I've been driving a 2006 Honda Civic Hybrid for awhile now. It's averaging 45mpg over 90k miles, but I have noticed that the ECU frequently makes some really wacky gear-ratio and clutching decisions with the CVT, especially while accelerating or climbing hills. I would like to know the pinout on the CVT and see what I can do to control it myself as I am very confident I can improve fuel economy further as well as improve response and smoothness with a simple circuit of my own design. I am aware of four connectors on the CVT: Two speed sensors, the PRNDSL mode switch at the top, and an 8-pin connector near the front. I believe this connector is the control connector but have no idea what the individual pins do.

Does anybody here know the CVT pinout or are able to share the pertinent service manual pages with that information? I would also like to know where those connections "come out" and go to in the cabin, if possible.

 

https://www.ebay.com/itm/OEM-Digital...0AAOSwFfJceLQN

 

I went ahead and bought the service manual you shared. What a wealth of information! I learned that the HCHII doesn't have an ECU; instead it has a PCM that runs the engine, transmission, and pretty much controls the entire car. I have since hacked the PCM and broken out the three solenoid control signals (in and out, so I can control their drivers instead of needing to provide my own), and another connector with pertinent signals (throttle, brake, modes, etc). I've connected it all to an ESP32 microcontroller which is monitoring all the signals, controlling the PCM's transmission solenoid drivers, and reporting back to a tablet computer. Right now I just have it copying the PCM's control signals so the car can drive and I can analyze what they're doing.

Some observations so far:

• All three speed sensors are an equal-duty 5v square wave, so you can count pulses each time their state changes (from low to high and high to low), thus doubling the effective resolution.
• Sensor A (drive pulley) follows engine RPM at Hz / 2. In other words, 1000 RPM is 500 Hz. With pulse doubling, 1000 RPM is 1000 pulses/sec. This gives you tens of RPM at 100ms.
• Sensor B (driven pulley) follows transmission ratio, which is a bizarre 1.433 ratio less from MPH. I don't see any special alignment with this number except when divided into engine RPM, it returns the advertised CVT ratio of approximately 0.4 to 2.4 (I invert the numbers so low is low and high is high; Honda's numbers are the other way around).
• Sensor C (clutch) follows road MPH at Hz / 50. In other words, 50 MPH is 2500 Hz. With pulse doubling, 50 MPH is 5000 pulses/sec. This gives you tenths of an MPH at 100ms.
• You have to compare Sensor B and C to determine if the clutch is slipping. The fact that they count pulses at different rates is bizarre and makes it hard to detect minute slips of the clutch since you always have rounding errors in the division.
• Minute slips of the clutch matter because the PCM only drives the clutch hard enough to stop it from slipping. When you romp on the accelerator (or regen/engine brake) the clutch slips briefly and the PCM upps the power to lock it again.
• When the torque is relaxed, the clutch power is immediately dropped again to a level just high enough to prevent slippage.
• Solenoid C (clutch) has a very erratic PWM signal at times, sometimes having a "PWM within PWM", shifting back and forth 10/255 several times per second—almost as if it's trying to "bump" the clutch to keep it from sticking at a lower power level.
• During normal, gentle acceleration, the clutch locks at about 10 MPH. Harder acceleration raises that number. The clutch normally lets go around 7 MPH.
• The transmission has two main clutches: The forward clutch at the input which is manually controlled by the mode selector paddle, and the start clutch at the output which is electronically controlled by Solenoid C.
• The transmission has stunningly low drag; Sensor A & B take a second or two to "spin down" when you take the transmission out of gear (P or N) when stopped at idle RPM.
• Sensor A drops to half speed in reverse mode due to the planetary reverse gearing.
• The PCM drives the solenoids at 12v with a 195 Hz PWM signal that never goes above 50% duty cycle (or 127 out of 255).
• At startup, the PCM drives one set of static values, and then drops very low after a few seconds. If any of the current sense circuits don't respond correctly, the PCM throws an error code (solenoid A/B/C pressure to high/low etc.) and starts flashing the 'D' mode indicator.
• When the PCM throws an error like this, unlike with other error codes, it immediately disables the IMA, which stops charging the 12v battery. Turning the ignition off and back on (after resolving the original fault) restores normal operation.
• The PCM seems very tolerant of unexpected behavior from the transmission. It also doesn't seem to mind receiving the ESP32's PWM which is not timing synchronized with theirs, despite having current sense circuits on each solenoid.

