Winter is here and my CX-5 turbo is now a 2.0L!

Hello, have been reading this thread with some interest. Just got a 2020 CX5 GTR to replace a 2016 CX5 GT that was totaled.

As some background, I am an automotive engineer that has worked at OEs and done calibration work on turbo engines. Not Mazdas in particular, but I know my way around engine controls and turbos in general.

The idea that Mazda arbitrarily decided that under 20F it could be slippery so power should be limited is absurd. With all of the capabilities of current traction and stability systems, traction is actively managed based on conditions, and you can certainly have more grip below 20F than above based on conditions. The same goes for picking an arbitrary temperature to limit power to prevent damage to the engine or driveline. This is all actively managed. It is more likely that certain conditions occur around 20F that are causing this issue.

When it dipped down to 15F this morning in Maryland I took the opportunity to go for a quick drive to see if I could replicate the problem and collect some data. I was a little unprepared as I did it on a whim, but I found a few things of interest. With more time there are other things I would have looked at, especially after looking at the first pass of the data.

The car was in my detached garage, so it was about 30F when I started out, despite it being about 15F outside. There is a lot of thermal mass in a car, so it takes some time for the ambient and intake air numbers to drop.

Here is the data from 4 runs that I made, in order. The approximate 0-60s are 8.0, 11.0, 10.0, 10.0 and the intake/ambient air temps were dropping (and the coolant temps were going up) as I went. All runs were reasonably flat but there was certainly some variance. Traction was not an issue.

1581827969702.png

1581828269185.png

1581828296200.png

1581828324625.png


Without getting into too long of an explanation, most modern cars use a "torque model" method of control where the throttle input is used as a "torque demand" from the driver, and the engine control comes up with the parameters to achieve that torque. The Load Value is a good measure of the torque output. That torque demand is sometimes more than what the engine is capable of, but can be less than what the engine is capable of for a number of reasons, usually to do with protecting the engine or driveline. If the limit is the driveline, it may be gear based and only apply in lower gears. On a N/A engine that involves varing things like throttle position, ignition timing, and cam timing. On a turbo engine, the boost pressure is added into the mix. Boost can be hard to control because turbos typically use boost pressure to actuate the wastegate. So you end up in a chicken-and-egg situation. If the turbo starts building boost too quickly, you "need the boost to stop the boost" and the delay in the system can make it hard to stop the boost from building quick enough.

I can't make any conclusions until I collect more data in warmer temps, but my theory based on this initial look is that when the temps get this cold the engine is capable of exceeding the torque demand at full throttle, and the control system is having a difficult time regulating it-- so it ends up taking aggressive measures to keep the torque within limits, and it is slow to recover from it.

This agrees with the thoughts other people had above about rolling into the throttle rather than stomping on in and seeing if there is a difference. Would also be interested in what would happen if you did some runs in the cold starting in manual/2nd gear.

Fortunately or unfortunately, it rarely gets under 20F here in Maryland (unlike when I lived in Detroit...)
 
Good info above.
When I had the extended rental, I found boost was minimal in 1st and 2nd gear.
I tried manual 2nd gear and low speed, but it didnt matter, still low boost.
I even tried rolling into the throttle and it still struggled in 1st and 2nd.
When I put it in 3rd gear, 8psi+ was easy on demand. It never lacked boost in 3rd gear when you got on it, even ~1/2 throttle.
Unfortunately I wasn't able to monitor anything else.
 
Welcome aboard, Jeff.

Glad you survived the demise of your 2016. It's encouraging to see an automotive engineer having enough confidence in the brand (and model) to replace kind-for-kind.

I have a Reserve and have not experienced this issue, but I the same climate you live in, just south of you in central Virginia (after having escaped NOVA).

So two (hopefully valid) questions come to mind:

1-From your experience, how can computerized control systems have such a hard time managing this (or any) issue, or be so slow to recover? Is it because there are limitations in the car's mechanics?

It just seems that with everything being monitored & analyzed instantaneously, and with every input being computer controlled by wire, near-perfection in every scenario should be possible...and this feels "inelegant," to say the least...sort if "In Case of Emergency Break Glass."

