Beagle Bone & KiWi grounding and ESD protection
Hi All,
I have posted a similar message (without some KiWi specific technical descriptions) on the Beagle Bone Green forum in order to see what the 'official' line is.
I have several Beaglebone Green's which are used in conjunction with the KiWi SDR cape.
I recently had a problem where the Ethernet ports failed on two BBG's.
I replaced the PHY chips on the BBG's and both boards are now working again.
However I've been installing the BBG's and KiwI's into metal cases and have realised that the grounding is split into:-
1. The Digital ground (DC supply 0v)
2. A chassis ground (Ethernet, USB port, and BBG fixing holes)
3. An ESD protection 'ring' around the Ethernet port, which is connected to the chassis ground by means of a small (0805) surface mount resistor (R136) of value 0.1R, which is mounted near the edge of the board under the Ethernet socket.
The problem is that when used with a KiWi, which may be connected to an antenna with an external ground system, it is possible for the various grounds to end up at different potentials.
For example my metal case is mounted by means of the fixing holes in the BBG which are at chassis ground.
The DC input is connected to the digital ground which may in turn be connected to my AC supply ground, if the DC power supply has an internal connection between the two. It also connects to the RF connector screens on the KiWi board.
The Ethernet port has a separate ESD protection ring which is used to suppress voltage spikes in order to prevent damage to the PHY chip.
In my case I have discovered that the 0.1R surface mount resistor which connects the chassis ground to the Ethernet ESD protection ground has blown open circuit (presumably due to a very large voltage spike) and as a result there is no longer any EDS protection on the Ethernet port.
Note that the DC input on the KiWi board has a common mode choke, so there is actually a further difference in RF potential between the Ground of the DC connector and the Digital ground on the Beagle bone and KiWi boards. The effectiveness of this choke will varr depending upon if the DC supply has a ground connection between the DC 0v and the AC mains ground (which in turn will depend upon the electrical path between the AC ground and your antenna ground which connects to the KiWi RF connector screens). It may also connect via the Ethernet port cable screen to the chassis ground and in turn to the Ethernet port ESD ground by means of the 0.1R resistor on the Beaglebone board (if it has not already gone open circuit)
I suspect this is why the PHY chips have blown.
I've been looking at the BBG schematics but I've been finding it difficult to trace any decoupling capacitors between the Digital ground and Chassis ground or a diagram showing the component layouts on the BBG (apart from an Orcad file which I can't open).
Has anyone else had similar problems ?
Is there an image of the PCB layouts showing component designations and positions ?
I have tested to see if the KiWi noise floor is affected but it all seems OK, in fact it may be slightly better as the chassis, digital and ESD grounds don't seem to have any RF decoupling capacitors strapped between them.
So I'm thinking of linking the Chassis, Digital and ESD grounds together, in order to avoid future problems - can anyone see any issue in doing this ?
Regards,
Martin - G8JNJ
I have posted a similar message (without some KiWi specific technical descriptions) on the Beagle Bone Green forum in order to see what the 'official' line is.
I have several Beaglebone Green's which are used in conjunction with the KiWi SDR cape.
I recently had a problem where the Ethernet ports failed on two BBG's.
I replaced the PHY chips on the BBG's and both boards are now working again.
However I've been installing the BBG's and KiwI's into metal cases and have realised that the grounding is split into:-
1. The Digital ground (DC supply 0v)
2. A chassis ground (Ethernet, USB port, and BBG fixing holes)
3. An ESD protection 'ring' around the Ethernet port, which is connected to the chassis ground by means of a small (0805) surface mount resistor (R136) of value 0.1R, which is mounted near the edge of the board under the Ethernet socket.
The problem is that when used with a KiWi, which may be connected to an antenna with an external ground system, it is possible for the various grounds to end up at different potentials.
For example my metal case is mounted by means of the fixing holes in the BBG which are at chassis ground.
The DC input is connected to the digital ground which may in turn be connected to my AC supply ground, if the DC power supply has an internal connection between the two. It also connects to the RF connector screens on the KiWi board.
The Ethernet port has a separate ESD protection ring which is used to suppress voltage spikes in order to prevent damage to the PHY chip.
In my case I have discovered that the 0.1R surface mount resistor which connects the chassis ground to the Ethernet ESD protection ground has blown open circuit (presumably due to a very large voltage spike) and as a result there is no longer any EDS protection on the Ethernet port.
