Frankenso injector drivers

[tomiata]

Some day I'd like to get a run like that with one of those thermocouples measuring the injector driver temperature and another measuring the ambient temperature. I honestly thought it was going to need a heat sink. I was expecting the injectors to shut off when it hit the temperature limit. I wonder how far off my prediction is. Or perhaps that would still happen if the ambient were to get hotter.

I can start some testing like this. I have a DMM with thermocouple input I will try to get some numbers.

[kb1gtt]

Any chance you have a DMM that can log to a PC? It would be handy to see the temperatures over time, such that we can see if we are at steady state or not.

[tomiata]

The DMM has a serial output that I've never used. I'll see what I can do.

[kb1gtt]

Oh, cool. What is the make and model of the DMM. Do you have software for it that you can use to capture the data on the PC?

[tomiata]

Oh, cool. What is the make and model of the DMM. Do you have software for it that you can use to capture the data on the PC?

Tekpower TP4000ZC

I got it a few years ago on amazon, and I see it's still there for $36 https://www.amazon.com/Tekpower-TP4000ZC-Interfaced-Multimeter-Computer/dp/B000OPDFLM

It came with some generic looking software, and I just tried it on an old laptop that has an rs232 port. The software is pretty basic, but looks like will do the job to log to a csv file, e.g.

Code: Select all

"DMM log file!!!"
"1","","27.9","C","12/17 15:37:25"
"2","","27.9","C","12/17 15:37:26"
"3","","27.8","C","12/17 15:37:27"
"4","","27.7","C","12/17 15:37:28"
"5","","27.6","C","12/17 15:37:29"
"6","","27.6","C","12/17 15:37:30"

[tomiata]

Check out this this setup before I do too much. I have injector 1 on PD3 to Q407. I tried to apply small blob of thermal grease on the thermocouple to contact the package. (It turned out be too much and made a mess )

IMG_20161217_171923~2.jpg (691.41 KiB) Viewed 16249 times

I did not set up the old laptop to do logging yet, but checked the temperature from the meter. The initial temperature was 17C. After starting the car, temp goes to about 38C in about 2 minutes at idle about 1100rpm. Reving up to 2500 rpm makes the temp go up to 45C within about 30 seconds.

[kb1gtt]

Looks good to me. I see thermal shutdown at about 150C. So that's the max. Also if you could measure the ambient that would be great, even if it's just a manual measurement once ever couple minutes. I suspect the car temperature does not change very much. Basically if we see your ambient is at say 20C, and the rises to say 85C, we are good, as the delta temperature is 65, and we are aiming for an ambient of 85C, so 85C + 65C = 150C which is the minimum shutdown spec. So basically we are looking for a delta temperature from ambient of less than 65C. So far you appear to have about a 28C raise. So far it's looking good, but what happens if the ECU gets some time to heat soak. Right now the board is cold so it's sucking up more of the heat. As that board gets warmer it will decrease how much heat you can pump into the board resulting in a hotter temperature. This is why logging is handy as we'll be able to see when it has gotten to a steady state and has sufficiently heat soaked the board.

[tomiata]

Testing while driving did not go very well. After a short time the thermocouple lost contact with driver package, from vibration I guess, and I did not get good readings. But before that I saw temperatures get up to about 53C cruising at 4000 rpm with ambient temp at 10C on the floor of the car. (It was very cold in Texas today.) And the temp reading dropped at lower rpm. I'll try again next week after I get a usb to rs232 adapter and can use a more modern laptop for this.

[puff]

10С ambient - you mean literarily on the floor, inside the car? could you please remind, where exactly in car you installed your ecu? I wonder, what happens if ambient temperature for the ecu (under the bonnet? the hood? whatever you call it) reaches somewhat 50-60C? these injector drivers are the hottest parts of the ecu, aren't they? or there is an idle stepper driver as well. then, there is a constanlty on driver for the relay, i guess, although with lower currents...

[kb1gtt]

I understand it's on the passenger floor. You call 50F (10C) cold. Last week we had -5F (-20C) with a -30F (-35C) wind chill.

So that's 10C ambient with 53C on the chip, so a rise of 43C. While this has not yet been fully heat soaked, it implies it could handle an ambient of about 150C - 43C = 107C (224F). In reality the 150C is catastrophic shutdown. There are various specs for the PCB. The most conservative I've found is UL's max temperature at 105C, but most PCB MFG's spec something closer to 125C to 150C. So we are best off if we can stay under 105C. So I would put your max ambient at 105C - 43C = 62C (143F). Granted the chip temp is not the same as the PCB temperature. However if you keep the chip below the PCB temperature you're not likely to exceed the PCB temperature. I used fat traces for these injector traces to prevent the traces from generating heat.

So if your ambient exceeds 143F (62C) it would fatigue the PCB adhesives, corrosion would be accelerated, etc. If you exceed something above 224F (107C), you would run the risk of the injectors shutting down.

Of course this is for a 4 cyl application. More cylinders will push that thermal limit. Also it's tentative as it has not been fully heat soaked yet.

