VEMS Community Forum
VEMS => Configuration => Topic started by: Sprocket on May 02, 2007, 11:40:32 pm
-
Where to start? This is the last major config area on my instal
I have read Robs guide on vems.hu but there is bits missing. Maybe some one can expand on the unfinished areas please :) The Wiki is doing my nut :(
As for a listening device, how can i achieve the best results DIY stylee ;D. Only to set up the sensor, I will be hitting the rollers for full mapping
As ever, thanks for any help anyone throws my way ;D
-
What bits are missing in the guide?
Take a look at Cliff's guide in the Calibration section, theres link to Jez Horsham's site where he shows the construction of a nice DIY det detector.
Rob
-
I've been using something like Jez's DIY det detectors for the last 5 years. They work very well.
I'm going to have to read Rob's knock guide again now... I've forgotten most of it. Should have some time over lunch.
-
The Wiki is doing my nut :(
I know what you mean. It can take a lot of filtering to get to the information you want. Unfortunately that is the problem with something like that which acts as a discussion board, ideas forum and information source all in one place. Hopefully this forum will overcom that by seperating those areas out (by different posts or topics)
Anyway, I don't quite understand your first question. What is it you need help with - the initial setting up of the sensor and knock frequencies? Sod it, here is my view on doing the whole lot.
Ok, well first, choose a decent knock sensor. The modern generation of Bosch or Denso sensors are good. Try and choose one with a flying lead as these pick up less electrical noise than the ones that have no lead attached.
Next find somewhere to mount it, preferably on the cool side of the block away from exhaust manifolds, near the top of the block, and for a single sensor in between cyls 2 and 3. Of course the chances of there actually being a mounting boss of the right size in that location is slim, so you may just have to go with whatever you can find. Generally the head will pick up more noise from the valvetrain etc, but it is possible to mount it there. Wherever you choose to mount it, make sure you fit it according to the manufacturers spec, particularly regarding bolt torque, as this affects the detection ability.
With the sensor fitted you next need to set up the detection side of VEMS by determining the correct detection frequency. Personally I don't subscribe to the 'hit the block wih a spannner' method, but Rob tells me he has had some success with this, so you will need to make up your own mind. The chart that was posted on the wiki gives you a good starting point for the frequencies you should expect from your engine depending on the bore size.
To be able to measure the frequency you need something that can band pass filter and then perform an FFT analysis on the signal coming from the knock sensor. The knock sensor should only be used in the 5kHz to 20kHz range normally, and there will be several frequencies generated in this range. The trick is to choose one that gives the best correllation to knock and a good signal to noise ratio under all conditions. Note that signal and noise increase substantially with engine speed, so accurate knock detection (i.e. detection without mis-detection of events that are not knock) at the redline becomes increasingly difficult.
In my opinion, to generate knock, you need to run the engine and make it knock in the region where knock will actually occur in normal service. Whilst you can force the engine to knock by advancing past MBT, this is not what you will do in reality (you shouldn't have mapped the ignition more advanced than MBT), so the true signal will be masked. I suggest running the engine at low speed (1500rpm) and WOT, and advancing the ignition to BLD (approximately 1 ping per second).
This is where your 'det cans' come in. They will allow you to correllate real knock that you can detect with your very sophisticated filtering and pattern recognition system (your brain) with the signals you measure from the sensor. From there, take similar measurements at 1000 rpm steps to as high a speed as you dare to run with DET. The way I see it, if i'm going to rely on a knock system at a given point, then I should have tested it there first to convince myself it will work. If you want your knock system to bail you out at 7000rpm, then you need to go there and measure the sensor signal to see if its possible or not. Its not nice, and it may seem risky, but in the long run its much less risky than relying on an unproven system to save your engine.
At each condition knocking, you will also need to take a noise log, typically at BLD-3 (or whatever retard gives you no pings) With all that data collected, tabulate the knock frequencies detected at each speed. You should be able to chose a couple of clear knock frequency bands at least. Then do a signal to noise calculation of the peak signal voltages of the knock and noise data having been filtered to your selected detection frequency ranges. This will allow you to make the best choice of detection frequency for all conditions.
Once you have selected your detection frequency, you can plumb it back into the ECU and validate it by looking for detection of knock events vs your det cans. You'll have to get Rob to explain how the window length and integrator bits work 'cos I'm not clear on that myself yet. However, once you are happy you can rely on the detection (or you have defined the speed range where you can and can't rely on it) you can start to calibrate the actions - the retard step, advance step and delay to achieve the level of control you require.
And I think thats about it really....... ;D
-
I'm getting started on setting up the knock sensor, how do I view it in the logger as looking at the data in excel I can see more columns than displayed in the viewer?
http://www.box.net/shared/dqvdmk1rrr (http://www.box.net/shared/dqvdmk1rrr)
1st log on was my trip out yesterday without the knocksensor disabled, it was a really smooth drive, only issue I'm getting at the moment is a slight misfire/hesitation around 2000-2500 rpm under load at about 1/4-1/3 throttle.
