Running Power over ethernet on 10BASE-T1L Single Pair Ethernet with SPEBlox Long

– Copied from a customer email –

I like the small physical size and lower power consumption of your 10BASE-T1L boards, and most importantly, the fact that they do not emit EMI in frequencies that interfere with the sonar (unlike VDSL, which requires shielding). So far, I have gotten them to link across 650m of 26awg unshielded cable that I designed for VDSL (but has an impedance of ~100 ohms), which seems encouraging.

I’d like to see if I can get your boards to link across my cable with power coupling networks also inserted (because I need to send power + data over the same twisted pair; I’m currently targeting 110VDC). I’m wondering whether you have designed a power-over-ethernet version of your 10BASE-T1L boards, or if you have any plans to do so?

For the power coupling network, I’m planning to build and test the circuit below (with one inserted at each end of the cable). It’s based on a Wurth design that was shown in this webinar (and application note). It looks like it should work with your board (although my L1 may be redundant), but if anything jumps out at you as incompatible, I’d appreciate you letting me know.

Good to know you got them working at 650m of 26AWG. For 24AWG, we’ll have to see how that performs. 10BASE-T1L is a bit more sensitive to cabling type compared to VDSL. The reason for this is the same reason that 10BASE-T1L is less radiative in terms of EMI.

Essentially, VDSL pumps a whole load of frequencies down the line to “see what sticks”; this makes it less sensitive to cabling but can cause EMI problems. On the other hand, 10BASE-T1L uses PAM-3 encoding with a fundamental frequency of around 7.5MHz. 7.5MHz is unlikely to cause interference problems but it does tend to mean you have less flexibility in terms of cable and connector type. Generally, all SPE variants tend to be used with the twisted pair variety of cable found in CAT5/6 ethernet cables. Typically this is 24-28AWG with some specifications on twist rate, copper type etc. If you deviate too far from that, you can’t expect to see the full 1km range.

Regarding power and data on the same SPE connection, this is an area of continued development. In official parlance, it’s called SPoE (IEEE 802.3cg), and it involves a autonegotiation between the Power Sourcing Equipment (PSE) and the Powered Device (PD). There isn’t a huge amount of silicon on the market to achieve this yet (though Analog Devices have some good options), but expect this to change as the market develops.

What I believe you’re talking about is just passively injecting voltage onto the line and extracting it at the other end. There are people doing that too with mixed results. When it comes to doing that on SPEBlox Long, there’s a few things to consider.

Firstly, the AC coupling capacitors on the ethernet lines are only 100V rated. So if you put 110V on there, those capacitors will likely blow. SPoE states using 60V, so you should probably stay around that level. Also note that given SPEBlox Long’s small size, the maximum tolerated clearance on the circuitry connected to the ethernet signal tops out at around 60V. Not to say that using above 60V won’t work, it probably will, but we’re starting to get outside the realms of what the board was designed for. See below for the output structure of SPEBlox Long.

In your proposed circuit I assume you would take TRD_P and TRD_N and connect it to ETHERNET+ and ETHERNET-. In this case, the signal from L8 on SPEBlox Long would connect to your transformer T1 (along with the TVS suppression diodes), and then you’d use two other CMCs past two AC coupling capacitors to combine power and data. The issue with this is that in the circuit you sent, it’s probably designed to be interfaced directly with the pins of a PHY chip. However if you were to use this with SPEBlox Long, you’d end up with some redundant circuit paths. It’d probably be fine but you’d see an increased attentuation and therefore a reduced range. Since you only need 600m though, you might be fine, but difficult to know without testing.

I think a better solution would be to omit T1, and just use L2 to inject the voltage. Of course, the benefit of transformer T1 is that it would allow you to use 110V because it would isolate the field level ethernet from the board.

SPEBlox Long does contain some unpopulated components that would allow it to inject its power supply voltage directly onto the lines. See the image below.

You could solder inductors here, and then use 60V, and test like that with your cabling. This means you don’t need to do any custom board work and would at least allow you to get some feel for how well your system will work at 60V on your cable.

Suggested part numbers for L6 and L7 are ASMPH-0603-R33M-T, and for L4 and L5, it’s 1239AS-H-6R8M=P2. Both are large footprints and should be easy to solder, just be careful of using hot air in that area as you may burn the screw terminal.

Let me know if you have any questions.


Also @petergjones if he has any input on this.

Hi Josh, this is very helpful, thank you for investing the time, thought and effort to provide such a detailed response!

Unfortunately, I think I am stuck using a relatively high (100V+) voltage on the twisted pair, due to a combination of the cable design (it has to be long and thin = high resistance) and power consumption on the PD end. But, even the 110V I’m contemplating is a significant reduction vs my existing VDSL-based system, which has higher power requirements, and a correspondingly higher voltage (250V from the PSE).

In the Wurth webinar and application note, they show that the transformer-based design has favorable performance in terms of both return loss and mode conversion loss, but given the redundancy you mentioned with SPEBlox Long, perhaps I should also try a transformerless design that uses coupling capacitors rated for a higher voltage (to replicate the transformer’s isolation capability; Wurth also tested a circuit with 2kV coupling caps, as a benchmark for their transformer-based PoDL solution).

Thank you again!!

Welcome. If you could put the higher voltage rated capacitors in there, after SPEBlox Long, then I think you could try the 110VDC.

Feel free to post a diagram of what you intend to try and I can give it a review before you commit it to copper.