1 2
oldeskewltoy Dork
3/3/13 9:08 a.m.

Where is a face in palm when you need one......

its the operator...... a tool is a tool.....

HOW IT IS USED is the difference.

erohslc HalfDork
3/3/13 10:25 a.m.
Keith Tanner wrote: The 5252 cross only happens with ft-lb and HP. Not "etc" Use different units and they'll cross at different points - or not at all. As long as you have RPM and torque, HP is a pretty simple calculation. The most common reason to see the lines not cross at 5252 is if the torque and power lines are using different scales. Happens more than you'd think for some reason.

1) I said 'ft-lb and HP, etc.' because those are the 2 convenient abstractions commonly used for the fundamental physical units, the definitions of which cause the 'magic' 5252 number. The 'etc.' implies different pairs of units. So of course in the MKS system, it's a different number (sorry, I don't know what it is). And if we used 'stone' and 'slugs' and 'cubits' and such, it would be a different number.

2) I specifically avoided the phrase 'where they cross', and said '...and the two numbers are different at 5252 RPM ...' for exactly that reason. It's visually convenient when the scales are the same, but not neccesary.

3) Yes, it's a simple calculation. The trick is how it gets implemented it in the real world.

Steady state operation, you can pretty much do the math in your head.
But a common dyno use case is to sweep the RPM, capture the results, and express them as a graph. Hundreds of data points and calculations, maybe thousands, performed over the span of a few seconds.

I'm a bithead, so depending on the language and platform I was writing the code in, I'd simply implement the formula using the values from my data acqusition framework.
But there are issues there, with sampling rates, signal filtering, synchronization, aliasing, latency, scaling, etc.

And what about the 'old school' dyno's, they had to do the calculation with analog, or even mechanical techniques.

We'll not even consider the correction factors like temperature, baro pressure, humidity, that have to be applied to get a number that's suitable for quantitative comparisons over time.

But I bet you knew all that, so I'd say that fundamentally we are in agreement.

Keith Tanner
Keith Tanner GRM+ Memberand MegaDork
3/3/13 1:48 p.m.

Calculation speed isn't an issue - for one, we're not talking about spectacular resolution here. 10 Hz would probably be more than enough. If you'd like, I can check to see what the rate of acquisition is on our dyno.

Depending on the dyno, you could be measuring torque directly (as we do with ours) or calculating it via the acceleration of a known mass (as with a big Dynojet drum). BTW, a typical Dynojet run of a 250 rwhp car takes about 12-15 seconds, to get an idea of what sort of time span we're dealing with.

So that's the first part, get a good torque reading. Once all that's done, you apply the simple formula to get a HP number. Implementation isn't an issue there, so the fact that HP is calculated and not measured is irrelevant. You can try to make it sound complicated, but it's not really.

Correction factors are a different discussion. Luckily, there are standards.

The primary purpose of dynos is to give people on the internet something to argue about, or so it seems.

wbjones UberDork
3/3/13 3:00 p.m.

+it can get you a re-classification for TT/PT competition with NASA

Jerry Reader
3/3/13 3:10 p.m.
Keith Tanner wrote: The primary purpose of dynos is to give people on the internet something to argue about, or so it seems.

Luckily my 2 questions got answered before any argument started.

erohslc HalfDork
3/3/13 3:53 p.m.

I agree that dyno's are too often used for bragging rights, rather than as analytic/diagnostic tools.

But the fact is, underneath, it IS complicated.

Let's look at your 10 Hz acquistion rate:

Assume a 2K RPM to 8K RPM in 6 seconds (Yes, I'm lazy, it makes the math simpler). So average delta RPM is 1000/second.

At 10 Hz sample rate, that's nominally 100 RPM per interval delta.

But wait, maybe it was 99 (or 98), and maybe it was 101 (or 102). So we have an error band of 1%, 2%, maybe 4%, depends on the implementation.

Depends on how RPM was determined. There are two primary methods, both depending on precision timeclocks:
1) Count the number of pulses in a precise clock interval (from say the edge of a magnetic or optical crankshaft sensor pulse).
2) Count the number of precise clock intervals between pulses (from say the edge of a magnetic or optical crankshaft sensor pulse).

These may appear to be the same, but in fact they can give very different results. What happens when the clock pulse edge and the sensor pulse edge occur close to, at the same time, or overlapping one another? It's possible to minimise the error, you cannot ever eliminate it.

Same can be said about torque measurement.

Assume that we have a load cell, which yields an analog signal. That signal will normally be converted into an equivalent digital value, which makes it very easy to use in calculations. Do we know what the open loop gain, the voltage offset, the input leakage current, etc. of the op amps used in the analog to digital converter? Do we know how many bits it can resolve, linearity, temperature drift, settling time?


We do know that precision and accuracy costs money. And manufacturers, particularly in a small volume segment like dyno manufacture, will seldom spend more money than needed to be competitive.

In calculation of HP, the errors of RPM and torque are multiplicative, not additive. So 2% on RPM and 2% on torque yield more than 4% error in the result.

If the errors are consistent, maybe that's OK, although one might be tempted to attribute a 4% HP gain to 'Super Swirl Electric TurboCharger Air Filter' or 'Bundle'O'Snakes Multiprong Spark Plug Electrodes'.
I'm sure that the results you get are good enough for what you need, for what 99% of us need.

I'm just suggesting that we don't lose sight of where the numbers come from, how they are generated, and the kinds of hidden innacuracies that thay contain.

As others have said, it's tool. Know your tool, know it's strengths and weaknesses, and use it accordingly.



erohslc HalfDork
3/3/13 4:02 p.m.
Jerry wrote:
Keith Tanner wrote: The primary purpose of dynos is to give people on the internet something to argue about, or so it seems.
Luckily my 2 questions got answered before any argument started.

I hope we are not arguing.
I'm sharing some insight and life experience, same as the rest.

I've read and appreciated Keith's posts for quite a while, I'll listen to anything he has to say.

Others may have different views, and I'm very OK with that, too much to learn in this life, and not enough time.

If I cross any lines feel free to tell me to STFU.

(I may choose to ignore you, but feel free).


Swank Force One
Swank Force One MegaDork
6/10/13 10:21 a.m.

Just to share another Dyno Dynamics vs. other dyno story...

Got my MSM re-tuned on Saturday. Previously made 245whp/238wtq on FM's Rototest.


Normally i'd be mad about this, but the car is SUBSTANTIALLY faster than it was previously. I still DO find the numbers pretty weird, particularly the torque vs. hp numbers, but the way the car moves is a night and day difference.

Will be going back in August to test a flex-fuel and ECU-controlled EBC setup. (I think the HP numbers were affected by a very mediocre MBC.)

1 2
Our Preferred Partners