As of yet, the most I've done was copy the live, estimated PWM values and synthesize them back. This introduced 100-200ms of lag and also reduced the sampling speed to 10 times per second. The PCM didn't mind at all, but the clutch was jerky, and the transmission's ratio occasionally oscillated up and down a bit. Accelerating from a later stop, the PCM abruptly clutched hard, completely let go, and clutched harder...and snapped the right axle! It never threw an error, flashed the 'D' light, or stored a code during all of this. I replaced the axle (which had nearly 400,000 miles on it and was badly rusted at the damper) and the transmission is fine. Plus the brake shimmy at 70mph is gone now—apparently that axle had been fixing to break and I'm glad it broke at this time instead of on a road trip. I was also very pleased to note that my Amazon brake pads and rotors have about 50% life remaining. I put them on about 100,000 miles ago! For now, I will let the PCM control the clutch while I focus on the pulley ratio.

I also have a few important, but unanswered questions:

• My intuitive assumption is that the higher the torque, the tighter the belt needs to be. Here's the irony: The greater the torque, Solenoid A & B both go down as Solenoid C goes up, and vice versa. I then guessed that the pulleys are spring loaded large to tension the belt, and that the solenoids drive hydraulic pressure to relieve spring tension and/or cause an imbalance in tension so the ratio changes. I also guessed that Solenoid A was the drive pulley and Solenoid B the driven pulley based on their ratio (lower Solenoid B = lower ratio). I also know that without horizontal pressure, the drive belt will fall apart into a couple hundred pieces. Spring tension would keep this from happening. Everything made sense. Then I dug deep in the service manual, and they make it sound like hydraulic pressure sets the belt tension, although that is vague. They also said that solenoid B controls the drive pulley! Now nothing makes sense... Why would you make the belt looser when the driver is flooring the accelerator or engine braking? Is this why it feels like it's slipping instead of having power at 5000 RPM? (Coming from a 5-speed manual, it has always felt that way; it should have much more power than it does, and the clutch is confirmed not slipping.) If so, why did they do that? I thought heat killed the belt. At least keep the tension the same, if not increasing it when greater torque is required! And why would you have pulley A be controlled by solenoid B and pulley B controlled by solenoid A?
• How sensitive/forgiving is the belt to tension changes? This is of even greater concern if it's tensioned by hydraulic pressure instead of springs—hydraulics could snap the belt in very short order if asked to do too much!
• How do you know if the belt is slipping? There's no belt position sensor to know what ratio it should be doing to compare speed sensors, like with the clutch.
• I wonder what the PCM will do if say, I set the transmission to a ratio that puts the engine at 1200 RPM when they want to needlessly rev it up to 1800 just because the tilt sensor detected the car on an incline or decline (I hate that stupid behavior with a passion!). Or if I set it to maximum ratio (2.4) so the RPM will be a steady 1600 at 55 MPH instead of 1800-2200, or 2000 at 70 MPH instead of 2200-3000. Will they just try to adjust the ratio so far and ignore the speed sensors reporting a different ratio? Or will they rush to minimum ratio to compensate, and then throw an error after a bit with no response? The latter behavior could potentially kill the project. I don't mind driving around with the 'D' indicator flashing, but I need the IMA to stay operational, and hacking the CAN bus to control the IMA myself is not what I had in mind with this project!
• If the PCM does over-compensate but doesn't throw an error, how quickly will it correct itself if abruptly handed control? For example, I wouldn't want the PCM to ask for 1st gear at 55 MPH for five seconds while it figures out how to readjust itself. If that's what it's going to do, I'll have to change my code not to fallback to the PCM's values unless the car is stopped.
• Does the PCM "learn" as it drives the transmission, which would mean my controlling the transmission would gradually screw up their lookup tables, rendering it unable to smoothly drive the car as a fallback?
• I feel that I need a faster sampling time than 100ms, especially if I'm going to operate the clutch at some point. I would consider the transmission's sensors as "high speed sensors", so pulse counting makes sense. But maybe I should do pulse interval instead for greater instantaneous resolution. However, the pulses are so fast it will be hard to calculate their interval accurately! The ESP32 has a nanosecond timer. I don't think that's anywhere near fast enough to time multi kHz pulses. Maybe I could somehow combine both methods...

