Teensy EMC sensibility (electromagnetic immunity)

Status
Not open for further replies.
On Windows when it shuts the USB down for power overload or problems you can see that in DevMan - to some degree - perhaps the same on Macbook. So if the PC drops the port it might be noted - first need to see if the USB drops at all from the lamp.

Yeah without using a cut Teensy - the battery power is odd - then you've got a new influence so the test and results could easily vary. But that is the kind of environment that is 'user created' as noted by Epyon that needs to be 'engineered out' - as long as the device is relatively stable - which the Teensy seems to be in my use.

I've got a twin teensy test I'm doing toward something FrankB is working on using 8 data lines for parallel data xfer - both on USB and powered by USB - I could go bring up a Fluorescent desk lamp and see if that bugs anything.
 
Hello

I made some tests and found interesting things.

1.)
It is irrelevant if the Teensy is powered via USB cable or via battery.
I can also reproduce it when the Teensy is powered via USB.

2.)
The notebook is running from battery

3.)
When I connect the Teensy alone I cannot reproduce the problem.

4.)
So what is the difference of my project with the Teensy on a breadboard and the Teensy alone ?
The longer cables make the difference.

I found out that if I ONLY connect a cable to the GND of the Teensy this seems to act like an antenna for the spikes from the neon lamp.
I have a big power supply that delivers +/- 0-50V and 5A.
ALTHOUGH this is power supply is turned OFF only connecting the Ground of the power supply with the ground of the Teensy is antenna enough that the Teensy loses USB connection EACH time I turn on the neon lamp!

The power supply is not connected to the earth. And it is turned OFF !
The cable between the Teensy GND and the power supply is approx 1 meter.
The USB cable is half a meter long.

Teensy USB.png

In the sketch the follwing code should be enough:

void setup()
{
Serial.begin(115200);
}

void loop()
{
Serial.println("Hello");
delay(1000);
}

This creates the COM port on the PC that is closed when the neon lamp is turned on.

If you don't have a neon lamp try to plug in and plug out any device that produces a spark or turn on/off a hairdryer.
 
Last edited:
Elmue said:
The longer cables make the difference.
Click to expand...

Can you be a little more specific about the cable lengths which reproduce this problem?

When I connect the Teensy alone I cannot reproduce the problem.
.....
So what is the difference of my project with the Teensy on a breadboard and the Teensy alone ?
Click to expand...

How could I know? I can't even see this. I can't know if the Teensy is just plugged into an empty breadboard, or whether it has wires connecting to the breadboard's power buses, whether the wires are short or long loops rising up into the air, or whether there's other circuits on the breadboard, or other wires connecting from the breadboard to other stuff, etc. Those details can really matter!


Is it possible this problem isn't related to the Teensy at all?

I mean, does that particular laptop lose communication with *any* usb device if there's a lengthy ground wire connected?
 
For me, it seems not to be Teensy related. I explained before how and why Neon lamps can cause EMP trouble in general, and as Elmue described, his huge power supply is not connected to earth, nor is his notebook, and he is using long cables. That is a setup which cordially invites all possible "over the air" troubles and no one with at least elementary radio engineering skills would dare complaining in public about a presumed "weakness" of a microprocessor, but would first and discretely fix his unprofessional development environment.
 
Theremingenieur said:
For me, it seems not to be Teensy related. I explained before how and why Neon lamps can cause EMP trouble in general, and as Elmue described, his huge power supply is not connected to earth, nor is his notebook, and he is using long cables. That is a setup which cordially invites all possible "over the air" troubles and no one with at least elementary radio engineering skills would dare complaining in public about a presumed "weakness" of a microprocessor, but would first and discretely fix his unprofessional development environment.
Click to expand...

totally agree
 
I'm going to try this later this weekend with a Macbook Air, using a lamp and hair dryer (if the lamp doesn't do it), and a turned-off lab bench power supply.

I'll let the USB cable power the Teensy, since it was stated "is irrelevant if the Teensy is powered via USB cable or via battery". Power by the cable is simpler to set up. :)

Really, I'm curious to see if anything bad will happen. Whether that old lamp and Robin's hair dryer can produce the same EM pulse is a good question. Likewise, whether the Macbook Air is similar to Elmue's notebook is a huge question. My gut feeling is the EMC sensitivity is on the laptop's USB port, since the Teensy doesn't reboot or do anything else unusual.
 
