Following a conversation over the Christmas meal yesterday (why do all my products end up relating to food ? ) my brain started thinking of another possibly useful product.

The discussion was about standalone aftermarket ECU's and the almost universal lack of knock control on anything less than £1K. The only one I've personally had experience of recently bacame nearer £1.5K with the knock control added and IMO the knock control was VERY crude and unlikly to offer good protection on road style usage, only with constant hard driving in the map area prone to knock.

Anyway, it would be relatively simple to interrupt the IGT line from ECU to igniter (5V pulse) and add a variable delay to retard the ignition by a few degrees. This line would also provide and RPM measurement and a trigger for 'knock window'. Add a single analogue signal from the map sensor for load axis and the signal from the knock sensor and you have all you need to provide a very comprehensive knock control system.

In short, a small box to go next to the ECU whick will have 2x power wires and 2x signal wires to connect in to the loom at the back of the ECU, plus cut one wire at the back of the ECU and join each end to 2 further wires from the small box. The box will have a USB port to link to a PC with windows software to set it all up.

Finger in air, direct selling price likely to be about £200

What are peoples initial thoughts ?

Is there such a device already on the market that I'm unaware of ? No point re-inventing wheels.

Would it be useful to make it a 4 channel device for coil on plug ?

What knock control strategy to use ?
Initial thoughs are to split a 2D area of RPM vs load up into zones (cells) with each cell containing a 'retard value'. When knock is detected, immediately retard ignition by a few degrees depending on level and add a small figure e.g. 1/2 degree to the current cell in the map. Areas of the map where knock happens will quickly build sufficient retard to stop knock happeneing.
Also, how about advancing the cells back up again ? Do we just clear to zero on switch off, or do we gently advance back up again so we're running on the edge of kncok ? Modern cars seem to do this as there are plenty of claims of significantly increased MPG from running high octane fuels.

I think it's a given that 'knock sensed' will be based on exceeding a threshold based on RPM and load in a 2D table. This could easily be done in a 'learn mode' with ignition retarded by a couple of degrees to ensure zero det. and then run through the RPM range with zero load, medium load and full boost.

_________________
If at first you don't suck seed, try drier grain.

Hi,

on my project i want to do the same thing and linked to the ecu too

the problem is you don't know the knock barrier of the engine which means you have to bring the engine to a lot of different knocking points
that means a lot of R&D, money, engines and time to work, easy for a big company but customers like use didn't realy profit

going with the tuners expierence was the way i choose, every thing other just burns money or isn't a well developed or working thing was my personal awareness

greetz

_________________
KEEP COOL UNDER PRESSURE

street legal?.....perhaps in tschibuti

Good idea, especially for those already with a stand alone ECU, but I would go for one of these. I will be investing in one for my project build.

http://www.ecumaster.com/en/products/en ... ement-unit

Very reasonable priced too at around £600.

_________________
ST205

At first glance it looks like exceptional value for money. Keep us posted as to how it turns out.

_________________
If at first you don't suck seed, try drier grain.

I am a dealer for these ecumaster emu. We use one of each on our cars and have installed a few now

Sent from my GT-P5110 using Tapatalk 2

When I looked into this a few years back the products offered by these guys seemed interesting - http://www.phormula.co.uk/

_________________
Kris
95 ST205 - slept for 10 years, now waking up...
Avensis
http://www.gtfours.co.uk
_______________________________________
Turbocharging - the replacement for displacement

I am certain I tried to respond to this 2 or 3 times!

A very similar item is JS knock Guard - I believe Don knows more than me on this as he considered it and had researched it probably 5 years ago.

I ended up with a HKS Knock AFr meter http://www.nengun.com/hks/a-f-knock-amplifier as and in particular I understand it gates the knock signal for crank angle and outputs an analogue 0-5V signal (ECU input for ign retard) , Audible warning and has a headphone out socket for erghh the dr beats (as if I would)

I really like your idea Chris but I would like to analyse a knock signal more to ascertain if it could be accurately detected in relation to engine noise or sump pan rattles.

