MSefi.com read only forum (Jan.15,2018)

Scott,
3.7 -> 3.25V is the opcamp struggling a little under load. Increasing R12 a bit will help, but I do not think it is a problem. I need to hunt down a better datasheet for that opamp. Current is what turns on the optical isolator, which helps to remove minor noise spikes that do not have enough energy to trip the device.

On #1 - why does the pulse width on the missing tooth not match?
I suspect that this has to do with the cross over settings + the RC effects described below. On the missing tooth, the slope of the waveform is probably 2X of the slope with a tooth. It will trigger on cross over from either the rising edge passing the cross point or the falling edge. One cross over will happen as the signal goes from the lowest Vr voltage (which looks like it is referenced to ground in the circuit). The other will happen as it goes from the highest Vr voltage down to the cross over point the triggering. If the Vr signal is 20V and the cross over point is set to say 3V, the 0-3v cross will happen earlier than the 20v -> 3V cross relative to the peak position of the Vr waveform. This is similar to raising and lowering a scope trigger to move a sinusoidal horizontally on the scope.
If the above is correct:
1. Swapping the Vr signals should cause the waveform to change a bit.
2. Adjusting the Zero cross over point should move the opamp output around a bit.
Probably a good idea to figure out if MS-II is triggering on the rising edge / falling edge and make sure that edge is stable. As shown in you picture, the falling edge is stable. When that gets run through the optical isolator, it will be the rising edge that will be stable.

As you are running on the stim, everything is stim voltage and the amplitude effects do not come into play, but the slopes are probably higher than they should because the opamp is having problems charging C30.

The ‘banana’ effect is an RC curve probably caused by the opamp's limited current output and filtering circuits and is to be expected. The optical isolator was designed to operate off the coil signal => for a V8 engine one revolution = 4 ignition events. With the 36-1 trigger wheel, the circuit is effectively seeing 36 ignition events per revolution. ~ about an order of magnitude more events. The filtering was designed for a much slower signal than what is being fed into it now. For now, try removing “Dave” capacitor C30 and see what happens. C30 is probably required when running off the coil, which is a really noisy signal. The opamp has to charge this capacitor, and is doing so at the limit of its abilities. This creates the RC curve rise that you see. Removing C12 should eliminate this. Also, changing R12 to ~ 2X (680 ohms) will also help limit the load on the opamp. That will change the RC constant calculations below, but it should not be a problem.

R12 and C12 form an RC circuit that is used as a filter:
Currently T = R12*C12 = 390 * 0.001e-6 = 0.39 micro seconds. The 36 wheel pulses are much faster than this, so at first pass, I think R12 and C12 are fine.
36 * 6000 R/min = 36 * 100 R/sec
1/(36*100) * 1000000 uSec/sec = 278 usec between pulses.
278 >> 0.39 => not a problem.

With R12 change:
278 >> 0.78

- Charles.

Charles-

You do a great job at putting the pieces together and articulating it to a guy like me. Thanks for the help.

I will try your suggestions, but it might be next several days.

I swapped circuit back to conventional VR circuit only, and I noticed that at higher RPMs, the regular VrOut signal (pin 1 on opamp) is MUCH cleaner than the inverted VrOutInv on pin 7. I'm not sure if it's a bad solder joint in the inverted stuff, or if it is actually the circuit design. I'd be interested in hearing from anyone else how different their 2 signals look coming out of pins 1 & 7. Anyhow, made me think that perhaps I should swap polarity on the sensor (stim signal for now), and use pin 1 to send through opto circuit, but I'm not sure if I would then be sucker to the early 1.4ms trigger.

I have always been setup to trigger on rising edge, as my VR signal is steep on the way down, and floats back up through the missing tooth (http://img231.imageshack.us/img231/5979/37vpq7.png). Using VR circuit, VrOutInv, and Opto circuit is like triple inversion.