Note that the DC input on the KiWi board has a common mode choke, so there is actually a further difference in RF potential between the Ground of the DC connector and the Digital ground on the Beagle bone and KiWi boards. The effectiveness of this choke will varr depending upon if the DC supply has a ground connection between the DC 0v and the AC mains ground (which in turn will depend upon the electrical path between the AC ground and your antenna ground which connects to the KiWi RF connector screens). It may also connect via the Ethernet port cable screen to the chassis ground and in turn to the Ethernet port ESD ground by means of the 0.1R resistor on the Beaglebone board (if it has not already gone open circuit)
I suspect this is why the PHY chips have blown.
I've been looking at the BBG schematics but I've been finding it difficult to trace any decoupling capacitors between the Digital ground and Chassis ground or a diagram showing the component layouts on the BBG (apart from an Orcad file which I can't open).
Has anyone else had similar problems ?
Is there an image of the PCB layouts showing component designations and positions ?
I have tested to see if the KiWi noise floor is affected but it all seems OK, in fact it may be slightly better as the chassis, digital and ESD grounds don't seem to have any RF decoupling capacitors strapped between them.
So I'm thinking of linking the Chassis, Digital and ESD grounds together, in order to avoid future problems - can anyone see any issue in doing this ?
Regards,
Martin - G8JNJ
Comments
R136 is the 0.1 Ohm resistor that provides the Ethernet ESD protection, which had gone open circuit on my board.
I changed R136 for a Zero Ohm resistor and added an additional Zero Ohm resistor between the Digital Ground (pin 1&2 on P9) and the chassis ground, which is around the adjacent board mounting hole and has a hex support pillar bolted to it when it is used with the KiWi metal case.
I used Zero Ohm resistors just in case there was ever a high current surge, as I thought it would be better for the resistors to blow rather than the PCB tracks.
Regards,
Martin - G8JNJ
It would seem to be a good practice to have all grounds at the same potential or as close to it as practical. The ESD ground to Ethernet Screen connection should be close to 0 ohms, I should think. I do not know why having the digital ground above or below the chassis or common ground would have a benefit. I've noted that the LZ1AQ amp feedline instructions describes a shielded 4 pair cable with the shield grounded at the receiver but not at the amplifier. I can see the chokes on the DC ground would raise the potential on the amp ground above the potential on the cable shield. So perhaps for common mode isolation there is a purpose in grounding at a common node and not at the equipment. Schemes of this type are beyond my understanding at this point. I'll watch how this discussion gets answered.
Ron
KA7U
>It would seem to be a good practice to have all grounds at the same potential or as close to it as practical.
>
Hi Ron,
That's my thinking too. I'm sure the Beagle Bone designers has their reasons, but when used in conjunction with the KiWi it's not a good combination and it becomes worse if you are using the SEED metal case.
Ethernet and USB ports (which are at chassis ground) poke through the case and may, or may not, make electrical contact with the case (which is also connected to chassis ground via the mounting pillars), depending upon the clearance between the connector shells and the case itself. This is not too bad, as at least the case and connector shells are connected together by various routes.
However the GPS and main antenna connectors also pass through the metal case but also may, or may not, make contact with each other depending upon the clearances.
I measured approximately 3.5K Ohm resistance and 1nF capacitance between the digital ground and the chassis ground. So they are not at the same RF potential (especially on lower frequencies).
So you can have the external antennas connected to an external ground and the digital ground and the case connected to a local (mains supply) ground (via the Ethernet or USB cable screen and external device such as a PC or Ethernet router) and also the DC feed connected to the external local (mains supply) ground on one side of the KiWi PCB mounted common mode input choke and the secondary side of the common mode input choke connected to the digital ground which may be connected to the external antenna ground.
Depending upon how the various cables are routed, where they connect to various external grounds and if the metal case is or isn't touching the RF and Ethernet connector shells, then you may end up with more (or less) unwanted interference.
I'd suggest that folks check the DC continuity of R136 to protect against ESD spikes on the Ethernet port and also between the USB or Ethernet port metal screen and the RF connectors or metal screening cans on the KiWi board.
I suspect this is could be why some folks have more interference problems than others (and variation between Beagle / KiWi boards), and also why the Ethernet port is prone to failure.
I think the best option is probably to connect the Chassis, Digital and ESD grounds together on the KiWi boards, so that they are all at the same potential, both from an RF interference perspective, and more importantly to help ensure better electrical safety. This could also help prevent damage when the various cables are plugged and unplugged from the boards.
