Broadband Related > Broadband Hardware

Modem power supplies and grounding

(1/3) > >>

IanG:
Hello All. [Being new, please tell me if I am posting this in the wrong section.]

While trying to track down occasional interference that causes ADSL dropouts, I got interested in the ADSL modem grounding arrangements. As far as I can see, Ethernet and DSL inputs are normally transformer coupled, while the DC power supply is also floating, having no earth pin. So the internal chassis of the modem is likely to have RF signals capacitively coupled from all those inputs, which is not ideal. Or so I thought.

Changing the 12V power supply for another (different make from a different modem) had no effect, but replacing it with a linear lab power supply gave a notable improvment. Connection rate improved from 17.6 to 18.5 Mbps; at the same time FEC errors were much reduced, and the S/N ratio improved by a few dB over almost the whole spectrum.

Using the lab power supply is not ideal. For one thing, its internal fan makes a racket. Instead, I thought of buying a class 1 switcher (i.e. with earthed output) to see if it gives the same improvement.

So I have a few questions. Can someone confirm that it is safe to connect an earthed supply to the modem? And have others tried an earthed supply, and found it to be effective? The modem under test, by the way, was the Netgear DGND3700v2.

burakkucat:
Welcome to the Kitz forum.  :)

Thinking about the internals of the average CPE, there is no real, classical, ground-plane that could be earthed. Obviously there is a logical signal "chassis" . . . if there are header pins for a serial console, then one of them will be the signal ground.

The xDSL connection will be via capacitors, one in each leg of the pair. Essentially it is an RF feeder, capacitance coupled to the analogue front-end.

As for the LAN Ethernet connections, I am not to sure how the two (or four) pairs and the circuitry are coupled together.

Finally a switch-mode power supply will have no earthing and would be doubly insulated.

I am not surprised that a change of power supply shows a difference in the modem/router's performance. Most switch-mode PSUs are inherently noisy and some of the "budget class", Chinese, made devices are really poor quality. The best test would be to arrange a temporary DC supply . . . if the CPE requires 12V DC, then a car battery (or a power tool battery pack) would be ideal.

sevenlayermuddle:
My inclination would be to not interfere with the manufacturer’s earthing/grounding arrangements.

My reasoning is... The ‘earth’ conducter in home wiring is provided not as a “noise reduction wire” but as a safety feature to ensure that in the event of a fault, a fuse will do its job, melting and breaking the circuit, before the house catches fire.   Or (hopefully), breaking the circuit before an occupant touches a piece of exposed metalwork that has become accidentally connected to mains potential.

Worth mentioning too that for many domestic supplies, the earth conductor is directly connected (by the supplier) to the neutral conductor.   With that in mind, it maybe becomes more intuitive that (1) the earth conductor may actually be electrically quite noisy, and (2) messing around with earthing arrangements may have unintended and dangerous consequences.

Many modern home appliances simply do not need an earth connection, as they have other safety arrangements.   For example, my powerful and sensitive home AV amplifier has no earth, and works perfectly, with no susceptibility to interference.   In contrast, my old amplifier from 1970s does have its chassis connected to earth, which renders it susceptible to all sorts of “ground loop” interference problems when connected to other equipment, unless strict rules are followed.

...just my take.   But whilst I have a scroll of paper somewhere that proclaims I once knew a thing or two about electrical engineering, I am no expert these days, happy to be set right. ;)

IanG:
Many thanks for all for your responses. You are right to be cautious.

To hopefully understand the hardware better, I made some measurements on two unpowered modems, one Netgear and one TP-Link. The attached diagram from an Analog Devices ADSL design guide is consistent with the measurements. On both modems, the input coupling capacitors (0.1 uF in the diagram) measured as 68 nF. I think they are present to stop the transformer shorting out the POTS signal at low frequencies. Also on both modems, the transformer inter-winding capacitance (measured from the DSL input to the modem 0V line)  was about 39 pF, which makes me happier about the modem surviving lightning surges. At a distance, of course.

Ethernet inputs similarly had no direct connection to 0V, but couple via a transformer. Input capacitance to 0V was 70 pF for Netgear, and 1 nF for TP-Link. With that measured, I felt happier that what I was proposing would not be unsafe, although it might well be pointless.

I completely take the point that mains earth is by no means an RF ground, and that all lines including the DSL common mode input are likely to have spurious RF on them. My concern is not with ripple voltage from the switching power supply, but with currents capacitively coupled from its mains voltage switching circuit to its output, which an earth return should loop back to the input, assuming the mains input has class Y capacitors to the earth line. Or so I hope. Anyway, I have ordered a class 1 switching supply, and will post the result here if it proves to be effective.

burakkucat:

--- Quote from: IanG on June 02, 2019, 10:46:13 PM ---On both modems, the input coupling capacitors (0.1 uF in the diagram) measured as 68 nF. I think they are present to stop the transformer shorting out the POTS signal at low frequencies.

--- End quote ---

The capacitors are the high-pass filters . . . passing the RF of the xDSL signal and blocking both the voice-band POTS at baseband frequency & the POTS DC component.

It will be interesting to see the results of your experiments. So please do post an update, in due course.

Navigation

[0] Message Index

[#] Next page