to driveline effiencies, inertia, traction/wheel slippage, but also such mundane things as tire pressure. By the way, that's an interesting way to fudge chassis dyno data.

Taking one percentage and applying it to all data from a given car is shaky, at best. Applying the same percentage to another car is less than shaky.

We witnesed a test at Superflow where they ran a big-block Chevy (~800 hp) engine on one of their engine dynos one day, then installed the engine back into it's Nova drag-car chassis and ran it on their AutoDyn chassis dyno the next day. They had fairly good corrections for the drag slicks and powerglide trans/convertor losses from many runs. When the curves were overlaid, with the empirically derived correction factor applied they agreed within 2%! We were all impressed.

The AutoDyn, in my opinion, is much more sophisticated (and more costly) than the Dynojet chassis dyno, and Superflow's software (WinDyn) is outstanding. I was impressed with the correlation. In another test, they ran a 450hp rated Viper on the AutoDyn chassis dyno, and in 5th gear, it pulled a corrected 450 +/- about 5 hp. at 203 dyno mph. Of course, the engine got very hot, even with mega fans blowing air on the car.

I'm new to this particular forum, but my friend LarryTT asked me to visit. My background is relevant. You don't have to believe what I said above, but I think Mike O. and zain have a good grasp of the topic; believe them.

Apples don't always equal oranges, unfortunately.

My highly-opinionated $.02.

Mike is right that there is a constant friction term. And if you start to include dynamic acceleration terms and viscous friction, and tire deformation, well it does become a science fair.

Most plausable tests of chassis/wheel dynos of stock cars have produced 20-25% less peak HP at the wheels than stock. Depending on vintage of A4 (the test I've seen are A4) then that's 20-30 HP to run the quattro driveline.

If we doubled the HP of the engine, would it take double the HP to run the drive line? (and let's just stay at the peak HP point for sake of discussion)

Mike would say no. Fixed losses. Fixed torque, they fixed HP loss at the RPM at peak. And to first order, he's probably close.

Unless the 0-60 time went down a lot, the torque needed to spin up all the moving parts isn't going to increase much. Doubling your HP doesn't exactly half our 0-60 time as we all know. Losses due to viscous terms and intertia will increase as a percentage, but relative to achieved acceleration rate (like 0-60 time ratios) not shaft HP.

So until your mods start making significant changes to your 0-60 times, saying 20-30 HP to run the driveline (at peak) (fixed) isn't a bad approximation.

I'd love to see some solid data to back up how much torque/hp it takes to run the quattro driveline however.

pay enough attention to temp/humidity/baro correction factors. Accurate inlet air temps are very important for accurate correction, and these can vary considerably during chassis dyno pulls. I have seen chassis dyno runs where the air temp wasn't measured even under the hood.

Engine dyno rooms are much more controlled environments and inlet air temp is measured at the carb or MAF. IMO, this is how all chassis runs should be monitored. With charge coolers, it's also important, IMO, to try to keep as close as possible to the 77F or so inlet temp which the OEM engine dyno folks correct to.

Additionally, manufacturers' dyno numbers are taken at steady-state (step) conditions. Ideally, we should dyno an engine in an acceleration mode which matches the chassis run. That would be something like 300 to 500 rpm/sec.

Between inlet air conditions and acceleration rate, you could easily 'misplace' 20-30 hp. Unless you have a chassis dyno capable of step-rpm runs (like the AutoDyn with eddy current absorbers), and accurately measure air conditions, correlation to anything other than runs on the same dyno with the same car is just a guess.

At the extreme end are race shops like Hendrick Motorsports (Jeff Gordon, Terry Labonte, etc) where they are looking for 2-3 hp changes on 750+ hp engines. A typical engine dyno repeatability is usually given at 1-2%. That's as much as 15 hp.
To get 1/4 to 1/2% repeatability takes an extremely anal approach. At a seminar given by Hendrick MS' dyno chief, we learned that they not only calibrate the dyno with weights before and after a test series (instead of monthly), but they control inlet air temp and even fuel temp as it enters the carb to a few degrees F. That blew my mind. I believed him when he said they really can see 2-3 hp changes. Geeesh!

Audi knows exactly how much the driveline losses are, but unless you have an inside contact, geting those #s isn't practical. I, too would like to know. Comparing Torsen to Haldex would also be interesting.

My $.02

I did a stock dyno pull (on a different dyno of course, so results aren't directly comparable) where my car was putting down 104.7HP @ the wheels (to the ground). That works out to ~30% driveline loss. Check my PP for the plot.

That a chassis dyno run done with a fixed load at WOT will really show up low. Because, compared with a variable load set of step measurements, the accelerating run test will cause the rotational masses to come into play.

