Figure out a SOM solution #9

Open
opened 2024-02-22 00:44:27 +01:00 by kreyren · 0 comments
Owner

To follow-up on tsvetan's original idea of teres having a replacible mainboards to change the main chips with the intention to make teres multi-platform..

Tsvetan's original idea is not sustainable and not flexible enough as evident by the lack of feature in the upstream so for teres-2 it's core idea is to implement a System-On-Module solution as described in the OSHW-System-On-Module project[1].

They tried to implement the solution as a part of the SOM204 project, but the implementation needs to be reworked for practical use in this scenario, but we are going to use it's components to explain the idea and the issues:

In short the problem is that fabrication of the current teres-1 board with the chip is very expensive in small quantities (~200 EUR per 5) due to the amount of layers the PCBs with the chip needs where the cost scales with the size and quantity.. So to optimize this for small quantities so that invidual users can adjust the design to their own computing without hassle we are going to make the separate chip board ("SOM") as small as economically acceptable and modular for mass-fabrication:

image
* OLIMEX's SOM204 project

which is only to have things that are impossible to implement through the connector due to the impendance of signals reaching the chips -> SOM boards are only to have the chip, DRAM and eMMC ideally with optional SPI flash for bootloader firmware and chip-specific features such as separete crypto engine, etc..

and place it into an evaluation board:

image

This then enables us to have these evaluation boards with 2~4 layers that are generally up to 30 EUR per 5 locally or self-fabricable through a CNC mill or fibre laser making it very hackable into different form-factors and environments.

In example real life usecase: You buy a SOM board with A64 for ~10 EUR and can then use an evaluation board to turn teres into a 3D printer, CNC, drone, tablet, etc.. for ~30 EUR or you can buy a different SOM module e.g. the TH1520 (riscv64, quad-core @ up to 2.5 GHz) for ~60 EUR and swap the SOM board to have riscv64

*Note that we are excluding the cost of invidual components such as the display which is elaborated in the link above.

The issue with SOM Size

With the OLIMEX's SOM204 using SO-DIMM 204 pin connector generally used for a RAM in laptops it makes the board always at a set size which is projected to complicate the design of e.g. phone form-factors as the need to integrate the SOM board into a small formfactor would result in the phone being significantly thicker than an alternative design.

One solution to this might be to fabricate a small non-SOM boards for this use, but that would make the mainboard again less economical to adjust by users.

Alternative solution might be to declare a standard for a pin array using a solder with a lower melting temperature than the components around it so that it's easy to replace with a hot air station or designing a socket to which it will be mounted.

References

  1. Kreyren's OSHW-System-On-Module project - https://git.dotya.ml/kreyren/OSHW-System-On-Module
  2. OLIMEX's SOM204 project -- https://www.olimex.com/Products/SOM204/
To follow-up on tsvetan's original idea of teres having a replacible mainboards to change the main chips with the intention to make teres multi-platform.. Tsvetan's original idea is not sustainable and not flexible enough as evident by the lack of feature in the upstream so for teres-2 it's core idea is to implement a System-On-Module solution as described in [the OSHW-System-On-Module project](https://git.dotya.ml/kreyren/OSHW-System-On-Module)[1]. They tried to implement the solution as a part of the SOM204 project, but the implementation needs to be reworked for practical use in this scenario, but we are going to use it's components to explain the idea and the issues: In short the problem is that fabrication of the current teres-1 board with the chip is very expensive in small quantities (~200 EUR per 5) due to the amount of layers the PCBs with the chip needs where the cost scales with the size and quantity.. So to optimize this for small quantities so that invidual users can adjust the design to their own computing without hassle we are going to make the separate chip board ("SOM") as small as economically acceptable and modular for mass-fabrication: ![image](/attachments/70187da6-6c1a-42a9-ad12-37e5a48865fa) _* OLIMEX's SOM204 project_ which is only to have things that are impossible to implement through the connector due to the impendance of signals reaching the chips -> SOM boards are only to have the chip, DRAM and eMMC ideally with optional SPI flash for bootloader firmware and chip-specific features such as separete crypto engine, etc.. and place it into an evaluation board: ![image](/attachments/fc50c7f9-c7bb-4ccb-bc8a-a2c5d1f45afc) This then enables us to have these evaluation boards with 2~4 layers that are generally up to 30 EUR per 5 locally or self-fabricable through a CNC mill or fibre laser making it very hackable into different form-factors and environments. In example real life usecase: You buy a SOM board with A64 for ~10 EUR and can then use an evaluation board to turn teres into a 3D printer, CNC, drone, tablet, etc.. for ~30 EUR or you can buy a different SOM module e.g. the TH1520 (riscv64, quad-core @ up to 2.5 GHz) for ~60 EUR and swap the SOM board to have riscv64 \*Note that we are excluding the cost of invidual components such as the display which is elaborated in the link above. ### The issue with SOM Size With the OLIMEX's SOM204 using SO-DIMM 204 pin connector generally used for a RAM in laptops it makes the board always at a set size which is projected to complicate the design of e.g. phone form-factors as the need to integrate the SOM board into a small formfactor would result in the phone being significantly thicker than an alternative design. One solution to this might be to fabricate a small non-SOM boards for this use, but that would make the mainboard again less economical to adjust by users. Alternative solution might be to declare a standard for a pin array using a solder with a lower melting temperature than the components around it so that it's easy to replace with a hot air station or designing a socket to which it will be mounted. ### References 1. Kreyren's OSHW-System-On-Module project - https://git.dotya.ml/kreyren/OSHW-System-On-Module 2. OLIMEX's SOM204 project -- https://www.olimex.com/Products/SOM204/
107 KiB
329 KiB
kreyren added this to the Teres-2 milestone 2024-02-22 00:48:52 +01:00
kreyren changed title from Figure out a SOM solution (teres-2) to Figure out a SOM solution 2024-02-22 13:38:14 +01:00
kreyren pinned this 2024-05-11 17:44:18 +02:00
Sign in to join this conversation.
No Label
No Milestone
No Assignees
1 Participants
Notifications
Due Date
The due date is invalid or out of range. Please use the format 'yyyy-mm-dd'.

No due date set.

Reference: KREYLIMEX/TERES#9
No description provided.