Shift Light

[tvitarelli]

Has anyone tried this shift light? It seems like a great product.

http://www.ecliptech-us.com/index.php?option=com_content&view=article&id=59&Itemid=109

[BAC5.2]

$146 doesn't seem terrible. I'd be interested to see it in action. The colors don't jive with the interior, and I don't know where it would discretely mount without being annoying.

 

I think I might get a TurboXS Knocklite and mount it in one of the vents. Clean and simple, plus it flashes at knock!

[TSiWRX]

^ Just remember that the knock-light is essentially a visualization of a listening device.

 

It just reports "noise" above a set threshold.

 

Remember that it's only one tool, one potential indicator.

[BAC5.2]

^ Just remember that the knock-light is essentially a visualization of a listening device.

 

It just reports "noise" above a set threshold.

 

Remember that it's only one tool, one potential indicator.

 

Right... But it operates off of the same sensor that informs the ECU that the engine is knocking. It flashes under the same conditions that cause the ECU to log knock...

[TSiWRX]

^ Sorta.....

 

What the ECU interprets as knock is based on a much more complex algorithm. The knock-light is pure audible-feedback based, and this "reporting" is further affected by where you set that threshold to be.

 

Basically, you can have situations where the ECU "sees knock," but because you've set your threshold too high, the light does not report. The alternative situation, where you've set the light's threshold too low, so that it "falses" reports without the ECU actually seeing knock events, can very well also present.

 

Here's some good discussion of the TurboXS, from way back in '05/'06:

 

http://legacygt.com/forums/showthread.php?t=17486&highlight=turboxs

 

Note that I can only guaranty the pinout schematic for '05s.

 

BTW, an important tech tip: what we've learned, collectively, is that the unshielded wires for this unit can cause problems. Be sure that you route the knock lead away from anything that has a potential to produce strong interference (i.e. stereo, HVAC, etc.), as well as to either try to shield the included wiring or replace with shielded wire. Enough interference can cause the knock signal to feed back to the ECU that it'll interpret false knock, and this can be severe enough that the ECU will actually pull the dynamic modifier and drop the car to wastegate boost only!

 

As you can see from that old thread, I've got my threshold set fairly high.

 

This is because I don't want to worry about false-positives. I datalog the car routinelhy enough that I'm not really worried about something just going awry, short of catastrophic mechanical failure.

 

Given that, my true goal was to have some ability to more or less instantly visualize any fuel-based issues (over the past 5 years or so, I've only gotten "skunked" gas, from my preferred station, 3 times - but each time, it was this little device that first alerted me to something not being quite right) - as well as, of course, to have a single-set-point shift light, which I desperately wanted before I got the cat-back on my car, as it was so silent (I came from a long list of Fast and Furious DSMs ) that I'd occasionally bump into the limiter in 2nd and even 3rd.

 

You'll see that, in the referenced thread, other brothers here have chosen to set theirs on a lower threshold to err on the side of safety. The effectiveness of that particular approach, aside from the false-positives, is also limited on a "per unique vehicle" basis - some are just "noisier" than others, for whatever reason, and such a low setting may not be viable.

 

Know the limitations of the device.

[BAC5.2]

I think it's more the limitations of the operator, rather than the device, right? You said yourself you have your sensitivity set really low (to prevent overactivity). So, in your case, the knocklite is completely unreliable for detecting knock, right? Light, or even moderate, knock will go undetected by the light. Luckily, the Knocklite is NOT intended for tuning purposes (nor did I ever claim it was).

 

The current revision of the knocklite records "no-knock" input (whcih can be done at cruise if desired), and compares current knock signal to the reference input for every few RPM. Significant variation turns on the light. Actually, if you watch the sensor output, the ramp up to significant variation is what triggers the light, not necessarily a passing of the threshold. I suppose, theoretically, any rapid variation would trigger knock, even if it doesn't pass the threshold.

 

The Knocklite has nothing to do with engine management. If it detects knock, it flashes. Whether or not the engine is actually knocking is irrelevant. The only result of the knocklite registering knock, is showing you. Sensitivity of the engine management is more important than sensitivity of the light. Obviously, whatever engine management you are using should be sufficiently tuned to react appropriately to knock. If the Knocklite flashes, it would definitely warrant looking into the ECU.