And there's a lot of mass to spin up. As you are "spending" 10 ft lbs per wheel (36 lb wheel/tire assumption) to accelerate them from 0 to spinning the equivalent of 60 MPH. That load comes right off the HP measurment you'd get.

Interesting. Makes it really hard to get any genuine data for all our mods.

For example putting lighter wheels on a test car would result in more HP "at the wheels" under an acceleration test. Which makes sense, cars to accelerate a bit better with lighter wheels.

Did you make more HP? No.
Did you make more HP get to the road? Yes.
Is the car actually faster as a result? Yes, I guess. A little.

Maybe we should all just talk about at the wheels HP, because (assuming you don't play games with test conditions) that's what really makes the car go. If you got more HP to the wheels via engine mods, or more efficient drivetrain, then what difference does it make?

maybe a Torsen to Haldex swap gets 10Hp to the wheels. Maybe it costs you 10 HP to the wheels. The engine is the same. But if one is more efficient than the other, then one car will be faster than the other...

Facts would be nice... But I don't know anybody at audi.

- Z

at 2000 rpm per second, like in first gear on a TT or maybe an S4. It plays a very little part when accelerating at 20 rpm per second like at 120 mph in 5th or 6th.

Sure, tire/wheel inertia is important for acceleration as you said, but in first the engine is turning 10-15 times faster than the wheels, and 3 times as fast (approx)in 6th, so engine inertia is much more critical. You should see more hp on a chassis dyno in 4th or 5th than in 2nd or 3rd just due to the slower acceleration in the higher gears.

There is a great engine design program, Engine Analyzer Pro from Performance Trends which can demonstrate the difference between step and accelerating tests. If one has good numbers for the inertia of engine parts, you can also see the effect of lightening flywheels, cranks etc. Of course that does very little for 5th and 6th gear acceleration, but quite a bit more for 1st-3rd.

Race engines which accelerate very rapidly (maybe as much as 3-4000 rpm/second in an 18,000 rpm F1 engine) have absolutely the least amount of rotating inertia they can legally get. That's why F1 uses 100 mm (under 4 in.) dia clutches.

BTW, there's talk that the 2002 Ferrari F1 car, which we may not see until the third race or so, has eliminated the clutch between the engine and the gearbox(!)and uses only the clutch packs on each output side of a conventional limited slip diff. (Racecar Engineering, Feb 2002 pg.4).
Standing starts would be handled by the electronically controlled differential. I like how these guys/gals think!

FWIW: Overinflating the tires on a chassis dyno run should increase the hp due to lower rolling resistance. Also, if you want accurate results, you should be monitoring drive wheel/tire rpm (usually with one white stripe on the tire and a light). Just using the dyno roll for mph doesn't take into account tire slip or rolling radius change. I've seen chassis tests run without tire rpm measurement. Have you guys seen your dyno operators use it?

If I were an unscrupulous tuner/dyno operator and I was selling you the 'trick of the week' or any mod for that matter, and doing back-back tests to "prove" to you that my mods work, I might increase tire pressure 10-15 psi for the 'after' test, and maybe direct more cooling air toward the air inlet or chargecooler inlet than I did on the 'before' test. Of course, the blatant way is to play with the correction factor, but assuming that is noted on the hard copy, you could find it.

I'm not a tuner/dyno operator, BTW, and I'm not accusing anyone, but it can be done easily. I've seen a lot of guys with more money than knowledge dump big bucks (Euros?) with less than honest tuners. I've also seen a reputable engine builder's 4 cyl 2+L Honda with bars and bars of boost twist a Superflow engine dyno (at Superflow) to 765 hp. That truly impressed me; much more than a BBC doing the same number.The engine builder was Javier Gutierrez, owner of JE Engine Dynamics. If I was into Hondas, he'd be my choice.

That dyno hooks to the hub directly, and the wheels are removed from the car. I guess that would reduce some drag and give a higher reading. Good call.

unless they were run on an AutoDyn. Even then, stabilizing for 5-10 secs. at each 500 rpm increment could take a couple of minutes which is a very long time for a car on a chassis dyno unless lots of cooling air is available.

While this is uncommon, I agree that's what we need to compare to OEM engine tests.

If you are trying to compare a given engine of yours both on an engine dyno and on the chassis dyno, figure the average acceleration on the chassis dyno (4000 rpm change in 10 seconds or 400rpm/sec, for example), then have the engine dynoed at that same acceleration. With good temp/humidity/baro corrections you'd have a fairly accurate way of getting driveline losses.

FWIW, most aftermarket/speed shop/racecar engine dyno pulls are not steady state, except perhaps for Daytona or Talladaga NASCAR or maybe IRL engines which don't change rpm much during a lap.
The wildest are probably Pro Stock engines which barely have 10 seconds of valve spring life when they are lifting 2-1/4 inch + valves about an inch at 9-10 grand.

simple ratio on the result, you hear statements like.

