On 4/18/2020 at 8:16 PM, nighthowlers said:
...
BTW, I am not sure what you mean by "ground level shifting". That is a term I've never come across and unfortunately unable to interpret. GND is simply a label reference for measurement purposes. In such a floating circuit what matters is the net potential difference, which should be 24V. Maybe what I interpret by ground level shifting is that when the power budget is exceeded, the power supply becomes current limited and is compensated by a reduction in voltage (i.e. 24V drooping).
...
I have come across this terminology in manuals of power supplies, especially high-power units with separate sense-lines. If the ground-wire is too long or too thin, you can get a significant voltage drop across this line. Let's say the ground line has a resistance of 0.1ohm, and the current is 10A. Then you get a voltage of 1V across the ground-line. So the "ground" at the electronics board is 1V higher than at the supply. This was the definition of "ground level shifting". Hense the sense-lines in the power supply, which needed to be connected as close to the board as possible.
The concept is also mentioned in manuals of analog to digital conversions, and precision measurements of National Instruments boards, but I don't remember if they used this terminology?
But this effect can also happen within boards, if the wires are rather thin, or the currents are high.
So, if there is a sensor that uses a common ground path with a power-eating source, then that ground level shifting could cause significant errors if the sensor works in the mV range. Power- and measurement-circuits need separate wiring, which should only connect at one point close to the supply (if using a common supply).
I don't know the layout of the UM-boards, so I don't know if this effect plays a role here. But it is a theoretical possibility. And the difference in quality with bed on and bed off, seems to point in this direction, although there could still be other issues of course.
And indeed, you can not measure this effect in a closed loop system by measuring the voltages in the system, since the regulation tries to adjust the situation to always read the same sensor-voltages. You need an independant temperature sensor with independant power supply (e.g. battery powered), parallel to the system under test. Basically a separate sensor or thermistor that you connect to the nozzle.
Power supplies with separate sense lines were required in the age of mainframe computers, because they would draw enormous amounts of current, and the supplies were located quite far away from the electronics boards. Power wires could easily be 3m long, or more, so you would get a huge voltage drop across these lines. Anyone working on the wiring of such old mainframe computers, had to be aware of this concept. For example, the Aesthedes graphics computer shown below from the 1980's had two +5V 40A power supplies, one +12V 15A, and one -12V 2A, all loaded close to their maximum, so that is almost 80A on the 5V line. But when PCs came along, the importance of sense lines dropped, so I don't know if they are still used today (I haven't done much electronics the last years).
Edited by geert_2
Corrected typos
- 1
Recommended Posts
nighthowlers 2
Dear @rcfocus: Thank you so much for the detailed and elaborate response. I completely understand what you mean by a non-optimal grounding path. I think it does have to do something with noise (as @Torgeir also alluded to). See one more data point below.
I turned the bed heater ON exactly in middle of the cube (so it was ON at 100% from middle to almost 3/4th way complete). No impact on quality, but as soon as the bed reaches it set_point of 60C and starts modulating ON/OFF, we see extrusion variations that correlate with the ON/OFF almost perfectly.
Link to post
Share on other sites