Auto Bed Leveling troubles

[printbus]

Both type 1 and type 6 are for 4.7K pullups.

Well, at least if you put any faith in the comments....

[Roxy]

Both type 1 and type 6 are for 4.7K pullups.

Well, at least if you put any faith in the comments....

OK! I didn't see that. Obviously... one of those entries should be removed.

[printbus]

OK, maybe more rant. Perhaps someone finds it amusing.

I think the lesson here is to put more faith in your relative temperature results than in the absolute values. Pay more attention to the results you are getting and adjust accordingly than what others are using for temperature settings and what filament manufacturers offer as recommended temperatures.

Some day, I may revamp what I use in the thermistor tables based on data from a manufacturer for a thermistor I've purchased from a trusted component supplier. These are all typically 1% parts, and IMO 1% manufacturer specs are going to potentially be far more accurate than what some guy (like me) measured in his basement or home oven with some cheap IR thermometer or other questionable calibration scheme. Frankly, I'd rather want to see and be working with true measurement data (and likely throwing out everything I had previously learned about temperatures on my printer) than temperatures that have been biased in an attempt to compensate for heating and/or measurement errors, whatever the source. But even this gets challenging. I've noticed most reprap suppliers indicate they are providing EPCOS part number B57560G104F thermistors, since that's the reprap "standard". As best I can tell, TDK/EPCOS made that an obsolete part number years ago. So, are these suppliers "helping" by translating to the current part number behind the scenes for us? Or are they really obtaining their thermistors from some low cost "EPCOS compatible" source and the parts are really of unknown specs? And of course, the code comments only say the tables are for an EPCOS 100K NTC thermistor. Jeez - I think they have about a dozen different ones.

Finally, having done my own temperature-control designs, I had to sort of chuckle when I first saw the 1K or 4.7K pull-up voltage-divider approach used on the reprap boards. This simple approach sets up the possibility for measurement errors due to self-heating in the thermistor (current flowing through a resistance dissipates heat). It might work "OK" for the higher temperatures we're using in the printers, but can be unbelievably inaccurate at measuring lower free-air temperatures. Feel free to search on "thermistor self heating" for more info.

[Roxy]

I think the lesson here is to put more faith in your relative temperature results than in the absolute values. Pay more attention to the results you are getting and adjust accordingly than what others are using for temperature settings and what filament manufacturers offer as recommended temperatures.

Agreed. It is kind of like when I was younger, my mother would always set the oven 5 or 10 degrees hotter because that was what was necessary to get things to bake right. You just knew that was the case and adjusted things.

[TopJimmyCooks]

Is there a defined testing methodology described somewhere? Hell, I've got a fluke multimeter with a thermocouple somewhere . . . I'm not saying for everyone to calibrate their own, but i would be happy to publish my Epcos clone thermistor and makerfarm 8" bed results. control a few variables like ambient temp and top surface of bed/type of glass and it should be close enough.

[printbus]

Agreed. It is kind of like when I was younger, my mother would always set the oven 5 or 10 degrees hotter because that was what was necessary to get things to bake right. You just knew that was the case and adjusted things.

Exactly. BTW, if that was an oven with the old mechanical type knob, there would have likely been a screw on the back of the knob that allowed the knob dial to be adjusted for calibration. I offered to set my mom's dial once, but she refused since it would have thrown off everything she knew about baking in that oven.

Is there a defined testing methodology described somewhere? Hell, I've got a fluke multimeter with a thermocouple somewhere . . . I'm not saying for everyone to calibrate their own, but i would be happy to publish my Epcos clone thermistor and makerfarm 8" bed results. control a few variables like ambient temp and top surface of bed/type of glass and it should be close enough.

There probably is, but it's likely buried in blogs or the reprap wiki. And you'd have to start with the fundamental question about whether the thermistortables file is meant to capture theoretical temperature measurement data for just the thermistor by itself, or is meant to include compensation for real-world losses affecting the net temperature gain from the heaters involved.

--------

I couldn't resist the temptation and spent some time digging into the data, starting with truth data direct from TDK/EPCOS. This was quite interesting. The amount of raw data is a bit overwhelming. If I think of a way to compile the info, I'll start another thread to share it. Here, however, is a summary.

The type 1 data in the thermistortables file closely follows the resistance-vs-temperature data provided by EPCOS, after converting for the simple voltage-divider approach feeding a 10-bit analog to digital conversion in the Arduino Mega processor. This suggests that the thermistortables data is intended to reflect data for just the thermistor by itself, not factoring for any real-world effects on the measurement. I'm not saying this is particularly good or bad - it's just the approach being used.

As Roxy pointed out, the type 6 data IN THE MARLIN BASELINE differs considerably for the higher end of the temperature curve. My conjecture is that Type 6 was originally someone's attempt at coming up with a set of data that did reflect real-world effects on the measurement. This is supported by the comment in configuration.h for Type 6 where it says "created using a fluke thermocouple".

And then there's the type 6 thermistortable data provided by MakerFarm. It turns out that this data is very close to being the same curve data as for Type 1. Here I conjecture that MakerFarm (or their firmware source) started using the Marlin type 6 data and found it undesirable. For whatever reason, the type 6 thermistortable data was altered instead of just changing configuration.h to specify use of the type1 data. Comments in the MakerFarm configuration.h file that imply the type 6 thermistor data is not as accurate as type 1 are wrong, since the comments refer to thermistortable data that has been changed in the MakerFarm baseline.

I doubt anyone would notice the difference between using Type 1, or using Type 6 as long as the MakerFarm thermistortable alterations go along with it. I would NOT, however, pull in the Marlin thermistortables.h file and stick with the Type 6 setting in configuration.h. That'll give you some wonky temperature results, as Pronus indicated earlier. You'd still be able to make it work - your "oven knob" would just be off more than it should be and it might take you a while to get used to adjusting the temps correctly. I can't help but wonder how many people have unknowingly struggled with that.

FOLLOWUP COMMENT: The compilation of data is now available here - http://3dprintboard.com/showthread.p...rtables-h-data