Live Dyno

[Fix_Until_Broke]

This started out as an idea quite a while ago from a discussion with Jfettig, but over the last few months, I was able to make it into a reality - Having the ability to monitor engine power on a vehicle, real time, anytime, anywhere. Chassis dyno's are good, but a real pain to take time to do, schedule, etc and they're also relatively expensive if you're doing part/tuning comparisons.

Ok, so how do we do this and not spend big $$ on wheel hub torque sensors or telemery on strain gauges put on the axles or flywheel, etc like the OEM's do? We need torque and speed to calculate power. Given the engine mounting layout on the MKiV chassis, if you remove the dogbone mount, the engine/transmission rotates freely on the two upper mounts. If we could put something in the dogbone that would measure force, then if we know the distance between the upper motor mounts and the dogbone (thanks A5INKY !), we can calculate torque.

Enter in a load pin, used for all kinds of force measuring things, made by a bunch of different companies. Perfect for this kind of application.

So, I took an old dogbone that I had, bought some clevis's and such from McMaster, spent some time on a lathe making everything fit with the OEM bushings and chassis interface and came up with this replacement dogbone assembly that measures force.

Here it is installed in my car. It's been there for a few months now, seems to be holding up fine. No additional noises/vibrations/etc (not that my car is a good reference for that kind of thing).

Ok, so now we have force and with a bit of math we have torque. This torque is not engine torque though since it's reacting the entire driveline through the transmission and final drive. Fortunately, this is all just more math through the transmission ratios to get back to engine torque.

Engine speed is relatively easy - just a sensor on the flywheel (same as I used for my cranking compression tests).

So, after all this, we've got force, engine speed, a bunch of math and then ultimately power. I've got a fancy data acquisition system that does all of this, but it's not required - a very inexpensive 2 channel (1 analog, 1 digital) system would do the same thing, you'd just have to post-process the data in Excel or Matlab or whatever tools you're comfortable with.

So I hooked all this up, and took it out for a test drive and got the below results from a 5th gear pull. I'm not sure how much I trust the absolute values (they seem low to me), but the ability to evaluate changes is more important than the actual numbers anyway.

And for those of you in the metric system - one in NM and Kw

Because pictures are a little boring, here's a video of how the data goes across the screen in one of the configurations

http://youtu.be/46I9IizjiVk

There's still a lot more to do to make it easier to use, more accurate, etc, but I'm pretty happy with it and see a lot of value to come from it as well.

Thanks for reading

 

[Enabled]

Hmm, nice, so this is calibrated to your previous dynos or how?

 

[Fix_Until_Broke]

Not calibrated - just nominal sensor scaling, math on transmission ratio's and a reasonable estimate on the moment from A5INKY's measurements.

 

[BlankThis]

I never knew something like this was possible. Amazing

 

[A5INKY]

Pretty neat idea FUB. I appreciate the out-of-the-box thinking.

Had I know what the measurement was for I would have used a large square to give you a measurement perpendicular to the axis of rotation instead of the ground. Oh well, I bet you were able to calculate the length pretty close anyway.

 

[mk1-83]

nice work really good.

 

[Layerz]

This is pretty damned awesome.

I'm struggling to understand where exactly the load sensor is getting it's load from but will have a better look later.

If I was to throw a question on this, any reason not to of used the but dyno formula way of this using engine speed, speed and weight?

Just thinking with a crank sensor like you already have and gps for speed the variable would be weight this could be run through an Arduino with a Oled screen/2x16 display etc...

I'm setting up an arduino boost gauge atm and thinking about adding EGT and was thinking about adding HP through what I have mentioned above and having a way to switch between full/half/low tank and passenger numbers as a variable.

I guess in summary of my initial question or what I am getting at is would the load sensor provide more accuracy, or just another way to determine the figures?

 

[bhodgkiss]

Nice work! Very neat. Would be interesting to know the dyno compare

 

[Joester]

Wow, this is great! Very impressive and good job thinking outside the box. A couple comments/questions.

#1 - Does this take into account the extra force the dogbone sees from simply accelerating the weight of the engine? For example, if you didnt have a dogbone mount installed and someone hooked up a tow strap to your car and did a 0-60 run, the engine would still sway backwards in the bay (dogbone in compression). This would decrease your apparent torque, and I would imagine by quite a bit, so this might be why your numbers are lower than you think they should be.

