Honda Insight Forum banner

Sorting out a Citrus

3977 Views 96 Replies 13 Participants Last post by  *sean*
My Citrus tends to stumble and lack power in lean burn when it is under increasing load. When it exits lean burn, there is a gentle burst of acceleration. I hope to uncover why. Items I'll be looking at:
  • EGR valve
  • EGR plate
  • EGR passages
  • Spark plug indexing
  • Spark plug age
  • Spark coil performance
  • LAF sensor
  • Manifold pressure
  • Fuel injectors
  • Fuel pressure
  • Engine deposits
  • Valve lash
  • Valve leakage
  • Valve guides
  • Exhaust system
  • Compression/worn rings
  • ECM control loop issues
Have I missed anything?

I won't look at all of these areas if I find the root cause before I get to them all!!!

I have already looked at several of these areas, but have decided that I need better data. OBD2 is too slow, so I plan to measure a number of sensors in real time, with many samples per cycle of whatever I'm measuring. This includes
  • Manifold air pressure
  • IAC valve position
  • EGR valve position
  • LAF sensor (A/F ratio)
  • Injector pulse width
  • Timing advance
  • Crank sensor
  • Throttle position
  • Camshaft sensor
  • Some OBD2 parameters I can't measure, such as fuel trims
Arriving are some new tools such as an in-car fuel injector cleaner and parts to build the "black box" that will read the values above. I hope to be able to do A/B tests in many cases, such as with different EGR valves, different LAF sensors, and more.

Any ideas for measurements are welcome. This will probably start in earnest between Christmas and New Years.
81 - 97 of 97 Posts

· Registered
Joined
·
2,358 Posts
Discussion Starter · #81 ·
(was composing this while Mario composed the previous post, let's hit send and get this out) So I do want to finish this LTO monitor before I return to ICE issues. It occurred to me that the things I built my NIMH channel monitor to find - weak (high IR) cells, pack balance, remaining capacity - are also what I want for LTO, and are things we are seeing pop up in the LiBCM threads with mismatched packs. Thought maybe I could build one monitor for either LTO or NiMH. Decided to look at the rate the LTO packs some data and guess what: once every 20 mS is the same rate as the BCM checks the NiMH channels!
Font Slope Technology Parallel Electronic device
 

· Registered
Joined
·
2,358 Posts
Discussion Starter · #82 · (Edited)
I have used RS-485 over ethernet cables before, it has a lot of advantages.
It's 100 ohm differential impedance, which is close to the ideal 120 ohm that RS-485 uses. You just terminate the ends with a 100 ohm resistor instead of 120 ohm. It uses slightly more power and has a slightly lower differential voltage, but modern drivers are good enough that it doesn't matter in practice. It's still in spec for the transceivers.
Ethernet cables are cheap and can be had in any length you need off the shelf.

I used one pair for power and one pair for ground (not power and ground in each pair, one pair only + and one pair only -. This will keep things from blowing up if you ever plugged actual ethernet into your system). One pair to transmit data and one pair to send data back.

This creates two RS-485 busses that supports any number of devices on it, where each device has two pass-through ethernet ports and you can just plug them all together. The master device sends data on the transmit bus, and all slaves read the message and decide if it's for them. One slave processes the command and sends data back on the receive bus. This requires a simple protocol and addressing scheme where each device has its own address and only one device transmits at a time (first the master on TX line, then one slave on RX line. Repeat over and over). This also requires having termination resistors on the master board and on the furthest slave board.

That's just one way to do the busses, you could also just use one pair and have it be half-duplex, and bring the bus back on the other pair so you only need termination on the master. But then you need a loopback on the last slave. But regardless, ethernet cables are a good way to go for this, IMO.
USB cables are 90 ohm differential impedance, so you'd need 90 ohm termination resistors. This is just a little too low, I think.
@Mario I really appreciate this info. I forgot about the little detail about the impedance of the cable.

I had put Ethernet cable out of my mind long ago because I have a thousand foot roll of CAT 5 but it is solid, not stranded. I tried using some and it just didn't seem like a good idea. Of course, I forgot that patch cable is typically stranded. And that an RJ-45 jack can't easily get pulled out. 24 AWG stranded appears to be a standard and the way to go.