Last edited:
IANAEE, but I if I had this problem, I'd try a few easy things:
  • Ferrite on the USB cable
  • Aluminum foil shielding around all the wiring
  • Over-voltage protection components, i.e. diodes, fast fuses, MOVs, etc
  • Optoisolation - this will partition the problem and allow the antenna to be identified, I imagine

Maybe the real EEs can weigh in on whether this is the right list and what order to try things in.

I'd also try to make the rig more sensitive by using
  • Longer wires everywhere
  • Unshielded wires
  • A bigger spark or a more controlled spark, say from a lighter or car spark plug
  • Some way of measuring the RF like a radio, especially a software defined radio setup so I could maybe see the frequency and power spectrum of the spike.
 
pictographer said:
IANAEE, but I if I had this problem, I'd try a few easy things:
  • Ferrite on the USB cable
  • Aluminum foil shielding around all the wiring
  • Over-voltage protection components, i.e. diodes, fast fuses, MOVs, etc
  • Optoisolation - this will partition the problem and allow the antenna to be identified, I imagine
Click to expand...

That's only trying to cure symptoms, not fixing the problem at the root.

pictographer said:
Maybe the real EEs can weigh in on whether this is the right list and what order to try things in.
Click to expand...

Some have already given their statements about that highly unprofessional setup above.

pictographer said:
I'd also try to make the rig more sensitive by using
  • Longer wires everywhere
  • Unshielded wires
  • A bigger spark or a more controlled spark, say from a lighter or car spark plug
  • Some way of measuring the RF like a radio, especially a software defined radio setup so I could maybe see the frequency and power spectrum of the spike.
Click to expand...

Just to make things still worse and more unprofessional???
 
theremingineur:
> For me, it seems not to be Teensy related.

For you it is not related because you have a theory that you have never proved in the pratice.

WMXZ:
> totally agree

Another one who has an opinion without ever proving it.


Hey folks!
Why don't you take a Teensy and try it before you post your opinions here?
All you write here is theory until you PROVE your theory in pratice.

This is what distinguishes a scientist from a priest!

____________________________

I made another test:
The same as with the Teensy but now with an Arduino UNO.

The Arduino UNO is not affected by the neon lamp!
I can even put the lamp at a distance of 10cm to the Arduino UNO and USB is never disconnected.

This proves that it is NOT my notebook that loses USB conection.
It is definitely the Teensy!

Before you post any more comments, please TRY it !
 
Ach, Elmue, the fact that the UNO is less sensitive to such a phenomenon is not a scientific proof since it's a kind of technical dinosaur. When everybody was still working with electron tubes, EMV was never an issue, these problems arose only when transistors and MOS technology found widespread use.

But all that does not change the fact that it is unprofessional from your side to work with 3.3V CMOS circuits in 90nm technology without taking precautions for proper grounding and shielding in your obviously rough environment.

Trying to copy your setup is potentially killing almost every CMOS IC because it is not "selon les règles de l'art" as we say in French, couldn't find a matching translation of that expression in English, sorry.
 
Last edited:
With regard to : "Another one who has an opinion without ever proving it."
> Would not this also put the burden on the one making such a statement that they found the weak link and isolated and found what component needed to be shielded to prevent the situation?

I suspect a noisy light fixture. I have a Chinese one made 2+ years ago that has certification [CSA C22.2 No. 12] of some sort on the base. In fact it is possibly the only Fluorescent in my house that doesn't affect an AM radio at 3 feet! (standard screw in CFL bulbs can create nearly house wide interference - as do some LED bulbs I've gotten.)

I have twin Teensy's on my laptop both on USB - they are talking to each other at something over 1.4M byte/sec on 8 parallel wires on a perfboard. My 27 watt Fluorescent desk lamp has NO affect on them when repeatedly switched on and off at 10cm. The reader is validating the data received and both logging their progress to USB with no loss of USB on either one. In fact the test completed 14K cycles when left alone - it usually deadlocks under 4K (the next run make 3.5K with no CFL lamp switching).

