Torque Log Analyzer - Improve Your Hybrid Driving

[CPN]

If it's fully powered down, then yes it is but just turning the screen off (momentary push of the power button) and then back on is not rebooting. Long press the power button and you should get a popup to "Power down". Turning back on from this state requires a longer press on the power button and is a cold boot. Worth a try. Unusual for the BT driver to be an "app"? Still, your phone has bluetooth 4.0 so it really ought to be ok... If those MG1/2 readings are still drifting after that then I would suggest another ELM327...

[Lesolee]

Well I bought the known working adapter (as suggested by CPN) …

http://www.ebay.co.uk/itm/271623595403

and it does the same as my original adapter so that the standard PIDs work, but the extended Hybrid PIDs don’t work properly. For example MG1 temperature reads room temperature or -40°C, not quite alternately, but wrong about half the time.

I deleted Torque Pro completely and side-loaded an older version (1.8.16). Exactly the same fault (regardless of adapter settings in terms of the auto-timing disable and the faster communication setting.)

My failure is complete.

[Alessandro71]

Have you tried with the proven PID set you find on the download page of the Torqueloganalyzer site?

Be sure to not use the Prius 2 version on the 3 and vice versa.

[CPN]

To use the terms "Prius 2" and Prius 3" is a bit confusing for Yaris and Auris owners and I would suggest using "Gen 2" and "Gen 3" instead as these are more relevant.

@Leslie: you are using the wrong set as what gave me a clue was you were talking about "MG1 Temperature" and that is from the "GenIII Prius 4-24-12 (Metric).csv" file and not all of those work for the Yaris/Auris. The set of PIDs you want and for anyone else who has a Yaris MY2014 and newer, is the one attached to this post... It's a smaller, less comprehensive, set but they all work well...

HSD3 PIDs subset (Metric).txt

(rename the .txt file extension to .csv before loading it onto your phone for Torque)

[Lesolee]

Although I haven’t got to the root cause of the problem (which is me being a dumb f*ck ) I have made a significant “breakthrough”. I went to Alessandro’s website and “viewed” the reduced set of PIDs.

The link is the HSD3 PIDs one.

http://torqueloganalyzer.blogspot.co.uk/p/downloads.html

When you click on it to view it I got this

https://drive.google.com/file/d/0B8xG59QDgvQ3MWRGaFJpNFV1dFk/edit?pli=1

Looking at the “header” field on the end you get numbers like “700” and “7E+2”. When I looked at the version I was using I had “7.00E+00” and “7.00E+00”. BUT, that was viewing the csv file in Excel. When I used Notepad++ the ACTUAL strings are “7E0” and “7E2”. The viewer (in both cases) is taking a three digit HEX field (within a string) and parsing it as a decimal in exponential format!

It is important to note that Excel reads the quoted text strings and converts them to numbers so that when it re-saves it they are numbers not text strings. Torque Pro is evidently very weak on its type checking and error checking.

So, editing the csv directly in Notepad++ is the way to go. I took the more interesting PIDs, moved them to the beginning of the file, but crucially added a “full stop” at the beginning of each long name. The purpose of this is to make the interesting PIDs appear at the beginning of the list in Torque. Even so it is a total pain putting over 30 digital gauges in via the Android interface. (I did try saving the partial setup as a .dash file in an attempt to speed up the process, but Torque has a pseudo random format for the .dash file and the file structure was not self-evident).

The ELM327 (ebay) adapter gives 31.0 reads/second and the (much) more expensive POWER2 (amazon) adapter has 34.6. Crucially however, the ELM327 version did not read all the Battery voltages correctly (at full speed) whereas the POWER2 one did.

In my reading of the GenIII Prius file I found a “reverse beep enable” command code, “3BAC00”. I thought this might “repair” my broken reverse beeper (which only beeps once in reverse now, having been corrupted by the built-in Torque scan tool). Sadly that command has yet to work, possibly more finger trouble on my part.

Alessandro and Colin: thanks for your help and support.

[Alessandro71]

Excel is the root of all evil

Don't bother with it

Be aware that renaming PIDs may cause the log analyzer to not recognize them any more.

[CPN]

Leslie, post screen shots of any Torque dashboards you develop and how you created them because I, for one, would be most interested...

[Lesolee]

Reversing Beep Fixed

Sorry for being a bit off-topic, but I said in my first post that my reversing beep was “broken” by the scan tool within Torque Pro. I found out how to fix it.

