What should be immediately apparent is that we are wasting a TON of our available duty-cycle in the turn-on and turn-off dead-times. This significantly reduces our control resolution. In some cases this can result in either an unstable system, with a small oscillation about the target (Iterm), or a system that never hits target. Making matters worse, the turn-on and turn-off vary dramatically with battery voltage. This will be super problematic on NAs, where the voltage regulation is embedded in the alternator. Furthermore, the turn-on delay is inversely proportional to battery voltage, but the turn-off is directly proportional to the battery voltage. The next two figures show the impact of battery voltage on the turn-on and turn-off dead times for this solenoid. Dead-time is expressed in terms of effective duty-cycle, which I am defining as the actual ON time of the solenoid divided by the period of the gate drive PWM. Commanded duty-cycle is the actual PWM duty-cycle as delivered by the controller.





Plotting the effective duty-cycle vs the commanded duty-cycle is shown in the following figure.




A few observations should immediately jump out. First, we have horrible utilization of the dynamic range of the controller. In order to account for voltage variations, we are only really able to use 50% of the available PWM codes. Second, there is a significant difference in the effective DC with respect to the commanded DC at different voltages. On my NA alternator, I routinely saw voltages ranging down to 12V and up to 14VDC. That leads to a greater than 10% uncertainty in effective duty cycle! All of that has to be made up with an overly aggressive I term. The third main observation is that the response isn’t linear. Equal changes in commanded duty-cycle do not result in equal changes in the effective dutycycle.



First, we are going to address the turn-off dead time. It is the most dominant issue observed so far. It is very common for solenoid driving circuits to employ a flyback diode. The flyback diode shunts current stored in the inductor back to the source at turn-off. What happens is the inductor develops a potential drop across it while current is flowing through it (on). When the current flowing through it stops abruptly, the voltage “flies” up to a voltage in excess of the battery voltage. This can be damaging to switches (FETs) that don’t offer adequate voltage ratings to handle it. Seems like a good idea right? Well, look what happens to the turn-off time when we remove the flyback diode. Notice that the switch voltage hits 50+ Volts with the diode removed.









The effect on the usable effective duty-cycle is astounding. All we need to do in order to make this change is select a switch that has adequate voltage ratings to handle the flyback voltage. Additional protection diodes that clamp the gate signal to the local 5V rail may also be in order, but these are simple, easy, and cheap modifications.

Our response is still not very linear, especially at low effective duty-cycles, so the next step is to go after the turn-on delay. Remember that turn-on is inversely proportional to coil energizing voltage. Increasing the voltage will result in a lower turn-on delay. Undocumented experimentation shows that voltages much in excess of 18V yield diminishing returns, so an energizing voltage of 18VDC with no flyback diode is tested.

 

Now we don’t really want to be driving the solenoid at all times with 18VDC. It will probably overheat, and while I don’t plan on compensating for solenoid temperature, here is a scope capture with infinite persistence that shows how the waveforms vary when I hit the solenoid with a heat gun. Astute readers will notice some other artifacts in these waveforms that are interesting, and will be discussed next.




Enter, the widget. This is a circuit I designed that supplies an energizing voltage of 18VDC to minimize turn-on delay while limiting hold current to around 350 mA in order to reduce thermal issues. The widget has a spanning voltage input, which means that the input voltage can vary above or below the output voltage. While it is unlikely that we will ever get 18V on the car battery bus, during hold, we see coil voltages of less than 12V. The solenoid voltage and current waveforms with the widget and no flyback diode are shown in the following figure. The final turn-on, turn-off, and effective DC vs commanded DC plots are also provided. We now have a linear relationship between commanded DC and effective DC, as shown by the trendline. A simple y=mx + b linear relationship can be applied to compensate the effective DC.

 

note: all of the battery voltage plots are at 50% commanded DC

 

The results of this experimentation were astonishing to me. Probably all available via google search, but I am very-much a hands-on learner. I suspect I will be compelled to take a closer look at the injector characteristics now.

Next up – Wastegate valves!

 

I only understood about half of it, but that half was very impressive!

 

Thanks!

 

This build is absolutely nuts! Subbd.

 

I'm a little late to the party, but yea those turned out well, glad it all worked for ya. Interesting build for sure!

 

Subbing. This was an interesting read... you're nuts Ted. Can't wait to see the running car eventually.

Looking forward to getting together with you to tinker with my Megasquirt, reading this has made me feel hilariously unprepared.

 

Stuff is happening. Going to hold off on posting about the manifold pressure control system, but that work is moving forward.

 

Figured I would post this, might be useful to someone.

Ever wanted to put firesleeve on a pre-made hose?

I use a series of aluminum tubes with matching/similar ID and OD. Start small and work your way up.

 

Like a glove
only 1 interference, tensioner bracket hits the sway mount. Should be a fairly easy fix.

 

this thing is nuts

lol

 

Most recent widget creation

this arduino based module accepts up to 4 different 0-5V pressure transducer outputs, selects one via rotary switch in the armrest console, and drives the stock OP gauge using calibrated current. Current drive should make the gauge reading system voltage agnostic. Right now I have it setup to display oil P, fuel P, turbo oil P, and fuel P on the tank side of the fuel filter.

overall progress note, I am working through the electrical system now, and still need to fab the charge piping. I can see the light at the end though!

 

Minor mod to the description above, will configure the last setting to display (10x) the fuel pressure drop from the pump head to the rail return.

 

 

Gauge adapter demo

 

Rough concept: if you're going to be able to change inputs for the oil pressure gauge on the fly, replace one of the incandescent bulbs in the cluster near the oil pressure gauge with a rgb led that you also control with your Arduino. Color change based on sensor input will help improve situational awareness as to what you're measuring at any given moment.

 

Actually a great idea, and something similar was originally in scope, but got removed because if you can believe it, I am descoping the project in the interest of schedule. I do intend to put the hooks in to support it (extra wires in the harness, functionality in the gauge adapter).

 

Makes sense. Might even be worth it to plumb in an LED now so you don't have to pull the brittle pod off again if you ever decide to implement.