Knock sensing and correction
#31
Inventor
Rennlist Member
Rennlist Member
The J&S does not need its own hall sensor for individual cylinder control. It is using the input to the coil(s) to determine which cylinder to retard.
The unit only uses one knock sensor though. The front sensor location is easiest to use on pre-S4 intakes.
See "Operating Principle" below.
The unit only uses one knock sensor though. The front sensor location is easiest to use on pre-S4 intakes.
See "Operating Principle" below.
Last edited by PorKen; 01-02-2024 at 05:31 PM.
The following users liked this post:
davek9 (01-03-2024)
#32
Rennlist Member
I have a different question. Can these knock sensors be tuned? I think I've read that the knock frequency of steel sleeved cylinders is different.
I ask because I have three S4 blocks I'd like to sleeve.
Apologies if this is too off topic.
Kevin
I ask because I have three S4 blocks I'd like to sleeve.
Apologies if this is too off topic.
Kevin
#33
Rennlist Member
"There are 3 leads to each knock sensor, there is a shield, a ground and the "live" signal. The sensors are a piezo type sensor, and generate a signal when "squeezed". Or in the case of the 928 when a shock wave compresses the piezo element. So the tightening torque is important.
This signal generated is of the order of a 100mV. They are high impedance devices, but I don't believe oil on them should impair their efficiency. Water is a different matter."
Last edited by hernanca; 01-02-2024 at 09:15 PM.
#34
Rennlist Member
Most knock sensors are now wideband and you set the a filter to only examine a narrow band.
You need something like the Plex Knock Monitor to analyze the frequency response of the knock. The Plex does both visual representation as well as audio capture.
I am sure you could probably rent one from somewhere while configuring everything:
PLEX KNOCK MONITOR V3 - Plex Tuning (plex-tuning.com)
The J&S is a bit simplistic in its filtering strategy so it may not be overly helpful in identifying frequency and amplitude of knock.
You need something like the Plex Knock Monitor to analyze the frequency response of the knock. The Plex does both visual representation as well as audio capture.
I am sure you could probably rent one from somewhere while configuring everything:
PLEX KNOCK MONITOR V3 - Plex Tuning (plex-tuning.com)
The J&S is a bit simplistic in its filtering strategy so it may not be overly helpful in identifying frequency and amplitude of knock.
#36
I design and build the J&S knock controllers. Yes, J&S DOES appear to have a simplistic filtering strategy. The CONCEPT is very simple, but most of it is in software. If I told you how it works you would say why didn't I think of that. Because it's proprietary, it's harder to market, as I can't explain why it works so well.
Most OEM and high end knock systems use a brute force approach, with several tables that must be filled with data after extensive dyno time. I came up with a better way.
In '92, we visited MoTeC USA and let them drive our test vehicle. They liked the system so much they offered to buy processors with the knock code pre-loaded. We wanted to sell them an encapsulated module, they insisted on the bare chip. We felt the security risk was too high so we passed. In 2010 they released their knock detector, which uses the brute force approach, and requires extensive setup by a MoTeC dealer.
Check out this video of the Vampire on a Mini Cooper, posted by Bob at AutoBahn Power, a couple years after he purchased the unit:
Note that with the water methanol turned off, the unit detected and controlled knock that was missed by the Siemens ECU. When the water/meth was turned on, no detections were made, demonstrating the unit had responded to real knock, not noise.
Most OEM and high end knock systems use a brute force approach, with several tables that must be filled with data after extensive dyno time. I came up with a better way.
In '92, we visited MoTeC USA and let them drive our test vehicle. They liked the system so much they offered to buy processors with the knock code pre-loaded. We wanted to sell them an encapsulated module, they insisted on the bare chip. We felt the security risk was too high so we passed. In 2010 they released their knock detector, which uses the brute force approach, and requires extensive setup by a MoTeC dealer.