2nd log was this morning for a very quick run up the road to see how the knock sensor recorded, looking at it in excel there are a couple of spikes.
Cheers
Alex
-
I use Phil Tobin's MegaLog viewer:
http://www.ideasandsolutions.biz/MSGraph/
Just add more data using the drop downs :)
-
I use Phil Tobin's MegaLog viewer:
http://www.ideasandsolutions.biz/MSGraph/
Just add more data using the drop downs :)
OOOoooooo thats a little tasty you can even drop your map in and see the VE entries its pulling.
Off to the pub but will hava a little play with that tomorow.
Regards
Agriv8
-
very cool, I can get two VE maps up at the same time, is it possible to bring in the spark table in the same way? if so how?
Cheers
Alex
-
Alex just been looking at your log.
You lambda looks strange to me ( but I aint bee doing this too long so probably talking ballcocks )
it seems to sit at 0.6 for a long time ( this is normal as it warms up but yours doesnt come in until 340s) think mine comes in arround 177s but suspect that depends on config.
It then runs up to all the way up to 2.79 and sits there I dont think that can be right.
Think that could be causing all sorts of funnies as your ECU is likley to think your engine is running lean. possibly a wiring or config problem as cant see that you engine would run happily at lambda at 2.79
Though would wait for a second opinion before changing things.
Regards
agriv8
-
Agriv8, good spot on the log, I should have mentioned before, but Lambda isn't connected yet, need to borrow a mates welder first.
I don't think it's causing anything to move around yet, although I could be wrong....
I'm still playing with the knock sensor, trying to change the noise level without much success, can't figure out how to get the teminal program to work, although as I'm getting figures coming back in the log do I need to change it by much?
-
Which terminal program? Bray terminal in the Tools folder you set up as
Baud: 9600,
Data bits: 8
Parity: None
Stop bits: 1
As for that thing in MegaTune... ???
-
Cheers Rob, Brays works a treat.
I'm trying to set the noise level, but I'm struggling with the instructions a bit, (okay possible a lot!), I've cut & paste phatbobs instruction and comented within them:
"The Values that you are interested in are:
D: xxxx is the signal in the Knock Window Phase, which shows us the noise level where knock is likely to occur.
R: xxxx is the signal value in the noise phase, the noise phase is where knock cannot happen."
I understand the above ;)
"The maximum value of D or R is 3FF" ???
"we suggest that you tune for a 1/3 of this value to be noise, to set this you will need to drive the car without it knocking and set the noise value to a value averaging 131 to 133." okay, I assume I keep hitting "mll" and see what figures come up as I drive.
"You shouldn't see too much of a difference between D & R at this stage" sounds about right
"measurements have shown that the values are around 1.1 to 1.2 times the D value" okay
": 14F to 170." lost me again
"Starting value is Manmttg99c0F To increase the intergrator time constant reduce the number OE to 00, and vice versa to decrease 10 to 1F." really lost me
"Type mll to see the D & R values after each constant change. " okay
"Tune the sensor until your R value reads 305 to 307 with normal driving." doesn't that contradict the bit above stating 131 to 133?
"Finally type mcs to save, and byebye to take VEMS out of its 'terminal' mode." I get this bit ;)
Cheers
Alex
-
"The Values that you are interested in are:
D: xxxx is the signal in the Knock Window Phase, which shows us the noise level where knock is likely to occur.
R: xxxx is the signal value in the noise phase, the noise phase is where knock cannot happen."
"The maximum value of D or R is 3FF" ???
"we suggest that you tune for a 1/3 of this value to be noise, to set this you will need to drive the car without it knocking and set the noise value to a value averaging 131 to 133." okay, I assume I keep hitting "mll" and see what figures come up as I drive.
"You shouldn't see too much of a difference between D & R at this stage" sounds about right
"measurements have shown that the values are around 1.1 to 1.2 times the D value" okay
": 14F to 170." lost me again
"Starting value is Manmttg99c0F To increase the intergrator time constant reduce the number OE to 00, and vice versa to decrease 10 to 1F." really lost me
"Tune the sensor until your R value reads 305 to 307 with normal driving." doesn't that contradict the bit above stating 131 to 133?
Nasty looking instructions there... The maximum value of D or R is 3FF - this is a Hex value 1023 in english. 1/3rd of which is 131 to 133 (or 305 to 307 in english). So an 14F to 170 is 1.1 to 1.2 times D - no english conversion there, use Calculator to figure it out ;)
Manmttg99cXX where XX starts at 0F, you can reduce the number 0E ... 00 or increase the number 10 ... 1F which will increase the difference between D & R
Is this any use?
Rob
-
Good stuff!!
However, i have been talking to a few people who run EFi on the A series and all have said what i had been thinking. The ancient A series has far to much noise as it is to be able to get a decent knock signal to work with. :-\
I'll leave it for now till i get the engine running and mapped and my stress load is reduced from other things.
