knock sensor output
#46
Addict
Rennlist Member
Rennlist
Site Sponsor
Rennlist Member
Rennlist
Site Sponsor
Klaus , so based on your comments the stroke of the engine is NOT a part of the calculation but bore size IS. How much change would occur going from a bore of 100 mm to 104.5 mm or is that insignificant ? or still with in the sensing range of the stock knock sensors? Since combustion temperature is part of the formula a boosted engine I would very simplistically think generates higher temperatures from a bigger bang ?
#47
Addict
Lifetime Rennlist
Member
Lifetime Rennlist
Member
Thread Starter
Its amzing the Knowledge that can jump out of the wood work! Great stuff thanks.
Some of you guys scare me!
However, anything that ends in "ithm" or has units of "ms" is selcetively tuned out from my freq of selctive listening..it was in Highschool and college atleast.
So ....would it be possible to hook up a device to see when it is occuring, that is my main objective. Id like to see if i can find out where boost, inlet temp, and fuel pressure, in a rough range, cross that Knock threshold.
Some of you guys scare me!
However, anything that ends in "ithm" or has units of "ms" is selcetively tuned out from my freq of selctive listening..it was in Highschool and college atleast.
So ....would it be possible to hook up a device to see when it is occuring, that is my main objective. Id like to see if i can find out where boost, inlet temp, and fuel pressure, in a rough range, cross that Knock threshold.
#48
Hi,
Yes, stroke does not come into play because the sound waves during knock traverse the cylinder bore. I doubt that going from 100mm to 104.5 does much to the detection. These early ECU's used relatively flat filters (mechanical or electronic). Boosting does change combustion temperatures somewhat, but AFR is more significant. Accoustic knock sensing is by nature very inaccurate. That's why I posted in our LM-1 forum that a simple headphone amplifier connected to the knock sensors is much better. The human 'sound processing system' has not been matched yet by any computer in regards to recognition of various sound patterns.
Also, knock from preignition is typically not detected by knock sensors. The reason is that typical ECUs 'window' the detection to between 10 deg ATDC to ~60 deg ATDC where normal knock is most likely. Preignition can happen BTDC. Because piston slap can easily mimick the same characteristic sound as knock, the signal is windowed to not detect that.
The reason knock sensing frequency curves are fairly flat is also because depending on which crank degree knock occurs the frequency can change. The piston moving downward expands the gas with a corresponding lowering of gas temperature. Because of the broad crank angle window where knock can occur, sharp frequency filters would limit the detection range.
Regards,
Klaus
Yes, stroke does not come into play because the sound waves during knock traverse the cylinder bore. I doubt that going from 100mm to 104.5 does much to the detection. These early ECU's used relatively flat filters (mechanical or electronic). Boosting does change combustion temperatures somewhat, but AFR is more significant. Accoustic knock sensing is by nature very inaccurate. That's why I posted in our LM-1 forum that a simple headphone amplifier connected to the knock sensors is much better. The human 'sound processing system' has not been matched yet by any computer in regards to recognition of various sound patterns.
Also, knock from preignition is typically not detected by knock sensors. The reason is that typical ECUs 'window' the detection to between 10 deg ATDC to ~60 deg ATDC where normal knock is most likely. Preignition can happen BTDC. Because piston slap can easily mimick the same characteristic sound as knock, the signal is windowed to not detect that.
The reason knock sensing frequency curves are fairly flat is also because depending on which crank degree knock occurs the frequency can change. The piston moving downward expands the gas with a corresponding lowering of gas temperature. Because of the broad crank angle window where knock can occur, sharp frequency filters would limit the detection range.
Regards,
Klaus
#49
Shameful Thread Killer
Rennlist Member
Rennlist Member
No FFT? I'm bummed, FFT is the way to go when vibration acoustics is the problem. We we're doing that with hardware in the mid '70s at Spectral Dynamics, but maybe it does just use amplitude of the piezio output.
BTW, I made another mistake, it's not a 2ms sample time, because we have two modules for the engine and each supports four cylinders, so the sample window would double to 4ms, and would also have an offset due to the V of the cylinders. Oops, I just crashed the probe into mars again. We are way off the topic of this thread. Did the poster get what he needed from any of this?
BTW, I made another mistake, it's not a 2ms sample time, because we have two modules for the engine and each supports four cylinders, so the sample window would double to 4ms, and would also have an offset due to the V of the cylinders. Oops, I just crashed the probe into mars again. We are way off the topic of this thread. Did the poster get what he needed from any of this?
#50
Did the poster get what he needed from any of this?
But the discussion is fun regardless and hopefully useful
Regards,
Klaus
#52
Shameful Thread Killer
Rennlist Member
Rennlist Member
Originally Posted by klatinn
I hope. What I wanted to convey is that there just isn't a simple generic way to extract a loggable signal from accoustic knock sensors. It's a complex subject and needs to be tuned to the engine characteristics to get any meaningful data.
