[JrZook]

royce wrote:

yeah but how bigs the coil

???

This ain't an ignition coil thread

[JrZook]

So now im at the stage where im logging the incoming pulses from the crank angle sensor and measuring the time in-between them. The following graph was logged at idle at roughly 1000rpm.



The flywheel physically has projections situated at 60-60-60-180degrees apart. Two of these correspond to TDC of each cylinder with the large 180degree spread as a homing pulse to the start of this pulse train. The long pulse can be easily depicted from the graph and as expected occurs every 4th pulse signalling one complete crank rotation. As the engine is stable and running it will be pretty easy to synchronise the code to find the long pulse.

Starting on the otherhand is a whole new ball game!! Next graph shows what the pulses look like when the engine is cranking over on the starter.



We can see the long pulse and if you count them they occur every 12th pulse but I assume that the small in-between minor pulses are just from electrical noise as the flywheel is turning over sooo slowly that the inductance pick up sensor is getting small distorted signals, then the long pulses occur every 6th.

Now since the engine is not stable running the crank has a jerky motion and the long pulses here are most likely showing the compression strokes of the v-twin.

I still need to do some more work to try and decipher this pulse train.

[Rhinoman]

I'm an electronics designer working on EFI systems for a tier 1 supplier.
My first thought looking at that is that you'll struggle to get clean running with that PCB, you haven't got a decent ground plane on it which causes all sorts of problems on an EFI system.
I'd be interested to see some circuitry, I would suspect that the gate drivers were killed by a current spike caused by the gate capacitance, did you have series resistors?
Cheers
James

[JrZook]

Hi James!!

Cheers for the input, good to hear from someone that is in the field .

I understand the need to keep noise down to a minimum on the pcb. In this case it was physically impossible to lay down a full ground plane on the board as I could not keep all the routing to a single side. Thus I opted to have no ground plane instead of a broken one (Ive heard that a broken one tends to cause more issue then none at all).

I regards to getting it clean running I did have a bit of an issue before with a very noisy 5V reg giving me some ridiculous noise on the supply rail (~1Vp-p!!!). I have replaced the reg and added a few more supply filter caps and now its down to ~10mV p-p while connected to the running bike which I thought was very acceptable. Where else should I be looking at in regards to noise?

I've tested the input to the circuit and the MCU logging it upto 8000rpm on a signal gen and it didn't seem to hiccup. The above graph while the engine is running to me seems quite stable with no false triggers, although this was at low engine speed (I will get another log at higher speeds to have a look).

Im running the ADC's with 8-bit resolution and they only tend to fluctuate 1-2 counts with the engine running.

What else do you think I should be checking?

Have you done much work in designing zero-crossing detection and filtering circuits for inductance based crank/cam sensors? Mine is fairly simple and seems to pickup and filter the signal quite well so far while the engine is running (only been tested from idle to ~4k so far). As shown above, at super slow engine speeds (starting) I get issues from the low amplitude trigger signal from the crank inductance sensor. This may just be an issue with the caps and resistor values I've chosen.

The crank signal is AC coupled and centered to 1/2 the MCU supply via a resistive divider. It is also clamped by diodes to keep it within ~0.7V of the supply. This is fed to the MCU internal comparator. Another resistive divider set to ~2.7V on the other side of the comparator. This then generates the input capture interrupts. Ill see if I can throw up a few schematics.

In regards to the UCC27324 4amp FET drivers I was burning, according to their data sheet they were rated to driving an input capacitance of 1800pf. The IRF840 FETs im using have a maximum input capacitance of 1200pf hence why I directly connected the driver to the gate. I was told that gate resistors only slow down the response of the FET and that the drivers alone should be able to handle them...

Keen to hear your response!!

