The 5890 is the next-generation successor to the 5851 and 5871. It is based on the M289xx chipset (M28945 + M28927) and supports all 3 of G.shdsl, SDSL/2B1Q and IDSL on the same box. At first it looks very similar except for the case being white instead of black, but even just a look at the rear panel shows that it's really different.
We have two 5890 units from eBay for experimentation and reverse engineering, and having finally found a power supply (see below) we've been able to play with them and make some observations.
The firmware on the 5890 has very much the same look and feel as on the older 5851/5871. The only really obvious difference is that whereas the 5851 needed to have different firmware images programmed into flash to support different SDSL flavors, the 5890 can be switched between SHDSL, SDSL and IDSL and between all supported flavors thereof with fairly straightforward configuration commands.
There are two submodels of the 5890 that we know of: 5890-004 and 5890-005. However, it appears that the only difference is in the default configuration profile: -004 defaults to SHDSL and -005 defaults to Nokia SDSL. Both can be reconfigured to any of the supported flavors though.
Whereas 5851, 5861 and 5871 took AC mains power directly and used an internal open frame power supply, 5890 takes 12 VDC power input. The almost-rectangular PCB takes up the entire room inside the enclosure; no more cutout for the power supply.
Whereas the previous routers used a mixture of 5V and 3.3V logic, this one
appears to be an all 3.3V design.
I would not be surprised if there is no +5V supply on the board at all.
All principal components use 3.3V I/O and even lower core voltages, and the
M28927 AFE also needs analog +12V
, probably for the line driver.
It thus appears that the 12 VDC input powers the AFE and 3.3V is made
from it with a switching DC-DC converter, powering everything else.
Core voltages are probably made from 3.3V with linear regulators.
(I'm going to show the board to my dad who is a power supply expert for
a verification of my theory.)
The 12 VDC power input connector is the most odd-looking one I've ever seen, totally non-standard. This oddity makes it incredibly difficult to find a suitable power supply.
Ethernet has changed from 10BaseT to 10/100. The Fast Ethernet switch chip used is Marvell 88E6060.
The rear panel features a block of 4 Ethernet jacks and a 5th standalone one. The latter looks like an uplink port to an uninitiated eye, but the reality is a little more complex. The Marvell switch chip does auto-crossover on all ports, so a separate uplink port is unnecessary. However, it is also an intelligent switch with VLAN functionality, so the CPU can see which port the traffic is coming from and can direct packets to specific ports. Siemens' firmware uses the 5th standalone Ethernet port as the DMZ interface.
The CPU has changed from MPC850DE to MPC859T. The latter is based on the newer MPC866 core.
The DRAM and flash capacity has been doubled to 8 MB and 4 MB, respectively. The flash chip is a 28F320, still a soldered TSOP.
The board supports two-pair configurations. There are footprints for 2 of everything: 2 M28945s, 2 M28927 AFEs, 2 transformers, 2 of all other line interface components. Only one of each is populated on the 5890. There is only one WAN jack though, so the 2nd pair must be on some other pins of the same jack.
The populated BGA CPU lives at reference designator U1, but right next to it is an unpopulated BGA footprint with reference designator U2. Is this for supporting two different CPU types, or for a dual-processor configuration? Or is that unpopulated footprint for something else altogether?
There is an Altera EP1C3T144C8 FPGA on the board. (Again no separate configuration device, so it must be in the main firmware.) Given that the M289xx chipset has some warts in its support for SDSL flavors, perhaps they are using the FPGA to augment it. Or it could be for some unrelated frills.