Regards,
Martin - G8JNJ
It's also worth noting that the SEEED metal case does not have a direct connection between the antenna ground, or any ground, that may or may not, be provided by the DC input cable.
So the metal case is effectively 'floating' at whatever potential it likes, relative to the other ground connections.
This is definitely not good practice and should be avoided.
Regards,
Martin - G8JNJ
I first read this as implying there was no connection between the antenna ground (SMA outer threads or green header block GND terminal) and the Kiwi, Beagle or enclosure grounds. Just to be clear, that notion is false. They are all grounded together. It is true that the ground of the DC input passes through the CMC which has a significant resistance from a grounding/ESD perspective.
What the Kiwi doesn't have is a ground loop that would occur if the outer RF SMA threads were attached to the metal enclosure. That's why there is that oversized cutout surrounding the SMAs. I actually measured a small increase in the noise floor during enclosure design with the RF SMA connected to the enclosure.
The question of establishing a proper system/ESD ground is very valid. Most of this has to be solved external to the Kiwi. Probably a stiff earth ground wire should be attached to one of the screws of the enclosure and a power supply and antenna connection used that does not provide a DC path to earth ground. It important to note the Ethernet RJ45 is transformer coupled for ground loop / safety considerations. But voltage spikes/pulses can still get through hence the ESD precautions.
The BBG schematic and RJ45 schematic are attached. The scrubbed values on the RJ45 are 75 ohms and 1000 pF.
Attachments:
https://forum.kiwisdr.com/uploads/Uploader/a2/b5b5300183580af4edd16c878e2634.pdf
https://forum.kiwisdr.com/uploads/Uploader/76/de0720ac0e52039fb87b0b80e5449e.gif
>
>I first read this as implying there was no connection between the antenna ground (SMA outer threads or green header block GND terminal) >and the Kiwi, Beagle or enclosure grounds. Just to be clear, that notion is false.
>
Sorry I tried to make this clear, but have failed.
You are correct the SMA outer threads are connected to the KiWi and Beagle 0v rail, which is referred to as digital ground.
The chassis ground consisting of the metal parts of the Ethernet and USB connectors and the holes used for the Beagle / KiWi metal case are not directly connected to the KiWi and Beagle 0v rail which is referred to as digital ground, other than via approx 3.5K Ohm and 1nF.
The ESD protection on the Ethernet port is connected to the chassis ground via a 0.1 Ohm resistor. This means that it does not have a ground connection via the 0v DC supply cable (via the on board common mode choke) if that is grounded, or any other path. So if there is an ESD spike on the Ethernet port it doesn't really have a path to ground, unless one is provided by means of an external wire link.
In my case I had provided an external connection to ground via the DC supply cable, but a voltage spike has at some point blown the 0.1 Ohm resistor open circuit leaving the ESD protection disabled, which in turn has resulted in failure of the Ethernet port when another spike has occurred.
In practice it's very difficult to separate all these grounds from each other, so I'd suggest that it would probably be better to connect them all together so that they remain at the same potential in the event of any of the cables experiencing a voltage surge.
This is a particular problem faced by radio amateurs who may have an AC supply ground provided by their electricity distribution company and a further 'local' ground by virtue of the antenna system in use.
In such cases it is generally best practice to bond the two together at some point wit ha suitably sized conductor, so that in the event of a fault the two grounds remain at the same potential.
I just wished to make folks aware of this grounding arrangement as I think it could affect the techniques used in order to minimise unwanted interference and help better protect the KiWi from ESD exposure due to nearby lightning strikes etc. (which is another problem when an external antenna is being used).
Regards,
Martin - G8JNJ
Without any great schematic checking I grounded my Kiwi through the SMA threads on the assumption that while testing noise here without a metal case I had grounded that part, with the most reduction in noise pick up (everything else seemed to survive). The heavy earth strap goes direct to the plate holding the SMA connectors.
I used non conductive mountings and even insulated the USB connector from the (tight) case to avoid ground loops.
As I use fibre, or wireless, for the network side that connection is not grounded but the total copper is about 1m so sould be hard to induce much current.
I noted the instructions about insulating the post by the flash socket as a track can be shorted so I figured insulate the lot (except SMA) then add in grounding later, I just never got round to it as the QRM didn't invite investing much time fine tuning.
I'm not saying any of this is correct or wise just what I did and in view of the comments above I'll probably play with another hard drive enclosure but ground it more thoroughly apart from the SMA threads.
