Dynamic Characterization Explained

The whole thing started when I first used an injector that was not listed in Motec's injector list, and so I did not have correct dead times, or current settings to program into my ECU. It was evident that this was important information, but there was no place to go to get the info. I stumbled around experimenting with different settings until it seemed more or less OK, but in fact this experimenting raised more questions than it answered.

With prompting from Motec USA's engineering department, and the willingness to lose some sleep, I designed and built our current test equipment. It wasn't exactly easy, and took nearly 6 months, but it turned out to be well worth the effort.

This article describes what it's all about.

A Fuel Injector is a Dynamic Device

When a fuel injector is doing its job on a running engine, it is not held in the wide open, or static position. It is being cycled on and off several times per second. To put this into perspective, a 4 stroke engine running at 9,000 rpm will cycle the injector on and off 75 times in one second. A 2 stroke, rotary, or batch fired 4 stroke engine at the same rpm will cycle the injector on and off 150 times in one second.

To take it a step further, consider this. Using the first example of a 4 stroke engine running at 9,000 rpm, we have only 13.3 milliseconds (13.3 thousandths of a second!) to open the valve, deliver fuel, and then close it again. If we want to give the injector some breathing room, and run it at say 90% duty cycle, we now have only 12 milliseconds to complete the operation.

This is no small feat, and as you probably guessed, the injectors response under these conditions is far from perfect.

It is this imperfect response that leads us to terms like dead time, recovery time, minimum repeatable pulsewidth, linear operating range, etc.

Dead Time

Oops! Did I just say that? Me, the Motec fan/dealer/nothing is better than Motec snob?

OK, enough BS. It is a misleading term, and you will see why if you read far enough.

As I said, these terms all describe the same thing. That is the injectors inability to respond instantly to the signals coming from the ECU.

When the injector receives the signal from the ECU a few interesting things happen. First, the coil must build a magnetic field strong enough to move the valve. Second, the valve must move anywhere from 10 to 30 thousandths of an inch to reach the fully open position, at which point its flow rate equals that of the static flow rating.

The time wasted charging the coil, and coaxing the slow moving valve to its fully open position reduces the flow, and has the same effect as shortening the pulsewidth. The end result is that a 1000cc injector does not flow 500 cc's per minute at 50% duty cycle as you would expect.

But that's only half the story. The same delays exist when we remove the signal from the injector. First, the coil needs to discharge to the point that the magnetic field collapses and "lets go" of the valve. Second, the valve has to move from the fully open to the closed position. These delays cause the injector to continue to flow after the signal has been removed, and has the same effect as lengthening the pulsewidth.

When we combine the opening and closing delays, and consider their total effect on dynamic flow, we have the dead time.

It is worth noting at this point that dead time is not something you can determine with an oscilloscope. You must measure the dynamic flow where all delays are combined to determine the dead time accurately.

I'm guessing your eyes are glossing over, and you're ready for some pictures. Since humans are basically visual beings, I believe that the picture below tells far more than I could possibly get across in boring white text.

The chart above shows the flow vs pulsewidth curve for an Injector Dynamics 1000cc injector. (The actual flow is 1020 cc's per minute) The test was performed at 100 Hz (Injector is cycled on and off 100 times per second) This is equivalent to 12,000 rpm on a sequential fired 4 stroke engine, or 6,000 rpm for a 2 stroke, rotary, or batch fired 4 stroke engine.

This is a convenient test frequency, because 10 milliseconds equals 100% duty cycle, 5 milliseconds equals 50% duty cycle etc.

The first thing to note is that at 10 milliseconds the flow roughly equals 1000 cc per minute. Between 9.5 and 10 milliseconds, the injector is in its "semi static" flow range. It's response here is determined by a combination of the injector characteristics, and the characteristics of the drive circuit in the ECU.

Since the injector will (hopefully) never be run in this range, we don't need to worry about it.

The important point to consider is that at 50% duty cycle, the injector is only flowing 414 cc's per minute. Those of you well versed in "internet injector math" will be scratching your head wondering why the injector is not flowing 500 cc's per minute as the online calculators say it should.

The answer is DEAD TIME

As described above, this is the result of delays in the electrical and mechanical response of the injector which cause its dynamic flow to be less than it should be.