#2 - I don't imagine this takes into account the moments created by the acceleration of the crank (and other rotating things) in the block? For example, if you didnt have a dogbone mount and you floored it from idle, the engine would tend to sway forwards (if I'm imagining things correctly). This would increase your apparent torque, but I doubt by much at all.

Hopefully that #1 question makes sense because like I said I think its a big source of error. It might be remedied by a simple F=ma, but that would require more dynamic sensoring. If the engine weighs 600lb and you accelerate at .25g's, then that dogdone will see 75lb of compression assuming the CG of the engine is midway between the upper engine mounts and dogbone.

All-in-all, this is awesome. I love seeing this type of stuff and it's what sets this forum apart from others. It seems that everywhere else its just "will these $5000 ten inch wide rims fit my raceland coilover setup?"

 

[TDIMeister]

Your readings are a little low because I'm guessing you're using the approach of multiplying the load cell reading with the distance to a pivot that gives the reaction torque. This method assumes there's a frictionless pivot. However, you do have other mounts elsewhere so that the pivot is not perfectly frictionless and maybe the pivot is not a perfect line axis on the full range of motion (just speculating on a range of possible factors here). I think this approach also requires knowledge of the suspended mass and location of the CoG of the engine.

It's always questionable to apply fudge factors to make your results correlate but this is what likely needs to be done here, validating your in-car measurement results with those on a traditional rolling road dyno to account for the deviation from the ideal case. Great job though and it's just small beans in comparison to go the rest of the way with the calibration.

I had wondered on your most recent post in your other thread: Are these results with the stock, scroll-type SDI or some other intake manifold?

 

[Enabled]

#1 - Does this take into account the extra force the dogbone sees from simply accelerating the weight of the engine? For example, if you didnt have a dogbone mount installed and someone hooked up a tow strap to your car and did a 0-60 run, the engine would still sway backwards in the bay (dogbone in compression). This would decrease your apparent torque, and I would imagine by quite a bit, so this might be why your numbers are lower than you think they should be.

You can apply data correction from an accelerometer placed in the car somewhere.

It seems that everywhere else its just "will these $5000 ten inch wide rims fit my raceland coilover setup?"

LOL, that's hilarious.

 

[Digital Corpus]

And an accelerometer would help to account for addition loading, positive or negative, due the how flat the road is. I'd be curious to see if the output of the algorithms when doing a pull up a hill.

 

[[486]]

I don't think the force of any engine inclination due to road grade is going to read as anything significant. When you've got a junk dogbone mount the force needed to rotate it along the axis of the motor mounts isn't that great, one hand on the intake manifold will rock it pretty far pretty easily, nothing compared to the force of even pushing a car around, much less moving it around at road speeds.

FUB, love it, such a simple approach to a dyno!

 

[Fix_Until_Broke]

Let's see if I can address the comments/questions

I chose a force measurement for a few different reasons.

1) For accelerometers to work well the road needs to be flat, you need to know your mass accurately and the wind resistance needs to be negligible. With a force measurement, most of those things are transparent. Uphill or downhill does not matter much, wind does not matter, weight of the car does not matter nor if you're pulling a trailer, etc. It's measuring the force that the driveline puts on the chassis.

The dynamic forces are a factor as you mention. The CG of the driveline is below the upper pivot points so the overall acceleration of the vehicle will have some effect on the force measurement (and opposite as you note). Practically speaking though, the runs will be done in 3rd or 4th gear since 1st or 2nd gear will be traction limited and 5th or 6th gear will be top speed/power limited. The accelerations in 3rd or 4th gear should be relatively minimal/consistent as long as there are not major changes in the system from run to run (pulling a 2000lb trailer up a 5% grade with a 30 MPH headwind vs a flat unloaded run for example).

The data in the first post is from a 5th gear run so the overall accelerations are relatively low. Reactionary forces to accelerate the internal components should be seen by the load pin as you say.

The pivot point is not frictionless, but it's pretty close. If you take the dogbone off, the assembly moves on the upper mounts very easily (3-4 inches each way). With the dogbone in, it does not move much - maybe an inch at most so the geometry is pretty stable.

I'm currently running the SDI Scroll style intake manifold.