This Spanish website had the answer ...

http://translate.google.co.uk/translate?hl=en&sl=es&u=http://auris-tac.com/viewtopic.php%3Ff%3D22%26t%3D482%26start%3D10&prev=search

The codes you send are …

AT SH 7C0
21AC
3BAC00
21AC

(notice that there is no letter O used above, they are all zeros; 0.)

Which are easily programmed into a custom button in Torque Pro.

Add display | Push Button

When asked to select the type of button use “Send OBD command”

Then go into the configuration of that button (“Display configuration …”)

“Set command to send”

Type all 4 lines as above. (I used all upper case, the Spanish post used all lower case. Both evidently work fine. I don’t know what all the commands mean but the “3BAC00” is the reverse beep enable command as found in the GenIII Prius PID file. (Using 3BAC40 will apparently turn it off). Just sending 3BAC00 on its own is not enough to do the job. I presume that the same “preamble wrapper” (all the stuff surrounding the 3BAC00) would be used for the other configuration commands such as enabling/disabling the seat belt beeping.

My success is complete

[Lesolee]

Finally on topic!

I deleted the extra full stops in my titles of my log file so the analyzer woudn't be unhappy. This was a 1 hour trip up and down some quite hilly terrain and for the whole circular route (start and stop from the same location) the MPG was 71.

Considering my Auris is 1.5 years old with only 8000 miles on the clock, the amount of red in the analysis sheet seems a bit excessive.

[Alessandro71]

Thank you for your results.

I'll add it to my database.

As for the red cells, they represent only the worst value of each column, not necessarily a faulty one.

As of now, this kind of analysis has proven its value in diagnosing faulty batteries, correctly identifying broken cells to be repaired at a fraction of the cost of a whole Battery replacement.

We don't have enough cases to develop a forecast of Battery life expectancy yet, but we're working on that.

Your Battery seems perfectly fine right now.

If I had to make an educated guess, I'd keep an eye on the two blocks with the lowest voltage, but it may take years before they'll fail.

If you have a Facebook account, keep in touch with the Torqueloganalyzer page for news.

[johalareewi]

Excel is the root of all evil

Don't bother with it

Don't know about the root of ALL evil, but Excel does an excellent job of trashing CSV files.

Never ever use Excel to edit CSV files. Notepad is OK but best is something like UniCSVed.

[Lesolee]

I have been looking at my recent trip log to see what information can be gleaned from it. Since Battery condition is a topic of interest I wanted to verify the PID which gives the max-min block voltages. This is easy to do in an Excel spreadsheet as we have all the individual block voltages and we just need an extra column to calculate the max-min value from the 14 blocks. I had assumed (falsely) that the two values would be equal to the millivolt level. In fact the attached graph shows that they are wildly different.

The way I am interpreting/understanding this response is as follows … A low cost measurement system would not sample all blocks simultaneously. It would use a multiplexer (selector) on the front of the acquisition devices (analog to digital converter) and would step through the banks one at a time. This means that the banks are not measured at the same instant in time and therefore the bank voltages will differ from each other during large transient current events. From the log it is clear that the current is changing by as much as 100A between successive readings and that constitutes a pretty serious transient!

The bottom line is that the built-in block difference reading is suspect during fast transient currents. We want to only see valid block difference readings. Alessandro is looking to see how many samples the block voltage is different for. I am approaching the problem in a slightly different
way. I am comparing each block voltage to the instantaneous average of all block voltages (the average across all 14 blocks for one row in the log file). This gives less excessive excursions, which to my mind are more representative. Additionally, I have filtered the difference values by averaging the present reading with the immediately preceding value (a rolling average of 2 points).

If a cell is damaged or weak it should be weak on both charge and discharge, so the max and min value (magnitudes) should be fairly equal. And they are mostly, the asymmetry probably being due to this non-simultaneous sampling issue. Averaging the absolute values of the deviations should give a robust consistent metric. This looks good on my (newish, good) Battery. The real test would be to try it on known bad Battery data sets.

[Alessandro71]

I have a bad Battery data set.

Contact me at the address you see in the about box of torqueloganalyzer

[Lesolee]

For this analysis I am modelling the Battery blocks as ideal lossless batteries in series with a resistor equal to the series resistance reported by Torque Pro. The blue line is for the 0.024ohm series resistance per block reported on my newish (8000 mile) Auris. 100A seems to be the maximum regenerative braking current possible, as seen from the log. That must correspond to the charge indicator on the console being at maximum.