I'm leaning toward going half duplex and using the second pair to keep the stub length short in order to maximize throughput.

Thank you again for making this suggestion.

Do you have a preferred pinout of the power, ground and data pairs on the RJ-45 jack? (actually, it will depend partly on how ordinary patch cables are wired. Looking at diagrams of eight-wire patch cable online, it seems that the outer pair are the orange and brown pair which are close enough to red and black for me. Maybe use the blue center pair towards the RS-485 controller end of the connection.)

Also appreciate the point of not damaging an actual ethernet device. I have a few Teensy 4.1 devices with Ethernet, a number of Ethernet Arduino shiields, and of course Pis... on some of these, one RJ-45 will be RS-485 and power and Ethernet on the other.
 

· Premium Member
Southern California
Joined
·
973 Posts
Either scenario has the same stub lengths, which are also extremely short since they're only as far as from the connector to the transceiver on the PCB (probably a couple inches at most). No problem at all for low speeds.

I forgot that in the system I did this, I actually did do the half-duplex approach. I don't remember why I picked it over the full-duplex option. It doesn't really matter either way, since you always have to wait for a response before sending another command (even with full duplex, if you send multiple commands, multiple slaves could try to respond at the same time). The main advantage is you could send multiple commands to one slave as fast as you want, and it will process them and return the data as fast as it can.
On each slave, I had a "loopback" switch that would connect the "transmit" pair to the "return" pair. The master had two termination resistors on it. So I'd just flip the loopback switch on the last slave and that would bring the bus back to the master through the other twisted pair to be terminated. But you could just as easily have a pair of "termination" switches on each board that would put resistors on the busses, for the two bus approach.
Or, instead of switches, just an ethernet plug with resistors or a loopback on it that you plug into the last device.
You could even just do a single half-duplex bus and not bother with the loopback, just termination at the far slave.
You could put two ethernet ports on the master so it could be stuck in the middle of the bus rather than only on one end. It would also require loopback or termination on its empty ethernet port if it was at the end of the bus.
Lots of ways to skin this cat!

Here, the green blocks are termination resistors.


For either approach, make sure all devices have their transceivers in receive mode by default and only switch to transmit while transmitting data. Even for the two busses scenario, if all slaves always had their transmit drivers enabled, there would be bus contention and no data could get through when any device tries to transmit.
 

· Premium Member
Southern California
Joined
·
973 Posts
Do you have a preferred pinout of the power, ground and data pairs on the RJ-45 jack? (actually, it will depend partly on how ordinary patch cables are wired. Looking at diagrams of eight-wire patch cable online, it seems that the outer pair are the orange and brown pair which are close enough to red and black for me. Maybe use the blue center pair towards the RS-485 controller end of the connection.)

Also appreciate the point of not damaging an actual ethernet device. I have a few Teensy 4.1 devices with Ethernet, a number of Ethernet Arduino shiields, and of course Pis... on some of these, one RJ-45 will be RS-485 and power and Ethernet on the other.
I would tend to go with a similar pinout to PoE (power over ethernet), even though PoE usually uses 24V or 48V, so your board would definitely die if it was plugged into PoE anyway.

PoE uses blue+blue/white as positive, and brown+brown/white for ground. The other two pairs are used for data.



But really, it doesn't matter too much if you and your end users are careful about where they plug their cables. :)
 

· Registered
Joined
·
2,358 Posts
Discussion Starter · #85 · (Edited)
LEARNING ABOUT KNOCK SENSORS

I captured more data over the weekend. I have been flummoxed a bit by what I'm seeing on the knock sensor, perhaps needlessly so.

The knock sensor is just a piezo microphone. It has a stub attached which appears to make it resonant around 14 kHz, and appears to help make it easier for the ECM to detect pings, somehow. You can see its resonance when you zoom into the signal. Perhaps detonation includes a wave that reflects inside the cylinder space around this frequency??? and maybe this is why it's not a good idea to use a sensor from another car? If a resonant sensor is used, perhaps this mechanically filters noise... could the ringing signal be rectified and the envelope used to detect the knock? Just guessing here; I'm probably wrong. (see the TI article linked to at the bottom of the post.)