Theremingenieur- maybe one of these variations:
http://mymemory.translated.net/en/French/English/r%c3%a8gles-de-l%e2%80%99art

Good point about the UNO - more different that similar UNO .vs. Teensy - the core chip (voltage and the internals), board layout, USB method, clock speed,
 
@defragster: Thank you for that translation link. I'm not an English native speaker and always thankful for any hint which allows me to improve my foreign language skills. I grew up in a fully bilingual (French - German) environment and I did lots of Latin in school which allows me to read and understand Italian and Spanish, but English was somewhat neglected, I had to catch up later when I studied mathematics and computer science at the (German) university and found that many books, especially technical stuff, existed only in English...
 
Theremingenieur said:
@defragster: Thank you for that translation link.
Click to expand...
:) - happy to keep my lack of anything but long ago, and unused, Spanish out of the spotlight ...

<edit>: BTW - I also powered up a 128x128 TFT display on a PJRC_OSH purple test board with USB connected and it also seems to be immune to the switching of my Fluorescent lamp. Even running at 120MHz OC with SPI at 60 MHz - It also ignored my nearby Piezo lighter strikes at 4 inches and keeps running with USB output.
 
Last edited:
Theremingenieur said:
Just to make things still worse and more unprofessional???
Click to expand...

Of course not. The idea is to identify which variable correlates most strongly with the effect. Sometimes it's easier to make things worse than to make them better. If 9 of 10 changes that should make the problem worse have no effect, but the the tenth makes the problem dramatically worse, explore that one.
 
@Emlue - Any chance you can test with a different laptop? As you can see in this video, I can't reproduce the problem here. If anyone else has experienced this problem, it's certainly very rare. I know you're personally convinced the problem is with Teensy... but now that I've made 2 videos, hopefully this 2nd one closely replicating the test you described, can you please try on your end with some other computers? I know you said Arduino Uno works with this laptop, but still, can you try testing with another machine? Is that possible?
 
@Elmue - so the problem goes away if the laptop and or the Supply is grounded?

When I did my test I didn't unplug the laptop - but I think my lamp is much less noisy than Paul's or perhaps yours.
 
defragster said:
@Elmue - so the problem goes away if the laptop and or the Supply is grounded?

When I did my test I didn't unplug the laptop - but I think my lamp is much less noisy than Paul's or perhaps yours.
Click to expand...

the big power supply connected only via GND but not used, could that not be the problem (antenna) (it has a coil that picks-up magnetic field generated by the coil (choke or drossel that elmue mentioned)) ?

I understand, without power supply connected, there is no problem

from elmue's post
3.)
When I connect the Teensy alone I cannot reproduce the problem.
Click to expand...
 
Last edited:
pictographer said:
Of course not. The idea is to identify which variable correlates most strongly with the effect. Sometimes it's easier to make things worse than to make them better. If 9 of 10 changes that should make the problem worse have no effect, but the the tenth makes the problem dramatically worse, explore that one.
Click to expand...

That is the procedure I (as a scientist) would take, determine cause and effect by experimenting, and making it worse is a good way if you can do it non-destructive, obviously.
 
Hello Paul

I see that you really have investigated.

Why is your result so different ?

I don't have another notebook to make tests with.
So if the cause was my notebook the explanation would be that the Arduino UNO has a stronger USB signal ?

I don't know.
My notebook is an ASUS, not one of these cheap notebooks.

I can reproduce the problem easily here.
But at the end this is useless if you cannot reproduce it.
 
Last edited:
Elmue said:
Why is your result so different ?
Click to expand...

Almost certainly something different between the Macbook Air and your ASUS PC notebook.

So if the cause was my notebook the explanation would be that the Arduino UNO has a stronger USB signal ?
Click to expand...

First, "Arduino Uno" has used 3 different circuit board designs over the years, and of course many clones are called "Uno" but use different circuity, and now 2 companies both claiming to be the official Arduino. Which of these you tested, I do not know.

In the case of Arduino Uno R3 (from Arduino.cc), the schematic shows inductors in series with the USB power and shield, and no low-ESR capacitor (as required by the USB spec) directly across the power lines. Perhaps those inductors, plus whatever's inside your notebook, are making a difference? But even that is just a wild guess. I simply do not know why this is so different.