The Battery is charged linearly by current, but the resistive loss increase as the square of the current. That means the charge or discharge efficiency drops off as the current increases. It is therefore apparent that if you only brake to 50% of the maximum regenerative capability you get more efficient regeneration. However, on a new Battery the change from 87% to 93% is not that great. On the other hand if the battery is old and cold, and the series resistance has gone up to 0.100 ohms, the efficiency improves from 61% to 75%.

I don’t (yet) have any information on how the series resistance changes with time and temperature but this should be a major loss mechanism within the hybrid system. There are other losses of course

• Motor/generator loss
• Inverter loss.
• Galvanic efficiency loss in battery (1 Ah in doesn’t give 1 Ah out)

It’s hard to imagine the motor/generators being better than 95% efficient, and the same goes for the inverters. That being the case the overall efficiency for stop start driving must be pretty horrible and therefore nothing for the manufacturer(s) to shout about.

generator to inverter 95%

inverter to battery 95%

into battery 87%

from battery 87%

galvanic efficiency 95%

battery to driver 95%

driver to motor 95%

That lot amounts to at best 58% restoration of the kinetic energy when braking and re-accelerating on a vehicle in newish condition. These efficiency values are not going to be readily obtainable; all we really experience are the mpg figures.

(I have attached the (evil) Excel spreadsheet so you can check my method/calculations if interested. It is ZIP'ed because this website apparenty agrees that Excel spreadsheets are evil!)

seriesR.zip

[Lesolee]

Alessandro kindly provided me with a Torque log file for a faulty Prius 3, labelled Bosco, which I assume is this one …

http://www.hybrid-synergy.eu/showthread.php?tid=55253

(read using google translate as it is in Italian)

This is known faulty Battery. The first thing to note is that the series resistances are all close to my readings, despite this one having 183,000km on the clock. It’s filtered block voltage deviation from the mean is 1.3V on block 4, showing a clear fault.

On the attached image there are two distinct characteristics visible.

1. The state of charge suddenly drops from 48.2% to 24.7% for no obvious reason!
2. The block voltages appear to jump up for no reason (highlighted in yellow)

What seems to have happened is that the current reversed from 7.3A discharge to 23.3A charge, but that was only registered on the next set of readings.

The current changed by 30A. The instantaneous change in voltage is the series resistance times the current. So Block 1 changes by 1.25V for about 30A so the resistance is 0.042 ohms. The system is displaying this as 0.027 ohms in a later column of the log. Block 4 however changes by 2.18V meaning 0.073 ohms. Again it reads a healthy 0.025 ohms in the log.

Series resistance in a cell is bad. The “direct” measurement of series resistance using a pulsing current does not always give a good answer. I assume that is what is happening here, but I have no actual information about how Toyota are delivering the series resistance value.

Obviously if yours does this under warranty you would get the Battery replaced. If out of warranty then my first effort would be to overcharge the Battery as much as possible. At the moment the only way I have managed to get all the bars in the charge indicator lit up is to coast down a LONG steep hill with the regenerative braking on. Otherwise the ICE will automatically cut out when there are still two bars to go. Having got all the bars lit up, I would then run it on electric at its maximum capacity until the battery was as flat as I could get it. The lowest it typically goes on mine is two bars. Again I don’t know any “tricks” (in the car) to drain the last couple of bars for this battery reconditioning step. The battery pack doesn’t have built-in cell balancers and from what I have read it seems Toyota have used such a well matched group of cells that they will all stay together throughout their life. Thus changing a single block is not recommended (by them). Overcharging them is the only way to rebalance a weak cell.

[Alessandro71]

thank you for you analysis

it's not entirely clear for me, but it's my fault as i lack some electrical background

you are referring to block voltage deviation and resistance values that i don't see in the attached image

does your method of identifying block 4 as faulty pinpoints the same block even in the first log i passed you?

[alesf76]

Having got all the bars lit up, I would then run it on electric at its maximum capacity until the battery was as flat as I could get it. The lowest it typically goes on mine is two bars. Again I don’t know any “tricks” (in the car) to drain the last couple of bars for this battery reconditioning step.

Just shift to "N" and turn on A/C. WARNING: doing this you can make the SOC going too much down and make the car to turn off and require an HV Battery recharger. Go back to "P" when just a bar is left on the SOC display.