But the discussion is fun regardless and hopefully useful
Regards,
Klaus
But the discussion is fun regardless and hopefully useful
Regards,
Klaus
Yes, I should have admitted that from the get-go. Modern data loggers are fine for multi-probe, strobing of volts, temp, pressure, whatever. But not suitable for acoustic pickup and analysis. The long stethoscope idea was cute!
These guys;
http://www.dspt.com/Products.htm
or;
http://www.valuetronics.com/Details.cfm?ProdID=2166
will do what you are looking for, but it's not cheap and not for the faint of heart. The instruments are more suited for axial spectronomy runout (like a turbine shaft) but can be used for recip if configured right.
Doc
#53
Speake said that EZKs don't break too often
My EZK blew a few chips I think. I sold it and found out from the buyer ;( It turned out to ben 86 board in an 89 box - very weird - and had some manual soldering.... SOmeone was in there and wasn't nice.
I'm of the thinking that the manuals are very often worded incorrectly, and though it could be possible, the 928S4 is not retarding individual cylinders. I really thing that was not the point nor the ability of the system.
#54
Addict
Rennlist Member
Rennlist Member
Join Date: Oct 2003
Location: Gone. On the Open Road
Posts: 16,508
Received 1,648 Likes
on
1,075 Posts
This is an excellent thread!
John has hit the nail on the head. Given the hardware on and about the 32v S4 lump there are a lot of interesting things than can be done - IF you have the compute cycles to burn and the acumen to encode (and test!) the software. The key question is what is actually in the software.
Perhaps. Perhaps not. MarketingThink is a mostly-foreign proposition to engineering folks.
Again perhaps and perhaps not. The extra bits of memory and code to store and output the fault codes might not have fit into whatever space was available. I assume John that you've pulled back the covers on these ECUs, so you're likely to have a very good idea of how much storage (EPROM one would assume) is available. 1k? 4k? There's a lot of bits to squeeze on to them.
I agree. I have some experience in puzzling out machine code. However, you'd need both the code and a full schematic. With those in hand it becomes simple work - not rocket science. Legend has it, however, that the 'code' in the LH and EZK has some basic protection against reverse engineering. Any insight on that?
Quite likely. One would expect some-sort of 8-bit MCU as the guts. With the markings on the bug one could look up the specs in a data book and know for sure.
Don't worry about it. BTDT too. So far, we're still flying in the simulator so no worries.
Originally Posted by John Speake
but the real unknown is if the software is there in the 928 EZK.
I think if the 928 had featured individual cylinder igntion control, Porsche would have mentioned it
If the EZK was capable if identifying which cylinder or group of cylinders were knocking, then there would have been no reason not to output a specific fault code to help fault finding.
Otherwise a full analysis of the software would reveal all, but that also is a significant task to undertake.
Originally Posted by klatinn
Hi,
I doubt very much that the 928 ECU uses 'fast fourier transforms' to extract the knock signal. The processing power to execute a FFT in realtime with engine RPM is quite significant and just not available in the 928 ECU.
I doubt very much that the 928 ECU uses 'fast fourier transforms' to extract the knock signal. The processing power to execute a FFT in realtime with engine RPM is quite significant and just not available in the 928 ECU.
Originally Posted by docmirror
Oops, I just crashed the probe into mars again.
#55
Originally Posted by docmirror
I don't know which, or maybe both artifacts the Porsche 'knock' sensor is intended to capture. They are different artifacts, but share similar signatures, and cause similar damage. My guess is that they are capturing 'preignition' which is the most destructive, and the easiest to fix through ignition retarding. Usually, detecting and curing detonation is much more difficult, and also much more costly to implement.
Originally Posted by Tony
Id like to see if i can find out where boost, inlet temp, and fuel pressure, in a rough range, cross that Knock threshold.
I may end up trying to do some individual cylinder ignition timing monitoring once spring gets here, to see if this individual cylinder timing retard question can finally get a definite answer. I'm thinking that I'll just keep advancing the timing slowly until some timing retard by the EZK occurs, and then see if it's only one cylinder, a bank, or all of them being retarded.
There's some information on a do it yourself ion sensing knock detection setup at:
DIY ion sensing
Originally Posted by Tony
However, anything that ends in "ithm" or has units of "ms" is selcetively tuned out from my freq of selctive listening..it was in Highschool and college at least.
Knock listener
#56
Hi,
Just so we are talking about the same thing:
1. Knock occurs AFTER the mixture has been ignited by the spark. The resulting burn causes a pressure rise. Pressure rise causes temperature rise. If the temperature rise is high enough, autoignition can occur in some unwanted hot part of the combustion chamber. This causes a secondary flame front and more and faster pressure rise, which can then cause an avalanche effect where all the remaining mixture burns very fast -> knock. Can be prevented by retarding the spark so that the pressure rise BTDC is lower.