Question #2 is more thinking out loud and not something that requires an answer: If the factory programmed control system is having a hard time managing this, what are the odds that a reprogram (3rd party tune) can address it at all, much less without introducing risk?

Here's another forum where members who do their own ECU tweaks have discussed this issue in other cars. One member said:

I just looked at the stock rom and found a boost control table called: Target Boost Compensation (IAT). It provides a percentage to reduce target boost based in intake air temps. From 68 to 104 degrees - there is no compensation. At 50 its -3.41%, 32 is -6.83% and 14 is -10.24. At 122 its -4.00 and 140 is -8.00 Again - that's percentage to reduce target boost, not the psi to reduce it by. So at my temp (25 degrees) it would reduce by about 8% which would be 12.7 psi for the target - exactly what I was seeing.

The rom from my '05 didn't have this table - so it must be something newer (probably as of '07 or '08).

I'm not sure how germane this might be to the CX-5 turbo issue, since it's not as radical as what we're seeing.
 
This jives with what I thought several pages back. Just putting the iron kabosh on things because cold+turbo=POWA! and the CX5's drivetrain can't hack it.
 
Glad you survived the demise of your 2016. It's encouraging to see an automotive engineer having enough confidence in the brand (and model) to replace kind-for-kind.
Actually, it was my 16 year old daughter that had her license for just over a month. Wasn't really her fault and the car wasn't that bad, but it took a hit in the door that damaged the rocker and door opening around the hinges and latch. Would have been an expensive fix. We actually have 3 Mazdas in our fleet. Also have an '11 3 Hatch (2.5 6MT) and a '01 Miata. I'm not fiercly brand loyal (own lots of other stuff, including some little British stuff ;)) but in the types of car we wanted at the time, I liked these the best.
1581870721829.jpeg
1581870751231.jpeg


1-From your experience, how can computerized control systems have such a hard time managing this (or any) issue, or be so slow to recover? Is it because there are limitations in the car's mechanics?

It just seems that with everything being monitored & analyzed instantaneously, and with every input being computer controlled by wire, near-perfection in every scenario should be possible...and this feels "inelegant," to say the least...sort if "In Case of Emergency Break Glass."
I agree. In general terms, the control system first has to make a good guess as to what would provide the results it wants, then use feedback of the results to fine tune it. The art of calibration is figuring out good first guesses, and tuning the response so it doesn't over or under react. This just seems like a sloppy area of the calibration that isn't very good. Think of how early ABS, TC, and stability systems worked compared to today. What used to be harsh and sometimes inappropriate interventions are now almost unnoticeable. This is unusually poor for an otherwise well done car.

Question #2 is more thinking out loud and not something that requires an answer: If the factory programmed control system is having a hard time managing this, what are the odds that a reprogram (3rd party tune) can address it at all, much less without introducing risk?

Here's another forum where members who do their own ECU tweaks have discussed this issue in other cars. One member said:

I'm not sure how germane this might be to the CX-5 turbo issue, since it's not as radical as what we're seeing.
It's kind of humorous that you reference that because I spent a lot of time struggling with Subarus. The struggle there was always that the "tuner" companies (EcuTek, Cobb, etc.) didn't really know how everything worked and what/where the parameters were. So even if you wanted to change something, you couldn't always actually do it. This is the same situation. I have no doubt that this problem could be fixed in the calibration, but highly doubt anyone outside of Mazda has the knowledge of the system to successfully do it.

But in the example you reference, that is exactly the kind of "first guess" parameter that can cause this. When it is cold, you need less boost pressure to make the same torque. That table is how it makes the first guess. If it doesn't take enough boost out at first, it exceeds the torque limit then over-reacts to it. So taking that table and adjusting it for LESS boost at cold temperatures could actually solve the problem.
 
But in the example you reference, that is exactly the kind of "first guess" parameter that can cause this. When it is cold, you need less boost pressure to make the same torque. That table is how it makes the first guess. If it doesn't take enough boost out at first, it exceeds the torque limit then over-reacts to it. So taking that table and adjusting it for LESS boost at cold temperatures could actually solve the problem.