Check out this video of the Vampire on a Mini Cooper, posted by Bob at AutoBahn Power, a couple years after he purchased the unit:
Note that with the water methanol turned off, the unit detected and controlled knock that was missed by the Siemens ECU. When the water/meth was turned on, no detections were made, demonstrating the unit had responded to real knock, not noise.
Last edited by John Anthony; 01-04-2024 at 10:22 PM. Reason: More
#37
hernanca asked in a PM about using one or two sensors, as the knock circuit has a single input.
Louie Ott did some testing on his own to see the effect of connecting two sensors to the input. He used a spring loaded center punch to generate a standardized knock event. He then measured the signal at the input of the Vampire, with one or two sensors. He said the addition of the second sensor decreased the signal level by 30%, but felt that was outweighed by (theoretically) a better signal to noise ratio using two sensors, considering the sensitivity adjustment has more than enough range to compensate for the signal loss.
I have found sensor placement is not critical with my approach. I've done several different V8's using a single sensor, mounted either to the side of the block, or in an unused accessory mounting hole on either the side or front of a cylinder head.
For the beta install on a Ford Lighting (supercharged F150), due to time, I temporarily installed a sensor on a stud on the front corner of the driver's side cylinder head. I brought along a scope and a knock headphone amplifier with two outputs, so the owner and I could both listen and watch for knock, to verify the Vampire didn't miss any detections. It worked great, then two weeks later the owner connected to the centrally located factory sensor. Operation was unchanged. Interestingly, Lightnings were equipped with a factory knock sensor but Ford decided not to use them. The speculation is they couldn't get them to work with all the "blower whine". The only time I saw "blower whine" was on the forums.
I posted a photo on one of the forums showing the knock signal at the input to the Vampire's A/D converter. At freeway cruise, it was a flatline, but goose the throttle and the knock pulse is readily apparent.
Once inside the micro controller, the signal is run through a proprietary digital signal processing algorithm. If a detection is made, software retards proportionally the next time the knocking cylinder is ready to fire. Ten increments of retard are available, each increment is either one or two degrees, depending on a mode switch. Software allows up to seven increments of retard with a large knock event. Software re-advances at the rate of one increment every twenty revolutions
For units made before April of '08, the software would retard one, two, or three increments per knock event. We found on the Lightning this was too slow, as there was audible knock for a couple seconds, until the timing was retarded enough. The test truck was tuned for 91 octane and 19 psi boost, but we were testing on 89, lugging the engine at 1500 RPM, no boost. With the audible knock, I told the owner the unit wasn't designed to handle this, and was ready to throw in the towel. I drove home dejectedly, thinking about it, and came to the conclusion it just needed more range of authority. Everything was wonderful a week later.
The unit is triggered by the coil signals. The Vampire control logic is very cool, doesn't care if it's one coil or eight coils, just set the mode switches for the correct number of cylinders. To install, no wires are cut or intercepted. Each "control wire" is both an input and an output. It's an input when waiting for a coil trigger signal, then switches to output mode when retard is needed. To retard, the unit extends the dwell time, then fires the plug. What's really cool is any control wire can go to any coil and it still retards the correct cylinders. The stock ignition remains in place, and the engine will run normally if you unplug the Vampire. No bypass plugs or tach adapters are required.
I recently upgraded a few things. I moved the main harness connector to the rear panel, added an adjustable rev limiter and a proportional nitrous retard input. Most rev limiters operate by depriving the coil of a dwell signal. That's not possible with the Vampire, as the stock ignition remains connected to the coils. To solve this, the Vampire uses a "pulse swallowing" rev limiter. It gulps down the energy stored in the coil, preventing the spark. And it's soft touch. At the selected limit it fires two, skips one, then repeats, resulting in a one third drop in power. If you power through the limit by 100 RPM, it will switch to hard limiting, cutting as many cylinders as needed.