One thing I noticed in the picture I shared of the TSEL signal at 6,600 RPM with the triple inversion.... the tops of the bananas were getting cutoff before reaching full high voltage (looks like dwell being too short). But notice the voltage. It was approx 3.0-3.2v, and it was firing. So kinda makes me wonder if voltage IS the problem or not, and whether increasing voltage through opto circuit will be a solution. Still worth exploring though, I'll let you know what happens with the other changes. If you have an opinion or idea on the differences between pin 1 & inverted pin 7, lemme know.

Scott.

Scott,
The link you referenced ( http://img231.imageshack.us/img231/5979/37vpq7.png ) shows exactly what I was trying to describe on the Zero crossing set point. Because of the slope change, if cross over were set for –4 versus 0, there would be a large difference on the missing tooth edge placement on one side relative to the behavior seen on all of the other teeth.
=> only ONE edge is really going to be solid.
=> want the MS-II IRQ to trigger off of this edge.
=> Set up Vr +/-, optical input, VrOutIvn/ VrOut to make this happen.
=> if the crossing is close to zero, the missing tooth edge placement on one edge will look “funny” relative to the other edges and will be very sensitive to minor changes in voltage / grounds. => again, one edge will be truly stable, use it for triggering, realizing that as it passes through different circuits, it may get inverted.

In this waveform:
http://img505.imageshack.us/img505/7060/vroutinvph2.png
Red trace is the stim VR input
Blue – VrOutINV (pin 7)
The zero-crossing detector is triggering accurately on the falling edge of the stim VR input. The rising edge of the actual Vr signal will have some slope / cross over timing dependency as described above. If you measure from falling edge to falling edge on the blue trace, the missing tooth is accounted for. The time at which VrOutInv rises is a “don’t care” IF the MS-II IRQ is set to trigger on the falling edge.

In this waveform:
http://img369.imageshack.us/img369/8164 ... anaxd7.png
red = VrStim input
blue = Optical output.
Shows most of this nicely. I’ll walk through it, and someone can sing out if I am missing something.

The circuit is set up to trigger accurately on the falling edge of Vr Stim (red trace).
VrStim travels through Vr zero-crossing circuit to VrOutInv.
Let VrOutInv be High, transitioning to going low (falling edge).
VrOutInv provides the discharge path for C30, but this take a bit of time as C30 is large. As the voltage on the input to the optical isolator slowly drops off, the transistor in the optical isolator slowly turns off, killing the grounding path of R13 => the voltage of OptoOut will start to rise according to the R13*C11 constant. Removing / lowering the value of C30 will allow the voltage at the input of the optical isolator to transition faster. The sharper falling edge on OptoOut is probably caused by the transistor on resistance in the optical isolator being much less than R13.
Transistor off => R13 charges C11.
Transistor on => transistor discharges C11, with on resistance low enough to overcome the 4.7K pull up resistor.
=> Decreasing the value of C11 will cause the transition to happen faster. This may be needed as it forms another RC type circuit and was originally set for much slower signals.
R13*C11 = 4.7K*0.01uF = 47uS a little large. Might want to drop C11 to 0.001uF.

It may be better to have things set up (Vr input polarity and use of VrOut/VrOutInv) so that MS-II IRQ is triggering on the falling edge as that edge is sharper, but it should not matter too much. If it were mine, I’d keep the cross over voltage detection away from the large slope area of the missing tooth.

The RC “banana” rise of the curve does not matter much as long as the voltage gets to a point where the optical isolator transitions before the input transitions again. Looking at the VrStim voltage = 1v there is a fixed delay from the VrStim signal (red) to the optical out transition (blue). As long as it is fixed, it is fine, it will show up as a trigger offset and will be accounted for with the initial timing light adjustments. Removing C30 and changing C11 to 0.001uF will probably clean things up a bit.

There is nothing in the design of the inverter circuit to explain what you see there – it is a very standard circuit. It is a little surprising to me that a difference is seen. Aside from the component problems / soldering problems you suggested, the only other thing I can think of that might cause a problem is the fact that one circuit is loaded and the other is not. If VrOut and VrOutInv were swapped and VrOutInv had the load of Tsel, I suspect that it would clean up. Also, a 5K resistor from VrOutInv to ground would put some load on it and may clean it up. The IRQ may have a pull up / pull down on it to hold it inactive. I’ll try to look that up, or someone else who knows off the top of their head can chime in.

