I was stimulated by replies to Vector's thread on repositioning the knock sensor to have a look at how they are actually monitored.
I couldn't believe that there is a simple threshold at which a knock event is declared by the ECM because that just isn't going to work. There's too much going on to say 'Yikes! 100 mV @4kHz - thats a knock that is'.
What (apparently) actually happens is
The ECU continuously calculates a background noise level over time.
This is a smoothed level of noise that places more emphasis on long-term changes to the noise level than short-term.
On top of the background noise level, a knock threshold adder is determined. This is a smaller value whose base is determined basically by the change in the background noise level.
The final knock threshold adder is determined by a weighting factor ("Knock Threshold Weighting Factor") which is based on RPM or RPM/load (depending on ECU).
This table appears to be one of the main mechanisms by which the OEM calibrators adjust the knock threshold for different engines (with some ECUs there are 4 tables, one for each cylinder).
The background noise + knock threshold adder determine the final threshold that must be exceeded in order for a knock event to be determined at any given time.
Basically, the faster the increase in noise and the larger the increase in noise, the more likely a knock event will be determined.

In other words this is an adaptive process of similar complexity to the fuelling computation.

 

Am sure there can be a variety of how ECU's deal with knowing when a knock has occurred (and I don't mean a large detonation, but as small as the ECU is calibrate for).

The text above you show isn't for a Jaguar, please see below some for a 4.2 Jaguar (please note that the 4.0 ECU might even be less advanced):

The V8 EMS has two knock sensors located in the V
of the engine, one per cylinder bank. The sensors are
connected to the ECM via a twisted pair.

The knock sensors produce a voltage signal in
proportion to the amount of mechanical vibration
generated at each ignition point. Each sensor monitors
the related cylinder bank.

The knock sensors incorporate a piezo-ceramic crystal.
This crystal produces a voltage whenever an outside
force tries to deflect it, (i.e. exerts a mechanical load on
it). When the engine is running, the compression waves
in the material of the cylinder block, caused by the
combustion of the fuel/air mixture within the cylinders,
deflect the crystal and produce an output voltage signal.

The signals are supplied to the ECM, which compares
them with `mapped' signals stored in memory. From
this, the ECM can determine when detonation occurs
on individual cylinders. When detonation is detected,

the ECM retards the ignition timing on that cylinder for
a number of engine cycles, then gradually returns it to
the original setting.

Care must be taken at all times to avoid damaging the
knock sensors, but particularly during removal and
fitting procedures. The recommendations regarding
torque and surface preparation must be adhered to. The
torque applied to the sensor and the quality of the
surface preparation both have an influence over the
transfer of mechanical noise from the cylinder block to
the crystal.


Considering that even the torque/surface for mounting the sensor is already of high importance, do you believe that mounting a knock sensor to the supercharger is a safe way, i.e. will yield the same results as compared to the stock location?

 

Irrespective of whether there is a fixed baseline, as in the case described by Avos, or a dynamic baseline as described by SteveInFrance, there is a static strategy map underlying the baseline.

The map has a dependency on known transmission quality. In other words, a known mechanical system which the strategy map was designed against. The mechanical system includes the factors of position and fastening.

Using epoxy to glue a knock sensor to some other location changes both position and fastening, as well as introducing a dampening medium.

Even if the system uses a dynamic background noise calculation, it has no way of determining that the delta above background has changed.

While it is true that a good knock sensor connected to the ecm and glued to the engine will not present codes, a good knock sensor connected to the ecm and placed on a pillow will also not present codes. In other words, the engine could be knocking like a blacksmith on an anvil and the ecm will not know the difference. That is generally not a good thing.

 

André - The text I quoted is for 'generic' ECMs. I doubt the Jaguar version will be less sophisticated.
Your quote says exactly the same - there's a map held in the ECM.
The signals are supplied to the ECM, which compares
them with `mapped' signals stored in memory. From
this, the ECM can determine when detonation occurs
on individual cylinders.

The ECM will also hold a long term averaged noise level as mentioned above.

I would imagine the torque is simply to ensure good physical contact with the head (no-one's mentioned fittng to the S/C as far as I know) without damaging the piezo-electric element.
A good gob of epoxy should give good acoustical transfer at the frequency ranges involved.
I don't think there's anything magical about the stock location - it's just a leftover from the N/A version.
The velocity of sound in aluminium is 6420 m/s so displacing the sensor from its current location to the extreme end of the head will introduce a delay of less than 70µsec while the cylinder impulses at 6000 rpm occur at 2.5 msec so that too makes no significant difference.
As I said before if/when I loose a K/S I'll put a scope on the replacement.

Plums, obviously installing the KS in the glovebox would not be productive. Gluing it to the cylinder head is a very different situation.
The difference between knock and normal engine noise is pretty dramatic and unlikely to be seriously attenuated by a millimeter of epoxy.
I believe all modern ECUs monitor background noise which they use both to assess KS function and as an 'automatic gain control' to give a baseline for the mapped RPM/KS voltages.
After all a KS mounted on the approved lug can go deaf and that needs detecting. The same tests will be applied to a KS mounted elsewhere and should produce a fault code too.
Fault codes 0327 to 0333 are generated for too high or too low an output from the KS. Incorrect torque of the mounting bolt can set both codes, low output from inadequate acoustic coupling and high output (presumably) from vibration.
Vector's glue job produced none of these codes so the ECU must be happy.
Some ECUs (don't know about Jag) also 'ping' the KS at key on by sending a string of square waves to it and looking for reflected pulses coming back to ensure it is working and acoustically coupled to something.
Complex little blighter isn't it?

 

Here is a simple test everyone can do who has a hammer and stethoscope.

Hit (not to hard please) on a certain spot on the heads (as constant as possible), then listen with the stethoscope to the noise at different places on the head, and last but not least listen also to the intercooler.

That will prove one thing that the amplitude of the sound wave will be different on different spots where you measure, which is the risk I am pointing out.

Then for fun, listen to the supercharger whilst the engine is running (and other spots on the heads), then you can only hope that these sounds do not skew up the info for the ECU even more.

The Denso ECU's Jaguar uses have evolved pretty big over the years, you can't compare one from 2010 to one from 2005 or even earlier, so a general statement about them is pretty tricky imho.

I can only say based on the knowledge I have, is not to relocate the KS as suggested, a risk not worth taking.

 

Two obvious improvements to the knock sensor:

1)Don't use wiring that will degrade at high temperatures. You'll be mounting it right next to the head, dumbass.
2)Don't leave that wiring sitting in a channel that will fill up with water when the coolant leaks cause the guy sitting next to you who will design the thermostat housing is even dumber than you, dumbass.

I'm wondering what the mileage in a small mod that would put a little tension on the wire and lift it away from there would be.