Because of the dead time which equals .972 milliseconds, the injector is only flowing the equivalent of a 4.028 millisecond pulsewidth. (5 msec minus dead time of .972 msec equals 4.028 msec)

The first thought to pop into your head is probably "WHO CARES, I'm going to tune the map to achieve the correct air fuel ratio, and I will just increase the pulsewidth to get the flow that I need."

If that's what you're thinking you're absolutely correct, but as usual there's more to the picture.

Take a look at the flow at 3 milliseconds, and at 6 milliseconds. At 3 milliseconds the injector flows 207 cc's per minute, and at 6 milliseconds the injector flows 516 cc's per minute.

What we have here is non linearity due to the dead time of the injector. We doubled the pulsewidth, but instead of getting 2 times the flow, we got 2.49 times the flow, for a whopping 24.5% error!

Again, you could ignore this because you will be tuning the fuel map to achieve your target air fuel ratio...but there's more to the picture.

The rest of the picture is all of the compensations in the ECU. In a typical configuration we have air temperature compensation, barometric pressure compensation, manifold pressure compensation, etc. All of these compensations are based on the assumption that fuel flow will change in proportion to pulsewidth.

If the fuel flow does not respond in proportion to the pulsewidth we can get into trouble very quickly. Let's consider how this might affect our air density compensations. If we build and tune a motor at sea level and then race it to the top of Pikes Peak, the air density will decrease as we climb the hill. If we are to maintain the correct air fuel ratios the ECU must reduce the fuel flow.

The ECU will attempt to do this by reducing the pulsewidth. So let's say the air density at the peak is 20% less than when the motor was tuned. The ECU will reduce the pulsewidth by 20% (If your compensation tables are correct.) but the fuel flow will be reduced by a much greater amount.

The result is that the engine runs lean. If you're lucky, you will just be down on power. If you're unlucky, you will run over your own crank shaft.

This non linear response affects nearly every aspect of your tune. Even some you may not have considered. Take closed loop lambda control for instance. Many people don't use it because it hunts around, never getting the air fuel ratio quite right. Considering the chart above, how could it?

If the engine is 10% lean, the ECU will increase the pulsewidth by 10%. Instead of getting a 10% increase in fuel flow, you may get 13%. Now it's too rich, and the closed loop control has to swing back in the other direction. Since our fuel flow is not proportional to our pulsewidth, the system will hunt for quite a while before it gets it right, and by then you have probably moved to a different point on the map and the required correction is entirely different.

As you can see, the dead time of an injector can cause quite a bit of trouble. UNLESS we deal with it properly.

Once again, a picture tells the story best.

What you see in the picture above is the same flow vs pulsewidth chart, but with dead time compensation added. Note that the X axis is now labeled effective pulsewidth.

In the ECU, the pulsewidth is calculated based on the fuel table, all the compensation tables, and then finally the dead time compensation is added for the final pulsewidth which is sent to the injector.

Every manufacturer has their own terminology, but in Motec speak, the initial calculations result in the effective pulsewidth, and once the dead time is added, it is called the actual pulsewidth. (In this case, Motec's terminology is perfectly descriptive.)

Chances are the terminology is different in your ECU, but all ECU's will make a distinction between the effective, and the actual pulsewidth regardless what it is called.

Note: Some ECU's will only report one of these two values, and it is important to know which value you are seeing.

Remember in the first example we had a 5 millisecond pulsewidth, but the injector was only flowing the equivalent of a 4.028 millisecond pulsewidth? So what if we just added our dead time of .972 milliseconds to our pulsewidth calculations?

The answer is, linear response.

As simple as this is, it is also somewhat confusing, so let's break it down. Let's say that based on the number in the fuel table and all the internal compensations we have an effective pulsewidth of 6 milliseconds. The actual pulsewidth will be effective pulsewidth, plus the dead time of .972 milliseconds, for an actual pulsewidth of 6.972 milliseconds. Looking at the chart which is based on effective pulsewidth, the flow at 6 milliseconds is 614 cc's per minute.

As was stated earlier, 100hz is a convenient test frequency because 6 milliseconds equals 60% duty cycle, etc. In a perfect world, a 6 millisecond pulsewidth should result in 60% total flow. In this case, 60% of a 1020cc injector equals 612 cc's per minute, and our actual flow is 614 cc's per minute for an error of 3 tenths of a percent. Not bad eh?