[Lesolee]

you are referring to block voltage deviation and resistance values that i don't see in the attached image

does your method of identifying block 4 as faulty pinpoints the same block even in the first log i passed you?

Fair enough. Let me go through it more carefully. Referring to BoscoPrius3.jpg

The yellow highlighted line of Battery block voltages is the starting point. Block 1 is 15.438V but was 14.188V on the previous measurement.

Change is 15.438 – 14.188 = 1.250V

Block 4 is 14.878V but was 12.698V on the previous measurement

Change is 14.878 – 12.698 = 2.18V

Current changed from 7.28A to -23.26A

Change is 7.28 - - 23.26 = 30.54A

OHMS LAW:

VOLTAGE = CURRENT times RESISTANCE

Slightly more difficult differential form

(change in) VOLTAGE = [(change in) CURRENT ] times RESISTANCE

Block 1: 1.25V = 30.54A times RESISTANCE

Therefore RESISTANCE = 1.25/30.54 = 0.0409 ohms (not shown in jpg)

(the log shows 0.027 ohms but this is not shown in my jpg)

Block 4: 2.18V = 30.54A times RESISTANCE

Therefore RESISTANCE = 2.18/30.54 = 0.0714 ohms (not shown in jpg)

(the log shows 0.025 ohms but this is not shown in my jpg)

My previous measurement method subtracted individual block voltages from the mean block voltage for that row of the log. Both the “error" and “no error” logs you supplied both showed Bank 4 is having a -1.3V maximum deviation.

I am now going to reanalyse all the data using a change for each block relative to itself, that is the change between successive readings. More later …

[Lesolee]

I had been having trouble when trying to re-analyse the Bosco log files and I discovered why. They were logged at 1 second intervals, but there are something like 50 PIDs. The read rate on my adapter is 30 something per second (from memory). These logs need to be taken with the Torque Pro option “enable synchronous logging” selected. The fact that this was not done means the reads have been broken up across rows of the table, making my new differential analysis fail.

The attached image, delta.jpg, shows the change of voltage on each bank from one row to the next. You will see lots of zeros where the readings didn’t change on that row. Very nasty from an analysis point of view. The position at which the zeros starts changes with time, which suggests to me that the Battery bank voltages are read according to an internal clock and they are being read out by Torque Pro without regard to that clock. This reinforces the previous finding that the values are measured one at a time and therefore are never all taken at the same time as each other, or as the Battery ampere reading.

It’s going to take quite a bit of thinking to figure out how to get a more meaningful analysis of the actual Battery characteristic, as compared to the distorted view we are currently getting due to the non-simultaneous sampling.

[Alessandro71]

For this analysis I am modelling the battery blocks as ideal lossless batteries in series with a resistor equal to the series resistance reported by Torque Pro. The blue line is for the 0.024ohm series resistance per block reported on my newish (8000 mile) Auris. 100A seems to be the maximum regenerative braking current possible, as seen from the log. That must correspond to the charge indicator on the console being at maximum.

seriesR.jpg

The battery is charged linearly by current, but the resistive loss increase as the square of the current. That means the charge or discharge efficiency drops off as the current increases. It is therefore apparent that if you only brake to 50% of the maximum regenerative capability you get more efficient regeneration. However, on a new battery the change from 87% to 93% is not that great. On the other hand if the battery is old and cold, and the series resistance has gone up to 0.100 ohms, the efficiency improves from 61% to 75%.

I don’t (yet) have any information on how the series resistance changes with time and temperature but this should be a major loss mechanism within the hybrid system. There are other losses of course

• Motor/generator loss
• Inverter loss.
• Galvanic efficiency loss in battery (1 Ah in doesn’t give 1 Ah out)

It’s hard to imagine the motor/generators being better than 95% efficient, and the same goes for the inverters. That being the case the overall efficiency for stop start driving must be pretty horrible and therefore nothing for the manufacturer(s) to shout about.

generator to inverter 95%

inverter to battery 95%

into battery 87%

from battery 87%

galvanic efficiency 95%

battery to driver 95%

driver to motor 95%

That lot amounts to at best 58% restoration of the kinetic energy when braking and re-accelerating on a vehicle in newish condition. These efficiency values are not going to be readily obtainable; all we really experience are the mpg figures.

(I have attached the (evil) Excel spreadsheet so you can check my method/calculations if interested. It is ZIP'ed because this website apparenty agrees that Excel spreadsheets are evil!)

seriesR.zip

I've finally had the chance to look at your excel.

