Marlin PID performance

Using Marlin's autotune on a J-head MK5B I get temperature swings of about +/- 1 C. Is this typical performance? I am sure it is adequate but a bit disappointing.

With my own firmware using bang bang I get about +/-3 C but it has much better transient response because if I say drop the temperature 20C with Marlin it winds up the I term waiting for the temperature to change and then overshoots considerably. Bang bang has no more overshoot on a temperature change than its normal oscillation.

With an aluminium heater block and the thermistor placed half way between the resistor and the barrel I can get +/-1C with bang bang.

Bang bang doesn't work well with Marlin because it doesn't sample often enough. I sample hundreds of times per second and the presence of some noise in the ADC reading then creates pseudo PWM with a small band of proportional control close to the set point.

So it seems all this extra complication doesn't actually achieve much or does anybody get better performance with PID and a MK5B?

---

[www.hydraraptor.blogspot.com]

10 ms is the minimum time required to get a sample from the built in ADC, isn't it? So you max out at 100 samples/second. I've heard Marlin does oversampling to compensate for the poor resolution due to the 4k7 resistor, so it performs even slower.

I think a good first step for better accuracy would be to replace the 4k7 with a 1k one and getting rid of oversampling to allow these 100 samples/second. I haven't investigated this closer yet, because "unusual" steps in firmware configuration tends to shy users away.

Also, does your electronics do noise canceling for the analog voltage supply?

---

Generation 7 Electronics	Teacup Firmware	RepRap DIY

Using Marlin's autotune on a Stoffel V1 and 3 selfmade Hotends I get temperature swings of about +/- 0.1 to 0.2 C !
Check the 5V generatet from der Arduino Mega most have to many noice when put Opto Endstops on the 5V

Hardwarekiller