_________________
JP
GT4DC Treasurer

Grey St185 RC
Silver BMW E46 M3
Multicolour yawn Honda VFR400
Silver Honda Civic Type-R

Thanks for the feedback & links. I'm not a believer in re-inventing wheels so worth knowing what else is about.

Unless I've missed something, the phormula equipment only detects knock and proveds a 'knock detected' / 'knock level' output, it still relies on an external device to provide an actual knock control strategy.
Ditto the HKS knock/afr setup.

The JS devices look the most interesting as it does provide actual direct control and provide some 'knock retard' strategy, although simple and along the same lines as the link G4 ecu I was looking at.
Essentially, they respond 'reactively' to knock in that they wait until knock happens then retard for a period of time. The problem being, that occurrence is likely to be in one small area of the load/RPM map which will only happen every so often, most likely whenever the driver accellerates hard. This means that either the 'retard' must be removed VERY slowly leaving the engine running retarded elsewhere in the map, or it means that every time there will be knock which is then responded to. Still better than knock that isn't responded to, but not ideal IMO.

I'm proposing to split the map into individual areas, and only adding/removing the 'retard' while in that area of map. This means that in areas of map rarely used, the retard will stay for a long time once added and prevent serious knock from happening every time the engine goes into that area of map.

For an engine that has been mapped 'very safe' and the corrections reasonably accurate, it probably doesn't matter too much as it will be a very rare occurrence, most likely due to a fuel or other problem.

As Diceman says, I would like to see more evidence that the knock system can detect very low levels of knock before wanting to rely on it and map much closer to the ragged edge - although as I understand it most modern cars do just this in order to achieve maximum power and ecconomy. I think a session on the dyno will be called for once it's up & running.

Having put a bit more thought into this over the last few days, I'm thinking of providing a 0-5V 'retard' output as an alternative to intercepting the ignition as most ECU's allow ignition retard according to an analogue input. This removes any 'risk factor' from this device triggering ignition at a dangerous time and will also inherantly make it suitable for coil on plug / wasted spark setups.

_________________
If at first you don't suck seed, try drier grain.

Applying thought to 'knock window' and when it starts & ends, I would have thought it would start just after spark and end once all the mixture is alight. Would this be a constant time rather than angle as most seem to use ?

_________________
If at first you don't suck seed, try drier grain.

Essentially I think the crank angle is used as it is much simpler to implement.

Flame front speed will vary with mixture ratio, cylinder pressure and therefore crank angle, boost pressure and temperature. Allowing for that lot and writing the algorithms for the many mathematic formula will not only be complex but require rigorous testing to reveal any bugs.

I had wondered about using the frequency filtered differential between the "out of knock window" signal and the "in knock window" signal so as to provide an indicative engine noise map for comparison. Again it would require testing so ascertain if useful or not as I suspect engine noise would change during ignition and combustion anyway.

Back to the knock angle, again it would need a mule engine IMHO to monitor at what angles knock can occur depending on cause ( is the angle always the same regardless of cause?)

Personally I don't think I know enough to consider developing a strategy for knock detection but a quick google will probably provide research papers on angle/time of knock.

_________________
JP
GT4DC Treasurer

Grey St185 RC
Silver BMW E46 M3
Multicolour yawn Honda VFR400
Silver Honda Civic Type-R

This is a very good starting point for info and dispels a lot of common myths :-
http://www.streetrodstuff.com/Articles/ ... /index.php

The more I read the more I would be interested in reading in cylinder pressures to ascertain when our engines become "knock-limited" and can no longer maintain peak cylinder pressure at 14 degs ATDC. Did the Piezo spark plug knock detectors ever get cheaper?

_________________
JP
GT4DC Treasurer

Grey St185 RC
Silver BMW E46 M3
Multicolour yawn Honda VFR400
Silver Honda Civic Type-R

I remember datajon posting an article about looking at induced voltage on the spark plug in between sparks to measure a number of things going on inside. The chances of finding it are slim though.

