a bit of evidence around the problems of nissan cam based triggering

[burn4005]

have now installed a 36-2 trigger wheel on my RB26 and for interest sake, I have installed a 24-1 wheel in the stock nissan CAS and am only using the single window as a phase reset.

I have an Emtron KV8 which shows something called sync position that tells you where the CAM signal comes in relative to the Crank signals (so 50% would  be half way between two crank triggering events)

this gives a very accurate display of how much variability there is between the crank and the cam signals. this is on an R34 gtr type CAS, HKS timing belt, kelford Beehive springs and tomei type B cams. I am using Hall effect triggers and a falling edge trigger threshold of 2v for both.

under some load (but not extreme, this is only 10psi of boost) you can see over the course of less than a second there is a variance of 36.6% in sync position. because there are 10 degrees between trigger events on the crank, this ends up being a variance of 3.6 crank degrees!

I will post more once I have data where the engine is under heavy load (30psi) to see if it gets worse.

 

so you may say, well the keyway in the cam is to blame, not the timing belt. well here is the result at very low engine loads (idle) if there was play in the cam keyway this should still show up here as variance. but the sync position is only 4.2% which is only 0.4 crank degrees!

 

I would like to get a very fast scope to verify the trigger edge slopes are remaining the same, as that is the only other factor here that I have not controlled for (but at a steady RPM, the slope should be consistent for each trigger event regardless.
 

Edited September 17, 2018 by burn4005

[Lithium]

Interesting data, thanks for sharing.  It'd be nice to be able to separate any error which is introduced by triggering resolution - I don't have the time now to do the math to work out what (if any) meaningful error would be introduced via a combination of any signal filtering vs rpm rate of change.   In your case there are 3x the amount of updates from the crank trigger as there are from the cam trigger in a given amount of time, which will create an increased amount of error even if there is absolutely no drift at the cam as the rpm rate of change increases.

Are you in a position to be able do any more testing of this?  I'd be interesting to do a brief steady state test at the higher rpm - ie, hold it at 5000rpm briefly to see the trigger variance.

 

Edited September 17, 2018 by Lithium

[burn4005]

I can certainly do that test easily, both under no load (revving) and under load (standing on the brake)

but at ~6000rpm, and rate of change being in the 100s of rpm/s, the cranks change in velocity over 10 degrees is going to be negligible.

as the reset event occurs at the same point in the 720 degree cycle, give or take a few degrees for reasons we are trying to work out, the cranks angular velocity and acceleration should be very consistent under reasonably steady state conditions.

 

Edited September 17, 2018 by burn4005

[Lithium]

12 minutes ago, burn4005 said:

but I'm not sure what you're saying regarding error. the more teeth you have on the crank the smaller the error will be due to the time between teeth getting shorter. this will mean interpolation between teeth will be more accurate, as the crank angular velocity will change less and less as you reduce the sampling window size. and as this reset event occurs at exactly the same point in the 720 degree cycle (give or take a few degrees) the cranks speed and acceleration should be the pretty much identical.

once you're at 36 teeth an up, the crank is spinning so fast and there is such a short gap between teeth, the interpolation the ecu does is crazy accurate

This is kind of the point I was getting at, the evidence you are presenting is partly comparing the ability to accurately estimate the crank position by comparing something that gets updated every 10 crank degrees versus something that gets updated every 30 crank degrees.   

Theoretically the error between the two should be negligible at steady rpm, but the error will increase the moment the rpm rate of change increases - it looks a bit like you selected a section when the engine was starting to go through a drag down after a dyno pull, which is much higher rate of change than the pull itself.

And yeah, it should be more stable under steady state conditions which is why I asked if you can do that - it should mostly eliminate anything that may (or may not) be introduced by comparing 10deg updates with 30deg updates

Edited September 17, 2018 by Lithium

[burn4005]

with timing locked at 25 degrees and a light on cylinder 1, crank timing is bang on (within 1/4 of a degree) whether steady or revving or rough idle, so the ECU certainly knows where the crank is.

rough idle is actually going to show up triggering issues the most as the crank is turning so slowly. (2ms per 10 degrees at 800rpm vs 0.2ms at 8000rpm.

or for a 12-1 it would obviously be 6ms/0.6ms per 10 degrees instead.

Edited September 17, 2018 by burn4005

[Lithium]

8 minutes ago, burn4005 said:

with timing locked at 25 degrees and a light on cylinder 1, crank timing is bang on (within 1/4 of a degree) whether steady or revving or rough idle, so the ECU certainly knows where the crank is.

rough idle is actually going to show up triggering issues the most as the crank is turning so slowly. (2ms per 10 degrees at 800rpm vs 0.2ms at 8000rpm.

or for a 12-1 it would obviously be 6ms/0.6ms per 10 degrees instead.