There is an other aspect which should be mentioned. As far as I know, AC mains networks in Europe supply power in one of three ways: TN-C, TN-S or TN-CS. The only difference is where the PE (protective earth) and Neutral wires are separated. The worst of these is TN-C where PE and Neutral are separated at the customer premises i.e. only 4 wires entering the house. If you switch on a load with large capacitors at the input e.g. a switch mode power supply, it will represent a momentary short circuit between Neutral and one of the phase wires. This is of course only for a short duration (perhaps a few us) until the capacitors are charged. The problem is that during this short period, the Neutral wire can have a potential close to half the mains voltage. As the PE and Neutral wires are separated at the customer premises, also the PE can momentarily be close to half the mains voltage. Having both PE and local ground/earth/soil connected to different parts of a PCB will therefore easily destroy it.
I have put my KiwiSDR into a metal box and connected antenna ground, GPS ground and RJ45 shield to the box. 5V power supply is floating relative both PE and local ground.
Regards
Mauritz / SM2BYC
Attachments:
https://forum.kiwisdr.com/uploads/Uploader/24/2b198c390301d8a451fd51fd1fcaa6.jpg
You make valid points.
I've been trying to figure out the exact circuit, but this is the best I've managed so far.
I've also included three of the possible external ground paths, which may, or may not, be at the same potential.
I think if you are using a metal case it has to be bonded to the outers of the RF connectors. If you don't do this, the case is at a different potential to the rest of the circuits and will not provide an effective RF screen. It's also better from a safety perspective.
The Ethernet screen should also be bonded to the case (as you have done).
The DC power is tricky, as an external ground to the 0v line will reduce the effectiveness of the on-board common mode choke. But in some situations an external ground may improve things. If the 0v of the DC supply does not have a path to ground, I would leave it floating. If it does have a connection to ground, I would bond the 0V DC line to the case at the point of entry.
Some experimentation is likely to be required in order to determine the best results.
Regards,
Martin - G8JNJ
The supply earth to station earth is something I considered and from comments heard on air I believe the supply companies don't worry too much about that being in place as long as the link is light enough that if a fault occurred on the supply return (if used as earth) the whole street didn't earth through your antenna.
Obviously depends on earthing type as mentioned by Mauritz but to me the principle "RF good, current light" seems wise.
Something like a shorted ribbon cable (with one end cut at an angle) or foil might be safe as that would be half decent at RF but go open quickly if the current was too high.
If there is enough potential difference between the station RF earth, and supply earth, that it cooks a light link I think most bets are off on the equipment front.
Stu
NO NO NO NO NO. This is completely and utterly false. Just measure it! All BBBs and BBGs have a direct connection between the digital ground (triangular symbol marked DGND on schematics) and the "earth/frame/chassis" ground (slanted 3-prong-fork symbol) that on the schematics is shown connected to all connector shields and the 4 mounting hole vias. The easiest place to find DGND is pins 1 and 2 of the expansion connectors P8 and P9. Just measure between there and the shield cans on any of the connectors. On the BBB (not BBG) you also measure a short to the negative DC input connector. For BBG, if you were to strip down a micro USB cable to the black (typically) DC ground wire you should also find a connection from there to the connector shells.
UNLESS Seeed has recently changed the design in which case we're all in a lot of trouble. But I doubt that because they would not be meeting the BBB spec!
Measuring 'it' was what caused me the initial concern.
I disconnected everything from the KiWi and measured between the outer metal body of the RJ45 and the screen of the RF connectors, I saw 3.5K Ohm.
This seemed to correspond with the schematic as the only link I could find between the two ground paths was via R136, which had gone open circuit, but which I have replaced with a 0 Ohm link.
In order to double check, I've just pulled apart my two other KiWi's and they measure 0 Ohm.
I've rechecked the other KiWi and it still measures approximately 3.5K Ohm between the same points, so it looks like something else has blown open circuit in addition to R136.
This is the problem with the separate earth paths, if there is ever a high current / voltage between the different grounds and only a thin connection between them (such as a PCB via or similar) then something like this could happen.
I'd like to know what else has gone open circuit, so further investigation is required.
Apologies for frightening you John, I'll let you know what I find.
Martin - G8JNJ
I appreciate your efforts toward understanding grounds and associated current paths. I've been poking around this domain as well and I think it is not a wasted exercise.
I have other thoughts I'm not ready to write but I do want to comment now about the idea of "hooking everything to one ground". Different grounds do indeed offer the possibility of ground currents both for PS and mains related signals and for RF signals.