That's where most of us are confused. LESS boost and still make same power makes sense, but based on my drive and your data, it makes ~0 boost, which is just terrible for power (and in some cases, safety).
 
I'm glad your daughter is OK.

I, too, had an "out-of-the-gate" accident when I was young. It was my brother's '67 Camaro 350 SS. He was so particular that when he went to pick up the car (factory ordered) and the dealer had put his emblem on the trunk lid against my brother's specific instructions (2 holes drilled through with speed nuts on the back), he refused to take delivery of it. So the accident was particularly traumatic. Served him right for loaning it to a teenager ;) And the rear quarter panels rusted through in no time, so his concern for his trunk lid was wasted emotion.

Regarding ECU tunes, I guess you've seen that the field for Mazdas has always been thin, and it's getting thinner. On a related note, I'm still waiting for ScanTool to roll out their PID package for the 2019 Mazdas. Apparently, Mazda was slow in releasing the data that model year, and Scan Tool's still working on them.
 
That's where most of us are confused. LESS boost and still make same power makes sense, but based on my drive and your data, it makes ~0 boost, which is just terrible for power (and in some cases, safety).
Well, it makes boost. Until it doesn't!

Regarding ECU tunes, I guess you've seen that the field for Mazdas has always been thin, and it's getting thinner. On a related note, I'm still waiting for ScanTool to roll out their PID package for the 2019 Mazdas. Apparently, Mazda was slow in releasing the data that model year, and Scan Tool's still working on them.
Can't say I understand this. OBD is a standard that all automakers follow by law, and they all use the same PID definitions. When a tool plugs into a car, it (should) use a diagnostic request to "ask" the car what PIDs are supported, and that is the list it works with. It can do this with any car. If you are talking about manufacturer specific PIDs, the manufacturers don't "roll out" anything. Some vendors may backwards engineer them, but Mazda isn't giving anything out. Either way, the standard PIDs cover most of the things you might want.

So, I did a few runs today with the weather in the 40s. This is probably the best run, 0-60 about 7.5s. Not bad considering it doesn't have premium fuel in it. A few notes.

1. I turned on a bunch more channels, which will make the logging rate for each channel worse. So note that there really aren't too many points for each line, and the lines are just a smooth approximation. You kind of have to squint and look at the whole picture. So the 0-60 is an approximation based on points that re more than 1s apart.

2. I kept the channels the same for this graph as the prior ones for consistency, but the purpose of logging the extra channels was to see if there are some that would be better channels to use. WIll look at that later.

3. This is more what I would expect to see. A flat boost curve (boost responds slowly so it isn't used for small adjustments), a varying thottle curve, and a load curve that follows the shape of the throttle curve. The RPM curve is kind of hard to interpret, but those dips in throttle and load at around 716 likely line up with a torque reduction for the 1-2 shift.

1581904431426.png
 
Here is an update graph of the same run.

Changes:

Swapped in Relative Throttle Position replacing Absolute. Seems to be a more stable value. Relative should just be a scaled version of absolute, not sure why they don't follow as they should. Probably more a factor of the slow data from the OBD2.

Swapped in Absolute Load for Calculated Load. Absolute Load is basically volumetric efficiency (which can be above 100% on a boosted engine), Calculated is more of a % of max available torque.

1581905084235.png
 
Love all the data that has recently been shared! @Jeff F, what app are you using to make these graphs? They kinda look like Virtual Dyno graphs.

@Odrapnew looks like you're still considering getting a CX-5 turbo. So I take it, the low power instances weren't as bad as you thought it would be?
 
Love all the data that has recently been shared! @Jeff F, what app are you using to make these graphs? They kinda look like Virtual Dyno graphs.

@Odrapnew looks like you're still considering getting a CX-5 turbo. So I take it, the low power instances weren't as bad as you thought it would be?