 

The BEST way to have a visual representation of Knock is with a UTEC. Flashie-Flashie of the CEL is opportunity knocking.

[TSiWRX]

^ Ah, I think that we're getting crossed-up in our approach to this.... We're both on the same page (although you do go into more technical detail, whereas I'm more approaching it on a lay level), but we're both thinking the same:

 

I think it's more the limitations of the operator, rather than the device, right? You said yourself you have your sensitivity set really low (to prevent overactivity). So, in your case, the knocklite is completely unreliable for detecting knock, right? Light, or even moderate, knock will go undetected by the light. Luckily, the Knocklite is NOT intended for tuning purposes (nor did I ever claim it was).

 

Yes and no.

 

Yes, you didn't suggest that the Knocklight is intended for tuning purposes - I did not mean to imply that, and if I came across as having suggested that you did, I truly apologize. I've been sitting here with my toddler most of the day, rushing her to the potty (we're both home with a little GI bug), so I may have become somewhat muddled in my thinking. And also, that's not what I meant to say, either, that, indeed, this device is not intended for such a purpose.

 

Yes, there's significant operator/user preference involved.

 

Set high, yes, light-to-moderate knock may not be visualized.

 

But set low, the threshold is so low that even "proper" engine noise will set it off as a light-show.

 

In my case, yes, it would be unreliable for detecting low-level knock, but that's not what I use it for - that's protection that I would expect from both the map that the car's tuned with, as well as the datalogging that I routinely undertake. I've got a good map from a reliable, well-known tuner who spent a lot of time with the car. I often datalog my daily commutes. My aim is to have the Knocklight as a fail-safe so that when I do get skunked gas, I know it, right off the bat.

 

I'm fond of saying that even one knock can kill, and, for me, this is simply one way to guard against one particular threat, that's all.

 

The current revision of the knocklite records "no-knock" input (whcih can be done at cruise if desired), and compares current knock signal to the reference input for every few RPM. Significant variation turns on the light. Actually, if you watch the sensor output, the ramp up to significant variation is what triggers the light, not necessarily a passing of the threshold. I suppose, theoretically, any rapid variation would trigger knock, even if it doesn't pass the threshold.

That I did not know - I was not aware that it was a delta which caused the light to trip, rather than an absolute threshold. Thank you for educating me on that one. I didn't realize that this is what they meant by "variable noise profile." I thought that was just fancy marketing speak. It's good to know that there's more there.

 

I'm not sure that variations sub-threshold would trigger the light, though.

 

TXS seems to have expanded their website description of this item quite a bit since the early days, but the graphical representation still seems to suggests an absolute threshold as being the trigger....

 

Not sure on that, though, from my end, either - I guess we're both left to wonder, on this one.

 

The Knocklite has nothing to do with engine management. If it detects knock, it flashes. Whether or not the engine is actually knocking is irrelevant. The only result of the knocklite registering knock, is showing you. Sensitivity of the engine management is more important than sensitivity of the light. Obviously, whatever engine management you are using should be sufficiently tuned to react appropriately to knock. If the Knocklite flashes, it would definitely warrant looking into the ECU.

Correct - the Knocklight is independent of EM, but the Knocklight isn't necessarily a definitive representation of what's going on.

 

Set low enough, it'll be a light-show.

 

Set high enough, it may not catch true low-level, or even moderate, knock-events.

 

You're absolutely right - when the knock-light thinks that it sees audible noise, via the knock sensor, that surpasses what it thinks to be "proper" for the engine at the preset leve, it will report.

 

But this is in no way reflective of what the ECU thinks, this device does not receive feedback from the ECU, only the knock sensor, and that's the point that I've been trying to get across.

 

The limitations of this device is one based heavily on the user-set threshold, but even when set as carefully as possible, the possibility of false-positives as well as false-negatives cannot be overlooked as it it solely a "listening device" based off of the factory knock sensor (if installed in the fashion described in the reference post above, without its dedicated sensor), and this is my sole point of caution.

 

My point in the above postings is not to discourage use of this device.