I think with the KiwiSDR we have quite a challenge. This is a good thing. We have both a very sensitive receiver and a relatively broadband one. Some of us may be using it over more than three decades. Efficient power supplies as well as myriad other devices also inhabit this realm, some at very high voltages and currents. This isn't news to anyone here, I'm sure. But I want to point out an additional issue. In a very real sense when one looks closely enough, and I think we have a tool that is getting us to this point, "There ain't no such thing as ground." I once had a co-worker who had exactly that sign posted at his work-bench.
There exists a fundamental issue known sometimes as "the attenuator problem". This shows up when one tries to place a very high value attenuation across a two-port network, typically a "box" with SMA connectors on each end and something-or-other inside. Because such a box has finite size and is made from non-superconducting material, there is ALWAYS a common current current path, even though it be constructed from copper or silver. Because the inputs and outputs have conductor between them, and even though shunted by (short) return paths that return most of the current back to the source/load as intended, finite size combined with imperfect conductor means that one can never get infinite attenuation. When an attempt is made to make a very high value attenuator, for example, the presence of the unwanted path provides a common mode coupling between the ends which eventually exceeds the intended coupling level of the attenuator inside.
I've been looking at using inexpensive 12V ---> 5V buck converters as a means of powering the several kiwiSDRs I use. One of the reasons is efficiency, at a remote site I can't necessarily afford to waste the energy lost in to heat were I to use a linearly regulated PS. It has turned out that this actually is working amazingly well. With a very inexpensive automotive "USB power converter", available for US$7, rated at 5V at 2.5A and switching internally at a 160 kHz I've been able to rid the supported kiwiSDR of the line. By using traditional shunt/choke techniques one can remove the QRN/M from a local 5 watt transmitter that is connected to a KiwiSDR to a point below the -150 dBm territory that the receiver can decode on WSPR from a quiet antenna. Perhaps surprisingly, I can do this while a very significant differential 160 kHz evident on the KiwiSDR input. In my opinion, the KiwiSDR actually exhibits pretty good power supply rejection ratio.
I hope more to come on this topic since though I do find that differential noise at the PS can be rejected, common mode noise is another matter I'm still working on. There are a lot of coupling mechanisms, certainly not all internal to the KiwiSDR. Common mode current on feedline is a very common one I've written about elsewhere. Near field sources of I and V have a lot of means of ingress and actually I think many of them are not radiated (inverse square) as common wisdom within amateur radio suggests. Getting a better understanding of these mechanisms seems a very worthwhile effort for many of us that offers hope of further improving these fine receivers.
Thanks again for pursuing this. I intend to stay tuned!
Glenn n6gn
Guess what......
R136 has gone open circuit on this board too.
So the digital ground and ESD ground are no longer connected.
R136 had the correct value marking, so it's gone open circuit whilst in service.
I revisited the other board that still showed no connection between the digital and ESD grounds.
I removed R136 in order to try and check the PCB tracks, but they all seemed OK.
Then I replaced R136 again with a 0 Ohm resistor.
Now the digtial and ESD grounds show 0 Ohms between them.
So whatever was open circuit now seems to have recovered.
I suspect that it could have been the tiny through board vias that are right next to R136 under the RJ45 connector.
All is now good again with both previously faulty boards :-)
I have a couple of theories.
1. A nearby lightning surge had blown R136 open circuit at some point.
2. If a power supply with a grounded outer on the DC connector is in use, and the center pin of the 5v DC connector accidentally touches the metal can of the Ethernet, USB, SD card or metal mounting pillars (or metal case if mounted via the pillars). Then it may burn out R136 (and possibly some PCB track).
Either of the above (or something similar) will result in the loss of ESD protection (and possible failure of the Ethernet port) and RF screening.
Regards,
Martin - G8JNJ
I was stupid, I connected a long wire to the Kiwi antenna input without grounding. Today just before rain I heard sparks, disconnected the antenna quickly and retested Kiwi. Long story short, ethernet port on Kiwi is dead. Using a serial cable, I was able to talk my Kiwi into using an ethernet port on an USB hub + ethernet combo. It seems to be working, however I had to switch the 10/100MBit ethernet port to 10MBit only, for some reason the ethernet adapter was unusable with the Beagle CPU loaded.
Now I would love to swap the LAN8710A-EZC-TR PHI chip, however these chips are unobtanium in the West, on order in more than a year. There are some chips available from China.
Now I have to finally build my voltage balun (sic!).
73, Vojtech OK1IAK