AEM and Knock - eg: Paul Nelson Datalog
#1
AEM and Knock - eg: Paul Nelson Datalog
Ok first off I've never used AEM but have some friends that do. I run Autronic and listen to knock via det cans. But I'm curious about how the AEM guys do things, and I think the most common thing is watching the knock sensor voltage datalog. Here is the datalog from Paul Nelson Evo. Now can someone tell me what you look for, is there any filters applied to the signal, do you use the stock knock sensor, etc. Basically tell me all your secrets
I'm curious because I just started logging the knock sensor input with my Autronic ECU and I'm curious if I might be able to set up some sort of knock control just from the raw signal.
I'm curious because I just started logging the knock sensor input with my Autronic ECU and I'm curious if I might be able to set up some sort of knock control just from the raw signal.
#2
Evolved Member
iTrader: (1)
Ok first off I've never used AEM but have some friends that do. I run Autronic and listen to knock via det cans. But I'm curious about how the AEM guys do things, and I think the most common thing is watching the knock sensor voltage datalog. Here is the datalog from Paul Nelson Evo. Now can someone tell me what you look for, is there any filters applied to the signal, do you use the stock knock sensor, etc. Basically tell me all your secrets
I'm curious because I just started logging the knock sensor input with my Autronic ECU and I'm curious if I might be able to set up some sort of knock control just from the raw signal.
I'm curious because I just started logging the knock sensor input with my Autronic ECU and I'm curious if I might be able to set up some sort of knock control just from the raw signal.
#3
EvoM Guru
iTrader: (41)
I've looked at the AEM software and it seems that the general approach is to log a baseline RPM x stock knock sensor voltage under conditions when knock would be extremely unlikely (low boost, etc.) to get a baseline for engine noise throughout the rev range. Then you tweak the calibration file to adjust the ECU's reaction to variances from that baseline, including how much of a variance will be treated as knock, how it will adjust timing, how it will adjust fueling, how quickly it will recover from a knock condition, etc.
I haven't done this personally and there isn't a huge amount of open dialog on here about setting up an AEM, but I think this is the basic idea.
It seems to me that this is actually a more reliable approach than the stock ECU for a highly modded motor, since it is custom tailored to your very engine noise characteristics, versus just monitoring "knock counts" on the stock ECU; provided, of course, that everything is set up properly to start with.
I haven't done this personally and there isn't a huge amount of open dialog on here about setting up an AEM, but I think this is the basic idea.
It seems to me that this is actually a more reliable approach than the stock ECU for a highly modded motor, since it is custom tailored to your very engine noise characteristics, versus just monitoring "knock counts" on the stock ECU; provided, of course, that everything is set up properly to start with.
#4
Evolved Member
iTrader: (12)
Like others said you need to figure your own base line to know the motor noise. I will set my timing low and even with race gas and make a full low boost pull and use that as my base gauge. Of course as the motor makes more hp then the noise can go up. If I see a place I think may be knock I will pull timing to see if there is a change. I feel we have very good feed back upto 8500rpm and past that its little bit of a guessing game. I can tell when there is noise by how the spiks get tighter and there are more but it does not pop out at you like knock below 8500rpm does. I know the Subaru sensors read much lower and some think work better. I keep meaning to play with one but have not done it. Also another strange deal to me is how we run the same Evo/2g sensor in 1g's 2g' and evos and they all seem to have different knock to noise volts. Like a 1g at 8.5k would be about 2-2.5v and above would be knock, a 2g seems around 2.5-3V and a evo will have spikes to 3.5-3.8.
Here is a log that is starting to pick up knock.
Here is a log that is starting to pick up knock.
Last edited by Lucas English; May 6, 2008 at 11:03 AM.
#5
Former Sponsor
iTrader: (56)
Join Date: Sep 2004
Location: DFW, TX
Posts: 2,701
Likes: 0
Received 0 Likes
on
0 Posts
using the aem raw knock sensor data, you can see knock as spikes. its exactly how you would detect knock using your ears. the knock sensor is nothing more then a micophone anyway, so its just listening. a real knock event is evident as the spike will be much larger then any others.
any filtering done post raw data is going to filter out the "offset" voltage which is where the mean voltage is. large spikes that go over the filter voltage show up in the "filtered" knock counts, and the filter is user adjustable (basically you can drag a line along the rpm band at what voltage you deem to be knock). say you define 3V as the knock if it goes above that. if you record the knock counts, any time you get a spike above 3v it will show up as a knock count in the filtered data.
typically the filtered data is easier to read, but after some practice it is far more important to read the raw data as the filtered data can disguise stuff.
i'll have to find some logs with knock and show the comparison.
edit* lucas beat me to it. although i know i have some with both raw and filtered knock counts to show how it works.
any filtering done post raw data is going to filter out the "offset" voltage which is where the mean voltage is. large spikes that go over the filter voltage show up in the "filtered" knock counts, and the filter is user adjustable (basically you can drag a line along the rpm band at what voltage you deem to be knock). say you define 3V as the knock if it goes above that. if you record the knock counts, any time you get a spike above 3v it will show up as a knock count in the filtered data.
typically the filtered data is easier to read, but after some practice it is far more important to read the raw data as the filtered data can disguise stuff.
i'll have to find some logs with knock and show the comparison.
edit* lucas beat me to it. although i know i have some with both raw and filtered knock counts to show how it works.
Last edited by KevinD; May 6, 2008 at 10:42 AM.
#6
Evolved Member
iTrader: (30)
In logging Lucas' car we have seen 3.5v spikes logging the knock sensor voltage with ZERO reported knock by the stock ECU, and 10 counts at 1.3v which shows why filtering is so important.
