Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Power interface and distribution #3

Open
christiankuttke opened this issue Jun 28, 2024 · 18 comments
Open

Power interface and distribution #3

christiankuttke opened this issue Jun 28, 2024 · 18 comments

Comments

@christiankuttke
Copy link
Collaborator

Design a power distribution solution that fits in the box, has 1 universal power input, and gives all devices their needed power.
@christiankuttke
Copy link
Collaborator Author

Proposed input:

220V AC Kaltgerätekabel (C14, Buchse C13) with fuse on the outside of the device.

image

Power switch should have a power indicator light but should be protected from accidental turn off.

C13 is available in a locking variant which is preferable, if available.

@christiankuttke
Copy link
Collaborator Author

christiankuttke commented Jun 28, 2024 •
edited
Loading

Devices to power, ordered by Power Nets

Estimates are worst-case and over-specced.

12 V

  • Our uc32_standalone board featuring TMC2202 drivers, ESP32, interfaces for illumination, encoders, limit switches, PWM output (3A or more if necessary)
  • Hard drives (HDD spinning) at spin-up 2A per drive
  • Jetson orin (4A)

5V

  • Raspberry Pi Screen (500 mA worst case)
  • Nvidia Jetson Nano (3A)
  • Raspberry Pi (3A)
  • GPS module: 1A
  • HDD: 2 A
  • SSD: 1 A

Totals

Raspberry Pi-based system, with 1 spinning HDD

12V 5A and 5V 6.5 A (92.5W)

All other devices can be powered by the boards and are probably covered by the generous estimates of those board's consumption.

@beniroquai
Copy link
Contributor

This would be a high-level wiring chart:

image

now:

minimize(number(cables))

@beniroquai
Copy link
Contributor

beniroquai commented Jun 28, 2024 via email

@christiankuttke
Copy link
Collaborator Author

christiankuttke commented Jun 28, 2024 via email

@beniroquai
Copy link
Contributor

Wait, you want 230V inside the box? Absolutely not. This is difficult for many reasons. Let's keep it 12V only. I would recommend to use something like a through-hole connector for the 12V, no?
image

https://www.amazon.de/RUNCCI-YUN-Adapter-Hohlstecker%EF%BC%88Mit-Stecker-%C3%9Cberwachungskameras/dp/B089DXQDS9
this would directly go to the UC2 Standalone Board

C13 is this ?
image

@christiankuttke
Copy link
Collaborator Author

We need about 100W. Meanwell has neat desktop power supplies (bricks) with 90 or 120W output. But I looked around what the stock is and I found: For the 120W model, the round plug version (Meanwell product number ends with P1M) is only available at higher voltage.

Most stock is of this model with 24V : GST120A24-P1M https://www.digikey.de/de/products/detail/mean-well-usa-inc/GST120A24-P1M/7703595

Second is this model with 20V (comparable to laptop power supplies): GST120A20-P1M https://www.digikey.de/de/products/detail/mean-well-usa-inc/GST120A20-P1M/7703593

Which means that the round plug is probably not suitable to carry above 6 amps.

The connector used for 12V 120W output is power DIN 4-pin locking a.k.a. KYCON KPPX-4P with per-pin capacity of 7.5A but run as a pair of 2 for each 12V and GND. The "shield" of the power DIN connector also passes through "Frame Ground", so probably Protective Earth of the power input. I assume we could connect metal parts of our system to that to "Ground" them for noise mitigation/ESD protection.

image
from datasheet of Meanwell GSM120A12-R7B

With that plug, there is Meanwell GSM series (medical vs. GST industrial) and there is this 120W 12V product with OK stock numbers: GSM120A12-R7B https://www.digikey.de/de/products/detail/mean-well-usa-inc/GSM120A12-R7B/7703447

The connector "locks" by virtue of being difficult to pull apart. I think that would be an OK connector on the outside of our box.

@beniroquai
Copy link
Contributor

beniroquai commented Jul 5, 2024 via email

@christiankuttke
Copy link
Collaborator Author

To answer how I came to this value, please see the list above.

To go from here, we could either spend 45€ on the 100W power supply and have that part of the box overengineered and overspecced, or we could spend 20€ on a no-name power supply with less wattage and then deal with errors or data loss in the case that everything does run at one at one unfortunate point in time and pull over the designed amount of power.

The 100W power supply would eliminate power as a error source and allows us to build in quick-fixes like fans or other hard drive types/more drives without having to worry about power again at that point.

@beniroquai
Copy link
Contributor

I would go for meanwell. I'm just scared about the 100W heat dissipation we may face in the box :D
Why should we go for DIN vs the ordinary round 12V jack we have in the UC2 board? I remember you mentioned that but I didn't get it ;)

@christiankuttke
Copy link
Collaborator Author

The Meanwell doesn't exist with 100W 12V because the amp rating of the barrel connector is too low. The amp rating of the PowerDIN connector is higher.