Actually, the low power in 1st and 2nd gears is probably worse than I expected. Everything seems to be good in 3rd gear and above. Also, I'm still trying to determine how big of an issue this will be for me for the couple months of cold weather. This morning was an example where I needed to get moving quick and if I had low power when I was expecting full power, it could have been a safety issue. Luckily my Explorer delivered the needed acceleration.

I am still seriously considering the CX5, but the power and seat comfort are 2 areas of concern.
 
I'm collecting the data with BlueDriver, but dumping the data to a csv and making the graphs in Excel. Not the best tool for the job, but it works.

So... who wants to help fix this? It doesn't get cold enough around here to have many opportunities to work on it. Plus I'm lazy. If someone wants to do the dirty work I'm happy to provide technical assistance. ;)

Here is how I would go about it:

1. There are at least 2 air temp sensors in the car, one on the airbox (pre turbo) and one in the manifold (post turbo). The ambient reading is often taken from the one in the airbox with some extra filtering, so not sure if that is a separate, 3rd sensor or not. Have to figure that out.

2. Figure out which sensor is causing the problem, by simulating the inputs with a resistor. Basically fool the car into thinking it's HOTTER than it really is and see if we can cause the problem when it is above 20F. If not, skip to step 3.

3. Once we have figured out which temp reading is causing the problem, wait for it to be cold enough and simulate the sensor to read COLDER that it actually is, in the hope that it's initial guesses will be for even less boost/throttle and prevent it from getting to the point that it has to pull it back. Basically we want it to be more conservative and work up to it from below rather than overshoot at the start. If we weren't successful in 2, we might have to try it with more than 1 sensor to figure it out.

4. If #3 works and it's not a big shift that is needed, come up with a circuit that offsets the temp reading of the sensor to prevent it from happening.
 
Hey Jeff, if you get a chance (in cold weather), try a couple of these tests.

Manual shift into 2nd gear, lower rpm and get into the throttle and monitor the same parameters.
Repeat the same for 3rd gear and 4th gears if possible

As I mentioned, when I had the overnight test drive, 1st and 2nd gears were weak, regardless of what I did. At the top of 2nd there may have been a slight surge.
Once I reached 3rd gear or above, even 1/2 throttle at low rpm, I was rewarded with boost (~8psi). I could do it on demand in 3rd gear an higher. Also, stomping or easing into it at highway speeds (as long as 3rd gear or higher) I could feel the surge of power.

I think you're onto something regarding simulating higher temps. Maybe it could be something easy as plug in for winter and remove for all other seasons? At least that could work for MN since we rarely get above 40F from December through Feb.

One other thought, for those that live in cold areas, what about blocking the intercooler in winter months? Something like diesel trucks do, but only block the intercooler, not the full front. Maybe that helps retain some warmth in the air intake. Just a thought.
 
Last edited:
I didn't mean to imply that oil temp was the only culprit here. I mentioned air intake sensors before and it's very likely that the computer is dynamically adjusting engine condition based on a whole range of sensor inputs including oil temp, air intake temp, static pressure and humidity, and so forth. In non-turbo engines the primary culprit is a retardation of ignition timing that causes power loss, but in the turbo models that behavior likely takes a backseat to boost attenuation, although I suspect timing is still playing a role here.

Also, @Odrapnew IMO you should be more worried about the seat comfort issue than this power issue we're discussing here. I had to sell my CX-5 because I simply could not live with the seats, they were causing schiatica in my hip. I'm still bothered by it, both because I had to take a $10,000 depreciation hit on the vehicle and because I liked it and would have preferred to keep driving it. The seats were just so unlivable that it wasn't possible.
 
Can't say I understand this. OBD is a standard that all automakers follow by law, and they all use the same PID definitions. When a tool plugs into a car, it (should) use a diagnostic request to "ask" the car what PIDs are supported, and that is the list it works with.
Just to clarify (without going down too many rabbit holes here), I was talking about the software. I don't believe that they all read every monitored data point on all cars, even though the OBD reader transmits the info. There was a thread here where folks finally found software that would read the tranny temp info that was being sent by the OBD reader.

