Programmable knock control systems
11-15-2007, 04:55 PM
#81
Registered
Join Date: Oct 2007
Posts: 57
Likes: 0
Received 0 Likes
on
0 Posts
Essentially a resonant sensor will only produce a significant signal in the bandpass area for which it is tuned but within that area it will provide a larger signal than a broadband sensor.
The broadband sensor, however, will provide a signal across the whole frequency range which is what you really want as then you can decide the frequencies of interest in the ECU by software controlling a dedicated knock chip such as the Bosch CC196.
http://www.semiconductors.bosch.de/e...ensorif-ic.asp
This is the better solution as then you can switch frequencies etc on the fly as a function of whatever you like eg speed, load etc etc. This will be how the modern EFI Mercs operate (and most modern automotive EMS for that matter). Resonant sensors are very late 80s early 90s when ECU processing power was at a premium.
Knock frequencies are very variable and are mostly best utilised at signiificantly higher than the 5-9kHz ranges of these sensors. For best discrimination, a frequency range of 11-14kHz gives best results for even large capacity engines.
The broadband sensor, however, will provide a signal across the whole frequency range which is what you really want as then you can decide the frequencies of interest in the ECU by software controlling a dedicated knock chip such as the Bosch CC196.
http://www.semiconductors.bosch.de/e...ensorif-ic.asp
This is the better solution as then you can switch frequencies etc on the fly as a function of whatever you like eg speed, load etc etc. This will be how the modern EFI Mercs operate (and most modern automotive EMS for that matter). Resonant sensors are very late 80s early 90s when ECU processing power was at a premium.
Knock frequencies are very variable and are mostly best utilised at signiificantly higher than the 5-9kHz ranges of these sensors. For best discrimination, a frequency range of 11-14kHz gives best results for even large capacity engines.
11-15-2007, 05:15 PM
#82
Registered
Join Date: Jun 2007
Posts: 11
Likes: 0
Received 0 Likes
on
0 Posts
Can someone send me that Delphi attachment? I don't have full privileges.
GM resonant knock sensors are not tuned to an exact frequency. GM calls them "broadband resonant".
In general, the larger the cylinder bore, the lower the knock frequency. GM only makes four different tuned sensors, though several thread variations and internal load resistors account for the many part numbers.
For example, GM's lowest frequency sensor is centered at 5.2kHz, with about a 1kHz bandwidth. An SAE paper by some of their knock sensor designers has a graph showing this sensor would work with bore diameters of about 100 mm (four inches) or larger.
Their next sensor is 6.0kHz, suitable for bores of about 85-100 mm. The next is 7.0kHz, and a new 8kHz sensor has been added.
You can see that GM doesn't tune to an exact knock frequency. Higher combustion temps will raise the knock frequency, so you can see that there is no "magic" knock sensor.
These tuned sensors produce 500mv/g, compared to 20mv/g for the non-resonant sensors.
Our knock controller uses a proprietary software algorithm, running on a Motorola processor.
The secured version of this processor is now hard to get, so all our systems are in redesign.
My first design with the new processor is called the Vampire, for coil on plug Fords.
No ETA on the Marine unit. First priority is an eight channel version for our customers using MoTec and Autronic ECU's.
If there is sufficient interest, I will try to get more of the older processors from my contact in Asia. I still have perhaps thirty blank circuit boards.
GM resonant knock sensors are not tuned to an exact frequency. GM calls them "broadband resonant".
In general, the larger the cylinder bore, the lower the knock frequency. GM only makes four different tuned sensors, though several thread variations and internal load resistors account for the many part numbers.
For example, GM's lowest frequency sensor is centered at 5.2kHz, with about a 1kHz bandwidth. An SAE paper by some of their knock sensor designers has a graph showing this sensor would work with bore diameters of about 100 mm (four inches) or larger.
Their next sensor is 6.0kHz, suitable for bores of about 85-100 mm. The next is 7.0kHz, and a new 8kHz sensor has been added.