I've got your current and efficiency relationship, what is not clear to me is how you've got the values in your sheet.

Were they collected from an actual log or are they just a theoretical modeling of Battery behavior?

[Lesolee]

I've finally had the chance to look at your excel.

I've got your current and efficiency relationship, what is not clear to me is how you've got the values in your sheet.

Were they collected from an actual log or are they just a theoretical modeling of battery behavior?

If you mean post #39, and the plot "seriesR.jpg" this is me modelling the Battery performance on the basis of the series resistance measurement from the Torque log. The degradation of series resistance with time or temperature was a "what if" analysis. I have no evidence that this actually occurs. Furthermore I no longer believe that the readout of series resistance from the Torque log has any validity, although actual evidence would readily convince me otherwise.

[Lesolee]

I haven’t (entirely) lost interest in the whole Battery health thing, but I do have another interest. I currently feel that the readout of “internal resistance” on the Battery banks is misleading, given that a (particular) faulty Battery had very reasonable (low) values for this parameter. If anyone has any data where the internal resistance value goes above 0.05ohms I would be very interested to hear about it.

My logging set now has all the internal resistance values deleted (apart from bank 1) to speed up logging on the useful values.

Today’s project was to find out how the power drawn from the battery varies with speed in electric mode. This requires a level road surface of quite some length, with no nearby vehicles, because it is necessary to concentrate on the speedo – which would be dangerous on a public road.

I started off in a car park trying to do a run in each direction to average out any slope and wind direction. I have attached all the data in (evil) excel format so you can see what I have done should you be interested. Power consumed is simply the sum of the battery bank voltages multiplied by the battery current.

I have averaged the speed over the relatively constant speed sections and averaged the power over the same sections.

The runs were intended to be 10mph, 20mph, 30 mph and 40mph. I missed these targets quite badly. Worse than that, the graph of the data looks hopeless.

[deleted graph]

If the car takes a certain amount of power at 10mph it could reasonably take twice that power at 20mph, 3x at 30mph, and 4x at 40mph. This would mean constant efficiency and no significant increase in drag with speed. Suppose 10mph required 1000W. 2000W at 20mph would be the same efficiency and wind loading. More than 2000W would suggest either reduced efficiency in the power train or increased drag at that speed. Less than 2000W would definitely mean the power train was more efficient at the higher speed.

I need a longer flat area for my test. Alternatively it would be great if some interested volunteers with access to long flat areas could log some data as well (although not by putting themselves or others in danger).

I also (ideally) need a way to recharge the battery at the start of an individual test run. I had thought I could put the car into service mode, pump the accelerator pedal, and charge it that way. Sadly, whilst the ICE does run, it doesn’t create a significant charging current.

[deleted attached data: New better data set captured, see next post]

[Lesolee]

I re-did the tests at a new location with much greater success.

auris.zip

[Lesolee]

This is a further analysis of the data from post #48. The steady power loss when stationary was measured as around 220W. This is with no fan, no radio, no aircon, no headlights. Subtracting this from the power used at each speed and then dividing the result by the road speed we get the blue curve. This shows an increasing W/mph which means less efficiency at higher speed. According to this simple view you get best efficiency at 10mph in electric mode. However, that is not the full story. If you now load up the steady power by using the radio, the aircon and/or headlights, the peak efficiency point will shift to a higher speed.

If that doesn’t quite make sense think of it this way; for a given trip, if you travel at twice the speed you are consuming the fixed power for half the time, using half the energy. So you minimise the effect of fixed losses by driving faster. The greater the fixed losses, the greater the optimum efficiency point will be.

The red curve includes the (minimum) 220W fixed losses and shows that there is then little difference between travelling at 10mph or 17mph.

The next challenge is to find the power consumed at 40mph and above. It is at present unclear how this can be done because the Battery is going to charge itself (consuming power) and the electric motor/generators will also be running, confusing things. It is also going to require a long flat road to get to a steady state.

[Alessandro71]

I had been having trouble when trying to re-analyse the Bosco log files and I discovered why. They were logged at 1 second intervals, but there are something like 50 PIDs. The read rate on my adapter is 30 something per second (from memory). These logs need to be taken with the Torque Pro option “enable synchronous logging” selected. The fact that this was not done means the reads have been broken up across rows of the table, making my new differential analysis fail.

I collected another log with the same set of PIDs on my car with "synchronous logging" enabled

log entries are still 1 sec apart