For me, the most important thing is that it's repeatable, the accuracy is less important as long as it's relatively close. When I had the moment arm at 50cm instead of 35cm the numbers seemed more accurate (150HP/300ftlbf). Maybe it's the dynamic forces? I'll go through the run I've got and make an estimate on what they are to get a feel for their magnitude.

 

[diffas]

What is the advance over the Virtual dyno ? Live data mode?

 

[Fix_Until_Broke]

diffas - what is a virtual dyno?

 

[Digital Corpus]

I think he means using an accelerometer inside a smart phone and an app being run to gather the information.

 

[diffas]

I've been using it for years now for comparison runs. And it's pretty accurate too but sure it's not real dyno.

http://www.virtualdyno.net/

Basically you need few inputs from logs to calculate torq and hp. rpm, time and tps (that is usually 100). Sure you have to know your car weight and ratios and so on.

 

[Joester]

I too have used virtual dyno with success and pretty believable numbers.

I'm always happy to see something done a different way though!

 

[Galactic_Warrior]

Constructive thoughts

A few observations. Not meant to be critical:

1. Is the load pin oriented correctly? These can be orientation sensitive.
2. What is the full scale rating relative to what you are measuring? If below 10% of FS, then you will be operating in a less sensitive potion of the strain gages capability.
3. Finally, have you performed a dead weight calibration on it to see where you are relative to absolute?

 

[Fix_Until_Broke]

Yes, it's oriented correctly - perpendicular to the force applied

The full scale rating is dictated by full torque in 1st gear vs traction limitations - the pin must withstand that force. The pin is either 4500 or 6750 lbf maximum (I believe it's the 4500lb version). So, quick math

400 ft-lbf @ 14 inches is 343 lbf x 3.818 1st gear x 3.238 final drive = 4240 lbf. Now it will not achieve this (statically) due to traction limitations, but I'm not way out of range. In the same run, I did some first/second gear pulls where the tires were spinning - I'll look at the data and see if I can determine a traction limit with my tires and LSD. Maybe I could get by with a 3000lb pin and improve resolution?

Now 5th gear load on the pin would be 343 x 0.970 x 2.615 is still 870 lbf on the pin (at 400 ft-lbf which I don't believe I'm at). A 4th gear run would be 343 x 0.972 x 3.238 for 1080 lbf which is better, but will have more inertial effects as discussed above. I have not done any work at trying to quantify that yet.

I have not calibrated the sensor with a known loads - just nominal scaling. The zero was off a bit so I adjusted that so it read zero in neutral on level ground after going forward/reverse a few times. Same slope, just slid zero to match the sensor.

Thanks for the feedback/questions

 

[Fix_Until_Broke]

I looked at the data and saw a peak force of ~2250 lbf so I could probably go down a range in pin size (though that would require new mounting hardware).

 

[JFettig]

In an older e-mail from you, you mentioned it is a 4500lb sensor.

 

[Galactic_Warrior]

I may not have explained my comment correctly regarding the load pin orientation. if you look at the pin from the end (looking at a circle), there is typically an arrow etched on the body of the pin to indicate radial orientation.

There is also (typically) a slot or groove where a clip or key is placed to prevent rotation (when these are used in scales).

I state this, because if the pin is rotated off measurement axis, it can effect the reading. Typically if it is off by 5 degrees this is about an 8% error.

 

[Fix_Until_Broke]

No, I followed you correctly (I probably didn't explain well either )

 

[Enabled]

So lets say you accelerate from 30mph to 80mph in 10 seconds in 4th gear. There's an average of 0.22G force (or roughly 2.2 m/sec^2) of acceleration force totally unrelated to your engine torquing force that you're measuring, but your sensor will sense it in addition to your engine torque.
That will throw your dyno results off.

 

[Fix_Until_Broke]

#1 - Does this take into account the extra force the dogbone sees from simply accelerating the weight of the engine? For example, if you didnt have a dogbone mount installed and someone hooked up a tow strap to your car and did a 0-60 run, the engine would still sway backwards in the bay (dogbone in compression). This would decrease your apparent torque, and I would imagine by quite a bit, so this might be why your numbers are lower than you think they should be.

