Page 1 of 2

Knock Sensor

Posted: Tue Jan 02, 2007 10:45 pm
by Ni-Cad Steve
Does anyone have any data on engine knock sensors? They have been standard issue on automobiles for the last 20 years, yet there is little data available on how they work and what they are "listening" for. I am trying to graft 21st. century technology onto a 1940's era engine design which I suspect of knocking (badly). There are 9 ignition events on every crankshaft rotation. I do not know if the number of cylinders has any effect on how the system works. Any feedback would be greatly appreciated.

Posted: Wed Jan 03, 2007 12:00 am

Posted: Wed Jan 03, 2007 12:18 am
by Ni-Cad Steve
i get an error 404 ... Page could not be found

Posted: Wed Jan 03, 2007 12:25 am
by rshayes
What kind of engine is that? It sounds like an 18 cylinder radial aircraft engine.

Posted: Wed Jan 03, 2007 12:32 am

Posted: Wed Jan 03, 2007 1:04 am
by Ni-Cad Steve
It is in fact an 18 cylinder Wright R-3350(-93) engine with mechanical fuel injection. After a $175k engine overhaul last year, we replaced two more sets of pistons/cylinders (@ $54k/set) due to detonation. The net result was a racer that could not race and a pile of unserviceable parts. The boss is still a little pissed! The solution looks like it could be had with a relatively small number of (cheap) parts. I was wondering if the solution was as simple as a knock-sensor. Unfortunately there is relatively little data available on what a knock-sensor does and what the wave-form of a knock looks like. I have read some data that suggests that the knock looks like a spike in the 3kHz region. The article did not mention the number of cylinders in the test engine. Is the phenomenon independent of the number of cylinders? I am trying to figure out what the 'event' looks like so that I can 'capture' it in some form when it shows up.

Posted: Wed Jan 03, 2007 2:00 am
by rshayes
The automotive knock sensors appear to be piezoelectric microphones mounted on the engine block. One form of knock is preignition of the fuel charge in the cylinder. This creates audible sounds that can be heard and picked up by a microphone. In automobile engines, these signals go to the engine control computer which takes corrective action to minimize damage to the engine. Knocking may have become more significant when engine mixtures were made leaner to comply with smog and fuel economy requirements.

Knocking might be fairly easy to detect. I would look for a signal that indicates cylinders firing before the spark occurs. Other indication might be sharp or erratic signals, or signals of exceptionally high amplitude.

In the sixties, Tektronix marketed a scope equiped with sensors for engine analysis. One of these may have been a microphone or vibration sensor. The use of this equipment may have been described in either application notes or possibly in one of the technical books that they published about that time. If such a book was published, you might find it in the used book market.

This engine may have to be derated from its original horsepower unless high octane fuel is available. Using fuel with too low an octane rating can cause knocking, and in extreme cases, blow the cylinder heads off.

On a radial engine, you might need multiple sensors, possibly three or more on each bank. This would still be pretty cheap compared to your engine cost.

Posted: Wed Jan 03, 2007 6:47 am
by dyarker
I'd bet that engine needs at least 125 Av gas, maybe 145. If you're feeding it unleaded from the local station, then you WILL blow the engine. Knock detectors won't help if you don't feed it right.

Posted: Wed Jan 03, 2007 7:27 am
by Clyde Crashkop
I work on a 1955 Chris Craft boat with a Cadillac engine and 2 4 barrel carbs. One high speed trip burnt most of the spark plug electrodes and a hole in 1 new piston. The original fuel pump was too weak so it was way lean at high RPM. A high volume electric fuel pump solved the problem.

Posted: Wed Jan 03, 2007 8:50 am
by Chris Smith
Knock sensors are Piezo sensors with a tuned pick up in the form of the electronics/ computer.

They hear specific sounds and discriminate them into a lead and lag sensor on the ignition or fuel delivery.

To substitute high octane gas Toluene can be added.

I don’t know what the regulations are in your place but we used to purchase it by the 55 gallon drum. Im sure all the regulations have changed drastically so this may not be kosher any more?

Posted: Wed Jan 03, 2007 11:33 am
by haklesup
Seems to me that the acoustic signature from an 18 cylindar aircraft engine would be quite different from a 4, 6 or 8 cylinder auto engine. So any data you find on knock sensor waveforms won't fit very well.

If you go this way , I would consider the tried and true electronics T/S method of signature analysis. You're just going to have to attach a sensor and record waveforms and compare until you identify the signature of damaging event (assuming it has an acoustic report).

For recording you might get away with good audio gear maybe a DAT recorder for example. If you have good budget then a more robust data acquisition system is appropriate. An any case, analysis probably won't happen on an oscilloscope though this would be useful once you knew what it was and was moving on to corrective action. I would use a spectrum analyzer and based on your comment on the problem being at 3khz, this is what you need to see that (maybe FFT functions on some scopes). Analysis in programs like mathcad and other specialized signal processing programs is also possible.

