EMI problems

Suitably modified for NEMA 6P, for an extra $5k per unit. ;)

I really like those Hammond boxes with the metal mesh EMI gaskets. Pricey though.

With that and good EMI shielding on the board, there are no SMPS spurs visible above the dark noise, even on a spectrum analyzer with a 100-Hz IF bandwidth. Without the gaskets it isn't nearly as good, even though the powder coating is masked off on the contact surfaces.

M1 BOX ALUM EXTRUDED EMI 120x103x32mm MECH 1457L1201EBK Hammond $32.14 @ qty 10

Cheers

Phil Hobbs

Welcome to my (former) nightmares. There are plenty of articles online that discuss proper shielding. The problem is that they don't go into enough detail for dealing with strong RF fields. (Incidentally, the Subject line should be RFI, not EMI). Some things that should be considered (in random order):

1. Avoid magnetic materials and dissimilar metals. These can create non-linear junctions which act as mixers. You want to be dealing with a few strong signals, not a comb line screen full of mixer products. The best clues can be found under PIM (Passive InterModulation) design articles. Plate everything with silver and use silver plated connectors. Don't use anything magnetic which includes nickel plated connectors and magnetic stainless connectors. Besides, nickel and steel have somewhat high resistivity: Metal Resistivity (lower is better) (nano-Ohm-meters) Silver: 15.9 Copper: 16.8 Gold: 24.4 Aluminium: 26.2 Zinc: 59.0 Nickel: 69.9 Iron: 100 Tin: 109 Stainless: 120 Lots of stuff for the cellular industry now comes with Low PIM connectors:

It's unlikely that you'll need low PIM hardware, unless it's expected to be used on an RF polluted mountain top transmitter site or share a coax cable used to transmit RF.

1. Watch out for corrosion problems. That shiny nickel plated steel sheet metal enclosure you soldered together from hardware store stock will work very well for a few months. However, when it starts to corrode or oxidize, the shielding characteristics can change. RF travels on the outside of conductors (skin effect) and an oxidized conductor is not as good as the original unoxidized metal.
2. Watch out for triboelectric effects in coax cables and connectors. Just moving the cables can produce noise. Check the tribolectric series tables: and avoid anything that might produce static electricity.
3. Check for ground loops in your connecting cables. It's very easy to create a ground loop if you have lots of cables running around. While this is usually a problem at 60 Hz, a full wave loop at FM/TV frequencies makes a dandy antenna. If part of the loop is the enclosure, an unshielded wire, a wire not bypassed, or anything that can radiate RF inside the enclosure, it will radiate RF.

You can eliminate a ground loop if you only ground one end of a coaxial cable. This may seem counter-intuitive. The idea is to keep the coax cable from acting as a one turn transformer. The grounded end should be the signal source end. I had a box of coax jumpers that had the shield broken at one end. Over that was some insulation and a layer of copper tape connected to the adjacent coax connector. The idea was to cover the gap in the shield to not expose the center conductor. If installing such a cable magically fixed a spur or noise problem, I knew that I had a ground loop.

1. Play with various ferrite beads around coax cables. They work very well at removing RF signals from the outside of a coax cable. For low frequencies, a toroid core with a few turns of coax through the donut hole.
2. Avoid exposed center conductors on coax cables. This is a common problem when soldering coax cable to a bulkhead connector. There's one in your cookie can photo in the upper right with some small parts hanging off it. The long coax shield lead acts like an inductor giving a not so good RF ground connection.
3. Try to use differential circuits. The common mode rejection of anything that leaks into the box is worth the added cost and complexity. You'll see some of that on CATV amps, where keeping RF from leaking into (ingress) the enclosure is critical.

There's probably more, but I can't think of anything major. None of the things I mentioned are critical or vital for decent shielding. Just keep them in mind when you build your device.

Good luck and remember, RF is magic.

I would have preferred a nice rectangular die-cast box with a screw lid, but we don't have any handy. I guess I should order a few, various sizes. Drilling and screwing stuff to the DBC can is sort of flimsy. On the other hand, it solders nicely.

My film cans solder well too. I bought, like, 1000 pieces of a very nice Russian 39 nF feedthrough caps, and use those for power and low speed stuff. They hold up well to slow soldering, too.

Cheers

Phil Hobbs

Have you considered installing an RF screen room? We had an RF pollution problem at one company. Only one bench (not mine) needed screening, so we build an elevated wood frame, and covered it with brass mesh wire screen. The mesh was wide enough that it didn't need its own HVAC system to prevent meltdown. Add a brass door frame, ordinary door covered with brass mesh, and finger stock all around. At another company, located next to SJO, we had a burnin room that needed to limit the RF leakage towards the airport. This one was similar, but made of soldered brass sheet stock. Expensive, but quite good. I have photos of these and others, somewhere, maybe.

There are also benchtop RF screen boxes:

I needed one to tune up pagers on the same frequency as an active 500 watt paging transmitter about 1/4 mile away. There was no money available to buy something, so I hijacked an old environmental chamber, added finger stock on the door, wire mesh over the window, and filtering on the power, making a rather nice looking screen box.