I need to print out all of the scope pictures and get a better handle on the 1.4ms you are referring to. As long as it is consistent, it will just effect your trigger offset and can be easily taken into account. You are right about the multiple inversions. The trick is to get the edge you care about to be the edge that is used for the trigger. There are lots of possibilities. Use VrOut vs VrOutInv => one inversion. Change MS-II setting to trigger on falling vs rising edge => another level, swapping the Vr wires themselves gives another inversion.

If the system were set up to trigger on the “wrong edge” of the vr signal (what I labeled as the “don’t care” edge) and the zero crossing setpoint was set near the zero transition point, then the edge transition for the missing tooth could occur in odd places, as captured in the second image I referenced ( http://img505.imageshack.us/img505/7060/vroutinvph2.png ) in this post. That would cause no end of difficulty in the wheel decoder code. I wonder if this is the source of your original problem.

If triggering on the "don't care" edge is the problem, then the opto-isolator addition probably is not necessary. However, the introduction of the Opto-isolator will do two things that may be useful - better control over the filtering of the signal & a higher voltage signal to the IRQ pin (5-0.5V) versus the Vrout (3.7 to 0v).

- Charles.

Hey Scott, have you tried my mod by chance? Jumping from U7P1 to the base of a 2n2222 transistor? Basically the same way the sequencer is now done. You'll get that 0-5V swing then and put your input capture setting in phase with how the sensor is wired.

krisr wrote:Hey Scott, have you tried my mod by chance? Jumping from U7P1 to the base of a 2n2222 transistor? Basically the same way the sequencer is now done. You'll get that 0-5V swing then and put your input capture setting in phase with how the sensor is wired.

I have not tried that.

I am really tempted to try an LM1815 conditioner. The signal looks really nice without any lag to the IRQ out.

i'm at a point in my project where it's time to move the motor from the break-in stand to nestle it in the car.

so, might be a couple weeks before i can test on the motor again and have anything conclusive on these other potential fixes. stim has worked fine, but misses only happen on motor. ultimately, i hope to run on the sequencer which might be available before my car is ready for the road, but i have 3 things to try in order to solve this for solving sake:

- modify opto circuit as per Charles' suggestion to trigger through VR & opto (reduce C11 to .001uF, remove C30, increase R12 to 680 ohm). XG1 & XG2 is currently jumped, should they be?

- add transistor as per Kris' fix to increase voltage of VrOut (mimic the sequencer schematic by adding 2n2222 NPN transistor at tail end of VR circuit)

- use LM1815 in proto area to condition VR (use schematic similar to the dual VR circuit avail from jbperf.com). this worked nicely for user Nyabinghi.

i have a question on the LM1815 conditioner. the output from LM1815 is pin 12, and others have jumped pin 12 to the 4N25 optoisolator pin 6. isn't pin 6 the 'base' of the transistor that connects directly to TSEL? so why not connect pin 12 output from LM1815 to TSEL directly? C11 & R13 are necessary? and would XG1->XG2 jumper need to be there?

thx, Scott.

Just reread the entire post and looked at the .msq file.
The .msq has the Inginition Input Capture set to rising edge. This is referenced to the pin on the MS-II IRQ input. From the traces described earlier, only one edge is going to be stable. In the case of:
http://img505.imageshack.us/img505/7060/vroutinvph2.png
it is the falling edge. As the different options are tried, I'd make sure that the stable edge (rising / falling) is used for the Ignition Input Capture. The other edge can be all over the place when the missing tooth comes around, it will probably stable elsewhere.

XG1 and XG2 should be connected. If it were mine, I'd look at the input capture first though - it fits most of the symptoms, although I think it was set correctly for the trace above (looking at VrINV while using VrOut to TSEL w/Rising edge = okay).

