That number is the fraction of a step variation you might see from one step to another for an unloaded motor. Yes, it is due to magnetic inhomogeneities, and could be quite irregular around the full circle. Note that since a full circle add up to exactly 360 degrees and (usually) 200 full steps, for every step that is too big, there have to be small steps to make it all add up still. They can't all be 102%, since the number of steps per full circle is exact and constant.

Note that I use this principle to calibrate ultra-high precision x-ray diffractometry systems. With no external standards, and a highly repeatable (bot not accurate) angular scale, one can use 'circle closure', which is this property of cancelling errors, to define angular scales of truly extraordinary accuracy.

See, for example, https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4911639/
which is one of the papers I have published on this technique. The technique has been used since the 1800's (!).

Just in case someone on this thread missed the link from the original thread (linked at the very top of this thread), here is a paper I wrote on the topic of measuring stepper errors and encoder errors.

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4911639/

It may be of significant interest. This link is to the non-paywalled pubmedcentral source. The original (paywalled) article is in Metrologia (the official journal of the Bureau Internationale des Poids et Mesures, which oversees the metric system worldwide).

Also, here is a second paper, which discusses periodic errors in interpolated encoders. However, I don't think any of the discussions above refer to encoders that interpolate between their reference marks. The long, ugly token at the end of the link de-paywalls it. (Note that this is legal; it is an official U.S. Gov't publication, and free of copyright, at least in the USA). (Changed link... de-paywalling wasn't really working). It seems that the PDF link on Google Scholar may de-paywall this correctly. Try this:

https://scholar.google.com/scholar?hl=en&as_sdt=0%2C21&q=An+algorithm+for+the+compensation+of+short-period+errors+in+optical+encoders&btnG=

and use the PDF link

@Corexy The advantage of the Pt sensors over thermistors for wide temperature ranges is that the temperature coefficient is very well defined, and fully standardized across all the sensors. Thermistors have varying beta values, and when used over a wide temperature range can give significantly different results between two nominally identical units. In my day job, I work in a ver highly temperature-controlled lab with 0.01C regulation, and logging to 0.001C. This is a US Government standards lab. For this, we use calibrated thermistors because of their very high sensitivity. They are perfect for narrow-range applications. However, to cover a few hundred C or more, they are far from ideal. Although a Pt sensor is less sensitive (0.3%/C at room temp vs. typically 6%/C for a thermistor), they are very linear and interchangeable. A thermistor has too big a coefficient for wide ranges; a 100k thermistor at 25C is < 100 ohms at 280C. That's a huge dynamic range to cover. One the other hand, a Pt1000 is 1k at 0C and 2k at 273C. This is an easy range to digitize, and the resistance is high enough that modest-length wires don't affect the value too much.
The Pt100 sensors are well liked for heavy industrial and high precision applications because they are fairly robust, and the very low impedance makes them relatively noise resistant. On the other hand, the signal levels are low, and you must use a four-wire Kelvin connection unless the leads are very short, since the wiring resistance will be a big contribution to the total resistance. For high-precision work you can't even correct for the wiring resistance in a 2-wire connection, since you don't know the temperature of the wires.

@elvisprestley
The VIN voltage at the terminal block fluctuates from 24v - 9.5v. The voltage at the bed heater terminal block is constant at 9.5 V.
Your power supply has a big problem. Are you sure it doesn't have a 120/240 volt switch, set to 240 volts, and you are running on 120. This results in a power supply that typically fades in and out. Your Vin should vary at most a couple percent.
If you don't have the input voltage set wrong, the supply probably is defective.

@ignacmc Yes, a PT1000 sensor will have far better absolute temperature accuracy than a thermistor, if you don't have a precise way to calibrate the thermistors.

Thermistors have very high resolution, and can be easily calibrated over a narrow temperature range. I make much of my scientific living working with system controlled to absolute temperatures within 0.01C. Thermistors are great for this, if you only need a 10 degree span and have access to a calibration lab. However, without calibration, tiny errors in the Hart-Steinhart coefficients add up to very large temperature errors when used over a wide span.