Reading the plugs in addition to the datalogs will also go a long way. The above posted datalog was right before we pulled plugs at the track. The plugs were all even tan with no signs of knock. If I had been thinking I would have taken pix before I tossed them, but they looked great. Every part of the plug looked like I could have dropped it in any Evo and they would have ran as good as new.
Aaron
Reading the plugs in addition to the datalogs will also go a long way. The above posted datalog was right before we pulled plugs at the track. The plugs were all even tan with no signs of knock. If I had been thinking I would have taken pix before I tossed them, but they looked great. Every part of the plug looked like I could have dropped it in any Evo and they would have ran as good as new.
Aaron
#7
EvoM Guru
iTrader: (8)
Reading the plugs in addition to the datalogs will also go a long way. The above posted datalog was right before we pulled plugs at the track. The plugs were all even tan with no signs of knock. If I had been thinking I would have taken pix before I tossed them, but they looked great. Every part of the plug looked like I could have dropped it in any Evo and they would have ran as good as new.
Aaron
I've pulled plugs from a Evo that was riding the knock sensor with 15+ knock counts across the whole powerband and the plugs looked perfect.
Just an FYI.
Trending Topics
#8
EvoM Guru
iTrader: (41)
Sorry if I'm missing the obvious.. .
#10
Evolved Member
iTrader: (30)
Knock voltage is just what the sensor hears, then filters, decides if its real, then reports. I am saying based on the above AEM log where we had 1 or 2 spike to 3.8 volts it very well could have been engine/turbo noise. I base this on the fact that the Lucas has the same bottom end as Paul and gets noise of the same level.
Bryan, that is good information to know. Even at consistent 15 counts it didnt show any depositing or pitting on the plug? The ones out of Pauls car were textbook (literally) tan and even colored with no heat distortion, white deposits, burn marks, etc. Clean plugs all the way around. I will take pictures of the next batch and start a thread in advanced or something.
aaron
Bryan, that is good information to know. Even at consistent 15 counts it didnt show any depositing or pitting on the plug? The ones out of Pauls car were textbook (literally) tan and even colored with no heat distortion, white deposits, burn marks, etc. Clean plugs all the way around. I will take pictures of the next batch and start a thread in advanced or something.
aaron
#13
Evolved Member
iTrader: (30)
Optrand for the flawless victory...too bad its like 2k a sensor(need 4 in our case) then you still need the rest of the hardware and an oscilloscope to convert the signal and then log that. Something that can sample once every six nanoseconds would be way cool though.
#14
Evolving Member
iTrader: (4)
Join Date: Apr 2005
Location: Nampa ID
Posts: 400
Likes: 0
Received 0 Likes
on
0 Posts
Knock voltage varies from car to car. Every car is different in there own little way. Best way to know if knock is real or not is to do some extremely low boost, low timing pulls, this way you know for a fact it is not knocking, do 3rd or 4th gear pulls and record knock voltage via the AEM (if you have one) Then you can determine what is knock and what is not. Of course the more timing you add to it and the more boost you add to the car the noise will get a little noisier in the upper RPM's. Consistency is the key.
Small spikes are not to really be concerned about, it's when those spikes are bunched together and stand well over the rest of the noise graph.
But again, knock voltage on one car vs another car are 2 completely different things.
Checking plugs is by far one of the best ways to really know and you can use the knock voltage as a reference.
Hell, I don't even run a knock sensor in my car these days. Past 9000rpm they are pretty much useless. I tune by what the plugs tell me. I pull the plugs on the return road at the track. Driving it back up in the pits will discolor the plugs and you won't get a accurate reading.
Plugs can tell you everything the motor is doing.
detonation
when timing comes and goes
fuel mixture
too hot or too cold of a plug
hot cylinder conditions
ect ect.
Small spikes are not to really be concerned about, it's when those spikes are bunched together and stand well over the rest of the noise graph.
But again, knock voltage on one car vs another car are 2 completely different things.
Checking plugs is by far one of the best ways to really know and you can use the knock voltage as a reference.
Hell, I don't even run a knock sensor in my car these days. Past 9000rpm they are pretty much useless. I tune by what the plugs tell me. I pull the plugs on the return road at the track. Driving it back up in the pits will discolor the plugs and you won't get a accurate reading.
Plugs can tell you everything the motor is doing.
detonation
when timing comes and goes
fuel mixture
too hot or too cold of a plug
hot cylinder conditions
ect ect.
#15
Evolved Member
iTrader: (7)
Join Date: Dec 2006
Location: on earth
Posts: 943
Likes: 0
Received 0 Likes
on
0 Posts
Crcain let me know how you go, doing this with the autronic.. I have the same ECU just PnP for the IX i would love to see if we can get this going...
Maybe by checking the knock voltage we can creat a table which can pull timing when voltage goes to high on knock sensor.
They were supposed to have one ready that was a plug and play board fitting to the existing ecu but it is 2 years in the making and have not heard anything about it.
Only thing that actually puts me off the autronic sometimes, now a good tuner will tell you if it mapped properly you dont need knock control.
but you never know
Maybe by checking the knock voltage we can creat a table which can pull timing when voltage goes to high on knock sensor.
They were supposed to have one ready that was a plug and play board fitting to the existing ecu but it is 2 years in the making and have not heard anything about it.
Only thing that actually puts me off the autronic sometimes, now a good tuner will tell you if it mapped properly you dont need knock control.
but you never know
Last edited by Whoop_ass; May 6, 2008 at 11:23 PM.