Worth a look back at some of my past research on knock sensing.

viewtopic.php?f=13&t=2901

The standard sensor is fairly resonant and provides good frequency filtering. Because of this resonance, 'knock window' time filtering will be less effective, although looking at the above traces will probably be of some benefit.

Looking at some google results, it seems fairly typical to sample from 0 to 70 degrees crank angle. I would guess detonation will be most likely at around peak pressure (14 degrees) or soon after (as end gas temperatures rise).
The problem I have in this application is that I have no direct measurement of TDC, only spark point which will be between 10 and 40 degrees BTDC typically. It would be much much easier to sample for a fixed time after spark, slightly less easy to sample for a fixed angle after spark (calculate time for angle as we know RPM) but would require extra trigger inputs direct from crank / cam or ECU output (if ECU can be configured for this) to start from a fixed crank angle.

The other consideration is whether I want to measure individual cylinder knock (which would require cam trigger anyway) and then retard sparks individually for each cylinder. IMO this extra complexity is probably not going to be of great benefit unless there is a 'rogue' cylinder which detonates much easier. This also brings in another point of whether it is worth including 'engine condition monitoring' in the function. A classic example is in the linked waveforms, with the ringland failure clearly visible. I suspect impending big end failure would probably show up similarly.

As regards the electronics and detection methods, for the 'knock window' approach there is a handy chip available which I used in my ECU design (not tested this bit yet) which is probably favourable as it allows frequency and gain to be software selectable. TPIC8101 http://docs-europe.electrocomponents.co ... e7a9d7.pdf
This doesn't provide an audio out for det. cans however, so I may consider adding some circuitry to provide a headphone socket and maybe non-windowed level for simpler applications where it's difficult to get an accurate trigger signal.

_________________
If at first you don't suck seed, try drier grain.

Seems google is my friend

http://delphi.com/pdf/techpapers/2013-01-0354.pdf

http://www.aces.edu/~parmega/efi/temp/D ... ensing.pdf

http://www.max-boost.co.uk/max-boost/in ... tation.htm

http://www.ieeecss.org/CSM/library/1998 ... eTuner.pdf

_________________
If at first you don't suck seed, try drier grain.

A couple of other brain farts....

1) rather than remove timing advance would it be better to reduce boost? I mention this as the Saab "automatic performance control" module used this method to great effect in a road going car and it was very impressive in my old 9000 turbo barge. I wonder if this could help to also reduce the possibility of pre-ignition caused by detonation by lowering charge/cylinder temps.

2) I havent read through the links yet but renewed my memory of the discussions we had back in 2009 - I assume from your data you are able to ascertain crank angle (or cam angle) when knock occured and the time period after TDC or spark? What I am a little confused by and maybe need clarification on is from your measured king lead there appears to be 2 spikes per ign event? I assume that none correspond to the knock event and the Datajon mentioned spark plug ionisation event?

_________________
JP
GT4DC Treasurer

Grey St185 RC
Silver BMW E46 M3
Multicolour yawn Honda VFR400
Silver Honda Civic Type-R

1 - I've often thought that cutting boost would be a useful safety, however it is far from instant as you're waiting for a wastegate to open and then the turbine to slow down. The other problem would be the difficulty in accessing the boost control system to 'reduce' boost rather than just cutting with a relay contact in series with the solenoid. Easy if integrated with ECU and ECU controlled boost, and I'll probably look at this in my own ECU when the time comes.

2 - the scope probe was dangling near the king lead so VERY loose capacitive coupling of voltage or more likely rate of change of voltage. You are looking at a small spike when power is applied to the coil to charge, then a bigger spike when released which contines to 'ring' for a period of time due to the inductance / capacitance. I suspect the apparent double spikes on a couple of them are down to sampling errors on the 'scope. I doubt there was knock as the engine was idling at the time of this trace (30mS between ign = 1000RPM)

_________________
If at first you don't suck seed, try drier grain.