The dyno guy used two pickups on my car - "to elimate noise" he said...

Worf - your idea of the car "hoping" on the drum a bit might have merit - that would account for the "smoother" lines we got in 3rd gear, and that dyno operator saidf this happens to others in 4th gear too.

The straps on the dyno are long and hold the car in position laterally - fore and aft. They do not have any straps holding the car down...

When I dyno'd my wifes stock 87 AT the lines were very squigly on the print out, just like the 90GT. When I open the data file with the DynoJet software they are smooth. I wonder why that would be? Could it be something in the software at the dyno?
Soon as I put my hands on the printout from the dyno I will post it for you to see the comparison.

Good! Then the RPM measurement should be solid - especially if each of the pickups was hooked to a wire from a different coil.

Perhaps. It all depends upon how 'smart' the DynoJet's engineers were. If the rear end is hoping it MIGHT cause at certain speeds a harmonic vibration that MIGHT effect the velocity measurement IF the sample speed is in phase with the harmonic and IF the number-crunching software is dumb.

Hmmm... that's very interesting. I'm going to think about that for a bit.

I suspect that the DJ's measurement system records many thousands of data points during a run. These data are then crunched by the software - reduced, filtered, smoothed, and presented. I suspect that a lot of options are available in the DJ software (and that the operators have a wide range of facility (or lack of) in its use.)

Without a doubt.

OK. I'm interested in putting this data together so that we can attempt to make some conclusions - or at least formulate some more-directed questions. My inkling at this point is that there is important information that is being effectively hidden from us by the disparate presentation methods available from the DJ software or information that is in evidence that we cannot properly interpret because we're not sure what the DJ does with the data it collects.

Folks have speculated (or perhaps I interpreted their remarks as speculation) that the squiggly lines are - in effect - a measurement abberation that can be safely ignored. Given the conceptual simplicity of the DJ - and the fact that measurement science in software is what pays for my 928 habit - I want to understand more about what's going on in the guts before ignoring what DJ operators have consistently indicated to me is a 'knock sensor' curve.

Carl's comment above - that the operator said something to the effect of 'we see this with cars in fourth gear' - makes me wonder if it isn't the cars that are hoping but the DJ itself? If the drum is rolling smoothly and the rest of the DJ is in proper working condition why would cars produce squigglies in fourth gear? They all have rear tires out of balance? Or bad rear wheel bearings? Occam's Razor points to the DJ and not fourth gear. Why would we see oscillations in actual rear wheel torque only in 4th gear from a number of cars?

Is there are problem with the DJ that was used? All metrology equipment (well, equipment with moving parts, or software or A/D convertors, etc. - anything more complicated that a ruler...) requires care and feeding. If it doesn't get it the equipment will very likely produce erroneous measurements. What PM (preventative maintenance) does the DJ require? Periodic replacement of bearings? How about periodic calibration? Is there a self-diagnostic that might identify sensor faults?

Hey? O&N? You there? Ask the DJ folks about PM if you get 'hold of them...

I have a message into them, waiting for response.

The WinPep software has a "filter" setting (1-5) which smoothes the line out. Obviously a common phenomenon.

I doubt they would change the bearings on the 1200 lb. drums very often, and consider how many miles our wheel bearings go in between replacement. Once I get their ear, I will ask them about PM and calibrations; I'll also ask the operator about these things next time I go over there. I have a feeling the only things they do are software calibrations.

He would not give an answer without a dyno file - I sent him one of mine. Update forthcoming.

Mine is pretty smooth compared to the one in question, though. Maybe Carl, or anybody with a bumpy chart please forward it to Will@Dynojet.com & ask him why it looks that way...

OK - I just emailed my chart to him.

I did some digging...
 

I did some 'net digging and came up with some more interesting info. Nothing yet conclusive - but the tale gets even more interesting.

This is a link to a thread on an Evo forum:

http://forums.evolutionm.net/showthr...ojet+artifacts

(There's some cruft on the first page. The second page gets meaty.)

So far, I am becoming confident that there are noise issues with the inductive RPM pick-up and since the pick-up's output is used in the DynoJets filtering (math crunching) a bad (noisy) signal can induce squigglies. More disturbing is the assertion of one tuner. He specifically programmed a torque dip in the ECU of a turbo Miata. The torque dip was indistinguishable from other noise in the most raw data. "Smoothing" of the data with the WinPEP software erased the actual torque dip. Lastly there's speculation - maybe - that the DynoJet's means of measurement makes it difficult to separate noisy RPM signals from knock-sensor system induced timing shifts. Ruh rooo...

(I am still absorbing though - and reserve the right to refine my thoughts...)

