thermistor input?

[jharvey]

I've sketched up a draft of the thermistor input. Thoughts?



I expect the MOSFET will be shorted for now. I wanted to have pads available so that you can turn it on and off preventing self heating, increasing the accurate of this measurement(s). However it's probably best to keep focus on the min set of software for now.

Attached is an xls sheet that shows the curve for a thermistor I intend to use. How does this match with a typical automotive sensor?

Thoughts?

[Fred]

http://i260.photobucket.com/albums/ii15 ... eCurve.jpg

You tell me :-)

Self heating you say? Care to produce some numbers for that to show how much of an effect it has?

[jbelanger]

If you calibrate your thermistor(s) precisely enough to measure any heating effect AND tune your engine enough for it to have an effect, I'll be impressed!

I haven't done any computations but the power in the thermistor will be in the few milliWatts so any heating will be lost in any practical measurement. But as I say above, impress me.

Jean

[Fred]

I'm with Jean, the error involved will purely come down to case to junction thermal resistance which on a good sensor should be pretty low. I could conduct some experiments like this on the vehicle while parked I guess.

measure resistance with a meter briefly, wait 5m repeat, wait 5m repeat.

Connect 5v source through 2.2k resistor, give it 15mins (generous) to heat soak internally and disconnect and measure, reconnect, 5m discon measure, recon, 5m discon measure

The post 6 numbers which should be fairly close.

Besides, what is important here is not absolute temperature (though that is nice to know) but rather consistent temperatures accurate in absolute sense to maybe 5*C but consistent to 1 or less.

Fred.

[Delta]

A little off topic, but whats the learning curve like for Kicad - I'm reasonable with most simulators - very good with ICAP and LTSpice, and I can layout on eagle. Usually the FIRST program I pick up is the free/open source one, but in this case I missed the boat. Are the libraries good? do they have all the common new packages for MOSFETs etc like IPAK I2PAK TO-251 as well as the older TO-220 and TO93/92 etc or are we stuck with designing them ourselves.

On the thermistor front, generally you will find that they are actually specified for continuous use, so if a curve is published for R vs T then it will be after some predetermined warmup time taking into account heating due to current flow. I wouldn't be so worried about this at engine temps anyway as your around the 70-100degree area anyway. Its at startup to warmup where this has most effect, and depending on what fuel you are using depends on how critical it is to measure the lower temps accurately.

[jharvey]

Agreed that typically it can be ignored and calibrated out of the system. That's why I think it won't make it into the first round of software.

The problem I'm facing is that I have a need to use a thermistor as the intake air temp sensor. So I won't have the typical large heat sink that conducts heat very well. When I found the thermistor can only go up to 2mW and dissipates up to 6mW at 120C, I got concerned. It's not as much of an issue with a large heat-sink that can suck the heat away quickly. But that's probably not what I've go here.

So I put in a MOSFET to plan for IO path that allows the average to come below 2mW and perhaps even less as this temperature doesn't change much. Then you won't have to calibrate it, or plan for letting the sensor heat up. Simply allows for a more accurate reading at nearly no PCB expense.

[jharvey]

A little off topic, but whats the learning curve like for Kicad

I found KICAD fairly simple to learn, once you know the typical flow of PCB design software. There are some quarks and such, but once you learn them, it's not that hard. The biggest one is the mouse is handled isn't quite normal. The help file is actually handy unlike most help files. Give it a try.

I haven't gotten to the PCB side yet, so I don't know what the foot prints are like there. We've had to make three symbols so far for KICAD, one for the techart's board, one for the hall sensor, and one for the PIP protected FET. I don't think any of these items would have been found in other software packages.

[Delta]

Actually, I have found that if you have a temp sensor in an air stream that you need to calibrate the OTHER way, ie wind chill factor tends to reduce the temp of the sensor more than it really is, hence why autronic have an algorithm that uses the distance of the air temp sensor to intake valve and average head heat to calculate air intake temp - it always reads low + you have heating happening after you take the measurement. The values from a fast temp sensor tend to be lower than actual by about 1-2deg in an air stream thats even moving at a moderate rate. People neglect the fact that fast moving air tends to move heat quite quickly, perhaps the only time where you _MIGHT_ see something like current heating problems is at idle.

[Delta]

A little off topic, but whats the learning curve like for Kicad

one for the PIP protected FET.

I'm surprised you bothered with the MOSFET, it only has 3 legs like any other mosfet in TO-251 form. and its symbol shouldn't look any different in the layout, only the part number should be indicated, so I supose technically you could just copy a TO-251 mosfet (if one exists), choose one with a protection diode as part of the symbol and all is done.

[Fred]

wind chill factor tends to reduce the temp of the sensor more than it really is

Wind chill only applies to things with heat generation inside them. I guess if you are saying the self heating is included in the graph then sure, it will bring the sensor closer to true air temp and un skew the pre skewed results. Wind chill doesn't apply to inanimate things though.(you know what I mean)

hence why autronic have an algorithm that uses the distance of the air temp sensor to intake valve and average head heat to calculate air intake temp

Sounds like some of our ideas for heat soak correction to me. Though, that is for sensors reading higher than reality which seems to be MUCH more common in reality.