Project

General

Profile

Actions

Feature #5624

closed

Bug #5542: Move hub to datacenter colocation

design / prototyping / testing of icE1usb RS485/422 daughterboard

Added by laforge over 1 year ago. Updated over 1 year ago.

Status:
Resolved
Priority:
Normal
Assignee:
Category:
hardware
Target version:
-
Start date:
07/22/2022
Due date:
% Done:

100%


Description

The idea is to have a board with 3x RS422/485 drivers/receivers.

For the "OCTOI hub" situation, we'd need
  • 1x Rx NMEA
  • 1x Rx 1PPS
  • optionally: 1x Tx NMEA (no use case for it right now)
For the "DUG/DUL GPS source" use case, we'd need
  • 1x Tx NMEA
  • 1x TX 1PPS
  • optionally: 1x Rx NMEA (not 100% certain it's needed, but recommended)

The ISL83485 (switchable direction) I used in the siu-clock-breakout is sadly not available right now, but I'd hope some other driver circuit is.


Files

20220825_214236.jpg View 20220825_214236.jpg 427 KB tnt, 08/25/2022 08:23 PM
icE1usb-rs422.bin icE1usb-rs422.bin 102 KB tnt, 09/05/2022 02:35 PM
Actions #1

Updated by laforge over 1 year ago

I don't want to put any pressure on you, tnt, but I was wondering what the estimated schedule would be for this? To avoid extra travels of ~500km one way to the data centre, I'd want to piggy-back the deployment on a trip on September 19, 2022. And prior to that, the system with the GPS02 shoul be tested at my home. Is that realistic?

Would it help if I did that daughterboard hardware design?

Actions #2

Updated by tnt over 1 year ago

Yeah sorry, that actually completely slipped my mind.

The ISL83485 is also non-isolated. I'd rather use something that's completely floating and allows high potential difference since they'll be very far appart and the GPS02 will most likely be grounded/referenced at the install site on the roof.

I'll have a look at what chip I can find that's available.

Actions #3

Updated by laforge over 1 year ago

On Wed, Aug 10, 2022 at 11:48:37AM +0000, tnt wrote:

Yeah sorry, that actually completely slipped my mind.

no worries. We were waiting for answers from the colocation for a long time earlier.

The ISL83485 is also non-isolated. I'd rather use something that's completely floating and allows high potential difference since they'll be very far appart and the GPS02 will most likely be grounded/referenced at the install site on the roof.

The GND situations is as follows:

  • GPS02 SMA shield, metal enclosure and the GND line on the RJ45 are DC-connected inside GPS02
  • GPS02 enclosure is mounted on DIN rail, which is connected to Earth/PE in the cabinet on the roof
  • The 12V which we supply from the rack side is of course free floating, no PE/Earth connection there
  • The 5V/GND powering the icE40 is provided by some PC, which typically connects GND to chassis/PE in the rack

So indeed, the GND = PE on the roof and the GND = PE in the rack might have
significant voltage difference due to the 100m in between. So the differential
signals in rx vs tx direction might center around some different voltage.

Not sure what DIN regulations say about voltage difference on PE within a single
building. Having isolation is of course good.

I'll have a look at what chip I can find that's available.

Actions #4

Updated by tnt over 1 year ago

Just ordered parts to build 3 of them.
Still need to do the schematic / layout, but the mains components/design is done.

It uses separate digital isolators chip and rs422 tranceivers.
The isolated section will be powered from the 12v pins through a 5V regulator.

Actions #5

Updated by laforge over 1 year ago

On Thu, Aug 11, 2022 at 10:29:28AM +0000, tnt wrote:

Just ordered parts to build 3 of them.

thanks!

It uses separate digital isolators chip and rs422 tranceivers.
The isolated section will be powered from the 12v pins through a 5V regulator.

that's fine. So from outside the icE1usb we'd then feed the same 12V to both the
daughterboard and the GPS02.

Actions #6

Updated by tnt over 1 year ago

  • Status changed from New to In Progress
  • % Done changed from 0 to 50

Board has been designed and sent for fabrication at Aisler.
Expecting 3 prototypes PCB back, hopefully on Aug 26th. If that's the case, I can then assemble at least one and test it on that weekend.

Actions #7

Updated by laforge over 1 year ago

Thabks.Please ping me (then) regarding shipping, can send you a ups label.

Actions #8

Updated by tnt over 1 year ago

One proto has been assembled

Actions #9

Updated by tnt over 1 year ago

Damn, found the first issue, I forgot I2C pull-ups on the isolated side :/

Actions #10

Updated by tnt over 1 year ago

So the identified issues on the boards :

  • Missing pull-ups on the I2C isolated side
  • Completely wrong pinout on the RJ45
  • I2C issue on the ice1usb (see #5664)

I reworked the board to add pull-ups and then I used a RJ45-jumper-box to connect to the GPS 02 and managed to talk to the GPS, get the PPS and get the GPSDO loop using it.

Now there are 3 options :

  • Fix the PCB design and have new boards made
    • Downside is time. I'd only get the boards at best right before the weekend of Sep 10 and I'm busy all that weekend ...
    • And also cost, needing to order more components if I want to have 3 boards that are "identical"
  • Use the current PCB, just do the I2C pull-up rework and either crimp a special RJ45 cable that does the swap, or use a RJ45 jumper box
  • Use the current PCB, do the I2C pull-up and also cut all the traces from fuses to the external header and then place 6 wires to make the right connections
Actions #11

Updated by tnt over 1 year ago

  • % Done changed from 50 to 70
Actions #12

Updated by laforge over 1 year ago

Crimping a special cable is fine, it is used only inside my 1U enclosure. Thanks for all your work!

Actions #13

Updated by tnt over 1 year ago

  • % Done changed from 70 to 90

Assembled 3 boards and did some basic testing on them.
I ended up reworking the boards so that we can use standard straight RJ45 cable without having to mess with custom ones.

All that remains is to do some longer term testing.

Actions #14

Updated by laforge over 1 year ago

  • Assignee changed from tnt to laforge

assigning to me for testing, 2 boards were received by me in Berlin today. Thanks!

Actions #15

Updated by tnt over 1 year ago

Attached bitstream for testing.

Actions #16

Updated by laforge over 1 year ago

tnt wrote in #note-15:

Attached bitstream for testing.

thanks, seems to be working fine in my [brief] testing yesterday AFAICT.

Actions #17

Updated by laforge over 1 year ago

  • Status changed from In Progress to Resolved
  • % Done changed from 90 to 100

We can mark this as resolved, it seems to work fine.

Actions

Also available in: Atom PDF

Add picture from clipboard (Maximum size: 48.8 MB)