 

It's also not to say that this device can, in any way, surpass various "Gold Standards" in the determination of knock.

 

I've had this device on my car, now, for well over 3 years.

 

I was among the first here to explore the installation, as you'll note above - even going so far as to be able to correct errors in the initial product documentation, as applied to the early BL/BPs.

 

I was also among the initial wave in troubleshooting its various concerns.

 

I still use, and love, this little device. It's still in that same location, as I photographed it so many moons ago (and yes, I do wish I had located it inside the center driver's side vent, that's a *very* good place for it).

 

I'm just trying to present the item with due caution, that's all.

 

All that I am trying to do here - and I hope I'm not coming across as being unfriendly, either - is to make you and others aware that it's not perfect, and that its use depends both on the end-user's manipulation of the device, as well as its technical restrictions.

 

The BEST way to have a visual representation of Knock is with a UTEC. Flashie-Flashie of the CEL is opportunity knocking.

^ Or ride around with a det-can, but then again, that wouldn't be visualization...and it would make conversation on the phone or with your passengers somewhat harder!

 

You know, I'm new enough to Scoobys that I've never used the UTEC, and only rarely have I seen the UTEC interface (a local's 600+ wHP '04 STi used to use a UTEC/ECUTeK combo, and another was a mildly modified bugeye WRX) - was that flashy-flash a "translation" of the ECU's determination of true knock? or was it also a visualization of the knock sensor?

 

FWIW, what I want is for someone ('cause I'm flat not smart enough to do it! ) to write-up some program to integrate into a datalogger that will allow near-real-time (as there'll still be a lag) visualization of a knock warning based on the ECU's determination of knock, not what the knock sensor may or may not be saying. Not a display of such parameters, which is already available with current dataloggers, proprietary or Open-Source - but something that just adds up all the factors, and just throws up a plain yes/no flag or idiot-light. But even then, I'd still want something that tied into the knock sensor, like a knock-light.

[BAC5.2]

*phew* that's a long post!

 

The UTEC flashes the CEL when it sees knock.

 

There is an obvious overarching flaw here, and that is that every engine management system on the market "listens" to the knock sensor to determine knock. Whether the engine is actually detonating, that is something completely different than whether or not the ECU determines that it is.

 

In the stock ECU (and by extension, all open-source/accessport cars) there are a lot of situations where knock activity does not correspond with knock correction. You can watch knock occur, and watch timing advance at the same time. That's the fundamental flaw I find with those EM "solutions". They allow active knock (something you can hear if you listen through the sensor, what I imagine a det-can is that you are talking about)

 

Again, whether or not the engine is actually detonating is almost impossible to determine.

 

Everything is best guess. So a Knocklite is not the be-all-end-all of knock detection. It's just a light, that triggers when IT thinks there is knock. Use it for situations like you mentioned, determining significant knock events, or for a warning that your map might be just a little bit off. Let the ECU determine "true knock" and let it correct for it.

 

If you want an instant-flag, the CEL flash on the UTEC is pretty good. If it logs knock that it corrects for, it flashes the CEL. It's a post-interpretation flash, not a knock sensor flash. So it accounts for knock events that the UTEC calculates that are independent of knock sensor activity.

 

Otherwise, I think it wouldn't be too difficult to write a program that monitors the checksum column, and triggers a light. But I suppose the problem there is that checksum doesn't always equal knock correction for any EM using the stock ECU. So you'd have a flash, but the ECU won't have necessarily done anything about it... which doesn't make any sense. That kind of light would probably just make you feel bad.

[TSiWRX]

*phew* that's a long post!

 

That's my problem - I often get diarrhea at the keyboard, as my wife likes to say.

 

There is an obvious overarching flaw here, and that is that every engine management system on the market "listens" to the knock sensor to determine knock. Whether the engine is actually detonating, that is something completely different than whether or not the ECU determines that it is.

 

In the stock ECU (and by extension, all open-source/accessport cars) there are a lot of situations where knock activity does not correspond with knock correction. You can watch knock occur, and watch timing advance at the same time. That's the fundamental flaw I find with those EM "solutions". They allow active knock (something you can hear if you listen through the sensor, what I imagine a det-can is that you are talking about)

 

Again, whether or not the engine is actually detonating is almost impossible to determine.