Excerpts:


Question:
Before we get into this any further, could you please answer the following question regarding Dynojet resolution and accuracy:

What is the cause of those jagged, seemingly random, high frequency dips and valleys that we see in just about every dynojet graph (especially when the "smoothing" feature is set to the minimum)? Does wheel output actually jitting up and down like that during a run? Or are we seeing, among other things, a measuring/sampling artifact inherent in the way the Dynojet measures output? And why should there be a smoothening feature to begin with? We're talking a high-precision measuring device here, right? Kinda defeats the purpose of measuring something if your connecting the dots with a "best fit" curve afterwards.


Answer (Sort of...)
But just so you don't start accusing me of being evasive again. Yes it is measuring/sampling artifacts inherent in the way the Dynojet measures output. I believe most of that is due to slight inconsistencies in the engine RPM pickup. It is a minor variation that does not introduce large errors in the data. Systems like yours do not rely on the actual engine RPM for the calculation, that could explain why the power/torque peaks come out wrong occasionally.
Just because you cannot adjust it does not mean there is not an internal software filter on your system also. I would be very suprised if there were truly no filters, not to mention that I have seen some of your plots looking a little jagged in spots also.

Some more:

Quote:
But just so you don't start accusing me of being evasive again. Yes it is measuring/sampling artifacts inherent in the way the Dynojet measures output. I believe most of that is due to slight inconsistencies in the engine RPM pickup.

Nope. Those artifacts exists without the tach pick-up as well (that is, just going off of wheel speed).
Quote:
It is a minor variation that does not introduce large errors in the data

When you can't distinguish between misfire or pick up the exact RPM point of knock sensor activity, this "minor variation" (and when is a 5ft-lb dip minor?) causes problems. There is no excuse for sampling noise in a measuring device. Especially when it interferes with tuning or diagnosis. But it can lead to higher numbers since it tends to elevate the peaks a bit.
Quote:
Systems like yours do not rely on the actual engine RPM for the calculation, that could explain why the power/torque peaks come out wrong occasionally

No. The Dyno Dynamics Dyno does rely on an inductive pick-up for computation. Of course, it offers 3 levels of pick-up sensitivy which makes means that you can get it to work with every ignition system (coil on plug, plug wires, etc,.) and not suffer from the same break-up problems you see with the Dynojet pick-up. For cars that have their ignition system buried under engine covers (making them hard to get to), you can get an estimated engine speed by going off of wheel speed and a gearing factor. When doing this, you do lose a little bit of accuracy since it does not account for tire deflection (and the resultant rolling diameter changes) that occurs during the power run. If the power/torque peaks come out wrong, it is because the user didn't calculate the gearing factor correctly. It's quite easy to do, in case you're wondering. Simply limit wheel speed to, say 50mph. Go WOT and look at the XEDE or a scan tool to see see your engine revs. In the case of the EVO with factory wheels/tires, it will go to 4600rpm, giving a gearing factor of 92RPM/MPH (4600/50). It's hard to mess it up. But it's possible, I guess.
Quote:
I would be very suprised if there were truly no filters

I did a demonstration last year at a dyno day showing the resolution of the dyno for those who were used to seeing jagged and noisey Dynojet results. I mapped a TEC3'd turbo Miata to have have a big ignition advance dip during a very small (10RPM) span of engine speed. Basically, I mapped in a 4000, 4010, 4020 and 4030 RPM break points. Between 4010 and 4020RPM, the ignition advance dipped 5 degrees. On the dyno, there was, as one would expect, a big 15ft-lb dip in torque during that narrow interval. I did this test on a Dynojet only to find that dip was nearly indistinguishable from the general noise. And smoothening out the curve erased that dip entirely. So contrary to what you said earlier, this noise and sampling errors do induce a significant error in data.

FWIW, this is why it takes me so much longer to tune a car with a Dynojet since it's hard to pin-point the *exact* point of knock correction activity.
Quote:
, not to mention that I have seen some of your plots looking a little jagged in spots also.

And I do my best to tune that out

Some other tidbits:
 

A few more items linked below.

Summary:

- The FI systems of some cars will result in a leaner-than-desired mixture if tuning is done on a an inertial Dyno.

- The DJ uses an automatic means of gathering ambient data for use in calculating 'correction factors.' There are several possible correction factors (e.g. SAE, DIN,...etc.) and the metrology engineer in me knows that the ambient data measurements could also be a source of measurement noise.