You can see that GM doesn't tune to an exact knock frequency. Higher combustion temps will raise the knock frequency, so you can see that there is no "magic" knock sensor.
These tuned sensors produce 500mv/g, compared to 20mv/g for the non-resonant sensors.
Our knock controller uses a proprietary software algorithm, running on a Motorola processor.
The secured version of this processor is now hard to get, so all our systems are in redesign.
My first design with the new processor is called the Vampire, for coil on plug Fords.
No ETA on the Marine unit. First priority is an eight channel version for our customers using MoTec and Autronic ECU's.
If there is sufficient interest, I will try to get more of the older processors from my contact in Asia. I still have perhaps thirty blank circuit boards.
11-15-2007, 08:24 PM
#83
Registered
Hi John,
Email me at [email protected] with the discontinued "secure" processor and I'll reply with the pdf file posted, it's just a blurb of the new delphi sensors. I buy hard to find chips from china all the time and can add whatever you want to one of my orders.
Email me at [email protected] with the discontinued "secure" processor and I'll reply with the pdf file posted, it's just a blurb of the new delphi sensors. I buy hard to find chips from china all the time and can add whatever you want to one of my orders.
11-16-2007, 08:42 AM
#84
Gold Member
Gold Member
Join Date: Sep 2005
Location: Missouri
Posts: 1,987
Likes: 0
Received 0 Likes
on
0 Posts
PM/Email me your Email address and I will send a copy.
Bill
Bill
Can someone send me that Delphi attachment? I don't have full privileges.
GM resonant knock sensors are not tuned to an exact frequency. GM calls them "broadband resonant".
In general, the larger the cylinder bore, the lower the knock frequency. GM only makes four different tuned sensors, though several thread variations and internal load resistors account for the many part numbers.
For example, GM's lowest frequency sensor is centered at 5.2kHz, with about a 1kHz bandwidth. An SAE paper by some of their knock sensor designers has a graph showing this sensor would work with bore diameters of about 100 mm (four inches) or larger.
Their next sensor is 6.0kHz, suitable for bores of about 85-100 mm. The next is 7.0kHz, and a new 8kHz sensor has been added.
You can see that GM doesn't tune to an exact knock frequency. Higher combustion temps will raise the knock frequency, so you can see that there is no "magic" knock sensor.
These tuned sensors produce 500mv/g, compared to 20mv/g for the non-resonant sensors.
Our knock controller uses a proprietary software algorithm, running on a Motorola processor.
The secured version of this processor is now hard to get, so all our systems are in redesign.
My first design with the new processor is called the Vampire, for coil on plug Fords.
No ETA on the Marine unit. First priority is an eight channel version for our customers using MoTec and Autronic ECU's.
If there is sufficient interest, I will try to get more of the older processors from my contact in Asia. I still have perhaps thirty blank circuit boards.
GM resonant knock sensors are not tuned to an exact frequency. GM calls them "broadband resonant".
In general, the larger the cylinder bore, the lower the knock frequency. GM only makes four different tuned sensors, though several thread variations and internal load resistors account for the many part numbers.
For example, GM's lowest frequency sensor is centered at 5.2kHz, with about a 1kHz bandwidth. An SAE paper by some of their knock sensor designers has a graph showing this sensor would work with bore diameters of about 100 mm (four inches) or larger.
Their next sensor is 6.0kHz, suitable for bores of about 85-100 mm. The next is 7.0kHz, and a new 8kHz sensor has been added.
You can see that GM doesn't tune to an exact knock frequency. Higher combustion temps will raise the knock frequency, so you can see that there is no "magic" knock sensor.
These tuned sensors produce 500mv/g, compared to 20mv/g for the non-resonant sensors.
Our knock controller uses a proprietary software algorithm, running on a Motorola processor.
The secured version of this processor is now hard to get, so all our systems are in redesign.
My first design with the new processor is called the Vampire, for coil on plug Fords.
No ETA on the Marine unit. First priority is an eight channel version for our customers using MoTec and Autronic ECU's.
If there is sufficient interest, I will try to get more of the older processors from my contact in Asia. I still have perhaps thirty blank circuit boards.