 

You've blown me away with your capabilities. I have nothing to offer other than...

The belt has 0% tolerance for tension. If you put it in ANY kind of tension, the belt will grenade. It's a pusher belt.



AND

This and many similar CVTs are all of Bosch design. They achieve up to 7% improved fuel economy by optimizing the needed power delivery at the optimal rpm maximizing efficiency. While you clearly have a lot of skills, I doubt you're more capable than the Bosch design team. Any changes you make to the normal operation of the CVT is likely to result in a net loss in efficiency.

Lastly, your 90K average of 45mpg is exceptional. Neither of my '06 HCH2 would do that without aggressive measures on my part.

 

I've seen those videos and many more. Once again however, I am receiving conflicting information: Many sources have referred to the extreme tension put on the push belts, citing hydraulic pressures upwards of 800psi used on the variable pulleys to tension the belt and keep it from slipping! Honda's own enthusiast news release for the HCHII even says "Double hydraulic piston used on variable pulley increases pressure by 170 percent" (3/4 way down the page). At the same time, if the belt was crazy tight, obviously it wouldn't take 1-2 seconds to spin down from 1000 RPM going from idle to neutral and would stop right away. The thing that stumps me is that both pulley solenoids go down with more torque load, not up. I just need to understand what's happening and why so I know which way to err and what to be on the lookout for, as I don't want to break anything—either by overtensioning or by slippage. I don't think I will be able to rely on the PCM's signal as it will be trying to compensate. Plus it has already demonstrated that it is irregular and unpredictable, which is why I'm even attempting this project. So far I have gotten a pretty close approximation to their tension signal by watching the solenoid C signal to calculate torque load. Belt tension is solenoid A + solenoid B, and pulley ratio is solenoid B - solenoid A. Solenoid A doesn't move much whereas solenoid B follows the CVT ratio. This however is an illusion: More torque lowers both solenoids. When asking for more torque, you usually also lower the pulley ratio, which lowers solenoid B twice. But solenoid A needs to go down for more torque and go up for lower ratio. So its movement is minimal and follows tension more than it follows ratio. My control values move similarly using the A + B and B - A calculations, and I know I am treating both channels equally.