Cheers Dan

[Rhinoman]

Most of the noise problems come from current spikes when switching the inductive loads, without a solid or well routed ground you can get quite severe voltage shifts. A similar problem can occur with the positive voltage rail if you're using flywheel diodes, current spikes can lift the rail voltage up causing problems elsewhere.
Slowing the FETs a little isn't usually an issue, we have to do it to reduce EMC emissions and often to reduce the current in-rush to the big caps we use on the injector supplies.
Its worth considering going to surface mount components, they aren't that difficult to solder and with careful routing you can usually get most of your tracking on the top layer leaving most of the bottom layer for grounding.
VR sensors are extremely difficult to design for, we have a circuit that has evolved over something like 10 years. Most of the designs work in a similar way, with a variable arming threshold on the positive edge and then triggering on the falling edge.

[31zook]

JrZook wrote:

Hi James!!

Cheers for the input, good to hear from someone that is in the field .

I understand the need to keep noise down to a minimum on the pcb. In this case it was physically impossible to lay down a full ground plane on the board as I could not keep all the routing to a single side. Thus I opted to have no ground plane instead of a broken one (Ive heard that a broken one tends to cause more issue then none at all).

I regards to getting it clean running I did have a bit of an issue before with a very noisy 5V reg giving me some ridiculous noise on the supply rail (~1Vp-p!!!). I have replaced the reg and added a few more supply filter caps and now its down to ~10mV p-p while connected to the running bike which I thought was very acceptable. Where else should I be looking at in regards to noise?

I've tested the input to the circuit and the MCU logging it upto 8000rpm on a signal gen and it didn't seem to hiccup. The above graph while the engine is running to me seems quite stable with no false triggers, although this was at low engine speed (I will get another log at higher speeds to have a look).

Im running the ADC's with 8-bit resolution and they only tend to fluctuate 1-2 counts with the engine running.

What else do you think I should be checking?

Have you done much work in designing zero-crossing detection and filtering circuits for inductance based crank/cam sensors? Mine is fairly simple and seems to pickup and filter the signal quite well so far while the engine is running (only been tested from idle to ~4k so far). As shown above, at super slow engine speeds (starting) I get issues from the low amplitude trigger signal from the crank inductance sensor. This may just be an issue with the caps and resistor values I've chosen.

The crank signal is AC coupled and centered to 1/2 the MCU supply via a resistive divider. It is also clamped by diodes to keep it within ~0.7V of the supply. This is fed to the MCU internal comparator. Another resistive divider set to ~2.7V on the other side of the comparator. This then generates the input capture interrupts. Ill see if I can throw up a few schematics.

In regards to the UCC27324 4amp FET drivers I was burning, according to their data sheet they were rated to driving an input capacitance of 1800pf. The IRF840 FETs im using have a maximum input capacitance of 1200pf hence why I directly connected the driver to the gate. I was told that gate resistors only slow down the response of the FET and that the drivers alone should be able to handle them...

Keen to hear your response!!

Cheers Dan

what









































Did you write?

[JrZook]

Rhinoman wrote:

Most of the designs work in a similar way, with a variable arming threshold on the positive edge and then triggering on the falling edge.

Like the LM1815?

This will be a prototype board to play with and try to sort out issues with before I go and make another one, hence why I opted to keep it thru so bridging and probing terminals as well as component replacement is easier. This is only my second board so I'm still learning how to lay it out and route properly.

Hopefully it will still be good enough to at least use for some data logging as im using it for now and testing my code even if it doesn't actually switch inductive loads.

Welcome to any suggestions

[JrZook]

31zook wrote:

JrZook wrote:

Hi James!!

Cheers for the input, good to hear from someone that is in the field .

I understand the need to keep noise down to a minimum on the pcb. In this case it was physically impossible to lay down a full ground plane on the board as I could not keep all the routing to a single side. Thus I opted to have no ground plane instead of a broken one (Ive heard that a broken one tends to cause more issue then none at all).

I regards to getting it clean running I did have a bit of an issue before with a very noisy 5V reg giving me some ridiculous noise on the supply rail (~1Vp-p!!!). I have replaced the reg and added a few more supply filter caps and now its down to ~10mV p-p while connected to the running bike which I thought was very acceptable. Where else should I be looking at in regards to noise?

