Hi everyone.

We are paying for our sins trying to couple a standalone device we designed 6+ years ago with a Windows-running tablet as a display/control device.
The device we have passes everything by itself just fine.

Now it was decided to use a tablet as a display (before it had none) plus to be able to download parameters etc.

During the CE certification while conducting conducting (no pun intended) immunity test it was observed that the tablet reacts to noise in 300kHz / 700kHz range (obviously no ferrites help at these frequencies). Even though this does not interrupt the cycle the device is running continuously and the tablet recovers nicely when the test is stopped with no operator intervention, for these tests (unlike for ESD / EFT where criteria B is used, recovery with no operator intervention) no reaction manifesting itself is allowed at all.

It should be noted that the device uses a 12V power supply with a regular 2.5mm barrel power DC jack where the minus is also the MAINS Protection Ground. This power supply is used because the same tablet would misbehave with the fully isolated Power Supply when subjected to ESD test (it would sometimes go dark and then after a couple of minutes would agree to be powered up by pressing the power button). It actually might probably behave nice during conducted immunity with it but then we'd fail ESD.

I am not sure what I could do here. I even looked for something that could be sold to be plugged in series with USB cables to clean the surges but nothing really small enough in size has been found.

Could appreciate some fresh ideas (short of selling the device and the tablet separately. Could be done but .... uncouth).

There is a PDF file attached showing the simplified diagram of the things connected.

 

I don't begin to understand what's going on here but I have a stupid question: Does the problem remain if the DUT is powered by a 12V battery instead of the power adapter? Just trying to isolate where the problem originates.

 

The DUT is not a battery-powered device. It is strictly an AC adaptor-powered.
If it were a battery-powered device, this test would not be conducted in the first place as there is, simply speaking, nothing to plug in the coupling device. And the cables are not required to be tested as they are under 3m (10'). The test that we are failing is the conducted type, not radiated.

 

Which USB version are you using? I would guess that the lower the frequency the less likely the tablet would suffer from interference.

 

USB 2.x.

I doubt USB 1 would help, nor do tablets come with it.

 

The DUT is not a battery-powered device. It is strictly an AC adaptor-powered.

I understand, I was suggesting a test for experimental purposes to identify the source of your noise. If it comes from the power adapter, you know where to focus your energy.

 

Yikes, you've got a tough one.

• Check your USB cable. There are plenty of cheap ones out there. I've had issues with poor crimps on the connectors, especially the outer shell (the shield) on cheap USB cables. It might be worthwhile to dissect the cable to confirm. We had good results from cablestogo.com. Don't take that as an overall endorsement, I don't know their QA/QC policy and it's been a few years since I've been involved but they are definitely worth looking at.
  
• Have you tried different power supplies? Assuming the conducted energy is entering there, I'd focus on that.
  
• Does the issue manifest itself if you ground the chassis of your device to earth ground (your MAINS ground)? I know you've already investigated this and chosen the ungrounded supply, but if all else fails, it might be worth revisiting.
  
• Is the issue manifesting itself in the tablet or the device operation? If it's in the tablet, is there something that can be done in software to mitigate? We had a device driver issue where the USB would lose connection requiring the user to disconnect and reconnect as well as restart the application. Believe it or not, a re-write of the device driver fixed the issue.

 

Lo_volt:

The chassis is effectively GND-ed to Earth as the Earth is electrically bonded to -12V and then to the chassis. Obviously it would be better to have separate Earth and -12V as in 3-pin cable but this device was designed before the tablet was introduced, the design was "on the cheap" and the barrel-type power supply and the jack which is board-mount was a much cheaper way to go, especially in low qtys.

From the two power supplies we have tried (only one has gone through testing so far) I intend to try the previous one, with -12V uncoupled from Earth GND. The disturbance clearly comes through the Earth.
I used that one due to the fact that the tablet behaved nicely with it during ESD test I did.
What I noticed the tablet was lying flat on GND plane.
Once raised the ESD issue with the supply with uncoupled Earth went away, and what I did not know in the lab they do test it on the stand we provide. As soon as I tested it on that stand the ESD issue became very mild to none (Criterion B is allowed in ESD).
So I intend to try the uncoupled Earth power supply, hoping this does not screw the other tests and maybe fixes the Conducted Immunity test too.

Costs money - will have to redo many tests as the new power supply invalidates most of them.

 

Costs money - will have to redo many tests as the new power supply invalidates most of them.

Understood, and that's the frustrating part about the process. It may be worthwhile to set up a test bench at your location to do some amount of testing before taking it to the test lab. We found a low cost piezo "anti-static" gun worked well as an ESD simulator.

Some interesting things to keep in mind:

• ESD jolts will want to find the lowest impedance path to ground. The lowest impedance path will also result in the largest current and therefore the highest coupling into your device's circuits. When you provide a ground connection try giving it some amount of resistance so that it bleeds off charge slowly.
• ESD current can leap across small gaps. This can increase the susceptibility of your device. Another device I tested had several snap together sheet metal parts. Although they each touched, there was no fixed conductive connection from one part to another. Until we tied the parts together with either a screw or a short wire jumper it would fail ESD testing.
  
• Conducted immunity is all about not letting the interfering signal reach your circuit leaving the solutions to be filtering, shielding and isolation. Look for paths where the signal can interfere and try to eliminate them. Keep in mind that wires and traces adjacent to noisy ground paths can couple noise into your circuit even though they are not physically touching.

Good luck! Keep poking at it and you'll find a solution.

 

I understand, I was suggesting a test for experimental purposes to identify the source of your noise. If it comes from the power adapter, you know where to focus your energy.

Taking this a step further, and not knowing your device this may or may not make sense; but does the tablet still exhibit problems if it's near but not connected to the other device? Or if necessary, use one tablet and device combination to run the test and put an identical tablet next to it and see if that tablet misbehaves. The goal here is to try to identify if the interference is coming by wire, or by EMR.

Just another thought to keep in mind; touch screens do have conductors in them, and conductors can pickup on EMR. I have no idea how much EMR it would take to cause a problem.. I had a digital bench power supply go crazy on me and it took a few minutes to realize that the problem was caused by a radio (walkie talkie) that was set on top of it.