Ideas on how to reduce motor temps

[Mark77a]

I'm positive that your VDS motor temperature readings are accurate, so when it was reading 103'C I'm certain it really was 103'C and not a mere 55'C given the hot PEM, ESS and ambient temps.

It is possible that there's an elementary programming cock-up in the temperature conversions used for the graph scale . . . Its also possible that the actual stored temperature readings are raw integers recorded in Fahrenheit because its more granular than centigrade. Here's why it could just be a bug :-


Correct conversion C = (F - 32) x 5 / 9


But its possible the programmer has munged this up and got the numbers back to front, ie :-

Dodgy conversion C = (F x 9 / 32) - 5



Why do I suggest this dodgy formula? Because it just so happens that if you plug the 103'C temperature into the dodgy conversion formula you get about 55'c!!

EG, supposing the 103'c temp shown was actually rounded up from, say, a correct integer Fahrenheit reading of 215'F. Plugging it into the dodgy formula would yield the faulty 55' C as shown on the graph plot ..

ie, ( 215'F x 9 / 32 ) -5 = 55' C !!



Of course, having more data samples would either substantiate this or merely show my debug brain is clutching at straws. But there does seem to be a problem with the motor temp recording / conversion ... and I have spent a lifetime debugging software from machine code upwards :biggrin:

Yes, well spotted a fine bit of reverse maths !!!!
I am beginning to agree with this ... (pity, as motor running at 55 deg c rather than 103 deg c feels, well more cool and reliable ) .. I am not sure if TeslaGLoP is still supported > TeslaFlux I am guessing Doug George is Doug_G ?

If so Doug, any suggestions ? is this a simple fix ? (I wish I could offer some help on this but I'm just a mere mech eng )

 

[PV4EV]

In my experience the motor temp in general use (and not going for it) will usually hover around 70'c +/- 10'c. But ambient temp will shift this up/down as well.

You really have to be doing fairly illegal stuff on non-motorway roads before the temp gets into the 2nd yellow bar or 'power reduced' and its not like I hang around! Even on the recent track day I did (3 full on laps of Dunsfold per passenger run, as documented elsewhere on TMC) the motor temp didnt get past one yellow bar. Although I did cheat a bit and let the car cool off for 20+ mins between each run, whilst I went out in some of the other exotica doing an icecap-melting 3mpg ..

It will be interesting if Doug can review the code to see if there was a conversion cockup.


Meanwhile, here's a useful temperature range table for your perusal (might have already shown this) ..

 

[Doug_G]

Yes, TeslaGLoP is still supported.

There's no math in the application to screw up. The (reverse engineered) documentation that I have suggests that the data is stored in 0.5 degrees C steps. Seems to work for PEM, ESS, and Coolant. Perhaps the motor temperature is stored differently.

 

[Mark77a]

Yes, TeslaGLoP is still supported.

There's no math in the application to screw up. The (reverse engineered) documentation that I have suggests that the data is stored in 0.5 degrees C steps. Seems to work for PEM, ESS, and Coolant. Perhaps the motor temperature is stored differently.

Hey Doug thanks for your comments .. sorry ! up to now I had not connected your Handle Doug_G with TeslaGLoP ... durrrr !
Must say that is an awsome tool ... I just love what it can show ... so cool to demystify some of Teslas info in OUR cars.
Frankly, I wish Tesla were less secretive about their/OUR cars, info and codes ... now they are sharing patent - maybe they will also share Roadster workshop manuals and codes ?? (here's wishing anyway

Is there anyway we can help check the difference in Motor temp values between the log files and Touch screen/OVMS ?

the table from Scott451 above seems to match touch screen screen/OVMS motor temps, but does anyone know what the different versions are eg 3.4.15 x 2 and 3.5.17 ??

If the Log file/TeslaGloP values for motor temp are correct and the Touch screen/OVMS are wrong, this is VERY GOOD news .. as it means our motors run a whole lot cooler than we thought

 

[Doug_G]

Unfortunately a lot of people seem to conflate the opening of Tesla's patents with the opening of all proprietary information. It's not like they released their battery pack design to all - far from it. Assume you can access only those things that are in the USPTO database, and nothing else.

