CAM's and Shunting.
Discussion
I have been trying to resolve a shunting issue on my 500 for a couple of years now. (prior post: http://www.pistonheads.com/gassing/topic.asp?h=0&a... After extensive testing on 14CUX and MS2Extra I would like to explore the correlation (if any) between CAM profile and shunting - in particular the "Idle vacuum". I think this particular CAM characteristic might direct us to a CAM's tendency to shunt?
For example my CAM has an idle vacuum of 51-53 Kpa (15 In.Hg) for an idle of 800-900rpm. This could be considered a comparatively 'poor' vacuum and a more usual vacuum for an engine in good condition might be 35-40 Kpa (17-20 In.Hg). After much tuning, I still have a light shunt between 1200-2000rpm on both 14CUX / MS2 and I am thinking that my vacuum (CAM profile) just happens to conflict with the physical design of the RV8 (ie. The idle [datum] vacuum is reflected across the load/rpm range and just does not suit the induction design?) There is no doubt the higher the vacuum (wilder CAM) the greater the problems with idle and possibly shunting? I have not been able to tune it out.
Can you help with providing data for the following table:
For those of you who have the vacuum data, you can assist me in creating a more complete table based on your own experience - it could be useful and of general interest in CAM selection.
So let's keep it simple - if you have a shunt, what is the idle vacuum and what CAM?.
Cheers
David
For example my CAM has an idle vacuum of 51-53 Kpa (15 In.Hg) for an idle of 800-900rpm. This could be considered a comparatively 'poor' vacuum and a more usual vacuum for an engine in good condition might be 35-40 Kpa (17-20 In.Hg). After much tuning, I still have a light shunt between 1200-2000rpm on both 14CUX / MS2 and I am thinking that my vacuum (CAM profile) just happens to conflict with the physical design of the RV8 (ie. The idle [datum] vacuum is reflected across the load/rpm range and just does not suit the induction design?) There is no doubt the higher the vacuum (wilder CAM) the greater the problems with idle and possibly shunting? I have not been able to tune it out.
Can you help with providing data for the following table:
For those of you who have the vacuum data, you can assist me in creating a more complete table based on your own experience - it could be useful and of general interest in CAM selection.
So let's keep it simple - if you have a shunt, what is the idle vacuum and what CAM?.
Cheers
David
Edited by DaveG on Sunday 24th July 00:43
I think you have things the wrong way around- a "wild" cam will have a very poor vacuum at idle due to the big overlaps preventing much time for an inlet valve open only part of the cam cycle - so you have low gas velocities and poor cylinder filling. Add this to reversion that's pushing burnt mixture back into the plenum when the exhaust is valve is open for long periods on other cylinders, (ever wondered why the plenum is full of brown crap?) and your inlet charge is both low velocity and now contaminated, so no wonder it wont burn correctly. Messing around with timing and mixture can make the best of a bad job, but I bet if you ran separate throttle bodies shunting would never be a problem. Expensive fix however!
BTW taken of another site on the 14CUX- "the 14CUX system has a hard time adapting to a cam with a lobe separation angle under 112 degrees".
BTW taken of another site on the 14CUX- "the 14CUX system has a hard time adapting to a cam with a lobe separation angle under 112 degrees".
Edited by blitzracing on Saturday 23 July 09:29
Hi Mark. I think you mis-interpreted my expression 'high vacuum'. Depending on whether you are expressing in pressure terms of Hg or Kpa? (Numerically 'High vacuum' = high Hg = low Kpa).
I agree with your description and I am trying to see if there might be a vacuum threshold which correlate to a tendency to shunt. With a std Cam it can be tuned out, but as the duration and Ica change (as does the idle vacuum) is there a point at which it cannot be properly tuned out?
I agree with your description and I am trying to see if there might be a vacuum threshold which correlate to a tendency to shunt. With a std Cam it can be tuned out, but as the duration and Ica change (as does the idle vacuum) is there a point at which it cannot be properly tuned out?
Its cyclic nature is interesting- you would think you would see spikes in the AFM voltage or weird lambda readings. A bad burn is fine, but why does this end up as a two or so second cycle? It would good to get my data logger on the sensors on a car that shunts to see if it gives any clues.