I'm afraid I can't do more testing within the next 2-3 weeks. But in June, I could try repeating this test with a HP Envy laptop.

Any chance you could try with a different laptop?
 
Just one of many small details to consider is mentioned in section 7.2.4.2 on page 177 of the USB 2.0 spec.

http://www.pjrc.com/teensy/beta/usb20.pdf

This part talks about the effect of inductance of the USB cable and capacitance on both sides. Every Teensy has 1 uF or 2.2 uF capacitor directly across the VUSB to GND connection, with short leads, because of section 7.2.4.2.

However, I've heard numerous stories (many years ago) of FTDI's recommendation for a series inductor circuit allowing boards to pass several tests, even though it doesn't conform to the USB spec and would make these other problems even worse.

Perhaps ASUS also used some approach like the inductor, which Arduino is doing?

But again, all guesswork. Neither Apple nor ASUS publish schematics, and even if they did, these problems are subtle and even elusive when reproducible, but almost impossible to figure out what you can't reproduce them.
 
Indeed USB hardware is a fine tuned thing it seems - When the laptop USB goes away did Device Manager say if it is reporting that the PC has shut the port down and why? My USB 3 port on this 4.5 yo Toshiba regularly does that - for no known reason - even when plugged into a powered USB 3 HUB, 2011 was in the early days of USB 3. It works on some devices - for a time. I haven't tested it now with Win 10 much as I've given up on it.

I'd be more suspicious of the desk lamp used given the noise levels I've seen across CFL/Flourescent. Also the startup behavior of the lamps. The lamp I used has a single 27watt bulb and starts much faster than the flicker/pause in PJRC video, the problem lamp may be worse or more noisy. Also is the lamp tested running on 240v and 50Hz? If so it seems that could cause different startup surges.
 
Because this is my employer's IP, cannot say much. But some generic thoughts.

Have done some radiated immunity testing per CISPR24 on stuff with T3.1 and 3.2 at 3V/m with no problems (or at least were easily fixed). [Note on ITE EMC testing. There are changes afoot to use pulse modulation for the spot frequencies. Requirement could be up to 30V/m. If you had intended to use the normal AM modulation per IEC61000-4-3(up to 54V/m for calibration).] The end-use construction incorporated the lovely Teensy modules in a manner to account for both CM and DM. The test conditions (see CISPR16-2) should dupe the manner in which the equipment will be used; that is, if battery powered, reference that 'port' and the USB port only. If AC mains powered, the ground bond is critical to the test results - very critical for Class II equipment.

Emissions are another story - where my box must operate in Class B environment, must restrict clock to 24 or 48MHz, and no USB. Emissions and immunity are related - wires and traces can be a transmission line or an antenna; the signal return affects which one. Wires are always antennas, PCB layers always form 'patch' antennas, and vents are slot antennas. And antennas work both ways - if it can get into your box, it will also come out the same way.

Florescent lamps can be very bad, to the point of making both conducted and radiated emissions test data look like you were testing a comb generator. Incandescent or battery + linear LEDs are preferred lighting in the EMC chamber or test area.

AFAIK, all Apple computers meet class B emissions limits. But that stuff is typically measured at 10m (5m for Canada), so if inside the near/far field transition, the computer could be putting EMI into your stuff.

Most common EMI problems
1. ground bond Z too hi.
2. ground bound connected to neutral at equipment.
3. low permeability material for enclosure.
4. large loop areas -> L too hi for hi-freq returns.
5. Vcc or Vdd too hi.
6. poor placement of noise filters.
7. data lines not using hi-pass.

My PCB layout rules
1. clock trace layout first, and keep short.
2. shunt/guard traces layout next.
3. sandwich hi-freq traces between ground planes.
4. minimal use of vias.
5. segregate and group hi-freq signal traces, and keep off bottom.
6. VERY SMALL LOOP AREAS, and NO 1/4 WAVELENGTH TRACES.

So both immunity and emissions are solved by
1. good PCB design
2. careful equipment layouts to control wiring harness routing
3. very lo-Z ground bonding
4. enclosures with effective shielding.
5. filters at point of entry/exit
6. good ale
 
Status
Not open for further replies.
Back
Top