I've tested the input to the circuit and the MCU logging it upto 8000rpm on a signal gen and it didn't seem to hiccup. The above graph while the engine is running to me seems quite stable with no false triggers, although this was at low engine speed (I will get another log at higher speeds to have a look).

Im running the ADC's with 8-bit resolution and they only tend to fluctuate 1-2 counts with the engine running.

What else do you think I should be checking?

Have you done much work in designing zero-crossing detection and filtering circuits for inductance based crank/cam sensors? Mine is fairly simple and seems to pickup and filter the signal quite well so far while the engine is running (only been tested from idle to ~4k so far). As shown above, at super slow engine speeds (starting) I get issues from the low amplitude trigger signal from the crank inductance sensor. This may just be an issue with the caps and resistor values I've chosen.

The crank signal is AC coupled and centered to 1/2 the MCU supply via a resistive divider. It is also clamped by diodes to keep it within ~0.7V of the supply. This is fed to the MCU internal comparator. Another resistive divider set to ~2.7V on the other side of the comparator. This then generates the input capture interrupts. Ill see if I can throw up a few schematics.

In regards to the UCC27324 4amp FET drivers I was burning, according to their data sheet they were rated to driving an input capacitance of 1800pf. The IRF840 FETs im using have a maximum input capacitance of 1200pf hence why I directly connected the driver to the gate. I was told that gate resistors only slow down the response of the FET and that the drivers alone should be able to handle them...

Keen to hear your response!!

Cheers Dan

what









































Did you write?

Translated: ECU is having a few problems

[Rhinoman]

JrZook wrote:

Rhinoman wrote:

Most of the designs work in a similar way, with a variable arming threshold on the positive edge and then triggering on the falling edge.

Like the LM1815?

This will be a prototype board to play with and try to sort out issues with before I go and make another one, hence why I opted to keep it thru so bridging and probing terminals as well as component replacement is easier. This is only my second board so I'm still learning how to lay it out and route properly.

Hopefully it will still be good enough to at least use for some data logging as im using it for now and testing my code even if it doesn't actually switch inductive loads.

Welcome to any suggestions

Very similar to the way the LM1815 operates, I'm told that at one time that was used here but it was designed out due to too many problems with noise. I'm sure that it could be made to work properly though.
The first ECU that I designed was for the Vitara and was a home project when I was working in the defence industry, it took me four revisions to get a good PCB design. There is a good document that was on Ford's EMC website, I'll see if I can find it.

[Rhinoman]

have a look at this document:

http://www.fordemc.com/docs/download/EM ... %20PCB.pdf

Its a bit heavy going in places but there are some good tips about general layout and grounding strategys. The ground grid was used by Denso with 2-layer PCBs in a lot of their ECUs.

[JrZook]

Cheers for the link James! Was exactly what I have been chasing for a while regarding good PCB layout practises. Will be starting on the second prototype board sometime soon.

In other news found and killed a bug that was giving me random RPM spikes. Been logging it for over 30mins with no spikes whilst varying the simulated RPM between 1-8k.

Made up some synchronisation code to find the and number the pulses on the flywheel accordingly. It also seems to work quite rock solid and only takes 2 revolutions to sync the crank!

Got some new MOSFET driver chips and they all work perfectly!! No more issues there. Was running the 2 injectors and an ignition coil off it and the MOSFETs and drivers barely got warm! So I think that's a win

I have some LM1815's on there way to try and get a descent signal of this VR sensor at cranking speed. Hopefully they shall do the trick!

[alien]

JrZook wrote:

hah! your silly scope is named after me =P (no, not the BS part)

[Teracis]

JrZook wrote:

So now im at the stage where im logging the incoming pulses from the crank angle sensor and measuring the time in-between them. The following graph was logged at idle at roughly 1000rpm.

BIG GRAPH

The flywheel physically has projections situated at 60-60-60-180degrees apart. Two of these correspond to TDC of each cylinder with the large 180degree spread as a homing pulse to the start of this pulse train. The long pulse can be easily depicted from the graph and as expected occurs every 4th pulse signalling one complete crank rotation. As the engine is stable and running it will be pretty easy to synchronise the code to find the long pulse.