Looking more closely you can see that 2 milliseconds equals approximately 200 cc's per minute, 3 milliseconds equals approximately 300 cc's per minute etc. More importantly, the flow at 3 milliseconds is 308 cc's per minute which is almost exactly half the flow of 614 cc's per minute at 6 milliseconds. Double the pulsewidth equals double the flow, and our fuel flow is now directly proportional to our effective pulsewidth!

Our compensations will now work accurately, and all this was accomplished by simply adding the dead time of .972 milliseconds to our effective pulsewidth calculations.

If all of that sunk in you are probably thinking "Wow, I get it now" But guess what...there's more to the story. Remember I said that Motecs term "battery compensation" was misleading? Notice that we haven't even considered battery voltage yet?

As you can see from the example above, dead time compensation is required even if the battery voltage remains constant. Because the dead time varies with battery voltage, we need to provide dead time compensation values at various voltages to insure that our fuel flow is linear regardless of battery voltage.

Now it's time to grab another beer, a comfy chair, and let this sink in. Just to be sure, I'll say it one more time.

Dead time compensation is required even if the battery voltage remains constant!

It's now quarter to four in the morning, and it's clear that I will not get this article finished tonight.

So...In the next few days I will detail the test process, and describe overall linearity, recovery time, lower non linear operating range, and a bunch of other things that you may not even care about.

Maybe I'll even find time to proofread and correct typo's!

OK, so it's nearly noon and I've had enough coffee to jolt my system. Here's a little more info on dead times before I launch into the next rant.

The delays that cause dead time are the result of many variables.

The electrical delays are affected by the inductance of the coil, and the voltage applied to the coil. Since we can't change the inductance of the coil we don't need to worry about that. What we do need to know is that as we increase the voltage applied to the coil, it reacts faster, and the dead time is decreased. As a result, the dead times will always be higher at lower voltage. This is why our ECU manufacturers ask us for dead time compensation values at various voltages.

On to the mechanical portion of the picture, the time required to get the valve off it seat and to the fully open position is dependant on the mass of the valve, and the fuel pressure which is trying to hold the valve shut.

We can't change the mass of the valve, but we can use the latest technology injectors which have a valve mass that is a small fraction of what we see in the older injectors. This goes a long way towards increasing control of the injector at low pulsewidths.

As for fuel pressure, all we need to know is that as we increase pressure, the dead time values increase. That being the case, the dead time values that are correct at 45 psi will be off by a mile at 75 psi.

And this brings up a rather amusing story. You've all heard about the 1000cc high impedance injectors on the market that are "The Same" as Injector Dynamics injectors. I go back and forth between being angry, and amused. The following story definitely falls in the amused category.

Tony P (The Sicilian Injector Guru? Injectus Maximus? Speedy Scungilli?) was contacted by a guy who bought injectors that were "the same" as Injector Dynamics injectors. During the conversation, the customer offered to send the data sheet that came with the injectors, and this data sheet supposedly listed dead time values. So of course, Tony sent it to me.

I fully expected to see the Injector Dynamics data, and was already contemplating buying a plane ticket for an impromptu "visit" to the offenders. What I saw was not Injector Dynamics data, but something completely different that had me scratching my head for a while. It looked eerily familiar.

And then it hit me, it was data from the Ford fuel lab for the unmodified version of this injector. Not only was it for the unmodified version, it was the result of testing at 39.15 psi which is Ford's standard.

I was laughing so hard I nearly fell out of my chair.

It's clear to those of you who have read this far that fuel pressure has a substantial effect on the dead time, but what might not click is the fact that the dead time also changes when the injector is modified for increased flow.

The electrical delays are unchanged, but think about the time when the valve is moving from the fully closed, to the fully open position. Because of the modifications, the amount of fuel exiting the injector tip during this partial flow period is different during both the opening and closing phase, and so the dead time value is also different.

The funniest thing about this is that if you interpret the Ford data properly, you can see that there is a compensation table which shows the effect of pressure on dead time. Apparently they didn't catch that, and instead just supplied dead time data at one pressure.

BRILLIANT!

The best bet for this customer would be to un-modify the injector, and then run it at 39.15 psi just like Ford did for their testing. In this case the dead time data would be spot on, and he could expect stellar results!!!

I have attached a pic of the Ford fuel lab data for your amusement.

So if you see this data related to a high impedance 1000cc injector, you know where it came from. Try not to laugh out loud.

That's it for now. I have to go do some real work.

Please come by later for the rest of the story, and thanks for reading my rant.

P.S. If anyone comes up with a good nickname for Tony P let me know.

PY