Copied from a customer email
just a quick question about power supply of the UbiSwitch :
Can you confirm the UbiSwitch can run on 5V (5.5V) ?
I will use a DC/DC convertor (5V 6A) and I will trim it +10% to 5.5V so all other voltage regulators on my board are running happy.
UbiSwitch can run at 5V, but it’s suggested to run it a bit above that to account for any voltage drop due to resistive power wires. 5.5V will be just fine if your wires are short (less than 30cm) and more than, say, 24AWG
Copied from a customer email
It could be just a typo (5 / 8).
If it would be 8V I need to change my power supply.
You’re absolutely right to check.
In some ways, it wasn’t really a typo. We know the boards work at 5V but we’re not sure we necessarily want all customers using it at 5V. Some customers may use a bad 5V regulator without enough headroom, and since the board runs at 3.3V, there’s a risk of power issues.
It sounds like you know what you’re doing with a 5V 6A power supply (certainly looks like you know what you’re doing from your circuit design skills), so happy for you to use 5.5V
Copied from a customer email
I have the possibility to change the DC convertor to 12V or maybe 9V in the same footprint, but then I have other issues with thermal dissipation in other linear regulators.
Another question about length matching of the diff-pairs.
Since I don’t have information on the length of de diff-pairs on the UbiSwitch Module, and I am currently working with length matching to the board connectors, but is this the right way ?
If I look at the image of the Ubiswtich it seems the 4 pairs of a port are not length-matched.
Copied from a customer email
It says :
7.2 Recommended skew
10G USXGMII
â—Ź Keep the intra-pair skew to less than 0.1mm (between a + and - of a single pair)
â—Ź Keep the inter-pair skew to less than 0.5mm (between a TX and RX in a signal USXGMII port)
I understand the intra-pair skew is important to match, but I don’t see any reason why inter-pair skew for TX and RX signals is important.
The physical path (optical) is kind of separated for RX and TX and there is no timing relation between RX and TX signals. Correct me if I am wrong.
This is different for GigabitEthernet signals where 4 pairs are used together to send and receive data symbols.
Can you elaborate on this ?
5.5V at 6A should be fine. Just keep your power rails fat and short so that you don’t lose much voltage on the way to UbiSwitch…
To put numbers to it, if you’re running at 5V on UbiSwitch, and UbiSwitch draws 6W of power, that’s 1.2A of current draw.
Now, if we assume that we are supply 5.5V right at the output of your regulator, we’ll want to ensure that after the tracks on your board, we give UbiSwitch around 5.3V or so, right at the pins of the module. That means we must not drop more than 0.2V on your copper traces.
If the current draw is 1.2A, and we can’t lose more than 0.2V, then the resistance of your traces must be less than 0.16 Ohms.
Using a tool like this, we can see that a 2mm thick, 100mm long 0.5oz-in copper trace would have a resistance of around 0.0486Ohms. In other words, that’s not a very thick power trace and would still work, so there’s decent margin for you here.
1 Like
You’re completely correct of course. There isn’t any real reason we can think of that the RX and TX lines ought to be matched to within that kind of tolerance.
However, in the absence of finding an official specification on the USXGMII that specifies exactly the maximum amount of inter-pair skew, we prefer to err on the side of caution. Certainly on all our boards, we end up matching interpairs to within around 0.5mm, but that’s usually a happy co-incidence. I’m fairly sure you could increase this skew quite a bit without any issues, but personally, I’d rather not risk it.
Regarding the lengths of the tracks on UbiSwitch, you don’t really need to worry about those specific lengths. Just ensure that the lengths of your tracks on your board up to the UbiSwitch are matched up until the point the hit the pins of the UbiSwitch module, and you should be just fine.