For those occasions where a screen room or box are unavailable, I have a sheet of aluminized mylar, which I convert into a tent of sorts. It was originally a high altitude balloon. It has plenty of gaps to leak RF, but it works reasonably. I haven't tried it, but a space blanket might work:

Another possible solution is to reduce the RF at the source. What you need is a supply of Orgonite delivered to the Sutro tower in a ceremony called "gifting". The seller positively guarantees that if you dump some of these near the base of the tower, they will "...diminish the harmful effects of electromagnetic radiation, detoxify bodies of water and generally enhance our environment". There's no mention of fixing ingress and noise problems, but if you ask, the seller will probably add it to their list.

Radio waves aren't electromagnetic? My customers seem to use "EMI"

Some things

I don't think PIM is an issue for me now. I'm talking 120 microvolts RMS as my test limit.

Leaving one end of the shield open will make the noise on the shield couple directly into the signal. Ground loops, if any, can be killed by a common-mode ferrite or some such.

Yes.

It works fine in my situation. If it didn't, I would have had to work harder.

The DUT is differential out, so we hacked our amp box to add a wideband transformer at the input, to convert from diff to single-ended. Luckily, it already had two SMB connectors on the input.

My measurement bandwidth is only 100 MHz. That's almost DC.

That's NOWHERE good enough! You need a feedthrough cap there, too. That wall-wart jack is the least suppressing port on the can, so most likely the primary way interference gets in.

Only if still filled with cookies. Instruction booklet should indicate the end user needs to eat all the cookies before installing the electronics into the can.

Jon

That's the upmarket version of Thor Labs' Lab Snacks. Some people get a lot of them:

Cheers

Phil Hobbs

RFI is a subset of EMI. I tend to use RFI as anything generated by a transmitter, while EMI is whatever else is left. That's wrong, but convenient. When you started talking about seeing the various signals from the Sutro tower, I assumed that the major problem was RF related.

True. Some PIM test sets have two high power transmitters (5 to 25 watts) inside the box. I mentioned it because it can become a problem if whatever you are designing shares a common coax cable with multiple high power signals. Just a precaution.

The jumper with one ungrounded end is useful for identifying if you have a ground loop problem. I don't recommend it as a permanent solution.

However, I beg to differ with your approach. If a ground loop induces more noise (1:1 xformer effect) than the shield might couple directly, then grounding at one end only can be beneficial. Google search finds: (lots of good stuff here)

1. Ground the cable at one end. This eliminates the potential for noise inducing ground loops.

Some references mentioning to ground only one end of an ethernet shield.

No spectrum analyzer handy? Just think of all the EMI/RFI junk above

You might want to calculate the resonant frequency of the Royal Dansk cookie can: If you see a noise peak in that area, it could be something oscillating or filtering using the can as a resonator.

But it is good enough.

You need a feedthrough cap there, too. That

But it doesn't. My noise floor is flat when the gear is in the can. And I'm seeing under 9 nV/rootHz RMS noise at the amp output in a 100 MHz bandwidth. That's almost exactly what I expected from simulation.

I sometimes buy stuff from ebay and get some sort of extra treat inside. And sometime a Bible verse.

That makes no sense at all. Great.

There's also a huge amount of lower-frequency noise here, probably from AC-line powered things in the neighborhood.

This motor controller was making 20 volt spikes in the lab, until we hacked it.

Here's the general spectrum:

-28 dBm with just a Pomona lead antenna.

We have a couple of SAs, but my requirement is net RMS noise voltage, and a scope measures that directly.

I'm OK at 100 MHz. It might make a resonant cavity filter or something at some higher frequency.

Grin, I only get a lousy bag of microwave popcorn from Advanced Circuits.

George H. (It's all about the prize inside.)

You can leave the ground intact and get a loop antenna, or you can lift the ground at one end and get a dipole antenna. Choose your poison.

Lifting a coax ground is almost always a terrible, terrible idea.

-- john, KE5FX

As far as the signal is concerned, if there are ground potential differences between the source box and the load, opening a shield will make them worse.

Right.

Your browser eats them like crazy!

The RF signal in a coaxial cable propagates in the insulation material between the outside of the inner conductor and inside of the outher conductor. Disconnecting the shield at one end, the coax acts more like a G-line.

One way around would be disconnecting the shield at one end, but bridge it with a small capacitance good quality capacitor. It will effectively block any mains and LF related ground loop currents, but the coaxial line works as intended on HF and low-VHF frequencies.

Is this a one off proof of concept design ?

In that case, why do you insist doing noise measurements in hellish RF environment ? Move the test equipment and DUT into a properly shielded room or better into a deep cave or mine.

If the final product is intended to be working in such hellish RF environment, the shielding should be designed from the beginning. Put everything (including the photodiode) inside a die cast aluminum or copper box. Since the photodiode needs to detect some outside light, make a small hole in the box.

To prevent RF penetrating into the box through the hole, put a small metallic tube around the hole. It will then act as a waveguide filter, cutting out any low frequency RF noise. As the maximum frequency required is only 100 MHz, quite large diameter tubes can be used, so it doesn't limit the photodiode field-of-view too much. If only a narrow f-o-v is required, a 1 cm diameter metal tube would be sufficient, having cut off at several GHz.

Ground loops are very low impedance at low frequency, and so can do weird things. We don't usually expect a 5-mV signal with a circulating current of an amp.

A parallel combination of a 1-ohm resistor and a bunch of capacitance in series with the shield, combined with a diff amp, can be good medicine for that problem.

Cheers

Phil Hobbs