Starting with something very close to linear in the first place, and with a temperature coefficient which is set by the properties of a (nearly) pure element, such as platinum, results in a a sensor which can be used over a very wide range with no calibration. A PT100 or PT1000 sensor has lower resolution than a thermistor (it's very hard to read to 0.01C), but a much wider reliable span of readout. It should be within 1C over the entire temperature range from room temp to 300C or higher.

You should get a decent quality calibration resistor (0.1% accuracy) of about 2000 ohms, to check that the duet readback is working right in PT1000 mode. If it works out right, you will be in great shape. The reason I bring this up is that, for example, my duet2 board (1.02 revision) seems to have a significant charge injection problem on the scanning ADC. Putting the PT1000 on channel zero shows a significant (5C) offset at room temp, but putting it on channel 1 is fine. This is likely due to residual charge on the sample/hold capacitor left over from whatever port the ADC scanned before reading channel 0. I haven't looked into the software to see in detail, but it is a classical scanning A/D problem.

My experience with PETG is that it likes the bed quite hot, to get good adhesion. I have an Ultrabase, which has 3mm glass on the aluminum base. The temperature is measured at the aluminum layer. A rough estimate of the heat flux gives me about 10 degrees lower at the surface of the glass than at the thermistor. If I set the thermistor to 90C, so the glass is about 80C, PETG sticks quite nicely. Much lower than that, and its adhesion isn't too good. One characteristic of PETG is that you don't want the nozzle too close to the bed on the first layer. If the nozzle is close, the PETG pulls itself along off the bed. It likes a bit of a drop height, as opposed to most filaments which like a lot of squish.

Also, I print PETG at a fairly high nozzle temperature, 250C. I am using a PT1000 sensor, so I am fairly confident of this number. The viscosity is fairly low when it is this hot, so one can print fairly fast (240um layers at 90mm/sec with a Hemera extruder and 0.4mm nozzle).

One could apply a statistical hack if one really knows the problem is every 20 samples, and either the first or last sample in a batch. One can just replace the (known) bad point with the average of the point before and after. This doesn't change the power spectrum significantly. This is equivalent to adding a random Dirac comb of small amplitude (the difference between the correct, missing point and the average) to the data, which has only a very weak effect on the spectrum. The noise floor at frequencies which are harmonics of (sampling rate / 40) will be slightly raised.

@amit-bandarwad I would go HTTP; you already have the network available, avoid extra wires, and (as mentioned elsewhere) ground loops. There is also a lot better handling of non-determinism in HTTP (well, in TCP really) that makes such communications very reliable. Communications hiccups are 'someone else's problem' in that they are handled by the TCP stack. USB/Serial you have to make sure you never get any surprises, such as cable jiggles causing disconnects, or extra characters due to diagnostic messages, etc. The HTTP protocol guarantees exactly what to expect.

@mrehorstdmd I think they were magnetized by a capacitor discharge, with the flux going through an armature which could be pushed out at the same time the real rotor was inserted, so the flux path never got a chance to open open. AlNiCo was sensitive to this. High-strength AlNiCo magnets required 'keepers' (flux shorts) to not lose strength when stored.

It's worth realizing just how much the performance of permanent-magnet motors has improved in 30 or 40 years, due to rare-earth magnets. They were really wimpy before, even if not accidentally demagnetized.

@jay_s_uk You and I said the same thing, about the negative temperature coefficient. I said resistance is higher at lower temperatures; you corrected to lower at higher temperatures.

However, your statement about the value for a thermistor being specified at zero degrees is wrong. Most thermistors are standardized at 25C. Pt100/Pt1000 are, indeed, standardized at 0C. But it is still correct that the 100000 is the right number to have put in the file.

@Manuel207 Are the parts really rectangular, or parallelograms? They would have the right size in x & y if there is skew, but the right angles will be very slightly off right. It is most noticeable on circles, since slight bumps on the corners of rectangular pieces make it really hard to measure the angle between the flat faces. What I did was printed a large, thin circular plate and then measured it carefully, and adjusted the skew until it was round. I now have a really well-calibrated printer.

@Manuel207 Are the oval's axes aligned along x & y, or at 45 degrees to x & y? If the axes are aligned along x & y, it's not the skew compensation, but a scale issue between the axes, or backlash in the belts. If it's at 45 degrees, the skew compensation would fix it.