 

Everything is best guess. So a Knocklite is not the be-all-end-all of knock detection. It's just a light, that triggers when IT thinks there is knock. Use it for situations like you mentioned, determining significant knock events, or for a warning that your map might be just a little bit off. Let the ECU determine "true knock" and let it correct for it.

 

^ EXACTLY!!!!!! :D

 

That's exactly what I'd meant - except you expressed it better, and more concisely, at that.

 

 

If you want an instant-flag, the CEL flash on the UTEC is pretty good. If it logs knock that it corrects for, it flashes the CEL. It's a post-interpretation flash, not a knock sensor flash. So it accounts for knock events that the UTEC calculates that are independent of knock sensor activity.

 

Ah, got it.

 

Otherwise, I think it wouldn't be too difficult to write a program that monitors the checksum column, and triggers a light. But I suppose the problem there is that checksum doesn't always equal knock correction for any EM using the stock ECU. So you'd have a flash, but the ECU won't have necessarily done anything about it... which doesn't make any sense. That kind of light would probably just make you feel bad.

 

Exactly, there's a lot of factors to consider, not just a simple checksum - the "Knock" thread, I found, was extremely educational...but looking at it and thinking about how to implement a near-real-time reporting system made me realize that I'm just not smart enough to get it right.

[BAC5.2]

The problem is, it's impossible to reference a single sensor for knock. The only thing you could do is have a knock-correction triggered light. That is, a light that ignites everytime the ECU corrects for knock.

 

That would be reactive, and only if the ECU chooses to react to knock. It's not a total indication of knock, just an admission that the ECU determined knock bad enough to correct.

 

You could probably monitor crank position, and map that against timing. Any impulse (which you can directly measure by the fourth derivative of the crank position) that is out-of-sync with timing would indicate auto-ignition.

[TSiWRX]

^ Exactly - I want to know when the ECU thinks that it sees knock - and that's not something that's possible with the monitoring of any one sensor/parameter, nor is it something that's going to be, in the precise sense, "real-time."

 

Learning View is retroactive. And even pulling specific parameters from RomRaider in terms of the various knock learning, timing map, dynamic advance, etc. (the factors listed are just what rolls off the tip of my tongue ) still paints a picture of what's happened already, not what's going to happen.

 

And no matter how sophisticated such a program can be, like you said, it's still reactive - and only reactive to the ECU "admitting" knock is there (although not necessarily that it's bad enough to "correct for," at least that's not what I'm after - I'm not after the scenario where I'm going to be running positive dynamic advance while the timing is still being pulled: LittleBlueGT, IIRC, popped his engine running 2 to 4 degrees of dynamic advance, I'm not just looking for -KC...I'm more interested in something that takes into account *all* the various parameters that makes our ECU determine that it's seen knock, kinda like how you, a live person, would scan a datalog for evidence of knock activity, and raise a flag, visually/audibly, when those parameters are being/have been met).

 

But then again, I guess, even a knock-light - heck, even a det-can, the true Gold Standard - is still something that's warning of ACTIVE knocking.

 

I guess all we can ask for are ways to let us know that knock is happening/has happened, and that it's time to back off, NOW, before something really bad happens.

 

Too bad that there's no real way of predicting knock.

 

Currently, I think that we do the best we can, with the tools that we have.

 

We have a good map for the car. We try to put in good fuel. We try to make sure that the car's in good mehanical shape. We datalog to find evidence of knock. We put in a warning device that helps us know what the knock-sensor is hearing.

 

 

You could probably monitor crank position, and map that against timing. Any impulse (which you can directly measure by the fourth derivative of the crank position) that is out-of-sync with timing would indicate auto-ignition.

 

^ See, like I said, I'm not smart enough to do something like that!

[BAC5.2]

You could actually use something like Matlab to write the code, and have sensory data recorded directly into Excel.

 

Have an input column in Excel, dump to Matlab for processing, and have Matlab spit out the information to a macro in Excel. If the numbers don't jive, have the Macro flash the laptop screen red. Since it is easy to determine when changes in crankshaft velocity SHOULD occur, it is equally simple to determine when things are out-of-spec. Any latency on the computer end would be masked by the relatively long period of time between "correct" firings.