"I got 225 at the wheels so adjusting for the 30% driveline losses I have a 321HP car."

Which is kind of goofy because:

1 )Percentage is a pretty crude approximation, not quite modelling the friction and intertia terms correctly.

2) It's not comparable to OEM specs, which are measured at steps (so they remove the intertia terms)

3) Who cares what's at the shaft, what makes it to the road is what moves the car.

The really odd thing about all of this is that a chassis dyno run at WOT simulating acceleration and intertia loads is probably a more realisitic way to assess the car. It's just so easy to game and get lost on, and can't be used for comparisons easily.

Fun.

one of many reasons why it seems this "data" is all over the place.

Check <a href="http://www.adrenalinmotorsport.com/">http://www.adrenalinmotorsport.com/</a> for more info about the shop that houses it.

This isn't exactly what Hamlet had in mind.

The things I find most exaggerated are 'dyno' hp, top speed, fuel economy, and 'personal' size/performance claims, not necessarily in that order.

For example, I had my car dynoed recently. The number at the wheels was 163HP, so assuming a 30% loss (just using 30% because it's a nice round number) which one of these calculations is MORE accurate?

Calculation A:

163 x 1.3 = 212 (assuming 30% loss)

or

Calculation B:

163/.70 = 233 (assuming 30% loss)

But as all the other babble in the thread says, it's a pretty crude approximation.

In addition 30% to too high IMHO. Casting aside all the high-brow math crap I've seen more sane results with:

15% if you have a FWD manual car - for example PES dyno plot for my G2 was on a passat FWD. They used only 15% to get the 295 advertized "shaft" output.

For quattro cars, I've only seen G-tech data (which is a whole 'nother can of worms). With at the wheels data from 240-270 HP. So that's what? A range of 10-20% A coutious G-tech suspicious person might use 20-25% for losses in quattro.

I dunno, all pretty crude. But it seems like believing that there is 30% driveline loss is not supported.

Besides. I like Mike's take on this better: It takes a FIXED HP to run the driveline (again an approximation). And the people who've done stock car wheel dyno tests get, what, 30-50 HP lower than OEM specs. Implying the driveline takes 30 ish HP to run (?) so you just take your wheel dyno data and add 40 HP (?)


- Z

<center><img src="http://pictureposter.audiworld.com/11167/pb172857.jpg"></center><p>This is from a stock 1.8TQ5spd. Peak HP was 104.1HP @ ~5300rpm. With 30% driveline loss, that would be 148.7HP at the crank. The car is rated for 150.

seems to prove that pretty well.

Perhaps, as all the threads below keep observing the percentage thing is going to lead you astray on this stuff.

MikeO would observe that the right way to use your dyno data is to conclude that the driveline takes about 45 HP (~45 ftlbs of torque at 5252 RPM) to run the driveline in this test (including the various polar moment terms spinning up). If you increase the engine power by 2X. The driveline won't take 2X more to operate.

So for the relativly small changes in HP our various mods create, ADDING the difference as opposed to applying the ratio is likely to be more accurate. Do a mod that gets this dyno to 150 at the wheels and you conclude 200 at the shaft, as opposed to 214 (150/.7)

After all your plot is just as good a proof that the driveline subtracts 45HP as 30%. And for big mods, like a PES G2, going with the ratio method gives some pretty untennable results.

But hey it's all approximations.

since dyno testing (especially the numbers at the wheels) is clearly not an exact science, your guess is as good as mine. I trust my sources and I'll stick with what they tell me.

Even if my car IS making 230HP or whatever, it's not like I'm going to print up a bumper sticker with my HP numbers or anything. And it's still a LOOOOOONG haul from a Stage III or PES SC'd 30v.

Mirror's dyno run shows a 45 hp loss, so it makes sense to me that the loss is always going to be 45 hp, as long as your are only modifying the engine, and not the drivetrain.

Someone posted an example earlier that made it all click for me. Here is my own example that makes sense to me:

Imagine you have a car which puts out 100 engine hp. You dyno it and it shows 80 wheel hp. That is a 20 hp loss from engine to wheels. Assuming you want to use the percentage concept, this means it lost 20%.

So now imagine you modified that engine so that it now puts out 200 engine hp. Assuming that same 20% loss, that means you'll get 160 wheel hp. That means you'll now be losing 40 hp from engine to wheels. That makes no sense! You only modified the engine... why would the drivetrain now be sucking up more power?

It doesn't! :)

Of course, this is a very simplistic example, as the previous discussions in this thread have shown that there are many more factors involved. But it's a straight forward enough example to make the point that the percentage loss concept doesn't quite cut it.