More "elegant" from a consumer standpoint would be USB C, but there is no meanwell power supply and that'll have to be 20V or higher voltage to comply with the amp rating of the USB C plug/jack.

12V input can directly be used by at least the motors. It is high-ish with the advantage of fairly low current for a given wattage, but not too far away from 5V needed by many applicances (the computer and its peripherals).

To make 5V, DC-DC converters have to make it from 12V

With the other parts design-locked, i spec the new power budget:

  • 5V 5A (20W) for the Raspberry Pi, covering:
    • Processor
    • Screen and touch
    • NVMe HAT with SSD
    • Portable HDD plugged in to offload data
    • GPS module
  • 12V for the uc32_standalone board with ESP32 and 2 motor drivers
    • a big/strong one for peristaltics - 2.8A peak 1.7A RMS, about 40W worst case
    • a small one for focus - maybe 1.5A max, 18W
    • Microcontroller, 3W
    • LED illumination

Rounded up that's in the range of an 80W power supply. This one has the barrel connector.

https://www.digikey.de/de/products/detail/mean-well-usa-inc/GST90A12-P1M/7703717

image

@beniroquai
Copy link
Contributor

Sounds great. Let's go for the GST90A12-P1M.
Power management (i.e. 12V -> 5V for the Raspi) is handled inside the box I guess, correct? Will there be an additional DC-DC converter like the one I gave you? Should be an off-the-shelf part I guess?

@beniroquai
Copy link
Contributor

Hey @christiankuttke Now we have on12V input (Meanwell 5A@12V) that gets converted to 5A for the raspi (2x 3A@5V) DC-DC converter. Are there alternatives to e.g. provide 12V and 5V within one power supply? Would that make sense? Still, we are not getting all the power we need. Raspi is complaining and SSD is not running

@christiankuttke
Copy link
Collaborator Author

@beniroquai We can't find the problem if we do not measure now.

Points where to measure

  • A: on 12V bus, at input to box
  • B: On 5V bus, at input to Raspberry Pi GPIO header
  • C: On USB bus of Raspberry Pi (Voltage or simulated load only)

Scenarios that can be confirmed by measuring

  • The 12V->5V buck converters don't have enough amperage. Then we would see more than the rated current at point B and voltage drop at peaks of high CPU load or USB device insertion at points B and C. No voltage dip but sustainable current fluctuations, within the rated amperage of the AC-DC psu at point A.
  • The Raspberry limits the current passed on to USB devices, for example to 600 mA total. Then there is stable voltage and current below the buck converter's rating at point B, but brownout voltgage at point C that recovers when USB devices are gradually unplugged. A line in the raspi-config.txt would fix that, but it is not officially documented, only on stackoverflow in an old answer.
  • The 2 buck converters work against each other. Then there would be a big power (watts) loss at input vs output of the buck converters, so at points A (with other loads removed) vs. B, for example with all USB devices pulled out.

We won't get around measuring and systematically approaching this to fix the power problems. We can supply 5V 5A with my bench power supply for testing. (at point B or C)

@beniroquai
Copy link
Contributor

Agree! Let's measure that on Friday. I ordered another 10A 5V DCDC converter to get more juice for the Pi.

On the USB3 side.. (the pi has 2 of them, blue). Does it have a 600mA limit, still? Was there a software setting to get more power on the lines?

@christiankuttke
Copy link
Collaborator Author

Another idea about power: It could be that the routing of the 5V trace is not good and has too much voltage drop if everything has to go through the Raspberry Pi 5 power management chip. The chip is positioned close to the USB-C power input, but far away from the headers. If USB-C PD power supplies work but not 5V input over the headers, even when it has enough amps, then there could be too much voltage loss on the way from the headers to the PMIC, then back to the USB ports.

@beniroquai
Copy link
Contributor

beniroquai commented Aug 14, 2024 via email

@christiankuttke
Copy link
Collaborator Author

Re AC-DC power supply:

We used a Meanwell OWA-60E-12 power supply in the end. It has fixed Schuko and standard ID 2.1 x OD 5.5 x len 11 barrel jack with center positive. It has 5A output at 12V rated for 60W, and ramps voltage down to limit current when too much current is drawn (CC mode).

Supply: https://www.voelkner.de/products/902519/Mean-Well-OWA-60E-12-Tischnetzteil-Festspannung-12-V-DC-5A-60W.html

Datasheet: https://asset.re-in.de/add/160267/c1/-/en/001439263DS01/DA_Mean-Well-OWA-60E-12-Tischnetzteil-Festspannung-12-V-DC-5A-60W.pdf

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Assignees
No one assigned
Labels
None yet
Projects
None yet
Milestone
No milestone
Development

No branches or pull requests
2 participants
@beniroquai @christiankuttke