Thanks for the help ;)
-
Nasty looking instructions there... The maximum value of D or R is 3FF - this is a Hex value 1023 in english. 1/3rd of which is 131 to 133 (or 305 to 307 in english). So an 14F to 170 is 1.1 to 1.2 times D - no english conversion there, use Calculator to figure it out ;)
Manmttg99cXX where XX starts at 0F, you can reduce the number 0E ... 00 or increase the number 10 ... 1F which will increase the difference between D & R
Is this any use?
Rob
"The maximum value of D or R is 3FF - this is a Hex value 1023 in english. 1/3rd of which is 131 to 133 (or 305 to 307 in english)." that makes sense.
"So an 14F to 170 is 1.1 to 1.2 times D" not sure where the 14F fits, or is that 14F=170? "- no english conversion there, use Calculator to figure it out ;)" that's 170*1.1 (or 1.2)?
I assume this is the command line to change settings "Manmttg99cXX where XX starts at 0F, you can reduce the number 0E ... 00 or increase the number 10 ... 1F which will increase the difference between D & R" I'm going to look into Hex values as I just don't quite get it.
I'll have a play again tomorrow night and post results, a dummies guide will probably be needed ;D
Cheers
Alex
-
In decimal 303 * 1.1 = 333 which converts to 14D in hex
307 * 1.2 = 368 which converts to 170 in hex...
I think I'd better use the convention of using 0x to preceed the hex value:
303*1.1=333 = 0x14D
307*1.2=368 = 0x170
Use windows calculator to convert hex to decimal and back again.
The value used in Manmttg99cXX we recommend starting at 0F (0x0F) as its the mid point between 0x00 and 0x1F, shift the value down and up to see the effect on the difference between R & D
-
People have said the same about the ancient Rover too... It just means you need to do more signal processing...
Electrical noise can be dealt with - I assume you're working from your own scratch built loom (A series minis not being known for EFI) and you don't have points etc for the ignition. And you'll have used shielded cable as appropriate. So as far as signals go, things should be as clean as they can be.
All that leaves is mechanical noise from the engine. Most of these should surely be proportional to RPM, so you should have some knowledge of what this will be at any given moment.
Record what a knock sensor "hears" on a PC and run it through a spectrum analyser and compare it with the RPM of the engine at those instants. Do a few tests at different revs and plot a graph of the main noise frequency peaks vs RPM.
Now compare these to your 1st harmonic knock frequency to see if they sit "in the way" at any RPM. If not, then you'll be able to do some good VEMS has a nice digital bandpass filter built in to the knock detection which can remove the noise.
If the knock frequency and the engine noise happen at the same frequency, then there's little you can do at this particular RPM with VEMS (remember the knock frequency is approximately constant, but your engine noise changes in frequency with RPM). With luck, this won't happen at high revs + high load where knock is most damaging.
There's a few games to play, but essentially once a signal is buried within noise that's similar in spectral characteristics it's the devils own job to get it out again.
-
dnb,
Are you saying rover v8s are noisy and rattly ;) ;) ;)
I am running rhods anti pump lifters and yup as warned they are erm - ticky ( unfortunatly the same ticky as a sticky valve ).
Anyway where have you mounted your knock sensor / sensors ? as a future plan is to deck the block to up the CR and would like to add the knock sensoring.
Kind regards
Agriv8
-
I'm using the thinner tin head gasket to up the compression on mine. From 10:1 to 10.78:1 :)
My block has all the bosses etc for cross bolting. However TVR never took advantage of this :( But it does mean I got the knock sensor bosses that are used on the later GEMS etc Range Rovers - one per side of the block.
If you don't have these, there's a handy bolt hole where a set of (now redundant) earth wires went to on my block, behind where the dizzy bracket sits on my Serp front end.
-
Thanks dnb.
Also running the Tin gasgets and skimmed heads CR is arround 9.5 on my 4.2. looked at the altenator mounting points as the heads can be run on either bank
but thinking that a central point on either bank wound be better will have a look at my block there maybe somthing there that needs drilling /tapping any chance of a photo if you get time.
many thanks
agriv8
-
This bolt hole was on the block - not the head.
I'll get some pics sorted as soon as I can. (I need to find my camera or make my phone talk to my computer properly...)
-
I now understand Hex which makes it a lot easier!
A couple of the figures I got while reving it on the drive (couldn't take it out as it was pissing down and I have no roof):
D F1=241
R B5=181
&
D DB=218
R D8=216
Next plan is to get some actual data from driving, I understand that Manmttg99cXX moves the ratio between D & R. two questions:
1. Is the target to get R to 303-307 and then get D to read /1.1-1.2 of the value i.e. 275, or D to 303-307 and R to read 333-368?
2. How do you get R (or D) to be 303-307 if Manmttg99cXX only changes the value between D&R (okay, you could adjust it up or down but then the ratio between D&R would be wrong)?
-
All the values will increase with higher gain.
Remember these are rule of thumb numbers if you need to concentrate more on getting the ratio between R & D right than the actual values - not all engines are as noisey as each other.
-
I merged these two topics because they're both along the same sort of lines.