Yes, I understand all this - this is my point.  To try and get this on the same page, what do you think will be more accurate - 36-2 or 12-1 on the crank?  What do you think would happen if you compared timing logs between both setups?

 

Edited September 17, 2018 by Lithium

[burn4005]

36-1 by a factor of 3

60-2 by a factor of 5

etc

the shorter the interpolation window between trigger events is the less error can accumulate over that period if the conditions (crank accelerates) differ from the assumption (crank maintains angular velocity)

Edited September 17, 2018 by burn4005

[Lithium]

OK, so what would happen if you compared a timing data between a car running the 36-2 and the 12-1 in exactly the same operating conditions?

Edited September 17, 2018 by Lithium

[burn4005]

it would depend where the crank trigger edges were placed relative to the desired ignition events.

if you had a falling edge at 20deg BTDC on cylinder 1,  you would have best case timing accuracy at a desired event at 19deg BTDC and worst case accuracy at 21deg BTDC.

so your timing would be excellent at 19 BTDC regardless of 1,4,12,36,60 teeth, but would get progressively worse at 21 BTDC as you reduced the tooth count.

but that is splitting hairs a bit.

 

I just re-read this

Quote

In your case there are 3x the amount of updates from the crank trigger as there are from the cam trigger in a given amount of time, which will create an increased amount of error even if there is absolutely no drift at the cam as the rpm rate of change increases.

I get the feeling we're trying to convince each other about totally different things here?

to clarify, I have not done any tests using 12 crank teeth.. i am using the 24+1 disc only so i can use the single window for my 720 degree reset, so CAS1 is wired to the 36-2 crank trigger, and CAS2 is wired to the pin in the nissan optical CAS that sees the single window of this disc. i am NOT looking at the other 24 windows in this disc at all, nor have i done any testing with it.

 

trying to find some correlation, the sync position seems to be some component of engine speed (moves the position around, but an sync position band of 2.5 to 4 degrees exists everywhere, but rpm rate of change doesn't seem be correlated.

 

 

Edited September 17, 2018 by burn4005

[Dale FZ1]

Excellent initiative to put up this info about what is the critical point for tuning an RB.

To me a few things jump out.

1. Crank trigger is going to be closer to the ultimate point of truth for actual crank position.

2. Cam mounted trigger always has to be the second rate choice (by how much being the point of this thread) due to drive slop, belt flap, harmonics, power pulse sizes, etc.  And probably the number of teeth involved.

3. Hall effect sensor is not the ideal choice for signal reliability.

4. Depending on engine spec, output, and speed (and what is the target achievement), a cam trigger system may be a pragmatic/practical solution for those punters without the time and $$ resources to go with a proper crank trigger.  I'd like to see the result if the re-windowed CAS was replaced with something like the NZ Wiring kit.

[Komdotkom]

I've got the same setup on mine, but Motec and mag sensors.  I made a single tooth trigger for the exhaust cam instead of using the factory CAS.

I'll grab a screen shot of the ref sync trace from the Motec if you want to compare the outputs of Bosch Motorsport mag sensors vs Hall.

 

[Lithium]

15 hours ago, burn4005 said:

I get the feeling we're trying to convince each other about totally different things here?

to clarify, I have not done any tests using 12 crank teeth.. i am using the 24+1 disc only so i can use the single window for my 720 degree reset, so CAS1 is wired to the 36-2 crank trigger, and CAS2 is wired to the pin in the nissan optical CAS that sees the single window of this disc. i am NOT looking at the other 24 windows in this disc at all, nor have i done any testing with it.

 

OK cool.  To clarify, what I was questioning (to be clearly, not calling it out as BS - I more trying to analyse the data presented as it *is* good data but it'd be good to feel comfortable with the conclusion taken from it ) was this statement:
 

so you may say, well the keyway in the cam is to blame, not the timing belt. well here is the result at very low engine loads (idle) if there was play in the cam keyway this should still show up here as variance. but the sync position is only 4.2% which is only 0.4 crank degrees!

I might be reading you incorrectly, you are only looking at a window on the cam disc - so checking it once every 720deg of crank rotation?  And you are using that to gauge error between crank triggering and cam triggering?   If so, then I'd be questioning your data if you DIDN'T see a big variation when the engine is not operating at steady state rpm... if you get what I'm saying?