Starting on the otherhand is a whole new ball game!! Next graph shows what the pulses look like when the engine is cranking over on the starter.

BIG GRAPH OF STARTING

We can see the long pulse and if you count them they occur every 12th pulse but I assume that the small in-between minor pulses are just from electrical noise as the flywheel is turning over sooo slowly that the inductance pick up sensor is getting small distorted signals, then the long pulses occur every 6th.

Now since the engine is not stable running the crank has a jerky motion and the long pulses here are most likely showing the compression strokes of the v-twin.

I still need to do some more work to try and decipher this pulse train.

Just a thought, but to prove the pulses when it's cranking, could you take out the plugs so it spins evenly (i.e. doesn't slow down due to compression) and see how the pulses read?

This may not help at all because presumably you are chasing the correct pulses to start the thing, just an idea that might show you a better pulse.

[JrZook]

Teracis wrote:

Just a thought, but to prove the pulses when it's cranking, could you take out the plugs so it spins evenly (i.e. doesn't slow down due to compression) and see how the pulses read?

This may not help at all because presumably you are chasing the correct pulses to start the thing, just an idea that might show you a better pulse.

Yep good idea! The issue I'm having at cranking is that the engine is turning over so slowly I cannot get a stable enough output from the VR sensor for my ECU to trigger reliably. Hence its also picking up noise and false triggers (real small pulses on the starting graph).

I mite not take the plugs out completely as I'm looking for the pulse-train including the effects of the compression strokes etc on the actual lengths of the pulses.

Maybe if I loosen them off to bled away some compression I will actually get a descent signal to base everything else off for now.

Ultimately the trigger filtering circuit needs more work and I have some new chips on the way for that.

Would be pretty lame having to run next to your bike and push start it to get the cranking speed up high enough

[JrZook]

alien wrote:

hah! your silly scope is named after me =P (no, not the BS part)

Sure about that?

Yea she's an old skool scope and temperamental. All I could get at the mo so it does the job .

[Damo]

I am pretty sure I have one or two LM1815 at home if you need em. I'll have a look for ya.

[JrZook]

Damo wrote:

I am pretty sure I have one or two LM1815 at home if you need em. I'll have a look for ya.

That would be sweet man!! Sooner the better, so I can actually get into writing some real code

[Rhinoman]

Can you rig up a HE sensor for design proving - that would give you a stable reference. You would need one with built in Schimdt trigger but any old car or bike sensor should suffice.

[JrZook]

Good idea!

I was thinking about just making up a simple test stand with a circular disk and magnets positioned the same as the engine flywheel and use a small ratiometric hall effect sensor and comparator setup get get nice clean reliable pulses so at least for the time being I can focus on the code its self until I sort out all the bugs associated with this VR sensor.

At least this way I will be confident in the code and make that rock solid for a start.

[TZAR]

This thread needs some dumbing down I am afraid


BOOOBS!!!!!!!

Thats better

Jr you are one smart little cookie. Never loose that urge to keep making things better. ( In other words never get married)

[Rhinoman]

JrZook wrote:

Good idea!

I was thinking about just making up a simple test stand with a circular disk and magnets positioned the same as the engine flywheel and use a small ratiometric hall effect sensor and comparator setup get get nice clean reliable pulses so at least for the time being I can focus on the code its self until I sort out all the bugs associated with this VR sensor.

At least this way I will be confident in the code and make that rock solid for a start.

That would work you'll get a phase shift with RPM but you could take relative measurements. Have you seen the Motorola Small Engine EFi demo board? the documentation shows a simialr set up.

http://www.freescale.com/webapp/sps/sit ... 812ECUEVME

[JrZook]

I actually have read about that freescale engine controller a while back. Seems to a very neat, simple piece of kit. The documentation is a pretty good read also! Went over it again last night to get some-more ideas!