The LM1815 should work around the problem by detecting a different part of the waveform (closer to the peak). I wonder of the same thing could be accomplished using a diode string simlar to D1 and D2 - this would reduce the the signal by two diode drops moving it closer to the peak and further from the 0 cross over points with its low slope point during the missing tooth.

Charles.

Regarding 'Ignition Input Capture', I think I have a solid understanding of all the particular nuances involved. The MSQ setting of 'rising edge' was with this inverted blue trace into IRQ: http://img231.imageshack.us/img231/5979/37vpq7.png

If I were to swap to an IRQ signal that was in phase with VR trace, like the one you referenced directly above, I would swap to falling edge, for the reasons you outlined.

The crazy part about my whole situation, that seems to be unique compared to all the other VR problems I have read about, is that I SEEM to be sending a good signal into IRQ, but the MSII loses sync and drops ignition events without any apparent problems prior to IRQ pin. For instance, I found an example where an ignition event just didn't happen, and others that had odd dwell times, or arrived too early then lost sync. Pretty sure this is why Al figured it's a good idea to compare against another ECU unit.

But of course, my curiosity has me still seeking for an answer before trying that. So the LM1815 will be an entirely different circuit feeding into IRQ. My VR signal has wavy amplitude and I wonder if the adaptive hysterisis feature of LM1815 will adapt where the fixed pots failed. Of course this still wouldn't explain the seemingly good existing signal into IRQ. My bag of tricks to try LM1815 out arrived a couple days ago, I've just been buried. Maybe this weekend.... but I've got some welding to do to get suspension set up, and that's way more fun for me.

I should be back on the road this coming week as my timing cover is getting treated at the Loctite porosity/impregnation facility. Once I get a new seal and reset the timing tab i'll have it running in an evening!

Scott, I can confirm that I didn't get 1 ignition drop/reset/anything after going for a 1mi drive. I'm about to go put about 50 miles on it now and dont expect it to drop either.

I'd recommend you putting in the same mod I did before jumping on the LM1815 boat (extra $$ and more components to worry about). It'll put your input capture in phase & switch 0-5V regardless

Time to go get some fuel & have fun, shame it's wet outside

Kris-

that's great news. really great. i will for sure try your mod first, as it's simple and seems to have worked well for you.

i am making progress in preparing the car though. i'll prob be swapping the motor from the break-in stand to the car before i get an answer on this. i'm wiring the injectors individually rather than bank , who knows if seq is available by the time i'm ready to fire up again.

Scott.

Ahh nice! So the car is really coming along now hey? Wiring up the injectors individually is probably something I should have done too in hindsight but with the sequencer i'd want to run an MSD DIS-4 in wasted spark mode with sequential fuelling - just for the cool factor - and with the different connectors i'd have to rewire it all anyway. I've proven the Megasquirt install here now so offloading the unit + wiring wont be a problem anymore or i'll put it into dads 68 Tempest Wagon?

I changed my ground wiring around a bit too, I had found that I wired the WB heater negative to my megasquirt/sensor star ground, so I moved the megasquirt & sensor grounds to the engine block and moved the heater ground to the chassis and now my signals are ultra clean! I have a slight glitch in my logfiles from my USB/DB9 cable but i've always had that since day one.

finally got a chance to pull out the soldering iron.

i tried Kris' mod, but not much luck. his mod is:

VROut through 10k resistor to base of 2n2222
emitter of 2n2222 to ground
collector of 2n2222 through 330 ohm to 5v
collector of 2n2222 to TSEL

VROut is giving the 0v low to3.6v high as expected
after the 10k resistor, at base of 2n2222, the signal is only 0v low to 0.6v high
signal from collector of 2n2222 out to TSEL is about 4.6v low to 4.9v high, which obviously does not trigger IRQ.

signal from VR sensor to IRQ is in phase now as expected.

any suggestions why circuit is doing this?

not sure if it makes a difference... the 2n2222 NPN that i used is from radio shack: http://www.radioshack.com/product/index ... 2062586&cp

Try lowering the 10K resistor to 1K and see what that does. Also make sure that the Collector / emitter did not get reversed. It is easy to do.