I have seen this cyclic behaviour before. In one case it was the VE table not being perfectly tuned, closed loop fuelling being on and acceleration enrichment settings being less than ideal for low RPM operation. The weird lambda readings didn't show up in the ECU data logs because the sample rate in the logging was too slow. I wondered at the time whether the ECU sensor sample rate was too slow and I was seeing an aliasing effect - ie taking samples too slowly to see what is actually happening and concluding something much slower is there.
I have a H404 cam in my Griff and shunting isn't a problem - I did work at it for a year and changed many things. (Basic reliability of the car is a problem, but that's another story
)
I have a H404 cam in my Griff and shunting isn't a problem - I did work at it for a year and changed many things. (Basic reliability of the car is a problem, but that's another story
)Hi Dnb, I was interested in your comments and your previous mapping response - which is hard to transcribe: ( http://www.pistonheads.com/gassing/topic.asp?h=0&a... ). No where do you seem to say what ECU you are using?
I take your point on the sampling rate, currently on 8 but also been through 12, 24, 32. I think my maps are pretty good, but still need to take a dyno run.
If you have cured your shunting with a H404 I would be very interested to see your maps (VE, Spk, AFR) - maybe email me.
I take your point on the sampling rate, currently on 8 but also been through 12, 24, 32. I think my maps are pretty good, but still need to take a dyno run.
If you have cured your shunting with a H404 I would be very interested to see your maps (VE, Spk, AFR) - maybe email me.
No, it appears I didn't say what ECU I was using. It's one most people haven't heard of - VEMS. It's been installed since the early 2000s and I believe was one of the first (non motor trade) people to DIY install an aftermarket ECU in a Griff.
You have slightly missed my point regarding the sample rate. I don't think I explained it as well as I could so will try again a different way below. I am also not sure my tables will help you (but that doesn't mean you can't have them) - my engine is one of the more modified variants out there so it would probably help you more to follow the process I followed. Hopefully some of the ramblings below will help you.
Let's say we can get 10 samples per second out of the various sensors. If some event is happening at say 25 times a second, you would only see 1 in 50 of these events, and if you didn't know better would conclude the event is happening 0.5 times a second. (Known as aliasing in signal processing). I have a hypothesis that aliasing might be happening in some ECUs during shunting since something high frequency (ie the charge robbing and exhaust gasses getting in to the intake) gives rise to a low frequency thing (the shunting behaviour).
My next hypothesis is concerning sensor response times: The response time of a Bosch LSU 4.2 wideband sensor is of the order of 100ms according to Google. This means that no matter how fast you sample things, there will be some latency in reading fuel mixture and the reading will be an average of everything that happens over the 100ms - meaning that spikes will tend to be low pass filtered. This is likely to cause a similar effect to above - spikes in the fuelling will go unnoticed and the engine will do something the ECU doesn't expect and can't react to until it's too late so the ECU ends up chasing its own tail.
Add to this the engine having what I expect to be a non-linear transfer function around the stoichiometric point (when lambda = 1) you have a difficult control problem. I found it was easier to get the car to behave nicely when I aimed for lambda = 0.97 when very slightly loaded since the engine behaved in a more linear way when it wasn't alternating between lean and rich. This is less than ideal for emissions testing but careful positioning of load bins in the area mean you should be able to set aside an "emissions zone" in your tables for the MOT that the ECU never normally visits. (No, I don't work for Volkswagen!)
Another fun issue I found was acceleration enrichment at part throttle causing grief. If you didn't have enough, then the engine pulled all the fuel off the port walls and caused a lean spot several tens of milliseconds after the acceleration impulse. The result was shunting again. Too much fuel resulted in a flat spot at the point of acceleration. Getting this right is all a bit tricky since there is so much lag and filtering in the lambda readings... This can't be properly tuned until the VE table is perfect.
I ended up writing a tool to analyse data logs and look for steady state fuelling errors and estimate and compensate for sensor lag over hours of test drives to fix the VE table. This was long before we could simply download tuner studio autotune etc.
You have slightly missed my point regarding the sample rate. I don't think I explained it as well as I could so will try again a different way below. I am also not sure my tables will help you (but that doesn't mean you can't have them) - my engine is one of the more modified variants out there so it would probably help you more to follow the process I followed. Hopefully some of the ramblings below will help you.