2. Preignition occurs when parts of the combustion chamber (or carbon deposits or hot exh. valve) are hot enough to cause ignition during the compression stroke. This happens independent of the spark and can't be prevented by retarding. Usually more damaging.
Can happen even at cruise or idle where it does not neccessarily lead to knock (as defined above) but makes for a poor running engine.
Regards,
Klaus
Just so we are talking about the same thing:
1. Knock occurs AFTER the mixture has been ignited by the spark. The resulting burn causes a pressure rise. Pressure rise causes temperature rise. If the temperature rise is high enough, autoignition can occur in some unwanted hot part of the combustion chamber. This causes a secondary flame front and more and faster pressure rise, which can then cause an avalanche effect where all the remaining mixture burns very fast -> knock. Can be prevented by retarding the spark so that the pressure rise BTDC is lower.
2. Preignition occurs when parts of the combustion chamber (or carbon deposits or hot exh. valve) are hot enough to cause ignition during the compression stroke. This happens independent of the spark and can't be prevented by retarding. Usually more damaging.
Can happen even at cruise or idle where it does not neccessarily lead to knock (as defined above) but makes for a poor running engine.
Regards,
Klaus
#57
Rennlist Member
I assume John that you've pulled back the covers on these ECUs, so you're likely to have a very good idea of how much storage (EPROM one would assume) is available. 1k? 4k? There's a lot of bits to squeeze on to them.
>>>>>>>>>>>>
Hi Dave C -
They are 16k EPROMs. There is some space still unused on the LH, and more on the EZK.
>>>>>>>>>>>>>>
Quote:
Otherwise a full analysis of the software would reveal all, but that also is a significant task to undertake.
I agree. I have some experience in puzzling out machine code. However, you'd need both the code and a full schematic. With those in hand it becomes simple work - not rocket science. Legend has it, however, that the 'code' in the LH and EZK has some basic protection against reverse engineering. Any insight on that?
>>>>>>>>>>>
The LH code is not deliberately protected, but it is not nicely arranged, and it is quite difficult to sort out what is going on, for some of the more obscure fucitions I have no knowledge of the EZK code (yet)
>>>>>>>>>>>>>>
"Z" do you have the equipment to modify the EZK mapping, as you suggest ?
Regards
>>>>>>>>>>>>
Hi Dave C -
They are 16k EPROMs. There is some space still unused on the LH, and more on the EZK.
>>>>>>>>>>>>>>
Quote:
Otherwise a full analysis of the software would reveal all, but that also is a significant task to undertake.
I agree. I have some experience in puzzling out machine code. However, you'd need both the code and a full schematic. With those in hand it becomes simple work - not rocket science. Legend has it, however, that the 'code' in the LH and EZK has some basic protection against reverse engineering. Any insight on that?
>>>>>>>>>>>
The LH code is not deliberately protected, but it is not nicely arranged, and it is quite difficult to sort out what is going on, for some of the more obscure fucitions I have no knowledge of the EZK code (yet)
>>>>>>>>>>>>>>
"Z" do you have the equipment to modify the EZK mapping, as you suggest ?
Regards
#58
Addict
Rennlist Member
Rennlist Member
Join Date: Oct 2003
Location: Gone. On the Open Road
Posts: 16,508
Received 1,648 Likes
on
1,075 Posts
Originally Posted by docmirror
The knock sensors of the older variety are piezio-electric crystal, mated to a hard phenolic plate that transmits vibration well(I guess a mechanical microphone is a good simile).
Originally Posted by Z
Some of what I've read on the subject says that preignition is not necessarily detectible by sound, but is what's much more likely to cause significant engine damage.
On a slightly tangential note: Since the sensors are piezio-electric would anyone like to opine on the effect on the sensors' functionality of over-torquing the bolts that fasten them to the block? I would think that over-torquing them gorilla-style would compromise their ability to produce signal.
#59
Addict
Rennlist Member
Rennlist Member
Join Date: Oct 2003
Location: Gone. On the Open Road
Posts: 16,508
Received 1,648 Likes
on
1,075 Posts
Originally Posted by John Speake
They are 16k EPROMs. There is some space still unused on the LH, and more on the EZK.
The LH code is not deliberately protected, but it is not nicely arranged, and it is quite difficult to sort out what is going on, for some of the more obscure fucitions
#60
Rennlist Member
QUOTE:
On a slightly tangential note: Since the sensors are piezio-electric would anyone like to opine on the effect on the sensors' functionality of over-torquing the bolts that fasten them to the block? I would think that over-torquing them gorilla-style would compromise their ability to produce signal."
>>>>>>>>
The W/S manual has some notes on correct bolts, washers and torque ..
On a slightly tangential note: Since the sensors are piezio-electric would anyone like to opine on the effect on the sensors' functionality of over-torquing the bolts that fasten them to the block? I would think that over-torquing them gorilla-style would compromise their ability to produce signal."
>>>>>>>>
The W/S manual has some notes on correct bolts, washers and torque ..