(update... oops. I forgot this is core-xy, so the syndrome would be different. since the belts already pull at 45 degrees, both belt slip and skew should make an error at 45 degrees. This still leaves the question as to which axis the error is on.)

Have you tried kapton tape (which is extremely heat resistant), or the off-brand Koptan that comes from China (and is MUCH cheaper). They are an absolutely marvelous print surface. The only trick is the initial application of the tape, to get it smooth an bubble-free. Once you have a nice layer, it lasts a very long time. The silicone adhesive on it sticks nicely to aluminum, and works fine up to beyond 300C. I bought a roll of 9" wide tape (the Koptan from China), I think a pound of it, a few years ago for maybe $40. It is a lifetime supply.

@zapta Make sure when you do the boiling water 100C point, that you look up the local barometric pressure, and then use a conversion table to get the correct boiling point. It varies quite bit, especially if you are at altitude. This is why an old explorers trick in the mountains was to keep a thermometer, and put it in boiling water to measure the altitude.

@zapta A class A RTD is probably the most reliable, affordable temperature measurement device you can use to cover a moderately wide range with no special calibration. A thermocouple of appropriate type can be used without any calibration, and fairly good accuracy, over a very wide range, but the resolution isn't so good. If you want to go first-class, an SPRT (standard platinum resistance thermometer) is ideal, but costs. These are RTDs manufactured to extremely precise specification (loosely wound, special platinum wire, so thermal stresses don't change it, etc.).

@deckingman @zapta The difference between using an NTC and either a Pt RTD or thermocouple is that tiny variations in beta on an NTC correspond to large variations when applied over a wide temperature range. RTDs and thermocouples have low sensitivity (resolution), but tend to be more accurate without secondary calibration over a very wide range. The problem with NTCs is that commercial uncalibrated ones are trimmed to give the right answer at 25C, typically, and without secondary calibration to determine specific values of beta and c (etc.), one is using faith-based thermometry.

Note: I make my living on experiments that require precision (<0.01C) thermometry. For narrow range work, calibrated NTCs are unbeatable. I have one calibrated to 0.01C accuracy from -30C to 70C. The cost of the unit and calibration by Fluke was $1800. To go outside that range, though, really is hard with that type of sensor.

This is a much harder task than you are expecting. The temperature measured by a surface probe will depend critically on the thermal contact made between the probe and the surface. The temperature of the surface may also not be terribly close to whatever temperature you really care about on the hot end (the melt chamber? the nozzle tip? what?). The probe itself, of course, may not be accurately calibrated in the first place.

The closest we probably get to good measurement of hot end temperature probably comes from either a Pt RTD (Pt100 or Pt1000), or a thermocouple. Either one must be seated deeply in a well in the heater block. At this point, you probably know the temperature of the core of the heater block to a degree or so. The actual nozzle temperature is probably much lower, and that may be critical for the melt characteristics. Of course, the difference between the heater block and the nozzle tip will depend strongly on the fan speed and filament flow rate.

So, what are you really trying to accomplish?

@hpiz @infiniteloop A convenient number to remember about the AWG scale is that a 3 gauge number decrease almost exactly doubles the area of the wire. (See https://en.wikipedia.org/wiki/American_wire_gauge). This is an ancient bit of lore, but it is pretty easy to remember, and useful for scaling wire sizes. The statement that CAT6 wire is inappropriate for running steppers is absolutely correct. Get thicker wires.

@deckingman @samlogan87 Actually, if you want, you can regenerate calcium chloride in your oven. However, the energy cost probably makes it a dumb idea. It is very cheap in the first place, and the brine is fairly environmentally harmless, so putting it down the drain isn't a problem. Note that CaCl2 doesn't get to as low humidity as silica gel, but it is plenty low for filament drying. It can absorb a HUGE amount of water compared to silica gel, which is nice for keeping a big box of filament dry. As long as there is any solid CaCl2 left in the brine, it is keeping the RH somewhere in the 10% range near room temperature.

@kuon run multiple 12mm belts side by side?