11-16-2007, 11:07 AM
#85
Gold Member
Gold Member
Join Date: Sep 2005
Location: Missouri
Posts: 1,987
Likes: 0
Received 0 Likes
on
0 Posts
Ruaraidh,
This is very interesting. Are the 11-14kHz harmonics? Why does 11-14kHz give the best results for knock detection apparently regardless of engine size?
This is very interesting. Are the 11-14kHz harmonics? Why does 11-14kHz give the best results for knock detection apparently regardless of engine size?
11-16-2007, 12:35 PM
#86
Registered
Join Date: Jun 2007
Posts: 11
Likes: 0
Received 0 Likes
on
0 Posts
Rage: I have no PM privileges. Click on the hot link on my contact page.
The knock frequencies are fairly high, compared to most things we hear.
For example, here's a 1kHz tone:
http://www.mediacollege.com/audio/to...6bit_05sec.mp3
A 10kHz tone:
http://www.mediacollege.com/audio/to...6bit_05sec.mp3
When we hear "knock", we don't actually hear the knock frequency. We hear the "envelope" of the knock event, as clicks or rattles.
The envelope is the basic shape of the audio signal.
Scroll down a ways to see an example:
http://www.mathworks.com/products/de...tml?product=DS
Rage: I just received the file by email.
The knock frequencies are fairly high, compared to most things we hear.
For example, here's a 1kHz tone:
http://www.mediacollege.com/audio/to...6bit_05sec.mp3
A 10kHz tone:
http://www.mediacollege.com/audio/to...6bit_05sec.mp3
When we hear "knock", we don't actually hear the knock frequency. We hear the "envelope" of the knock event, as clicks or rattles.
The envelope is the basic shape of the audio signal.
Scroll down a ways to see an example:
http://www.mathworks.com/products/de...tml?product=DS
Rage: I just received the file by email.
Last edited by John at J&S; 11-16-2007 at 12:40 PM. Reason: more
11-16-2007, 04:02 PM
#87
Gold Member
Gold Member
Join Date: Sep 2005
Location: Missouri
Posts: 1,987
Likes: 0
Received 0 Likes
on
0 Posts
John,
What I would love to see is a frequency spectrum graph sweep of an engine's running background noise as the knock sensor signals it without engine knock and then with engine knock. Anything like that available?
What I would love to see is a frequency spectrum graph sweep of an engine's running background noise as the knock sensor signals it without engine knock and then with engine knock. Anything like that available?
Rage: I have no PM privileges. Click on the hot link on my contact page.
The knock frequencies are fairly high, compared to most things we hear.
For example, here's a 1kHz tone:
http://www.mediacollege.com/audio/to...6bit_05sec.mp3
A 10kHz tone:
http://www.mediacollege.com/audio/to...6bit_05sec.mp3
When we hear "knock", we don't actually hear the knock frequency. We hear the "envelope" of the knock event, as clicks or rattles.
The envelope is the basic shape of the audio signal.
Scroll down a ways to see an example:
http://www.mathworks.com/products/de...tml?product=DS
Rage: I just received the file by email.
The knock frequencies are fairly high, compared to most things we hear.
For example, here's a 1kHz tone:
http://www.mediacollege.com/audio/to...6bit_05sec.mp3
A 10kHz tone:
http://www.mediacollege.com/audio/to...6bit_05sec.mp3
When we hear "knock", we don't actually hear the knock frequency. We hear the "envelope" of the knock event, as clicks or rattles.
The envelope is the basic shape of the audio signal.
Scroll down a ways to see an example:
http://www.mathworks.com/products/de...tml?product=DS
Rage: I just received the file by email.
11-16-2007, 04:42 PM
#89
Registered
Join Date: Jun 2007
Posts: 11
Likes: 0
Received 0 Likes
on
0 Posts
http://www.ni.com/pdf/csma/us/collins.pdf
I have many systems installed on Porsche. Some have raced at Sebring. I don't know this guy. Too bad, since he says he lost an engine running Pike's Peak. Detectors are ok, controllers are better.
