Hi all
Right then, this knock thing - whats it all about? I hope you're sitting comfortably......
There is a good chance that you will pick up engine noise, and that it will be in the knock frequency range, so don't be too surprised.
The way to tell the difference between the two is to scope the knock sensor along with a trigger - something like an injector or a coil. If possible, also scope the crank signal (makes resolving things into crank angle easier for the next bit)
Knock generally only occurs in the region from combustion peak pressure until 40 deg after peak pressure. Depending on where you have your ignition timing set, peak pressure should occur anywhere from about 5 degrees after TDC (slightly too advanced of MBT) to about 25 deg after TDC ( quite retarded).
So with your scope, log a number of cycles (somewhere between 50 and 300 is ideal). Trigger each cycle off the injector signal. Then look through each cycle and look for the signal on the knock sensor, corresponding to the 0 to 60deg region after TDC for the cylinder of interest.
First point - anything outside of that window is either noise, or another cylinder knocking, which you are not interested in at the moment.
As you go through the cycles, look for the knock sensor signal in the same place each cycle. If its there nearly every time then its noise (unless you are unlucky enough to be so far into knock that it occurs every cycle - but then the engine probably wouldn't last very long
If you are knocking, then you will get a signal in the 0-60 window every so often. It will be random in size and occurrance. That is likely to be knock.
Now you've found it, you can zoom in on just that area and perform your FFT - and you will be rewarded with the knock frequencies of your engine.
Now, you noticed that the signal changes with speed. There are a few possibilities. First, because you are not triggering the scope, those signals you've analysed could be coming from anything (injector, valve closing, piston rocking etc).
Second, the noise amplitude increases with engine speed anyway - so the signal strength increases quite a lot between say 600rpm and 6000rpm.
Next point - how to get knock. You can drive the thing into knock at lower loads, but this means you need to go advanced of MBT to do it. This means you're in an unrepresentative condition. Most engines that are knock limited are worst at high loads at low speeds, since the piston stays around TDC for longer, increasing the chance of knock occurring. Fortuantely, this is also the least damaging time to allow light knock, since generally the loads are not as high as higher speeds, and there is more time between combustions to allow the pistons and chamber walls to cool so that bits don't start to melt. A good place to look for knock initially is around 2000 - 3000rpm, as this is often where cams start to work so VE increases quite sharply, whilst the speeds are still fairly low. On engines with cams tuned / set up to work higher up the rev range, increase the speed accordingly.
So in order to calibrate knock for an OEM, what we do is start with a WOT powercurve, optimising ignition to MBT or BLD at each speed site (note we will have alrady completed a base mapping exercise which has set the fuelling and ignition timing to a BLD-1 condition as a 'safe' running condition). For the knock data collection, 1000rpm steps will do to start with, though later you will increase that to 250rpm steps for validation. Note here that my idea of a power curve is a series of steady state points - you take the engine to x000rpm at WOT and hold it there whilst you adjust the ignition and record the data before moving to the next speed.
When you have found the BLD point take 2 scope logs. One at BLD, and one at around BLD - 5 degrees. Repeat this at each of your engine speeds. Don't forget to adjust fuelling at the higher speeds for the BLD-5 point to keep EGT's sensible (850ish max for NA and SC's, 950ish max pre turbo for, er, turbo's)
Then for each speed, find the knock event as described above. Compare the peak voltage of the knock event to the peak voltage at the same crank angle on the BLD-5 log ( the noise event). From this you can calculate your signal to noise ratio. It needs to be greater than 1.5 to have any chance of acceptable detection and control - prefereably more like 2.5 to 3.
Then you can also perform your FFT of the signal and work out the knock frequencies. Note that they may change in strength at different speeds (higher speeds often exite the higher harmonics more)
With the whole data set you can now choose an appropriate detection window (the crank period when you are looking for knock) frequency, and decide on where you will be able to accurately detect, and where you should map safe.
With WOT covered you can basiclly repeat the process at lower load points until you reach the load at which BLD = MBT. There is not point going lower in load than this since you will need to be advanced of MBT to generate knock, which is a condition you should never be in unless you have mapped the base ignition table wrong, so don't waste the effort.
Now you're probably thinking this is a lot of work - you're right. Its also hard on engines, and I probably wouldn't recommend doing it this way in a car since its unlikely to control the air and coolant temps very well, but it the 'correct' method, and one that you should aspire to, even if you have to modify the procedure in order to suit your resources.
Things you can do to make it easier mapping in the car:
Do this as part of your base mapping seesion when you will set ignition timing on a rolling road. In this case you will start mapping at the lowest loads, setting MBT spark, and increase loads until you reach the MBT=BLD threshold. be sure to note down fuel type, coolant and air temps, as these have a big effect on this threshold. From here take you knock logs as you map the ignition map, and you can process them later.
Here's a useful plot - the kind of values you should expact for knock frequency. This is not exact - its a general rule of thumb, but your engine may vary based on physical design, knock sensor location, knock sensor type, bolt torque and so on. Typically however, these frequencies are within about 500Hz of being right.
Note that the range of modern sensors is about 3kHz to 22kHz - so analyse in that range but no wider as its of no use.
Right thats an introduction. Have a think and a play, and then ask the next lot of questions