- Charles.

well, i was certain i read the labeling correct, and i sure checked over my connections after it didn't work.

thinking others might have had a problem with these transistors from radioshack, i just read through customer reviews, and found 2 interesting comments that the orientation may be mislabeled on the package, or the contents may not all be the same. how do i identify the part number of a transistor, is it printed on it? is there a way to test for emitter, collector orientation on the transistor?

comments:

- The E and C labels are reversed on the TO-92 package. This is no problem when using the TO-92 because one can simply rotate the component;

- These transistors are great in electronics projects and definitely a great choice for amatures. The only problem I found with this product is that website lists the wrong parts. It DOES NOT contain 15 2N2222 transistors but rather:
(5) 2N2222, (5) 2N3904, (5) 2N4401 NPN transistors.

For those who aren't looking for specificity, these transistors have almost the same characteristics, however for those who need specifically 2N2222 watch out, as 2N3904 can only support 1/ 10th the current of 2N2222.

it's possible these 2 comments are related... datasheet for 2N4401 shows collector / emitter reversed on the TO-92 (compared to 2n2222), so perhaps guy number 1 above actually used one of those out of the package, and guy number 2 figured out there were different transistors in there. either way, i'll take a look at them tonight.

transistors are labeled on the back. duh. the radioshack website lists the package as 15 2n2222 transistors, but sure enough, it was a grab-bag of 3 each as described in the user feedback from a disappointed user.

i had hooked up a 2n4401 which would have reversed orientation. i swapped the transistor to a 2n2222. signal out of the circuit was from 2.5v low to 5.0v high. hmm. i swapped the 10k resistor between VROut & base of transistor to 1k, and now it is getting a full 5 volt swing!! shazam. signal looks great. no bananas or anything.

hopefully i'll have some info this weekend if it rids of the misses. remember though... i still had missed ignition pulses on the recorded scope data that never showed up on the datalog as an ignition miss. so, if a datalog comes up clean, i think i'll still haul the desktop computer out to the garage to record TSEL signal to IRQ and compare with ignition outputs.

remember this picture? this had red trace = JS10 ignition pulse, blue = TSEL... this never showed up as a miss at all on the datalog, but it certainly missed an ignition pulse. hopefully all these will disappear?

Scott.

i think this may have done it. i didn't record with the scope, but i ran an 11 minute datalog, wound RPMs all around and there were zero drops on RPMs on the datalog, and trigger +/- sat at zero the whole time. most importantly, i didn't realize it before, but when i wound it up to a couple thousand RPM before, it was not so strong like it didn't want to keep winding up. now... at a couple thousand RPM, the motor sounds like it wants to rip your head off and go for broke. this is still on the bench with no load, so i maxed at 2,600 RPM.

i'm not sure that i'm going to haul out the equipment to record the scope readings, as that would be a couple hours of work that i really can apply on some other wrenching. i'll sort through details when it comes time to tuning with a load.

you guys... thanks so much for your help. i'd really like to buy a few jars for everyone, but for now, settle for this happy dancing guy:

Woohoo!!!! Awesome news man. I only copied the 10K resistor going to the base of the 2n2222 because that is what was in the sequencer and if Bruce used it that's good enough for me. When I saw how low it triggered the transistor voltage wise I thought it'd need some tweaking, say around 4k7 to start with but if 1K works and is clean then i'll experiment with mine.

We're planning a trip to the states probably next year so i'll hold you to those beers LOL

Today is the first really clear sunny day we've had in weeks so i'm going for a drive!!!

krisr wrote:..We're planning a trip to the states probably next year so i'll hold you to those beers LOL..

For sure! I'm in Minnesota though which is pretty far from any destination/tourist stuff. Midwest is not usually first stop for visitors. nice in the Spring and Fall. oh wait... we are halfway through Spring and it's supposed to snow tomorrow !?!?!