Thanks for doing this. It's good to get some real world validation that it has upheld design intent. It's one thing to create a spec, then design to that spec. It's another thing to validate it and see that it upheld the spec.

@puff, when under the bonnet (hood) I would suggest doing a temperature analysis, similar to what @tomiata is doing. When I designed the board, I used a generic thermal resistance, found in an MFG's datasheet. I understand this thermal resistance assumes no air flow, with all conduction being through a 1oZ 2 layer FR4 board, which would be mostly true for a potted board. However we have air flow, 4 layers (more copper) and it's screwed to the housing fairly close to the injector drivers, so it has external heat sinking. This means we do not know the thermal resistance from those chips to the ambient. While @tomiata is getting some data which indicates what this thermal resistance is in the real world, it's based on some amount of air flow. Under the bonnet would have a different amount of air flow, as well under the bonnet would probably be potted to prevent humidity issues. So it would have a different thermal resistance, and should be analyzed. However it would also likely survive this environment with reasonably small considerations. AKA provide it with a heat sink.

[puff]

I am still with my frankenstein board)
Speaking of ambient temperatre, I believe, it is the ambient inside the enclosure which counts. And for this very purpose we are probably once again returning to the idea of a small thermosensor - installed inside the enslosure, if not onboard.

[tomiata]

10С ambient - you mean literarily on the floor, inside the car? could you please remind, where exactly in car you installed your ecu? ...

The 90 Miata ecu mounts to the fire wall at the passenger's feet. I have the frankenso on the floor not bolted in yet. Case was open for this test. Cabin temp at the floor was 10C. Outside temp was about -2C, air intake was about 5C.

What options are available for doing temperature sensing in the case?

[AndreyB]

What options are available for doing temperature sensing in the case?

Do we want a 0805 thermister in the prototyping area?

http://www.vishay.com/docs/33008/nths.pdf
http://www.digikey.com/en/product-highlight/v/vishay-bc-components/ntcs-smt-series-smd-thermistor

[puff]

or better through the hole - to measure ambient air, not the board. but do we have an extra channel for that?

[AndreyB]

or better through the hole

part#?

I have not seen anyone using all 12 analog channels.

[kb1gtt]

Something that is much more accurate would be a TMP36
https://octopart.com/tmp36gt9z-analog+devices-93082

Also the LM35 is accurate and easy to interface to a MCU input pin.
https://octopart.com/lm35dz%2Fnopb-texas+instruments-24814031

[AndreyB]

Just a reminder that we have MCU own temperature which is always above ambient obviously - https://sourceforge.net/p/rusefi/tickets/302/

[AndreyB]

My vote for LM35DZ/NOPB since analog inputs are much more disposable resource for us and probably less code to add. I doubt we need that much precision.

[puff]

just ordered the tmp36 thing - as far as i got it's almost the same as lm35?

[AndreyB]

36 "Temp Sensor Analog(Voltage) Serial (2-Wire) 3-Pin TO-92 Tray" vs 35 "±0.5°C Temperature Sensor with Analog Output with 30V Capability 3-TO-92 0 to 100"

[puff]

are you sure we are talking about the same thing? try googling it once again?
"Qualified for automotive applications"

Снимок экрана 2016-12-21 в 2.10.25.png (74.13 KiB) Viewed 16178 times

[AndreyB]

I did not go further octopart

[kb1gtt]

I was mildly familar with the LM35, the TMP35 is new to me. I stumbled on it when I was looking for the LM35. I like the range of the TMP36, and would suggest that one.
https://octopart.com/tmp36gt9z-analog+devices-93082

[AndreyB]

I am lost. Does it have 3 pins? How does it implement BOTH analog and serial digiral ? two-wire? output?

[kb1gtt]

For the TMP36GT9Z, the world "Serial" is not in the MFG datasheet, as well they show the internal transistor arrangement, it does not have serial capabilities. So this is a mistake in the octopart database. I see the Mouser and Digikey descriptions do not note serial.

[puff]

I wonder, what the wiring should look like? probably, wiring it up from 3.3V, with direct connection to ADC pin? without opamps, protection diodes, voltage dividers?

[AndreyB]

I wonder, what the wiring should look like? probably, wiring it up from 3.3V, with direct connection to ADC pin? without opamps, protection diodes, voltage dividers?

On Frankenso channels #9 and #11 have an easy way to bypass the op-ams and hook up right to 3.3 analog inputs.

[puff]

any suggestions for inputs on frankenstein? and how do we add this TMP36 characteristics curve to a rover config file?

[AndreyB]

any suggestions for inputs on frankenstein?

Please check the schematics since there is probably same on Frankenstein. If not just remove one of the volgate dividers resistors and a add a wire?

https://sourceforge.net/p/rusefi/tickets/326/ will cover the code side of things, and I still need to order TMP36GT9Z for myself.

[puff]

Technically, these are not thermistors. Why do we need two? (conicidentally I ordered a couple)
Do we need opamps and protection, if it sits right there on the board? may be just cutting an output from the opamp?