So, I did some quick math on the acceleration in the first post

mph sec
53 34.8
81.87 45.4

28.87 MPH in 10.6 sec

12.9060448 m/sec
1.217551396 m/sec2
0.124113292 G

So ~1/8th of a G overall acceleration

Does anyone know how much the complete engine and transmission weigh? I'm guessing 1000lb/454kg ?

So, on one end of the spectrum, if the upper mounts are on the horizontal Cg axis, it won't matter because there's no inertial moment.

On the other end of the spectrum if the dogbone is reacting the whole 454kg x 0.125g = 53N (~12lbf), but it will be quite a bit less than this because the dogbone is not reacting the full 454kg, just the ~15cm (guess) moment from the Cg to the pivot point.

I'll have to break out a book to do the math on the above, but I know enough to say that it will be quite a bit less than 53N/12lbf out of 391lbf which is 3%

This factor will be larger in lower gears (higher acceleration rates) and less in higher gears (lower acceleration rates).

A chassis dyno won't account for it since the vehicle is not moving (though it will correlate power and torque in the same static conditions).

I think that uphill vs downhill will not effect the accuracy (other than differing rates of acceleration) since gravity will be effecting the chassis and driveline the same.

I think Joester's idea of pulling the car with another vehicle at different rates of acceleration would be best to empirically quantify this factor. I can get acceleration rate from the vehicle speed sensor and record load pin force vs acceleration. Once I have this for a few different rates of acceleration I can factor in a correction factor to the load pin based on the overall vehicle acceleration (which I'll know from the engine/VSS speed signals). This will then account for different acceleration rates in different gears.

 

[Enabled]

So, I did some quick math on the acceleration in the first post

mph sec
53 34.8
81.87 45.4

28.87 MPH in 10.6 sec

12.9060448 m/sec
1.217551396 m/sec2
0.124113292 G

So ~1/8th of a G overall acceleration

Does anyone know how much the complete engine and transmission weigh? I'm guessing 1000lb/454kg ?

So, on one end of the spectrum, if the upper mounts are on the horizontal Cg axis, it won't matter because there's no inertial moment.

On the other end of the spectrum if the dogbone is reacting the whole 454kg x 0.125g = 53N (~12lbf), but it will be quite a bit less than this because the dogbone is not reacting the full 454kg, just the ~15cm (guess) moment from the Cg to the pivot point.

I'll have to break out a book to do the math on the above, but I know enough to say that it will be quite a bit less than 53N/12lbf out of 391lbf which is 3%

This factor will be larger in lower gears (higher acceleration rates) and less in higher gears (lower acceleration rates).

A chassis dyno won't account for it since the vehicle is not moving (though it will correlate power and torque in the same static conditions).

I think that uphill vs downhill will not effect the accuracy (other than differing rates of acceleration) since gravity will be effecting the chassis and driveline the same.

I think Joester's idea of pulling the car with another vehicle at different rates of acceleration would be best to empirically quantify this factor. I can get acceleration rate from the vehicle speed sensor and record load pin force vs acceleration. Once I have this for a few different rates of acceleration I can factor in a correction factor to the load pin based on the overall vehicle acceleration (which I'll know from the engine/VSS speed signals). This will then account for different acceleration rates in different gears.

Yes, exactly, so that's why I said earlier, if you were to log data from an accelerometer, and subtract that value from the force on the dogbone, you would have live correction factor.

If not, the more powerful the car is, the more the results are skewed using the current method.

 

[Fix_Until_Broke]

Enabled - I'm not sure I follow you? Where would you put the accelerometer and would you take the data with the car powering itself or being powered?

 

[Enabled]

What sort of raw data does your dogbone sensor provide? Force in what units?


You have a ton of options here's an easy battery powered bluetooth one http://www.blueradios.com/hardware_sensors.htm to stream to your phone, computer, etc..

also you can probably find one that logs internally, then you download the data...



As far as an accelerometer, you can put one anywhere in the car to read the axis that the car is traveling. It will be recording the total force of acceleration that is applied to everything in the car, due to the car accelerating.

That force of total car acceleration is added to the force your load pin is reading from the centrifugal force on the dogbone mount. Your load pin is reading the force of car acceleration (plus the engine torque force) because it's in the same axis.

So you need to subtract the force of car acceleration from the force read by your pin in order to get ONLY the torque force from the engine.