In principle, the motor can run a lot hotter than the electronics; it's made of metal and insulators instead of semiconductors.

IIRC there is a diagnostic mode on the Roadster, which can be engaged from the hidden screens in the touchscreen. It shows actual temperatures on the bar graph display. That will provide the answer.

 

[Mark77a]

Thanks Doug,
yes as in Picture, above HERE the diagnostics show motor temp of 103 degC vs Log temp of 55 degC.

It would be good to nail this discrepancy - is it in TeslaGLoP or in the interpretation of underlying data from the Car ?

 

[slcasner]

There's no math in the application to screw up. The (reverse engineered) documentation that I have suggests that the data is stored in 0.5 degrees C steps. Seems to work for PEM, ESS, and Coolant. Perhaps the motor temperature is stored differently.

In the once-per-minute drive record, only the PEM temperature and coolant temperature are 8-bit values in 0.5 degrees C steps, with the PEM temp based at 0 and the coolant temp based at -30. The motor temp and ESS min/max temps are 16-bit values with the binary point between the two bytes. If TeslaGLop is using only the most significant byte of the motor temp and dividing by 2, that would result in temps half of what they should be. But 55 times 2 is not 103. Does the VDS photo correspond to the same time as the data points highlighted on the TeslaGLoP graphs?

 

[slcasner]

I got all 3, Motor, PEM and battery into the last blue bar at peak .. with temps of motor: 103C (217F) , PEM: 45C (113F) & Battery: 40C (104F).

How did you get the numerical temperature values to show on top of the bar graph? I looked all through the diagnostic screens and did not see an option to turn on the more intelligent display. My car is a 1.5 -- perhaps only the firmware for 2.x has that option?

 

[hcsharp]

How did you get the numerical temperature values to show on top of the bar graph? I looked all through the diagnostic screens and did not see an option to turn on the more intelligent display. My car is a 1.5 -- perhaps only the firmware for 2.x has that option?

That's correct - only the 2.x version will display the numerical temps on the bar graphs. For those who have a 2.x, you select debug mode from the diagnostic screen. Your favorite Ranger can do this for you. Debug mode comes with a small price. You occasionally get warnings on the VDS that aren't very serious but sound like they are.

 

[Doug_G]

In the once-per-minute drive record, only the PEM temperature and coolant temperature are 8-bit values in 0.5 degrees C steps, with the PEM temp based at 0 and the coolant temp based at -30. The motor temp and ESS min/max temps are 16-bit values with the binary point between the two bytes. If TeslaGLop is using only the most significant byte of the motor temp and dividing by 2, that would result in temps half of what they should be. But 55 times 2 is not 103. Does the VDS photo correspond to the same time as the data points highlighted on the TeslaGLoP graphs?

Sounds like the definition of the record that I got was incorrect. Do you have the correct format for that message?

Thanks.

 

[slcasner]

Do you have the correct format for that message?

Coded as C++ accessor methods:

float pemTemp() { return data[7] / 2.0; }
float motorTemp() { return (*((short *)&data[8]))/256.0; }
float coolantTemp() { return data[10] / 2.0 - 30; }
float essTempMin() { return (*((short *)&data[16]))/256.0; }
float essTempMax() { return (*((short *)&data[18]))/256.0; }
int minTempBrick() { return data[20]; }
int maxTempBrick() { return data[21]; }

 

[Mark77a]

How did you get the numerical temperature values to show on top of the bar graph? I looked all through the diagnostic screens and did not see an option to turn on the more intelligent display. My car is a 1.5 -- perhaps only the firmware for 2.x has that option?

I tried to pick point where Log matched bar graph ... but as they change pretty fast i'd add a tolerance of say +/- 5 degrees .. which would make the
assumption of approx 50% not far off .... but PV4EV's reverse maths seemed more impressive.

Diagnostic mode is discussed .....
Vital Stats in the Diagnostics Menu
Follow links there eg speakev.com :smile: or if all else fails PM any roadster owner who regularly posts on here .. I was in same position as you a couple of months ago until it was explained.
Another reason for not posting here is Tesla may change it .. which would spoil our fun (for a while).

BTW thanks for posting code .. didn't mean a thing to me, but if it helps demystify these great little cars then ... ACE !!