This first screenshot was taken while at full open throttle as the car was passing through 6000 RPM in first, just before the downshift to second:
Rectangle Slope Font Line Parallel


There is a LOT of analysis still to do (more on that in a second) but what was concerning me was that the knock sensor signal was not even. There is a ton of noise surrounding ignition TDC of cylinder two, while it is maybe 6 dB quieter around ignition of cylinders 1 and 3. You can see this at idle, too:

Rectangle Slope Font Screenshot Parallel


One possible explanation is that the knock sensor is directly beneath the intake port for cylinder 2. But see the annotations above. Why don't we see anything in the same position under cylinder 1 and just a little thing under cylinder two? (note that the knock sensor amplitude is scaled way up from the high RPM screenshot before; this is not a huge signal).

Now I'm wishing I could see some examples from other Insights.

Below we see something similar in lean burn under significant load at around 1700 RPM:

Colorfulness Rectangle Slope Font Parallel


I don't know how the knock sensor is positioned relative to the cylinder wall (maybe someone who has a block and cylinder head separated can tell me???) but perhaps the knock sensor is set up to be a stethescope into cylinder 2, and it is just relatively deaf for cylinders 1 and 3.

But note also there are a number of pulses, and they don't all line up with the ignition signals. (Well, there is a delay from spark ignition to maximum pressure...) But still, there are more pulses, and at different places.

This is an area that I'm unfamiliar with, but it seems we could learn a lot from, and perhaps even use the knock sensor signal as an indicator as to the quality of engine operation.

Trying to learn more about the operation of knock sensors and how these patterns may appear, I found this article providing useful background on the causes of knock and some factors that influence its onset.

This TI article talks about different knock sensor types, as does the above article. From the ringing characteristic seen in the Insight sensor, my guess is that it is a tuned or resonant sensor, not a broadband sensor; tuned for resonance with the knock. Thus it should be easy to tell if an arbitrary sensor is compatible with the Insight: rap it (gently!!!!) against the bench, and see if the frequency of the vibrating element inside matches that of a genuine Insight sensor?

The TI article notes that background noise that increases as engine speed increases can mask knocking, and that some systems stop detecting knock above 4000 RPM.

Or is the omission of signal a sign of misfiring? The TI article mentions this too.

Thoughts are welcome.
 

· Registered
Joined
·
2,358 Posts
Discussion Starter · #86 · (Edited)
I am still getting some slight shudder in lean burn with near max lean burn throttle, particularly below 2000 RPM. I'm not sure why. I need to write some Python code to see if the injector pulse width varies. I think the idle air control valve is active as well. I can't see this easily in the Saleae captures.

I am also seeing power drop off as I press on the gas, even when flat. I have seen this before - with a LAF sensor that was the wrong type for my 2004. Now, the 2004 and 2000 use a completely different LAF sensor. And I thought I'd put my new NTK LAF sensor into the car. But I'm not posistive. So a new one is coming from Rock Auto, and it's going to be very well labeled before it gets near the car.

Data to extract from the capture:
  • long and short term changes in
    • fuel injection pulse width
    • IAC valve duty cycle
    • timing advance
  • for cylinder 2 (which is closest to the knock sensor)
    • location of ignition TDC
    • location of large knock sensor impulse relative to TDC
  • and more subtle, to detect imbalance between cylinders
    • changes in period between CKP pulses - looking for slowdown of crankshaft due to misfiring, poor combustion, detonaton BTDC
    • LAF sensor output
On those last two subtle points, it may be valuable to create "virtual cylinders" - break the data up into bins associated with each cylinder - to see if one is performing differently from the others. Does the crank slow down more with one than the other two? Is the amount of oxygen read by the LAF sensor overall less or more in one cylinder than the other two?

And other questions: what may change if we swap the plugs and coils between cylinders 1 and 2? (Already done; need to measure.) If the analysis above reveals imbalnce between cylinders, what will fix it? Replacing plugs? coils? Cleaning deposits off the inside? or might something else be happening? (The engine sounds normal at idle and other than lean burn, runs well.)
 