Regarding MPG, I'm not dinging Bosch. It's Honda's programming of the PCM that I don't like. My first choice would be to modify the PCM's internal programming as this would allow for small, incremental modifications of something that already works and would avoid error codes. But the programming tool is expensive and I don't have access to the source code. I imagine it is a guarded industry secret and the programmers are used mostly to flash updated precompiled firmware files. Anyway, I have nothing good to say about Honda's programming of the PCM regarding the CVT. The one thing they have demonstrated is that the CVT is very forgiving and that the engine is quite efficient at a wide range of RPMs and power levels! I recently drove a couple newer, budget CVT cars and they operate the CVT just like I'd expect: Engine RPM follows the accelerator position consistently and nothing else. Once cruising at 35-75 MPH and you back off the accelerator to stop increasing speed, their CVTs go to maximum ratio and stay there like a manual transmission unless you press the accelerator significantly to increase speed, at which point they rev up to go from maximum efficiency and start heading towards maximum power. I can tell because when I press and release on the accelerator, the RPM stays rock solid unless I go past a certain point. Not the near constant RPM elevation, crazy gyrations, sluggish response, and weird tilt-based revving that Honda uses on the HCHII CVT.
Looking at the control signals, I can verify that the PCM is asking the transmission to do this; it's not doing this on its own due to age or malfunction. For example, sometimes it'll cruise through town at 1000-1400 RPM. Acceleration will feel tight and I barely have to touch the accelerator to maintain 25 or 35 MPH, and the MPG gauge is at 50-100. Other times (usually on a slight incline or decline) it will rev up to 1800 RPM at 35 MPH and I immediately feel the car lose power. Acceleration feels sluggish like a traditional slush-drive automatic instead of tight like a manual. Sometimes if I back off it'll drop back down and I can then reengage at lower RPM and better performance. Also sometimes if I let off, the RPM drops back down, and when I reengage the accelerator, the car drags back briefly before surging forward as it revs the engine back up. If it simply stayed at the lower RPM and increased fuel/air to the engine, the car would behave as expected.
Or I'll be coasting downhill at 45-70 MPH and they will start engine-braking with the CVT. No foot on the accelerator, every time the car noses down, it revs the engine up to 2000-3000 RPM and I have to start using the accelerator to maintain speed when it could've coasted with no gas if they dropped it down to maximum ratio (45-70 MPH=1300-2000 RPM) and stayed there. And then, when I start applying gas, it suddenly revs down, and the car stops engine braking. And if I need more power to maintain speed, because the car is at a tilt, it revs the CVT back up again and it feels like its slipping/engine braking again and I have to use 25% throttle when it should be at 5% at the lowest RPM! And no, it's not even charging the IMA. This is at full charge! It has always done this. It could help with overspeed when using cruise control, but it's really not strong enough of an engine brake to accomplish much besides lowering fuel economy. Which is another thing I'd like to address: I want cruise control to use engine braking if needed, all the way to 5000 RPM, to keep the car from speeding down hills when cruising on the interstate. It's what I already do manually with the mode paddle, but there's no reason it can't do it smoothly and automatically!
Speaking of cruise control, when the HCHII is in cruise mode, the PCM's CVT control behavior is modified in a strange way that further lowers fuel economy. First off, the cruise control maxes out at 50% throttle. But as it starts reaching 25-30%, the PCM starts aggressively revving the CVT, causing the car to scream up hills at 3000-6000 RPM when I could manually drive it at 2600-4500 RPM and maintain the same speed! This craps out the fuel economy very quickly, which causes me to often not use cruise control unless I'm driving in a flat area. I would rather the car max out at 50% and lose speed. I can always manually give it more gas if I want to go faster. Furthermore, once the PCM revs the CVT in cruise, it brings the ratio back up (and RPM back down) very slowly, like over 5-10 seconds once the power isn't needed. With all that engine braking, it has to continue to use more gas to overcome the engine load. Thus I find the car often gassing down the top of a hill that was so steep it should've been able to freewheel and recoup some of the MPG lost climbing the hill. Crazy stupid engineers wrote that code—why they had to make the CVT ratio decision so complicated and erratic is beyond me.
Additionally, I'm going down the road at 55mph, it's doing 1800-2000 RPM, and the PCM is reporting that the engine load is 50-70%. It should be doing 80-100% engine load at 1600 RPM (the maximum ratio of the CVT) instead of wasting that energy spinning the engine faster. Every time it revs the engine, even slightly, the live MPG meter goes down and the car loses a bit of power and I have to press the accelerator further. There is literally no road speed on a flat surface that the CVT ever goes to maximum ratio. You only see maximum ratio briefly when coasting slightly downhill after letting off the gas. And that's only before it revs up and starts engine braking. I know for fact that the CVT can do better. Potentially much better in some situations. I guess I get this sense of feel from driving a stick-shift too much! The questions are whether the PCM will let me control the CVT without throwing terminal error codes and whether I can do so without damaging the CVT by mis-tensioning the belt.

The logic of the whole belt tension thing I don't understand. Do the solenoids open or close the hydraulic valves? Does the hydraulic pressure push the pulleys together for increased tension/ratio, or are they spring-loaded and the hydraulic pressure works against the springs to relax the tension? If either are inverted, that would mean less solenoid equals more tension and it would make sense. But if my interpretation of the service manual is correct, and solenoid B controls pulley A (drive pulley), and the valve increases hydraulic pressure on the pulley to increase tension, then what is the reason we're lowering belt tension when we increase torque? Isn't that just going to make the belt slip, which destroys the belt? If the belt truly somehow has zero tolerance for tension, then how do we know when tension is being inadvertently applied so the solenoids can back off?