 

Or, you could mount cylinder pressure sensors in the heads, and map pressure waves of a no-knock situation. Then have a computer compare active pressures with known-good pressures. Assuming proper operation, you would be able to see det as spikes in a single cylinder's pressure at times incompatible with valve opening (i.e. pressure spikes when the valves are open would indicate post-ignition, and spikes before the valves are open would indicate pre-det).

 

Obviously, the right way to do it, is to assume that the engine is ALWAYS auto-detonating. In that scenario, follow Rao's advice about how to prevent engine damage.

[TSiWRX]

^ That's it, I'm no longer driving.

 

I'm going into the rickshaw business. Either that, or just go for teh SpecB, which is obviously immune to det.

[BagheeraLGT]

...wow, what a great conversation. i really enjoy stumbling onto threads with a wealth of knowledge like this one. its just too bad almost everything said here was beyond my realm of comprehension.....someday, i shall learn

[NSFW]

^ Exactly - I want to know when the ECU thinks that it sees knock - and that's not something that's possible with the monitoring of any one sensor/parameter, nor is it something that's going to be, in the precise sense, "real-time."

 

I run software on my CarPC that shows the load and RPM where knock has happened, in a table like this:

 

http://i3.codeplex.com/Project/Download/FileDownload.aspx?ProjectName=ssm&DownloadId=45074

 

(image from the project web site - http://ssm.codeplex.com/Wiki/View.aspx?title=Pictures )

 

It doesn't do anything to get my attention the instant knock happens, but it tells me where knock has been happening. The stuff around idle/cruise is uninteresting. I get familiar with how many knock counts there are in the high-rpm / high-load area and I can see when those have changed at a glance.

 

My car knocks rarely enough that I'm not really interested in being alerted the instant it knocks, it's enough to know where knock has happened, so that I can fix it the next time I tweak the tune.

 

I've looked into spark plugs with built-in pressure sensors but I'm told they cost around $1k each, so this will have to do for now.

[BAC5.2]

I've looked into spark plugs with built-in pressure sensors but I'm told they cost around $1k each, so this will have to do for now.

 

And only last 10,000 miles... if that...

 

You could just drill and tap the head (at each cylinder), and have a pressure sensor threaded into it. But then you run into topics only discussed in quantum mechanics. Observation impacts what you are observing, and the presence of a sensor could change the flow dynamics in the cylinder, and thus change the potential for det (the cylinder could be more or less prone to detonation with or without the sensor).

 

I like my crank velocity technique. It seems reasonable. You know when an impulse is to be delivered on the crank. If one isn't, you know it's a misfire. If it happens prior to, or after, it should, you know it's auto-ignition.

 

I can think of more possible ways to monitor, but that is DEFINITELY off topic.

[TSiWRX]

...wow, what a great conversation. i really enjoy stumbling onto threads with a wealth of knowledge like this one. its just too bad almost everything said here was beyond my realm of comprehension.....someday, i shall learn

 

^ It's not that hard, don't worry .

 

BAC5.2 approaches it from a more technical standpoint, but I approach it from a more lay manner - and both of us are basically saying the same thing: that the knock-light (or similar device) has its use, but that use is limited both by the way the light works, as well as by how the end-user chooses to set up the device.

 

My posts also specifically address some of the installation issues of the TSX Knock/Shift-light, and is product as well as vehicle-specific in terms of its content and tips. Those referenced posts you'll quickly understand when you have the actual product in front of you, with the installation task at-hand.

 

 

----

 

 

I run software on my CarPC that shows the load and RPM where knock has happened, in a table like this:

 

^ That's pretty darned cool!

 

I really should do a carputer project...maybe this will give me the necessary kick-in-the-ass, like outahere's posts on the rear fog.

 

My car knocks rarely enough that I'm not really interested in being alerted the instant it knocks, it's enough to know where knock has happened, so that I can fix it the next time I tweak the tune.

 

That's true, and that's, I think, all we can really ask for, practically speaking.

 

 

I've looked into spark plugs with built-in pressure sensors but I'm told they cost around $1k each, so this will have to do for now.

