There's is a another version of the i5 V7.0 module, which features different components and a changed pin mapping compared to the one described in this repository. It is probably a newer version of the i5 (maybe a V7.1 or V8.0, but wrongly labeled ?) and can be seen on the top here (below is the 'regular' i5 V7.0 module):
Notable differences encountered so far:
- Both modules carry the
V7.0 label, but only the top module has actually written i5 on it
- The top module has two more mounting wholes in the corners (the right corner hole is the same size & location as on the i9 module)
- The top module is missing the 3 TI multiplexers CD4051 (chip marking CM051B, one is on the back side)
- The top module features the smaller 256-pin ECP5 BGA package (LFE5U-25F-6BG 256 C)
=> the bottom, 'regular' module uses the a larger 381-pin BGA (LFE5U-25F-6BG 381 C)
=> both got the same number of I/Os, the smaller chip has fewer VCC & GND pads and of course a different pin mapping scheme!
- The top module features a different SDRAM chip: M12L64322A-6B
=> it is compatible with the other one below, EM638325-6H
- the top module features a different FLASH chip: Winbond 25Q32
=> it has the same (Q)SPI interface but features twice the size compared to the lower one (GD25Q16)
Since the main chip differences are hidden below the heat sink plate, it took me a while to understand that I am assuming the wrong pin mapping for this newer i5 module due to the different ECP5 package!
Surprisingly, a lot of things like the Ethernet phy 0 work just as usual when using the regular litex-boards i5 platform pin map, which configures the bitstream for a 'regular' i5 V7.0 module target: LFE5U-25F-6BG 381 C.
I am assuming most physical BGA pin identifiers that exist for both chips have been routed on both PCB modules to the same connections on purpose. However, since the smaller 256-BGA only features pin coordinates from A-T and 1-16, certain pin identifiers just don't exists on the top module.
That also explains sneaky problems I discovered using LiteX's regular colorlight_i5 platform pin mapping (the one also documented in this repo):
- the SDRAM cannot be properly accessed,
only every second word = 32-bits (most likely due to A17/B17 pins being replaced) (seems to be more complicated, see below)
the FLASH chip cannot be accessed from the ECP5 (most likely due to W2/V2 pins being replaced)
= > flashing itself via the on-board CMSISDAP JTAG adapter works just fine- several pmod/pin header I/Os cannot be reached (certainly pin ids above >T,>16, didn't yet try the serial pin finder approach)
Does anyone have already the correct pin map for this presumably newer i5 V7.0 module featuring a LFE5U-25F-6BG 256 C?
The pictures on Alixexpress always show the bottom i5 V7.0 module. Another store, selling the top module is here.
Interestingly, this webpage of a Colorlight partner features a very similar looking module that's differently labeled: i5**_V8.0**. A few minor layout differences can be spotted around the heat sink.
I've received 3 of the top modules shown in the image here, purchasing from Muse Labs' Aliexpress store, in the beginning of this year.
Let me reference a few people who contributed colorlight module stuff to Github:
@chmousset, @kazkojima, @DaveBerkeley, @Peter-van-Tol, @chenzhuoyu, @kittennbfive, @tomverbeure, @zyp.
There's is a another version of the
i5 V7.0module, which features different components and a changed pin mapping compared to the one described in this repository. It is probably a newer version of the i5 (maybe a V7.1 or V8.0, but wrongly labeled ?) and can be seen on the top here (below is the 'regular'i5 V7.0module):Notable differences encountered so far:
V7.0label, but only the top module has actually writteni5on it=> the bottom, 'regular' module uses the a larger 381-pin BGA (LFE5U-25F-6BG 381 C)
=> both got the same number of I/Os, the smaller chip has fewer VCC & GND pads and of course a different pin mapping scheme!
=> it is compatible with the other one below, EM638325-6H
=> it has the same (Q)SPI interface but features twice the size compared to the lower one (GD25Q16)
Since the main chip differences are hidden below the heat sink plate, it took me a while to understand that I am assuming the wrong pin mapping for this newer i5 module due to the different ECP5 package!
Surprisingly, a lot of things like the Ethernet phy 0 work just as usual when using the regular litex-boards i5 platform pin map, which configures the bitstream for a 'regular'
i5 V7.0module target: LFE5U-25F-6BG 381 C.I am assuming most physical BGA pin identifiers that exist for both chips have been routed on both PCB modules to the same connections on purpose. However, since the smaller 256-BGA only features pin coordinates from A-T and 1-16, certain pin identifiers just don't exists on the top module.
That also explains sneaky problems I discovered using LiteX's regular colorlight_i5 platform pin mapping (the one also documented in this repo):
only every second word = 32-bits (most likely due to A17/B17 pins being replaced)(seems to be more complicated, see below)the FLASH chip cannot be accessed from the ECP5 (most likely due to W2/V2 pins being replaced)= > flashing itself via the on-board CMSISDAP JTAG adapter works just fine
Does anyone have already the correct pin map for this presumably newer i5 V7.0 module featuring a LFE5U-25F-6BG 256 C?
The pictures on Alixexpress always show the bottom i5 V7.0 module. Another store, selling the top module is here.
Interestingly, this webpage of a Colorlight partner features a very similar looking module that's differently labeled: i5**_V8.0**. A few minor layout differences can be spotted around the heat sink.
I've received 3 of the top modules shown in the image here, purchasing from Muse Labs' Aliexpress store, in the beginning of this year.
Let me reference a few people who contributed colorlight module stuff to Github:
@chmousset, @kazkojima, @DaveBerkeley, @Peter-van-Tol, @chenzhuoyu, @kittennbfive, @tomverbeure, @zyp.