I have heard that it is possible, but I'm not familiar with the math involved. I'm assuming it involves the ET, trap speed, and weight of the car in some way?

They build a motor to a set spec for a set hp rating. Lets say it makes 700 hp but then they dyno the car and it is making 500 at the wheels. How do you take off 45hp for that car then since it is still losing its % rate anyway. Since heat is energy, the more energy the drive line takes the more it will lose. The drive line can not just put down double the hp/tq without losing more power then it did when it as running stock hp.

the loss consists of a constant portion, and a portion that is related to loading (hp), and a portion that is related to speed (3rd vs 4th).

the part related to load is shown by heat ...friction at tires, gear meshes, bearings. when at the track averaging 70mph at mostly wot, trans and diff are hot as a pistol. easy cruise at same speed, modestly hot. industrial gear boxes figure 1-2% loss per mesh.

coast down data (if done right, in neutral) does not pick up most of the load dependent losses.

These corrections, for est'd engine hp, do get lower at high hp levels. They are non-linear in nature.

'mirror' plot shows 30% losses. Another stock on another dyno might be 20%, due to diff tires/pressures, gear used, oil type/heat, dyno condition, IC heat soak, etc. But repeat visits to same dyno is good way to quantify % change due to mods, if u keep these other factors consistent.

for cars near stock hp, dyno % loss factors are roughly 10% for transverse fwd, 15% for rwd, 20+% for awd.

btw, standard SAE dyno elevation correction should NOT be applied. this overstates the power at high elevation for turbos.

A constant loss is also an approximation. Just a better one for the low HP and Hp deltas caused by mods we talk about here. 500 and 700 HP engines and drivelines are something else...

It is true, as I pointed out some number of threads ago, that HP loss is more than just a constant term. It includes the friction (static) and dynamic friction terms. Dynamic friction goes up with RPM. And Intertia and viscous terms which go up with the rate of change in RPM

But notice those other terms go up with rate of change in *RPM* not HP. If you double the HP of the car you do not 1/2 the acceleration time on the dyno. But by using a percentage ratio, you are in fact making that assumption.

power is heat is nice to say. But not a mathematic approximation of the situation at hand. For small HP and HP changes, where your 0-60 times are not changning much, the constant offset approximation will do better than a constant percentage as far as accuracy is concerned.

If I used your approach, I'd conclude my car is something like 350 at the shaft, and I just don't believe that.

I'm afraid I don't have some racing buddies to fall back on. Just some MIT degree stuff. So YMMV.

- Z

so at that point (rpm) losses are fairly constant, but they will still increase some as hp is increased.

Of course, overall vehicle performance isn't determined by peak hp, or even peak torque. Some can't see that. Those are easy numbers to bat around, however, so that's what we hear.

IMO: Answers to complex questions usually aren't simple. If a person doesn't understand the mechanics and physics of something, they either have to believe those who seem to, or they adopt the DCWF--MMMU ("Don't Confuse Me With Facts--My Mind's Made Up") approach. That doesn't make them bad people, just difficult to discuss this stuff with. You obviously understand "this stuff".

To be fair, almost all of us have some field in which we haven't a clue. Personally it's classical music, great art, opera and politics among others. I'm mostly lost there. "How things work" is much easier for me.

My $.02

When tuners like APR and GIAC come up with HP numbers for their various products, how do they calculate their numbers? Are they actually using engine dynos, or do they use some variation of the many "opinions" that I've seen reflected in the numerous posts on this thread (all of which are beyond my understanding).

various innacurate approximations used here. Which is quite bothersome. Especially for the smaller "look 10-15 more HP" products.

Decent vendors will have "stock" and "after" dyno plots. So you can at least see the at-the-wheels difference achieved. And if you are so inclined you can divide by 0.7 or add 45 or whatever is implied by the stock versus after plots to conclude shaft HP.

PES posts 295 HP for the G2. The "Stock" plot peaked at 180. The G2 at 256. They used a 15% factor to get to 295. Which is cool. But if the same factor is applied to the stock plot you get 211 HP before modification. And the car was a 99 or 2000 Passat FWD. 190 HP right? Seems like errors in the old "it's a percentage" arguement.

Or you could approximate as the car was 190, and thre reading was 180. So the FWD stick passat took 10 HP to run the driveline. So you could add 10HP to the G2 plot and conclude that kit achieved 266 HP. Or add more based on other people's plots and get higher numbers.

Or you could go back to the percentage appoximation, and go 190/180 = 1.055 (5.5%) and then take the G2 plot and scale it to 270 HP.

So. Is the G2, 270HP or 266HP or 295 at the crank?

People on this board have argued each of the above methods for calculating this.

- Z

It is *extremely* informative. :) It's been very refreshing to see some good technical discussion that hasn't resulted in name-calling and flaming.