 

[BAC5.2]

I still want to do a dual-touch-screen carputer. Maybe one day. I DID figure out how to permanently load windows on a specific monitor, which is cool. I think the release of Win7 (and it's faster boot times) will probably get the ball rolling a little. My biggest problem with a carputer are things that haven't been sufficiently remedied to my knowledge (though I don't really keep on the lookout). Those are sound-output, radio, GUI, auto-dimming, steering wheel control integration, and lag-time/boot time. Those are things I've only seen done acceptably by OEM's, and they have more money than I do.

 

The things I want are very specific, and I haven't had the time or funds to chase down the perfect setup.

[BagheeraLGT]

^ It's not that hard, don't worry .

 

BAC5.2 approaches it from a more technical standpoint, but I approach it from a more lay manner - and both of us are basically saying the same thing: that the knock-light (or similar device) has its use, but that use is limited both by the way the light works, as well as by how the end-user chooses to set up the device.

 

My posts also specifically address some of the installation issues of the TSX Knock/Shift-light, and is product as well as vehicle-specific in terms of its content and tips. Those referenced posts you'll quickly understand when you have the actual product in front of you, with the installation task at-hand.

 

Well, I do understand your concerns about how the light sensitivity can completely change how effective it is, for the user. Much like a clap light in a persons living room, so to speak, that if the clap function is set on high, so that it only reacts to the loud claps (or knocks), then many subtle claps (but knocks nonetheless) will happen, yet remain undocumented. And if the clap light settting is too low, then subtle footsteps, or conversation will have your living room lamp essentially act like a strobe light- the knock sensor will be triggered by things other than knocks..... mabye not the best analogy, but i couldnt think of a better one. i think my flaw is that knocks are all the same, one cant be 'louder' or more significant than the next- a knock is a knock. (As far as I know)

 

What I dont really get is, what is a knock to begin with? Why are they caused?

 

you mentioned that you really keep an eye out for skunked gas... will bad fuel, cause your Carputer to knock? From what i understand, a 'knock' is a term for when a cars computer throws up a red flag about some operating condition. So, as you can see, I really dont seem to know much about knocks (as I dont even know what one is rly). But, hey, thats why im here and not sitting in a toyota dealership while my car is getting an oil change;)

[TSiWRX]

A quickie on knock.

 

http://en.wikipedia.org/wiki/Engine_knocking

 

I only scanned through the article, but I think it answers all your questions.

[mikeislord]

So what about this shift light? LOL.

[BagheeraLGT]

thanks TSi... much appreciated, ill giver a looksie. Idk if u remember when i first joined the subie online community, but it was over the summer on .org. I think you posted on my member journal there. Anyways, back to the topic at hand.....shift light

[NSFW]

I like my crank velocity technique. It seems reasonable. You know when an impulse is to be delivered on the crank. If one isn't, you know it's a misfire. If it happens prior to, or after, it should, you know it's auto-ignition.

 

I have a little theory that this is how Subaru's misfire detection works. That's just pure speculation on my part though.

[TSiWRX]

So what about this shift light? LOL.

 

^ I actually really like this one - if I'd have more space on my dash....:lol:

 

It seems that it would handle the task of being a pure shift-light better than the TXS unit.

 

Seems feature-rich and well designed, particularly for the price.

 

 

----

 

thanks TSi... much appreciated, ill giver a looksie. Idk if u remember when i first joined the subie online community, but it was over the summer on .org. I think you posted on my member journal there. Anyways, back to the topic at hand.....shift light

 

Of course I remember!

 

 

----

 

 

RE: carputer.....

....and lag-time/boot time

 

Oh, good point - I'd honestly never even though about that one.

 

Given my typical commute (very short), that's gonna be a drag.

[BagheeraLGT]

So I ended up picking up a TurboXS knocklight off of the classifieds and would like to install it in the near future. After re-perusing through this thread (which I totally forgot about since 09 seems so long ago). Are there any install threads floating around? Im set on putting it in the driver side center vent (which is the easy part). I wanna make sure I know a safe wiring path which would avoid any unwanted interference, as spoken about earlier in the thread.

 

Thanks guys!