They use the maxim adaptive threshold chip which seems to do the same thing as the LM1815. So I'm hoping the LM will suit my needs.

While sitting in the garage thinking about how else I could clean up this signal and make it reliable I remembered I had a busted sierra dizzy laying around in my parts pile. They also use a VR sensor so I disassembled it and took out the ignitor module, figured out the wiring and connected it to the bike.

It worked out pretty well! Used a pull-up resistor in place of the ignition coil and the C.R.O displayed perfect square wave output pulses from 0-12V even at the slowest of cranking speeds! It gives me a nice sharp falling edge that corresponds to the negative going zero crossing of the VR input wave which is exactly what i wanted.

Will see how this performs as the input condition circuit for my ECU. Mite even use it if works really well.

Zuk parts to the rescue!!!

[Damo]

Good idea! I forgot the dissy has a VR sensor.

[Rhinoman]

JrZook wrote:

It worked out pretty well! Used a pull-up resistor in place of the ignition coil and the C.R.O displayed perfect square wave output pulses from 0-12V even at the slowest of cranking speeds! It gives me a nice sharp falling edge that corresponds to the negative going zero crossing of the VR input wave which is exactly what i wanted.

I think that if you take the trigger from the coil side then you will get advance with RPM, I don't think that the ignitor uses a zero-crossing detector, just a simple comparator. Maybe you can test that with a very slow simne wave input. If that does happen you could use a look up table to correct for this against RPM. I did use an SJ dizzy once before but I hacked it around so the ignition controller was connected between the VR sensor and the trigger unit

[Rhinoman]

Heres an interesting thread on the Maxim chip:

http://www.miataturbo.net/showthread.php?t=47243

I'll try tapping up the Maxim rep for a few samples, we certainly wouldn't go back to the LM device but the MAX seems like its worth evaluating.

[JrZook]

Mmmmmmmm you might be right there, i'll have to do some more experimentation with different frequencies on the sine wave input. All I thought was that the SJ ignitior module conditions the VR signal, pulls the ignitor output low upon a zero crossing and then holds the output low for a fixed dwell time.
With all the actual timing being dialled in using the centrifugal and vacuum advance (mechanical) mechanisms. So with your project did you find a noticeable phase drift with the module conditioning the VR input?

So not to keen on the LM1815's. What was there particular downfall? Noise? The MAX in that link seems to be working beautifully! Even on such a small amplitude input signal!! If you do get a few to play with keep us posted on your evaluation of them.

Cheers!

[Rhinoman]

I didn't do too much experimenting with the SJ dizzy, I picked up a late 8V dizzy with HE sensor and integral coil, it was cheap so I switched to that instead.
The LM chip was used here about 10yrs ago which was long before my time, from what I'm told missing pulses were the biggest problem. Searching the internet at the time it seems that most DIY projects used the chip in non-adaptive mode. Noise is always going to be a problem with a VR, something like +/-10mV of noise gives a degree or two of error with a 0.5V signal.
I've always steered away from the LM1815 because it wasn't available as an SMD in small quantities, I have got a couple os samples so I might try a back to back test.

[JrZook]

A back to back test between the two would be interesting. Since you do know quite a bit about the suzuki ECU's do you know what they used to filter the VR signal?

[Rhinoman]

Most of the Suzuki ECUs that I have played with used a HE sensor. I do have an early one on the bench at the moment, I'll take a look at it.

I've had a very quick glance at the two datasheets, there look to be two main differences between the Maxim device and the LM, one is the differential input on the Maxim which is always good. The second is that on the Maxim the arming threshold is 33% of the peak, on the LM its 80%. Thats what the guy on the other thread was demonstrating, with the LM if the input signal drops more than 20% between pulses then you miss one. The LM is a fixed threshold, the MAX chip can be varied by adding a bias voltage.

[JrZook]

Well that differential input would definitely a winner in terms of rejecting noise and false triggers. I guess that's why they can also drop the arming threshold down to 33% and still have a reliable output.

Hurmmm wonder where I can find a few of these in Oz.