 

[slcasner]

I tried to pick point where Log matched bar graph ... but as they change pretty fast i'd add a tolerance of say +/- 5 degrees .. which would make the assumption of approx 50% not far off ....

Thanks for the clarification, so the actual value of 111C (55.50 * 2) is not invalidated by your observation of 103C on the display.

but PV4EV's reverse maths seemed more impressive.

Perhaps, except that there is no F->C conversion going on because the car records the temperatures in Celsius. Also, I believe that the decoding of the records that I showed is correct because when scott451 and I were working on the reverse engineering several years ago I was lucky enough to cajole a then-employee of Tesla to give me a spreadsheet of log data for my car output from Tesla's own database of downloaded log data so I could see my stats for the 2009 REFUEL track day. That meant I was able to compare the reverse engineered results to the correct data until I had it right.

Diagnostic mode is discussed ..... Vital Stats in the Diagnostics Menu
Follow links there eg speakev.com :smile: or if all else fails PM any roadster owner who regularly posts on here .. I was in same position as you a couple of months ago until it was explained. Another reason for not posting here is Tesla may change it .. which would spoil our fun (for a while).

Oh, I have known the code for entering diagnostic mode for a long time. I have also visited Tesla Motors at each of its HQ locations since 2003, so I also know the origin of the number. hcsharp confirmed above that reason I could not find an option in the diagnostic mode to enable the numerical temperature display is that I have a 2008 Roadster 1.5, which has different firmware than the 2.x Roadsters.

BTW thanks for posting code .. didn't mean a thing to me, but if it helps demystify these great little cars then ... ACE !!

The code snippet is from my own command-line log decoder that I use mostly as a C++ programming exercise and to explore aspects of the data that interest me. I should have mentioned that the snippet is for the once-per-minute driving record (11) and that the data array begins after the 10 bytes of preamble, length, type, subtype, and timestamp.

 

[Perrin21]

How about some modifications to the AC system to send cooled air into the motor?

 

[djp]

How about some modifications to the AC system to send cooled air into the motor?

I don't think cool air is going to help much. The motor overheats just as fast on a cool spring day as a hot summer day, so I doubt dropping the air temp by 10C is going to make a difference when the motor is over 100C. We need a liquid cooling system, or a hydrogen cooled rotor!

Hydrogen-cooled turbo generator - Wikipedia, the free encyclopedia

 

[Mark77a]

How about some modifications to the AC system to send cooled air into the motor?

Not a bad idea.

I DO notice a big difference in motor temps when ambient temp is 30C vs 20C .. wouldn't be too difficult to physically try (the quickest, and most BS free, way to prove, or disprove )
Just put a 38mm flexi pipe from A/C OUT to motor or fan IN and test on favourite route. .. why dont you try it and report back ?

 

[hcsharp]

Not a bad idea.

I DO notice a big difference in motor temps when ambient temp is 30C vs 20C .. wouldn't be too difficult to physically try (the quickest, and most BS free, way to prove, or disprove )
Just put a 38mm flexi pipe from A/C OUT to motor or fan IN and test on favourite route. .. why dont you try it and report back ?

Where are you going to run that flexi pipe? I've been thinking about this for a different reason. I'd like to use the waste heat from the motor in the winter.

 

[Perrin21]

Where are you going to run that flexi pipe? I've been thinking about this for a different reason. I'd like to use the waste heat from the motor in the winter.

In theory that would also have a cooling effect on the motor if you extracted the heat using a fan rather than leaving it to heat up other parts of the system.

 

[Perrin21]

Heres another idea, How easy is the trunk box to remove? I wonder if it would improve airflow around the motor and PEM?

 

[TEG]

I heard that someone tried to do chilled gas (nitrogen from a canister) injection system when motor temps got high but it was only somewhat effective as the rotor gets too hot anyways and doesn't reject the heat well. I know (at least with early roadsters, not sure if all) that there was an electric insulating component to keep the rotor from becoming grounded through the shaft, but apparently this insulates the heat too and leaves the rotor hot even if you cool down the external parts of the motor. This insulating 'feature' was apparently done to allow 'reductive charging' where parts of the motor can be (double purpose) used in the charging circuit.