RichardCranium572 sent me the file on the Delphi sensor. Looks very much like a Bosch broadband, which is what I used on the Porsche, more recently on the Ford Mustang, with our new Vampire system for coil on plug Fords. Click the Vampire link at the top of my home page.
Snow Performance included us in their article in this December's issue of Muscle Mustangs and Fast Fords (on stands now). Article begins on page 180, written primarily to showcase Snow's $100 water injection flow meter alarm, which alerts the driver of a clogged nozzle, etc.
Instead, they wired the alarm into the switched retard input of the J&S Vampire, forcing it to retard whether it was knocking or not, if the water injection stopped flowing under load.
I asked the writer about doing a test of our unit by itself, and he said the editor had already shot that down, saying the product didn't have enough oomph (forget what he actually said) to warrant its own story.
I think Snow buys ads. We don't.
I have many systems installed on Porsche. Some have raced at Sebring. I don't know this guy. Too bad, since he says he lost an engine running Pike's Peak. Detectors are ok, controllers are better.
RichardCranium572 sent me the file on the Delphi sensor. Looks very much like a Bosch broadband, which is what I used on the Porsche, more recently on the Ford Mustang, with our new Vampire system for coil on plug Fords. Click the Vampire link at the top of my home page.
Snow Performance included us in their article in this December's issue of Muscle Mustangs and Fast Fords (on stands now). Article begins on page 180, written primarily to showcase Snow's $100 water injection flow meter alarm, which alerts the driver of a clogged nozzle, etc.
Instead, they wired the alarm into the switched retard input of the J&S Vampire, forcing it to retard whether it was knocking or not, if the water injection stopped flowing under load.
I asked the writer about doing a test of our unit by itself, and he said the editor had already shot that down, saying the product didn't have enough oomph (forget what he actually said) to warrant its own story.
I think Snow buys ads. We don't.
11-17-2007, 04:38 PM
#90
Registered
Join Date: Oct 2007
Posts: 57
Likes: 0
Received 0 Likes
on
0 Posts
Generally, combustion (as well as knock) noise is in the region of up to 9kHz.
Higher frequencies than that are multiples but they are also directly related to knock rather than valvetrain clatter or combustion noise (ie rate of pressure rise even when not knocking).
Last engine I did with broadband knock sensors and bosch knock control (>4" bore p'rod V8) turbo had great response in the 9kHz region whether or not the engine was knocking. ie it was difficult to discriminate combustion noise from knock. At higher filter frequencies (which you can select via the software), you can precisely identify knock irrespective of other noise. ie higher filter frequencies were more discriminating. It's not the fundamental kncok frequency that's important, it's the one that's most robust.
This is what most OEM knock control systems will do. They are calibrated by logging in-cylinder pressure and knock sensor output at very high data rates simultaneously. You can then compare both across the whole speed and load range at various states of knocking and non-knocking combustion and then analyse the ideal calibration approach.
Takes a lot of time to do on the dyno with specialised equipment. From what I've heard (from a colleague) the J&S system works very well and is easy to set up.
Higher frequencies than that are multiples but they are also directly related to knock rather than valvetrain clatter or combustion noise (ie rate of pressure rise even when not knocking).
Last engine I did with broadband knock sensors and bosch knock control (>4" bore p'rod V8) turbo had great response in the 9kHz region whether or not the engine was knocking. ie it was difficult to discriminate combustion noise from knock. At higher filter frequencies (which you can select via the software), you can precisely identify knock irrespective of other noise. ie higher filter frequencies were more discriminating. It's not the fundamental kncok frequency that's important, it's the one that's most robust.
This is what most OEM knock control systems will do. They are calibrated by logging in-cylinder pressure and knock sensor output at very high data rates simultaneously. You can then compare both across the whole speed and load range at various states of knocking and non-knocking combustion and then analyse the ideal calibration approach.
Takes a lot of time to do on the dyno with specialised equipment. From what I've heard (from a colleague) the J&S system works very well and is easy to set up.