· Super Moderator
Joined
·
9,830 Posts
I am still getting some slight shudder in lean burn with near max lean burn throttle, particularly below 2000 RPM. I'm not sure why. I need to write some Python code to see if the injector pulse width varies. I think the idle air control valve is active as well. I can't see this easily in the Saleae captures.
Good luck finding the last remaining little bit of shudder near max lean burn throttle. I've seen that in every Insight I've driven with over 150,000 miles. New or very low mileage Insights don't seem to have it, but some combination of slight wear or "opening" tolerances seem to produce the effect. I drove lots of cars at the 2016 National and they all had the effect. I've invested hundreds of dollars in new parts and many hours on my own hypermiller trying to ferret out the cause and have never found it, though the lack of a graceful drop of LB does not seem to have much or any effect of hypermiling performance. Honestly, I do wish you luck. I'd love to know what causes this. Your more scientific approach may turn something up - we can hope:)
 

· Registered
Joined
·
2,358 Posts
Discussion Starter · #88 ·
Good luck finding the last remaining little bit of shudder near max lean burn throttle. I've seen that in every Insight I've driven with over 150,000 miles. New or very low mileage Insights don't seem to have it, but some combination of slight wear or "opening" tolerances seem to produce the effect. I drove lots of cars at the 2016 National and they all had the effect. I've invested hundreds of dollars in new parts and many hours on my own hypermiller trying to ferret out the cause and have never found it, though the lack of a graceful drop of LB does not seem to have much or any effect of hypermiling performance. Honestly, I do wish you luck. I'd love to know what causes this. Your more scientific approach may turn something up - we can hope:)
Thanks for the encouragement. My 2004 didn't really do this. It may have had a slight bobble at the edge of lean burn, but it also could accelerate in lean burn on flats and keep speed on slight uphills.

I am able to see variations in the MAP sensor (its envelope, that is) with a period of several hundred milliseconds that might correspond to what I am feeling. A quick look at pulse widths of the IAC valve - which does appear to be actively used during engine operation - and the fuel injectors - along with the LAF sensor and EGR sensor data, may prove quite interesting.
 

· Registered
2000 MT
Joined
·
216 Posts
It's amazing to see map voltage fluctuate off scale (would be curious to align mV with pressure) with the pumping action of the engine in lean burn. I will be poking a bit more into the posts and see what's gone down so far. I don't have any power in lean burn so this will likely give me the push to do some over due tlc 😂.

Also would be curious to see the iacv overlayed on that graph if it really does operate off idle. As well as oxygen sensor voltage (as well as voltage to ∆ equivalent).
 

· Registered
Joined
·
2,358 Posts
Discussion Starter · #90 ·
Been busy with work and had to set this aside for a bit. The problem right now is that I want to see the PWM and pulsed signals not as pulses but as smooth waveforms so I can quickly see how much they vary over time. In other words, I need some virtual oscilliscope channels. Saleae's Logic app does not support this (you can manipulate some data but the results are at best numeric, not graphed). So I looked at exporting to CSV and importing to Pulseview, an open source oscilloscope and logic analyzer display app. Well, Pulseview crashed while loading the Saleae output and a friend informed me that this is the norm for Pulseview? Thought about gnuplot, but gnuplot is crushed by large data sets. Maybe MatLab which I've never used, but I don't have a license. So, looking for an open source alternative (or maybe project # 92398 is to load Pulseview into a debugger and find out why it crashes.) ... Datashader??????

A friend asked, "have you checked the compression?" which my answer was, "of course not, it runs fine". But he has a point - leave no stone unturned. You know you need to clean the basement when, while trying to find your compression tester, you find that you actually have a leakdown tester! I must have bought that years ago when first sorting out the red Insight....

(no need to use it on the red Insight any more... pretty sure the cylinders leak through the bent valves and the holes in the piston heads, lol)

oh I remember now. When I went to hook it up to the compressor that dad bought in 1977, I found that a hole had rusted through the tank. That was the end of that.

Anyhow, I do have a small portable tank which might do the trick if I don't dally around when measuring. And I ordered a borescope camera with three cameras on the end !!!! that looks pretty nice and will give me a much better idea of whether or not there's a ton of build-up that might be causing detonation or who knows what.

This is observational work that needs to be done and will not affect measurements.

I am considering ditching the Saleae for the measurements involving pulses and employ a Teensy instead. The Teensy 4.1 runs at 600 MHz and its USB serial is not a TTL serial converter but a true high speed USB interface. This means that unlike an Arduino, where logging takes up precious CPU time, you can log gobs with the Teensy and it won't blink. Hats off the Paul the guy who brought us the Teensy. He's really contributed a LOT to make it a solid choice for serious Arduino development (though some may argue otherwise, that is NOT an oxymoron with the right product and libraries.)