Let's say we can get 10 samples per second out of the various sensors. If some event is happening at say 25 times a second, you would only see 1 in 50 of these events, and if you didn't know better would conclude the event is happening 0.5 times a second. (Known as aliasing in signal processing). I have a hypothesis that aliasing might be happening in some ECUs during shunting since something high frequency (ie the charge robbing and exhaust gasses getting in to the intake) gives rise to a low frequency thing (the shunting behaviour).
My next hypothesis is concerning sensor response times: The response time of a Bosch LSU 4.2 wideband sensor is of the order of 100ms according to Google. This means that no matter how fast you sample things, there will be some latency in reading fuel mixture and the reading will be an average of everything that happens over the 100ms - meaning that spikes will tend to be low pass filtered. This is likely to cause a similar effect to above - spikes in the fuelling will go unnoticed and the engine will do something the ECU doesn't expect and can't react to until it's too late so the ECU ends up chasing its own tail.
Add to this the engine having what I expect to be a non-linear transfer function around the stoichiometric point (when lambda = 1) you have a difficult control problem. I found it was easier to get the car to behave nicely when I aimed for lambda = 0.97 when very slightly loaded since the engine behaved in a more linear way when it wasn't alternating between lean and rich. This is less than ideal for emissions testing but careful positioning of load bins in the area mean you should be able to set aside an "emissions zone" in your tables for the MOT that the ECU never normally visits. (No, I don't work for Volkswagen!)
Another fun issue I found was acceleration enrichment at part throttle causing grief. If you didn't have enough, then the engine pulled all the fuel off the port walls and caused a lean spot several tens of milliseconds after the acceleration impulse. The result was shunting again. Too much fuel resulted in a flat spot at the point of acceleration. Getting this right is all a bit tricky since there is so much lag and filtering in the lambda readings... This can't be properly tuned until the VE table is perfect.
I ended up writing a tool to analyse data logs and look for steady state fuelling errors and estimate and compensate for sensor lag over hours of test drives to fix the VE table. This was long before we could simply download tuner studio autotune etc.
Thanks for putting together your comprehensive response. My experience seems to support your hypotheses in general. I appear to be seeing some positive feedback in the control system, but cannot seem to identify the main contributors because of the spectrum of occurring conditions. (rpm 1250-199)(MAP 30-45Kpa)(AFR 14-15)(TPS 9-13)etc.
It is difficult to know what the internal sensor sampling rates might be. For data logging I think all channels are 20 per sec. I think I see your point concerning aliases. The rpm oscillation I see which is superimposed on the steady state rpm during a shunt is 60-100rpm at about 1 cycle/sec. Just out of interest here are a couple of random snapshots:-
In relation to AFR's, I use a Tech Edge LSU4 which has a pretty linear characteristic and for the most part I have kept my Accel Enrichment (AE) fairly constant, only varying the MAP/TPS percentages (Now using Speed Density, previously TPS=100%, now MAP/TPS=75%, however previous AlphaN was 100%TPS)
It really is a conundrum when there are so many variables, very time consuming changing one parameter at a time and trying to determine the effect, especially when it is probably going to be a combination of changes to resolve. I need a fellow tuner and a private road !!
Thanks for your input, it all helps.
It is difficult to know what the internal sensor sampling rates might be. For data logging I think all channels are 20 per sec. I think I see your point concerning aliases. The rpm oscillation I see which is superimposed on the steady state rpm during a shunt is 60-100rpm at about 1 cycle/sec. Just out of interest here are a couple of random snapshots:-
In relation to AFR's, I use a Tech Edge LSU4 which has a pretty linear characteristic and for the most part I have kept my Accel Enrichment (AE) fairly constant, only varying the MAP/TPS percentages (Now using Speed Density, previously TPS=100%, now MAP/TPS=75%, however previous AlphaN was 100%TPS)
It really is a conundrum when there are so many variables, very time consuming changing one parameter at a time and trying to determine the effect, especially when it is probably going to be a combination of changes to resolve. I need a fellow tuner and a private road !!
Thanks for your input, it all helps.
Gassing Station | Griffith | Top of Page | What's New | My Stuff