A couple reasons my MPG is above average: I always have the A/C switched off unless I need cooling, I've been running the engine on 0w16 oil instead of 0w20, I do a lot of highway miles, and I run my awesome Continental True Contact tires at 38-42psi. The hybrid battery is shot though—only one or two bars of charge or discharge before a recal. I am considering replacing it with a larger LiFePo4 pack.

 

"Not the near constant RPM elevation, crazy gyrations, sluggish response, and weird tilt-based revving that Honda uses on the HCHII CVT." - Your CVT operation does not sound normal to me based on the 5 CVT Hondas I've owned (2X 06 HCH2, 03 HCH1, 02 G1 Insight, 05 G1 Insight) UNLESS the hybrid battery is weak or shot.

If the HV battery is shot, you don't get normal CVT behavior since you're not getting the expected torque from the IMA motor. Weak batteries and recals absolutely destroy performance. PCM also takes battery state information from BCM/MCM and the '06-11 also maintains a "usable capacity" value for the battery. New is 75%. Yours is probably less than 20% based on your described symptoms.

In short, the PCM behavior you're seeing is partially governed by the battery state input to the PCM.

I was able to consistently get 50+mpg on one of my '06 by omitting A/C use and hypermiling during the summer here in Phoenix. I destroyed the battery by September. Bridgestone EP422 @ 44psi.

I assume you're 100% capable of designing a LiFePO4 battery from scratch including hacking the BCM/MCM to regulate current. If you're just going to go *****-nilly and just drop a battery in, you need to allow for a peak charge voltage of 211V (59S LiFePO4) as the car will regen to that level. You will gain no benefit from a larger pack unless you also hack it to force it into constantly pushing the IMA motor above a given voltage, but you may need to go larger anyway to accommodate the 100A discharge and 50A charge rates.

I know you've got a lot of miles under your belt, but I suspect you don't remember how it drives when the battery isn't full of turDcells. IMHO, take a step back from your current approach. Pull the IMA back from the car. Blow air through it with a box fan and charge it with 2X APC-35-350 power supplies in series with a diode in the output to protect the PSUs for 24 hours OR until voltage peaks for 8 hours OR 48 hours max OR temperature of the air coming out of the pack is perceptibly warm. Discharge it with light bulbs to 105V and recharge using the same criteria.

Install it and re-assess performance.

 

Thank you for the input. I bought this car at 280,000 miles, and thus have never driven a HCH with a good IMA battery! I did the grid charge and discharge on the original pack, which was in similar condition as this one is now. I was able to push it to about 195v. Unfortunately, when I put it back in the car, although it didn't drag down the IMA with continuous charging like it did before, it had a fraction of its original capacity. The BCM was recalibrating as often as every minute, often not losing or gaining a single bar, just alternating between 2 and 8 bars! After about 30 minutes of driving, one of the cells shorted out and the fan went to full speed. A couple hours later, I got home and removed the battery pack. It was still hot to the touch and remained so for another day (everything looked fine when I took it apart, however). So I retired that pack and put another old pack in, which has since deteriorated over the winter. The PCM's worst clutching (riding the clutch and abruptly slamming it into gear), or revving to very high RPM seems to always happen when the IMA battery is unexpectedly empty when it thought it was full. So fixing the IMA battery situation may very well make the PCM's CVT behavior more tolerable.
I have a number of these LiFePo4 battery packs (40v, 18aH, 12s8p), with a built-in balancing circuit. Four would put its voltage at 160v nominal (173v peak), and five would be 200v nominal (216v peak). The recommended fast charge per cell is 3.6v @ 10a, so unlike the NiMh battery pack that Honda is brutally charging and discharging, 80a charge and 500a discharge would be a walk in the park for these packs. I know the OEM IMA battery is rated at 158v, and I was able to push it up to 195v. Obviously, I would like to use four batteries instead of five, because the "empty" voltage would be far too low and would damage the cells if I went with five. It may even be very close with four—I don't know how low the voltage has to be before the BCM cuts out! I know that LiFePo4 batteries are quite forgiving with overvoltage, so that may be the way to go if it's not much higher than 173v. Obviously, I need to hook up a meter and find out what voltages the IMA is actually running at as it charges, discharges, and recalibrates. I will also need to either "fool" the BCM with a resistor array to simulate the battery taps, or I would need to string together varying amounts of series diodes to drop the LiFePo4 taps to the expected NiMh tap voltages. I would also need to put an under-temperature cutout for winter time, as LiFePo4 batteries can get damaged if you charge them while under freezing. Perhaps temporarily messing up the tap voltages or skewing the temperature probes to high temperatures would inhibit charging.
Regarding battery capacity, here's my thoughts: I know the BCM uses a watt-hour meter to determine the SoC of the pack. This is why it starts regularly recalibrating when the pack deteriorates, because the pack is unexpectedly at full or empty voltage before the watt-hours have been put in or taken out. So, I expect that the IMA will short-cycle these batteries, which will make them last a very long time. Additionally, as this pack is several times the capacity of the original, the BCM may simply start running down through its bars with assist, and start charging briefly only to recalibrate back to full battery again because the voltage is already high enough! And if the battery somehow gets low, it will do the same thing with charging (gradually charge up to full, and then recal back to empty and charge again). Or if I used a BCM fooler, I could dynamically manipulate the voltage at the top of the resistor array to make the BCM charge more (by lowering the voltage to force an empty recal) or discharge more (by giving full voltage to force a full recal). This would allow greater utilization of the battery's capacity when driving in the mountains without having to completely redo the BCM.