Louie Ott did some testing on his own to see the effect of connecting two sensors to the input. He used a spring loaded center punch to generate a standardized knock event. He then measured the signal at the input of the Vampire, with one or two sensors. He said the addition of the second sensor decreased the signal level by 30%, but felt that was outweighed by (theoretically) a better signal to noise ratio using two sensors, considering the sensitivity adjustment has more than enough range to compensate for the signal loss.
I have found sensor placement is not critical with my approach. I've done several different V8's using a single sensor, mounted either to the side of the block, or in an unused accessory mounting hole on either the side or front of a cylinder head.
For the beta install on a Ford Lighting (supercharged F150), due to time, I temporarily installed a sensor on a stud on the front corner of the driver's side cylinder head. I brought along a scope and a knock headphone amplifier with two outputs, so the owner and I could both listen and watch for knock, to verify the Vampire didn't miss any detections. It worked great, then two weeks later the owner connected to the centrally located factory sensor. Operation was unchanged. Interestingly, Lightnings were equipped with a factory knock sensor but Ford decided not to use them. The speculation is they couldn't get them to work with all the "blower whine". The only time I saw "blower whine" was on the forums.
I posted a photo on one of the forums showing the knock signal at the input to the Vampire's A/D converter. At freeway cruise, it was a flatline, but goose the throttle and the knock pulse is readily apparent.
Once inside the micro controller, the signal is run through a proprietary digital signal processing algorithm. If a detection is made, software retards proportionally the next time the knocking cylinder is ready to fire. Ten increments of retard are available, each increment is either one or two degrees, depending on a mode switch. Software allows up to seven increments of retard with a large knock event. Software re-advances at the rate of one increment every twenty revolutions
For units made before April of '08, the software would retard one, two, or three increments per knock event. We found on the Lightning this was too slow, as there was audible knock for a couple seconds, until the timing was retarded enough. The test truck was tuned for 91 octane and 19 psi boost, but we were testing on 89, lugging the engine at 1500 RPM, no boost. With the audible knock, I told the owner the unit wasn't designed to handle this, and was ready to throw in the towel. I drove home dejectedly, thinking about it, and came to the conclusion it just needed more range of authority. Everything was wonderful a week later.
The unit is triggered by the coil signals. The Vampire control logic is very cool, doesn't care if it's one coil or eight coils, just set the mode switches for the correct number of cylinders. To install, no wires are cut or intercepted. Each "control wire" is both an input and an output. It's an input when waiting for a coil trigger signal, then switches to output mode when retard is needed. To retard, the unit extends the dwell time, then fires the plug. What's really cool is any control wire can go to any coil and it still retards the correct cylinders. The stock ignition remains in place, and the engine will run normally if you unplug the Vampire. No bypass plugs or tach adapters are required.
I recently upgraded a few things. I moved the main harness connector to the rear panel, added an adjustable rev limiter and a proportional nitrous retard input. Most rev limiters operate by depriving the coil of a dwell signal. That's not possible with the Vampire, as the stock ignition remains connected to the coils. To solve this, the Vampire uses a "pulse swallowing" rev limiter. It gulps down the energy stored in the coil, preventing the spark. And it's soft touch. At the selected limit it fires two, skips one, then repeats, resulting in a one third drop in power. If you power through the limit by 100 RPM, it will switch to hard limiting, cutting as many cylinders as needed.
#38
Rennlist Member
Thank you for this reply John Anthony!
I especially appreciate the insight on Louie's considerations for using two knock sensors which is what I am hoping to do. And good to know that one knock sensor is sufficient as davek9 has set up with his supercharged 86.5 Auto and, as PorKen indicates, is the most straightforward set up for 85-86.5. (I have to be different. )
I especially appreciate the insight on Louie's considerations for using two knock sensors which is what I am hoping to do. And good to know that one knock sensor is sufficient as davek9 has set up with his supercharged 86.5 Auto and, as PorKen indicates, is the most straightforward set up for 85-86.5. (I have to be different. )
Last edited by hernanca; 01-09-2024 at 10:27 AM. Reason: Credits
The following users liked this post:
davek9 (01-09-2024)