D-Jet Bench Setup, Let the experimenting begin (again).... Options

[pbanders]

See attached photo, this is my D-Jet bench setup. The black box on the left is an EFI Associates 1400 D-Jet tester (circa 1972) that measures the injection pulse width, simulates the trigger contact points over a range of engine RPM's, and provides reference values for the CHT and air temp sensors, as well as the throttle switch operation. The ECU is an 044 that I've used in the past for all of the oscilloscope traces on my web page, I can access any circuit in the ECU with it. I've got a TPS connected to the harness in the middle of the picture, as you turn it it you can watch the injection pulse width increase, then when you stop, you can see the pulse with relax back down (there's a delayed effect as well as an immediate effect). I have an air temp sensor plugged into the harness, and instead of a CHT, I have a variable resistor (half of a pot) that lets me simulate the engine temperature. That's a NOS 043 MPS in the upper right, with my vacuum pump attached to simulate engine load.

I've also got an interposer box that I can put between the ECU and the wiring harness that can be installed into the car so I can monitor all of the ECU pin connectors while driving the car. My current box is a PITA to use in the car, my old design sucks. I've got an idea for a better one that I'm going to have to build.

I've also got a Fluke calibrated pressure sensor that I can independently monitor the manifold pressure, or the pressure at any other point in the vacuum system. I can datalog the output from my DMM to get any of the pressure values as a function of time.

Now, just gotta do something useful with it all. BTW, working on the 914 is one of about a million things I'm trying to do simultaneously, so I tend to get to it irregularly, just hoping that's more than "never". I plan to look into the hot start problem first. I'm going to take some data when it happens on the car (i.e. measure the voltage at the CHT through the sequence of events), bench simulate it, then try to come up with some simple solutions. The idle stabilizer will take longer to do.

[pbanders]

I've posted in the past about the issues with replacing the MPS with a modern electronic pressure sensor, I'll briefly summarize. It wouldn't be a simple task, because the two coils in the MPS are an integral part of the multivibrator circuit in the ECU, and also because the MPS itself is the system element that handles mixture enrichment and transition under full load conditions. Now that rebuild kits are available, it's easier to rebuild an MPS that to build an active microcontroller and sensor system to simulate its operation. If anyone disagrees and wants to prove me wrong, I'll probably the first guy to buy your new replacement system!

[stugray]

I've posted in the past about the issues with replacing the MPS with a modern electronic pressure sensor, I'll briefly summarize. It wouldn't be a simple task, because the two coils in the MPS are an integral part of the multivibrator circuit in the ECU, and also because the MPS itself is the system element that handles mixture enrichment and transition under full load conditions. Now that rebuild kits are available, it's easier to rebuild an MPS that to build an active microcontroller and sensor system to simulate its operation. If anyone disagrees and wants to prove me wrong, I'll probably the first guy to buy your new replacement system!

I thoroughly read your explanation of how the system works, and I believe trying to simulate the MPS output is overkill.
(posting your very own picture here)

(IMG:http://www.914world.com/bbs2/uploads_offsite/members.rennlist.com-10819-1448134653.1.gif)

As you explain, the MPS outputs a wave form (SC+IM bias) that 'returns to zero' in some amount of time (based on manifold pressure and temp) and controls the width of the injector pulse.

To trick the system, we don't necessarily need to simulate the MPS output exactly.
All we need to do is detect the beginning of the injection pulse (TL trigger goes low), and swing our simulated SC_IM_bias low (to below the trigger threshold = ~-0.7VDC ).

Then the simulator waits until it wants the injector pulse to end, then swing back to zero.

The fact that the simulated output does not look like the real MPS output is irrelevant.

So all the MPS simulator needs to output is a negative going square pulse that swings to below the trigger threshold, then returns to zero when the injector on duration has been reached.

So the MPS simulator output would just be a inverting opamp with a -1.0 VDC negative rail and a 0VDC positive rail.

You could even drive it with a discrete output and don't need to even mess with an analog signal.

It is entirely possible that I am missing something, but I believe it is that simple.

So the MPS simulator would have an analog input connected to a modern 0-5VDC Manifold pressure sensor, an intake air temp sensor, and an input trigger connected to the MPS input from the ECU.

It would use a lookup table to determine the desired injector pulse width.
When the MPS input pulse is detected, it just drives the MPS output to ~-1.0 VDC and holds it there until the injector pulse duration has been reached then drives the signal back to GND.
(in fact the signal could mimic the waveform in your diagram labeled 'Injection pulse duration')

The ECU will see the signal cross the threshold and end the injector pulse.

Am I close? (IMG:style_emoticons/default/blink.gif)

[pbanders]

I've posted in the past about the issues with replacing the MPS with a modern electronic pressure sensor, I'll briefly summarize. It wouldn't be a simple task, because the two coils in the MPS are an integral part of the multivibrator circuit in the ECU, and also because the MPS itself is the system element that handles mixture enrichment and transition under full load conditions. Now that rebuild kits are available, it's easier to rebuild an MPS that to build an active microcontroller and sensor system to simulate its operation. If anyone disagrees and wants to prove me wrong, I'll probably the first guy to buy your new replacement system!

I thoroughly read your explanation of how the system works, and I believe trying to simulate the MPS output is overkill.
(posting your very own picture here)

(IMG:http://www.914world.com/bbs2/uploads_offsite/members.rennlist.com-10819-1448134653.1.gif)

As you explain, the MPS outputs a wave form (SC+IM bias) that 'returns to zero' in some amount of time (based on manifold pressure and temp) and controls the width of the injector pulse.

To trick the system, we don't necessarily need to simulate the MPS output exactly.
All we need to do is detect the beginning of the injection pulse (TL trigger goes low), and swing our simulated SC_IM_bias low (to below the trigger threshold = ~-0.7VDC ).

Then the simulator waits until it wants the injector pulse to end, then swing back to zero.

The fact that the simulated output does not look like the real MPS output is irrelevant.

So all the MPS simulator needs to output is a negative going square pulse that swings to below the trigger threshold, then returns to zero when the injector on duration has been reached.

So the MPS simulator output would just be a inverting opamp with a -1.0 VDC negative rail and a 0VDC positive rail.

You could even drive it with a discrete output and don't need to even mess with an analog signal.

It is entirely possible that I am missing something, but I believe it is that simple.

So the MPS simulator would have an analog input connected to a modern 0-5VDC Manifold pressure sensor, an intake air temp sensor, and an input trigger connected to the MPS input from the ECU.

It would use a lookup table to determine the desired injector pulse width.
When the MPS input pulse is detected, it just drives the MPS output to ~-1.0 VDC and holds it there until the injector pulse duration has been reached then drives the signal back to GND.
(in fact the signal could mimic the waveform in your diagram labeled 'Injection pulse duration')

The ECU will see the signal cross the threshold and end the injector pulse.

Am I close? (IMG:style_emoticons/default/blink.gif)

I've thought through the same thing, it should be possible, I lacked the skills to implement, though I've been working on that! I assume the lookup table would encompass the full-load enrichment function, too? If you can mock up a prototype I can test it on my simulator and do a direct comparison to a stock MPS.