If you're spending $175k on a rebuild, I might think $15k on a data Acq/logger and required sensors to monitor multiple performance parameters is not out of the question.

At 3kHz, I would expect the pilot or tech the hear the problem if it were intermittant. If its always been that way, he may not have the data to hear the signature either. Once you see it on a scope and hear it at the same time, you should spot it in an instant in the future.

In any case, on an oscilloscope, I would expect each spark event to have a sharp peak of a particular shape. Preignition (knock) would tend to widen and depress the peak or at least otherwise deform it possibly splitting it into two seperate ignition events (one from compression heating, and one from the spark). To synch this with something comprehensable, connect a timing light sensor (AC current sense xfmr) to a spark cable and display this on channel 2 (or the trig input) of the scope so you know precisely when the ignition should occur. Any sign of ignition before this time indicates preignition. With a 4 channel scope and 3 timing references from 3 equally spaced cylinders, you should be able to figure out which pulse is from each cylinder relative to the spark plug pulses.

It could also be as simple as using better fuel as suggested.

BTW, with restrictions from the EPA on VOC solvents and other restrictions from the DEA on solvents in quantity (thanks to meth cooks). Toluene is quite hard to get without good reason.

Posted: Wed Jan 03, 2007 1:47 pm
by Chris Smith
The knock sensor is capable of telling the difference between an ignition and a detonation coming from the cylinder.

The Knock of a 18 cylinder will be different from a 4 or 8 cylinder simply because of the material, thickness, and mount position of the sensor. It’s a sonic wave front.

However, a detonation and a combustion are quite different and one is almost accompanied immediately by the spark or ignition, the other occurs well before TDC happens.

This detonation is what destroys such things as the cylinder head, the bearings, or over heated valves due to the rise in immediate temperatures as well as the shock of detonation.

You can even set knock sensors to measure only one and not the other through active filtration.

Posted: Wed Jan 03, 2007 2:29 pm
by rshayes
A varient of that engine was used on the B-29 bomber during World War II. The B-29 version probably operated on 145 octane gas to get maximum output. Operation on lower octane might require modification to reduce the compression ratio. This may have been done for post-war versions of the engine.

Some of the technical work done on aircraft engines during World War II may still be available. The National Advisory Comittee for Aviation (NACA, which later became NASA) that did research and development work in this area. Their technical reports may still be available from some archive. I would try asking the librarians in one of the technical libraries at the University of California (Berkeley) if they can find where these reports are being stored and if copies can be obtained.

A college library may also have bound or microfilm copies of the journal published by the Society of Automotive Engineers. Some of the papers may document the development work on aircraft engines during the war. Several automotive production facilities were used for aircraft production at this time.

Posted: Wed Jan 03, 2007 3:02 pm
by Chris Smith
The easy way to modify the compression ratio is to add in extra head gasket thickness, then dome shape heads with swirl techniques, and finally fuel injection and electronic ignition.

Also the use of the second spark plug hole often opens the cylinder prematurely to evacuate the start of each cycle just slightly,.... as does a change in valve timing or configuration. [new cams]

Computer fuel delivery systems and electronic ignitions help as well as many other well known ideas.

NASA does have a wonderful data bank and a free subscription to the “NASA Tech Briefsâ€

Posted: Wed Jan 03, 2007 9:27 pm
by Ni-Cad Steve
Wow! there are so many good ideas I do not know where to begin. To "flesh-in" the detail ... here goes ...
The 3350-93 is not the same engine that appeared in the B-29. The -93 is mechanically fuel injected (like the Daimler DB-603) whereas the the bomber motor was carbuerated (but still fed a supercharger). The fuel in WWII was typically 145/160 octane (purple color), enhanced primarily with benzene. The fuel consumed for normal flying in the race plane is 100LL which is adequate for the task. For racing, we use the VP Racing Fuel with a performance number (octane) of around 160 (at $6.00/gallon - ouch!). This gets "hot-rodded" with the addition of CI3, a very high manganese solution which the Air Force abandoned in the 1960's because it was too hard to mix in the normal fueling process. This yields a performance number around 200. This allows manifold absolute pressures (MAP) in the range of 65 to 100 in-Hg. This high MAP increases (virtually guarantees) the problem of detonation; which is the reason I got on this board. We are currently using anti-detonation injection (ADI) to combat the detonation. The ADI is a mixture of methanol and water injected into the induction system downstream of the supercharger. The latent heat of evaporation of the ADI sucks the extra heat out of the combustion charge before it enters the cylinder. The problem with this engine installation is that you cannot hear a knock like you can in my '84 VW Rabbitt. So we are back to figuring out how you can listen for these events.