In a world where we have unlimited lifetimes, I would plug it into a Pi connected to a display in the dash and do the analysis in the Pi in real time, and draw the results directly into virtual gauges on the instrument cluster made from an HDMI monitor (yes, something appropriately sized does exist...)

"Its not the destination, nor the journey, but imagining the journey..."
 

· Registered
Joined
·
2,358 Posts
Discussion Starter · #92 ·
I performed my first leakdown test ever tonight! And I managed to do it without a proper air compressor (since mine has rusted and must be retired). I did use a small portable air tank and a 12V compressor, and worked quickly, and was able to get usable readings.

Cylinders 1 and 3 had a leakage of less than 2%, cylinder 2 had a leakage of 8%.

Blue Gauge Measuring instrument Motor vehicle Clock

Above: cylinders 1 & 3 were virtually identical (this is #1)

Blue Motor vehicle Vehicle Gauge Measuring instrument

Above: #2.

I tested at 75 PSI rather than 100 PSI because of limitations due to the pressure tank. It is easy math to make the correction but the tester case had a chart.

So, is 8% leakage a big deal for an Insight?

(next: photos of the cylinders with a hint as to why #2 might be low)
 

· Registered
Joined
·
2,358 Posts
Discussion Starter · #93 ·
I will have to wait until I buy or have access to a proper compressor to listen for where the leak in cylinder 2 is happening. However, 8% isn't the end of the world??? Anyone else have leakdown data?

There is some nasty scoring in cylinder two. Gotta wonder how this happened:

Water Wood Tints and shades Automotive tire Metal

Water Automotive tire Wood Grey Formation

Above, cylinder 2. What's going on with the clean spot on the piston head? Every piston had one of these.

The white specks on the wall have me concerned that I introduced dirt into the cylinder during the process. Looks like I will now invest in an air compressor after all to blow this out. I wonder if my jank air compressor solution was not properly filtered when it drew in air and blew that junk into the cylinder. LESSON: be sure that your air source is CLEAN and FILTERED!

I wonder if the scoring was introduced during earlier servicing of the plugs. The car was missing its belly pans and had efvidence of red clay mud being sprayed inside the engine compartment. This was likely the cause of the steering sensor failure which was fixed earlier. If there was dirt around the coil packs that fell in during plug servicing (and they were found to not be the correct index) this could have done it. Message: Thoroughly clean and blow dirt from the area before servicing the upper engine. Put a table near your work area and clean it and your tools before starting work to keep stray dirt out. It appears that it only takes one grain to score the engine!!!!
 

· Registered
Joined
·
2,358 Posts
Discussion Starter · #95 ·
Getting ready to perform fuel injector cleaning as part of peeling the onion, I tested new tools with a fuel rail and injectors obtained from another member, and with alcohol instead of cleaner while sorting out leaks (and there were a few).

Automotive tire Tire Wheel Hood Wood

Above: air line, air regulator (replaced), cleaner tank, fuel rail and injectors (from a different car), injector activator powered by car battery.

Glad I did this instead of sending them out. I learned a lot. All injectors had a clean - and split! - pattern. I had no idea that was even a thing, and no mention of it in a search of IC.
Drinkware Automotive lighting Liquid Glass bottle Bottle

Close-up of nozzle spraying

Automotive tire Automotive lighting Rim Bicycle part Gas

How do they get an even split pattern with an odd number of orifices?

Wood Flooring Gas Tints and shades Hardwood

Maybe is isn't that even? This is ejecting alcohol, not injector cleaner.

Drinkware Rectangle Material property Silver Cylinder

Again, not cleaned, just alcohol. Not bad for junque box injectors.
 

· Registered
Joined
·
2,358 Posts
Discussion Starter · #97 ·
Lunch break: Another measurement, with more volume and with cleaning solution. One is down 10% compared to the others. Have not ruled out other possible issues such as vial volume differences, air regulation variances, incompletely bled air, etc. I need to characterize these.

I do have two more sets of injectors to try.

Drinkware Liquid Tableware Food storage containers Mason jar
 
81 - 97 of 97 Posts
Top