 

I would assign all your CVT woes to an inop IMA battery. You've had failed cells since day one based on the recalibration behavior. You're driving 20hp short. The default behavior when the IMA motor craps out is to rev the ICE higher into the power band to compensate.

158V "rated" doesn't tell you anything about the working voltage or provide a target. You'll never see one at 158V unless it's under load. RESTING voltages are usually near 176V after sitting for a few hours. If you ever see a pack at 158V resting, it is COMPLETELY DEAD.

Your understanding of the source of recalibrations is inaccurate/incomplete. The HCH2 DOES NOT count (no watt-hour meter) current like the prior models. It has a lookup table of tap voltages, pack current, temperatures and SoC based on a total pack voltage/temp. Recalibrations are hard coded based on limits, e.g., if a TAP drops below XX.XXV at Y°C with current > ZA, then recalibrate. Similar for the top side as well. I do not have this information. I know someone who has partially mapped this information. No. you can't get it. PACK voltage, current and temp are the source of the SoC computation, not the taps. Pack voltage has no influence on recalibrations.

Recalibrations are the result of hard coding the SoC. A negative recalibration (forced charging, 2 bars) is the default pack behavior when SoC is about 20% (I don't know the exact number). BCM detects recal condition, set's SoC to 20%, forced charge. Positive recalibration (march at 1 bar per second to 8 bars after a negative recal) is the flip side - max tap voltage triggers SoC reset to 85% (or so, again, don't know the exact number).

SoC dictates IMA behavior, high SoC favors discharge. Low SoC favors charge. "neutral" SoC is about 70-75%.

Given your choices, you need 60S to account for peak charging.

Built in balancers are usually horribly puny. You should test your modules thoroughly and ensure they really are in balance before deploying them.

You will need to provide 11 nearly identical stacked voltages to the BCM.

If you could spoof high pack/tap voltage, I expect you could force it to favor discharge. In operation, the voltage should be centered around 180V. NiMH chemistry is kinda wonky. I think of it in "active" and "inactive" states. Moving current in and out with no net change of capacity will drive voltages higher while operating, but once activity has ceased, voltage will drop.

One factor will be the likely much lower resistance of the LiFePo4 pack. With that many cells in parallel, you're going to narrow the operating voltage range.

Since you have two batteries and can grid charge/discharge, I'd do the following (ALL CHARGING WITH ACTIVE COOLING):

Charge at low current (<500mA) to 8800mAh input
Discharge to 106V
Charge at low current (<500mA) to 8800mAh input
Discharge to 66V
Charge at low current (<500mA) to 8800mAh input
Discharge to 66V
Charge at low current (<500mA) to 8800mAh input
Let sit for 7 days
Measure subpack voltages. Highest voltage should be around 16. "Good" subpacks will be within .05V of the highest.
Install pack in car
repeat process on second pack.
remove 1st pack.
Build a pack with the best 11 subpacks.
Grid charge to 8800mAh input
Install
Drive