ScanTool offers what seems like dozens of upgraded vehicle-specific PID sets that works with their software @ $9.99/set. They were waiting for Mazda to send them 2019 technical data so they could create that software. Maybe I'm using "PID" in the wrong context.
 
Last edited:
@CatsPaw I appreciate the comments. I wish I could get more seat time in one before I purchased. I tried searching for cx5 rental, but not having any luck. I should bring my wife's CX9 in for some work (whatever it may be) and get a CX5 loaner. :unsure:
 
It took 3 weeks before my hips started aching after I bought my '17. Nerve pain is weird like that. The effect was cumulative and continued to worsen with time. The side bolster was pinching my sciatic nerve into the side of my hip bone. Kind of just bad luck on my part really, I don't expect most people to encounter this issue. I just wasn't compatible with the seat anatomy-wise. But, complaints about these bolsters is fairly widespread on the internet from people of all body types so clearly Mazda made an oops on the design here.

I took a LONG test drive of the car with the specific goal of trying to detect body pain issues before buying, and I still got burned. It may not really be possible to know for sure with just a test drive. The experience has made me pretty leary about whatever I buy in the future.

I sat in a new Mazda 3 and they have changed their bolster design considerably in that car. The bolster is still hard, but it kind of floats and deflects out to the side instead of being rigidly attached to the center seat frame. There is still pressure on the hip, but it's not nearly as bad as the sitting on the edge of a 2" x 4" feeling the CX-5 seat gives.
 
Hey Jeff, if you get a chance (in cold weather), try a couple of these tests.
I likely won't see under 20F where I live until next winter.

One other thought, for those that live in cold areas, what about blocking the intercooler in winter months? Something like diesel trucks do, but only block the intercooler, not the full front. Maybe that helps retain some warmth in the air intake. Just a thought.
Not going to do any good. This all goes awry long before any appreciable heat is built up.

Just to clarify (without going down too many rabbit holes here), I was talking about the software. I don't believe that they all read every monitored data point on all cars, even though the OBD reader transmits the info. There was a thread here where folks finally found software that would read the tranny temp info that was being sent by the OBD reader.

ScanTool offers what seems like dozens of upgraded vehicle-specific PID sets that works with their software @ $9.99/set. They were waiting for Mazda to send them 2019 technical data so they could create that software. Maybe I'm using "PID" in the wrong context.
That isn't how it works. OBD works via PIDs where the tool has to request the specific information and the controller replies with the data. So the tool has to be configured
to request what PIDs are wanted, it isn't just reading from an existing stream. There is also a function to ask the controller which PIDs are supported. If it was requesting all the supported PIDs all the time, it would be reeeeeeealy slow. Transmission temp is not an OBD parameter, it would be a manufacturer specific PID.
 
On a separate note for people who are having problems with the seats-- have you considered going to an automotive upholstery shop to see about having the bolsters reworked? Might cost you a few hundred bucks, but if that is the only thing keeping you from having the car you want (or keeping the one you have) it might be worthwhile. They probably wouldn't want to mess with the seat back since it has airbags, but the bottom should be an easy job for any decent shop. They could swap out the foam in the bolsters for something softer so it still fits the cover properly, but doesn't put up a fight when you put pressure on it.
 
That isn't how it works. OBD works via PIDs where the tool has to request the specific information and the controller replies with the data. So the tool has to be configured
to request what PIDs are wanted, it isn't just reading from an existing stream. There is also a function to ask the controller which PIDs are supported. If it was requesting all the supported PIDs all the time, it would be reeeeeeealy slow. Transmission temp is not an OBD parameter, it would be a manufacturer specific PID.
Thanks.

Maybe I'll start another thread rather than hijack this one.

My out-of-the-box software (Torque Pro and Scan Tool) has a short list of available PIDS for my car. I think I understand that it's the software that contains the PIDs that ping the OBD reader to make requests of the car's controller. It's those Mazda PIDs that I'm waiting to get to upgrade my software.

I should go to an OBD forum and get myself more up to speed on this rather than waste folk's time here.
 
Last edited:
Back