- More issues with the RPM measurement.


http://www.germanmotorcars.com/Dyno_...0inertia_1.htm

Problems with Inertia dyno test procedure and fuel injected vehicles: A Sweep Test (hold throttle wide open and sweep from low rpm to high rpm) will often trigger the Acceleration Fuel Map, along with the Main Fuel Map, causing the fuel mixture readings to indicate dyno operator that the motor is overly rich. This would cause the tuner to lean out the main fuel map. Of course, in the real world, upper gears, the acceleration rate of the engine is much slower than what they tested, doesn't trigger the Acceleration Fuel Map, and the engine ends up a lot leaner in reality in top gear. It's not that common of a problem, since most people never drive that fast for that long to cause engine damage. Work around: Tune full throttle fueling in real world usage at dragstrip (to best trap speed) or in Steady State Mode on different dyno.



http://www.idavette.net/hib/dynojet/

During manufacturing, Dynojet Research figures the mass equivalent of each pair of drums to four places and bearing drag to five places. Those proprietary figures are figured into the computation the dyno computer makes. If the mass equivalent of the drums is known, the drum bearing drag is known and the rate at which a vehicle’s drive wheels accelerate the drums is accurately measured; then the "thrust force," in pounds, at the rear wheels can be computed with a high degree of accuracy.


In the real world, the measurements and computations are not quite that simple, but the complex methods Dynojet Research uses to apply these laws of physics and their mathematics to accurate measurement of rear wheel power is a proprietary secret.

This CPU and additional peripherals are called the "DynoWare EX+". It consists of: the CPU Module, a Dynamometer I/O Module (sends and receives data from the dyno and its hand-held controller); a RPM Module, (processes RPM data from an inductive pick-up clipped on a plug wire or other engine speed sensor) and an Atmospheric Sensing Module (measures absolute pressure, air temperature and humidity and transmits that data to the CPU). These four modules stack together, allowing easy upgrades or expansion with options such and Air/Fuel Ratio Module, for exhaust gas analysis, or an Optical Pick-up, for engine speed input. Additionally, WinPEP supports both the Pi and CDS data acquisition systems popular in motorsports which allows graphing of parameters such as boost pressure, fuel pressure, intake air temperature (IAT) and spark advance against the power and torque data.

Always ask what correction factor was used to derive the data when judging various performance products on the basis of dyno test results. For example: a car I tested recently for Vette Magazine generates 412rwhp, SAE and EEC corrected, but the same car puts out 423 rwhp, standard corrected and 425rwhp, DIN corrected.


One complaint we hear is about the RPM pickup. It seems to cause some users problems with waste-spark ignition systems. For that, Dynojet suggests the user make sure the dyno computer is grounded to the vehicle and that it’s set to read spark data every 360° of rotation rather than every 720°. The RPM pickup is also incompatible with newer ignitions that have no spark plug wires or short, inaccessible wires. The optical RPM pickup is offered as a solution for that but when Jim Bell interviewed for this story, he told us both optical and magnetic pickups are time-consuming to set-up such that they provide accurate data. Clearly, a more simple way of setting up and obtaining RPM data is something Dynojet Research needs to look at.

And lastly....

The winPEP software can produce more raw data - measured drum velocity, uncorrected HP data, etc. All very useful in finding a noise source.

I asked:

"Most charts I have seen exhibit this type of "sqiggle", and I thought you may have a general answer to it.

Attached is a run file from my last session which exhibits some of the behavior in question, but not as extreme as the first chart I sent.

I'm not looking for a specific answer as much as some ideas about what *might* cause this type of behavior."

I received this response:

""Squiggles" can be cause by a few things, such as timing, detonation, rear tire imbalance and even air flow issues. With extreme jaggedness, you can perform a coast down/negative horsepower test to rule out any rear tire imbalance. This test is performed the same way as a roll-on except that when you roll off the throttle you keep sampling and place the vehicle into neutral gear and let it measure the coast-down. You will have to choose "display negative values" in the graph options menu in order to view it."

(He suggests that actual physical conditions may cause the erratic chart reading.)

Again, my chart's sqiggles have much less amplitude than Carl's, so let's see what reply Carl gets...

Carl, did you notice any out of balance condition?

Update - just noticed that I got the answers from a guy in the "Marketing & Sales" department. ...may affect credibility.

Nice work. I can't wait to get my hands on a 32 valver for twins. By comparison my 4.5 liter 16 valve made the same torque on 12 psig of boost and only 30 ft-lbs less on 8 psig of boost. True proof of the turbos ability to make massive mid range torque.

I'm sure that car is a fun ride now.

John,
For comparison with my 32 valve S4:

Your car at 8psi = 301 HP 353 TQ
My S4 car at 8psi = 463 HP 442 TQ = +162HP +89TQ

Your car at 12psi = 366 HP 382 TQ
My S4 car at 12psi = 556 HP 495 TQ = +190HP +113TQ

Just goes to show what 16 more valves and .5 liters can do for you.
You need to get your system on a 32 valve car, should make for some interesting results.