Hopefully this helps clarify what I am trying to get at, and at least we can get on the same page and work out how best to use the data - putting this kind of thing on the microscope is a worthwhile venture

[burn4005]

I did the maths on trigger errors at high rates of crank acceleration:

one thing I haven't been able to account for as I don't have any data is changes in crank velocity over a cycle due to combustion pulses, if anyone has some please do share it!

but in a reasonably steady state engine, the tooth window where the reset comes in should have the pretty much the same angular velocity and acceleration cycle to cycle. (and we're talking about a complete cycle time of between 60ms at 2000rpm, and 15ms at 8000rpm)

so if you smoothly accelerate an engine from 2000-8000 in 1 second (a tough ask!)

that is a constant rate of 100rev/s/s or 0.1rev/s/ms

so at 2000rpm the engine is rotating at 33.3333rev/s

with 10 degrees per tooth, a single tooth takes 0.83333333ms to pass.

so the  speed of the crank by the next tooth will be 33.3333+0.1*0.8333333 = 33.4166rev/s

at this speed a single tooth (on a 36 tooth wheel) takes 0.831255138ms

so for a 10 degree tooth, at 2000rpm the change in time between teeth is 0.00207813ms. and at 8000rpm the change in time is 0.00013ms which equates to:

degrees at 2000/8000rpm:

90deg per tooth: 1.98044/0.12726

30 per tooth: 0.22/0.014

10 per tooth: 0.02/0.00156

5 per tooth: 0.00624/0.000391

 

 

Edited September 18, 2018 by burn4005

[Lithium]

Cool, yeah I hadn't done the math and just did a quick run through of my own and it does seem to add up with what you're saying - which is interesting as there is some pretty enthusiastic debate from some pretty respectable names about the likes of 12 @ crank not being enough etc... if your math (and the quick bunch I tried) add up then in theory there shouldn't really be too much error if you were to assume the crank velocity maintains a linear rate of change.  I don't have any data on the realistic behaviour, though as you say... that would be VERY interesting.

Still curious to see what you see if you maintain steady rpm at 5000rpm or so, just for interests sake - I've not got enough time to go fully through my math to make sure it's sane but it is starting to indicate that some things are overstated, which I guess explains how stock Evo triggering doesn't cause big issues haha.

If that does work out as being the case, I am pretty surprised that the difference would be that big.

 

[burn4005]

12 on the crank is plenty. honda K series engines were only 12 tooth, and they're a pretty revvy engine!

that debate can be settled very easily with a timing light, video camera and a high power pull on the dyno.

Edited September 18, 2018 by burn4005

[burn4005]

22 hours ago, Komdotkom said:

I've got the same setup on mine, but Motec and mag sensors.  I made a single tooth trigger for the exhaust cam instead of using the factory CAS.

I'll grab a screen shot of the ref sync trace from the Motec if you want to compare the outputs of Bosch Motorsport mag sensors vs Hall.

 

I'd certainly be interested!

the Mag is certainly the ideal, and capable of keeping up with a very small tooth pitch/module. I don't think i'd get away with a 60 tooth at the wheel diameter I'm using, but would have been fine with a mag.

I am using the motec 36-2 wheel and Motec supplied hall sensor (its actually a ZF/Cherry).

Edited September 18, 2018 by burn4005

[Komdotkom]

I ended up with the mag sensors after a long chat with a Motec guru who races and maintains a lot of IPRA cars. He did loads of research and was having issues with the hall sensors with missing teeth so moved over to the mag sensors and hasn't looked back.

The downside is that you have to muck around with the sensor trigger levels in the software a bit to get really good resolution, but to date they've been excellent. The sensor I'm using is actually for a 60-2 on a GT2 Porsche, they are pretty cheap ~$100.

You can download a fairly good spec sheet from here. They are dimensionally quite similar to the GT101 so I was able to use an existing GT101 style bracket with some basic machining.

https://www.bosch-motorsport-shop.com.au/inductive-speed-sensor

I've got it on the dyno next week so I'll grab a capture under load

[RICE RACING]

Anyone using the cam for crank position is either lazy or a rookie.

When using an ECU with quality knock control you will see this first hand. HALL sensor is only ever used on phase single tooth as its more applicable here for obvious reasons.

Real world using 36-2 on crank with VR sensor you reliably use much more correct ignition timing at 8000+rpm than any cam based 'sloppy' system.

[burn4005]

Yes, I think that's now accepted wisdom. The point of this thread is to show some evidence as to what is happening, and how bad it actually is.

[Lithium]

15 hours ago, burn4005 said:

that debate can be settled very easily with a timing light, video camera and a high power pull on the dyno.

Yeah, next time I've got a dyno handy this will be worth a try to get a gauge of the NZWiring kit solution.   To be clear, I've never debated that the crank trigger setup isn't the ideal solution *but* this data makes the difference look much worse than I expected... so far.

42 minutes ago, burn4005 said:

Yes, I think that's now accepted wisdom. The point of this thread is to show some evidence as to what is happening, and how bad it actually is.

Yeah going by the data so far it looks worse than I expected, albeit still would love to do this testing with something like NZWiring's kit.

Edited September 18, 2018 by Lithium