Forums

EMC Screening Query

Started by Syd Rumpo October 2, 2015
I'm working with some equipment which contains a commercial SMPS which 
has been enclosed in a stainless steel box to screen it.

However, we still get some noise from it, experiments suggest that this 
is radiated.  The noise interferes with some sensitive signal 
conditioning and manifests as noise on a channel at half the switching 
frequency which is about 130kHz, although messing about with lengths of 
wire suggests the real problem is in the tens of MHz region.

[Connect screening box top to a nearby ground (although it's already 
grounded) with a 1m piece of wire and the noise reduces. Great! We're 
getting somewhere!  Make that a few cm so it can be permanent.  Oh, the 
noise has become worse.  That's repeatable, long wire good, short wire 
bad. That's why I think it's tens of MHz.]

I don't have the expertise or equipment for this, and we'll probably get 
someone in who specialises in the black arts, but in the interim, I 
suspect that we could improve matters by using something other than 
stainless steel for the box - this was only used as there was some in 
the workshop.

The SMPS is about 120mm x 150mm x 50mm.  Would copper be much better? 
How thick?  Tinplate?  Suggestions?

Thanks
-- 
Syd
On Friday, 2 October 2015 10:06:37 UTC+1, Syd Rumpo  wrote:
> I'm working with some equipment which contains a commercial SMPS which > has been enclosed in a stainless steel box to screen it. > > However, we still get some noise from it, experiments suggest that this > is radiated. The noise interferes with some sensitive signal > conditioning and manifests as noise on a channel at half the switching > frequency which is about 130kHz, although messing about with lengths of > wire suggests the real problem is in the tens of MHz region. > ... > I > suspect that we could improve matters by using something other than > stainless steel for the box - this was only used as there was some in > the workshop. > > The SMPS is about 120mm x 150mm x 50mm. Would copper be much better? > How thick? Tinplate? Suggestions? >
If the interference is being coupled magnetically, then tin plated steel might be better than stainless steel as it may well have a much higher permeability. It also has the advantage that the seams can be soldered. However, no amount of shielding is going to help unless the inputs and outputs of the smps are suitably filtered at the point where they pass through the shield. If the magnetic filed from the SMPS is very high, spacing the shield away from it by a few mm can help avoid magnetic saturation of the shield. Feedthrough capacitors are available, either with solderable flanges or with nut and bolt fitting. Combined with an enclosure that has soldered seams you should get a useful improvement if the capacitor values are well chosen. If the psu is pcb mounted, then the shielding can be soldered to a ground fill surrounding the PSU which also extends as much as possible under it. That ground fill must be bonded to the main ground plane with many vias. Surface mounted capacitors immediately adjacent to the places where the power inputs and outputs penetrate the shield and soldered between the power rails and the ground fill will perform similarly to feedthroughs. John
On Friday, October 2, 2015 at 5:06:37 AM UTC-4, Syd Rumpo wrote:
> I'm working with some equipment which contains a commercial SMPS which > has been enclosed in a stainless steel box to screen it. > > However, we still get some noise from it, experiments suggest that this > is radiated. The noise interferes with some sensitive signal > conditioning and manifests as noise on a channel at half the switching > frequency which is about 130kHz, although messing about with lengths of > wire suggests the real problem is in the tens of MHz region. > > [Connect screening box top to a nearby ground (although it's already > grounded) with a 1m piece of wire and the noise reduces. Great! We're > getting somewhere! Make that a few cm so it can be permanent. Oh, the > noise has become worse. That's repeatable, long wire good, short wire > bad. That's why I think it's tens of MHz.] > > I don't have the expertise or equipment for this, and we'll probably get > someone in who specialises in the black arts, but in the interim, I > suspect that we could improve matters by using something other than > stainless steel for the box - this was only used as there was some in > the workshop. > > The SMPS is about 120mm x 150mm x 50mm. Would copper be much better? > How thick? Tinplate? Suggestions? > > Thanks > -- > Syd
Weird, is the SS box solid, or one of those wire mesh type enclosures? Usually when weird things happen like that it's ground related. (I guess you already knew that.) George H.
On 02/10/2015 15:05, George Herold wrote:
> On Friday, October 2, 2015 at 5:06:37 AM UTC-4, Syd Rumpo wrote: >> I'm working with some equipment which contains a commercial SMPS which >> has been enclosed in a stainless steel box to screen it. >> >> However, we still get some noise from it, experiments suggest that this >> is radiated. The noise interferes with some sensitive signal >> conditioning and manifests as noise on a channel at half the switching >> frequency which is about 130kHz, although messing about with lengths of >> wire suggests the real problem is in the tens of MHz region. >> >> [Connect screening box top to a nearby ground (although it's already >> grounded) with a 1m piece of wire and the noise reduces. Great! We're >> getting somewhere! Make that a few cm so it can be permanent. Oh, the >> noise has become worse. That's repeatable, long wire good, short wire >> bad. That's why I think it's tens of MHz.] >> >> I don't have the expertise or equipment for this, and we'll probably get >> someone in who specialises in the black arts, but in the interim, I >> suspect that we could improve matters by using something other than >> stainless steel for the box - this was only used as there was some in >> the workshop. >> >> The SMPS is about 120mm x 150mm x 50mm. Would copper be much better? >> How thick? Tinplate? Suggestions? >> >> Thanks >> -- >> Syd > > Weird, is the SS box solid, or one of those wire mesh type enclosures? > Usually when weird things happen like that it's ground related. > (I guess you already knew that.) > > George H. >
It's solid, with springy fingers making sure the lid makes contact - I tried copper tape around the edges anyway. John suggested, inter alia, making the box taller to move the lid away from transformer fields which sounds good and is achievable. Cheers -- Syd
On 02/10/2015 12:03, jrwalliker@gmail.com wrote:
> On Friday, 2 October 2015 10:06:37 UTC+1, Syd Rumpo wrote: >> I'm working with some equipment which contains a commercial SMPS which >> has been enclosed in a stainless steel box to screen it. >> >> However, we still get some noise from it, experiments suggest that this >> is radiated. The noise interferes with some sensitive signal >> conditioning and manifests as noise on a channel at half the switching >> frequency which is about 130kHz, although messing about with lengths of >> wire suggests the real problem is in the tens of MHz region. >> ... >> I >> suspect that we could improve matters by using something other than >> stainless steel for the box - this was only used as there was some in >> the workshop. >> >> The SMPS is about 120mm x 150mm x 50mm. Would copper be much better? >> How thick? Tinplate? Suggestions? >> > > If the interference is being coupled magnetically, then tin plated steel > might be better than stainless steel as it may well have a much higher > permeability. It also has the advantage that the seams can be soldered. However, no amount of shielding is going to help unless the inputs and > outputs of the smps are suitably filtered at the point where they pass > through the shield. > > If the magnetic filed from the SMPS is very high, spacing the shield > away from it by a few mm can help avoid magnetic saturation of the shield. > > Feedthrough capacitors are available, either with solderable flanges > or with nut and bolt fitting. Combined with an enclosure that has > soldered seams you should get a useful improvement if the capacitor > values are well chosen. > > If the psu is pcb mounted, then the shielding can be soldered to a > ground fill surrounding the PSU which also extends as much as possible > under it. That ground fill must be bonded to the main ground plane > with many vias. Surface mounted capacitors immediately adjacent to > the places where the power inputs and outputs penetrate the shield > and soldered between the power rails and the ground fill will perform > similarly to feedthroughs. > > John >
Thanks, that all makes sense - I'll see if we can try this before a guru can be enlisted. In fact the box lid is quite close to the main transformer core and there's plenty of room for it to be taller. We can use feedthrough caps for power in and out, but there's also a 15 way ribbon cable with D type carrying control signals which goes through a cutout to plug onto the board - I guess a filtered box mounted connector would be good here? Cheers -- Syd
Syd Rumpo wrote:

> I'm working with some equipment which contains a commercial SMPS which > has been enclosed in a stainless steel box to screen it. >
I ran into similar problems on a sensitive circuit. I was able to identify where on the SMPS the radiation was coming from. Hook a small coil of wire to a scope probe and ground clip, and move around the SMPS to find the hot spot. Then, rig all wiring to be as far from that area as possible. I used .062" aluminum sheet metal for the shield box. The filtered power entry module comes into the shielded area, and mains wires do not go anywhere through the rest of the equipment (like to a front panel switch.) Then, if that doesn't fix it, you can get "feedthrough capacitors" that bolt in a hole in the shield wall. These ground by bolting to the hole, and have a terminal sticking out each side of the wall. You connect the SMPS to the terminal on the inside of the shield, and the equipment being powered to the terminal on the other side. This greatly attenuates any conducted noise on the power wires. Jon
If the box is sealed tight, then it's coming out the wires.  Ground those 
to the box, as well as possible (FBs and bypass caps, or feed-thrus as 
another mentioned).

Tim

-- 
Seven Transistor Labs, LLC
Electrical Engineering Consultation and Contract Design
Website: http://seventransistorlabs.com

"Syd Rumpo"  wrote in message news:mulhau$crm$1@dont-email.me...

I'm working with some equipment which contains a commercial SMPS which
has been enclosed in a stainless steel box to screen it.

However, we still get some noise from it, experiments suggest that this
is radiated.  The noise interferes with some sensitive signal
conditioning and manifests as noise on a channel at half the switching
frequency which is about 130kHz, although messing about with lengths of
wire suggests the real problem is in the tens of MHz region.

[Connect screening box top to a nearby ground (although it's already
grounded) with a 1m piece of wire and the noise reduces. Great! We're
getting somewhere!  Make that a few cm so it can be permanent.  Oh, the
noise has become worse.  That's repeatable, long wire good, short wire
bad. That's why I think it's tens of MHz.]

I don't have the expertise or equipment for this, and we'll probably get
someone in who specialises in the black arts, but in the interim, I
suspect that we could improve matters by using something other than
stainless steel for the box - this was only used as there was some in
the workshop.

The SMPS is about 120mm x 150mm x 50mm.  Would copper be much better?
How thick?  Tinplate?  Suggestions?

Thanks
-- 
Syd 


On Fri, 2 Oct 2015 10:06:32 +0100, Syd Rumpo <usenet@nononono.co.uk>
wrote:

>I'm working with some equipment which contains a commercial SMPS which >has been enclosed in a stainless steel box to screen it. > >However, we still get some noise from it, experiments suggest that this >is radiated. The noise interferes with some sensitive signal >conditioning and manifests as noise on a channel at half the switching >frequency which is about 130kHz, although messing about with lengths of >wire suggests the real problem is in the tens of MHz region. > >[Connect screening box top to a nearby ground (although it's already >grounded) with a 1m piece of wire and the noise reduces. Great! We're >getting somewhere! Make that a few cm so it can be permanent. Oh, the >noise has become worse. That's repeatable, long wire good, short wire >bad. That's why I think it's tens of MHz.] > >I don't have the expertise or equipment for this, and we'll probably get >someone in who specialises in the black arts, but in the interim, I >suspect that we could improve matters by using something other than >stainless steel for the box - this was only used as there was some in >the workshop. > >The SMPS is about 120mm x 150mm x 50mm. Would copper be much better? >How thick? Tinplate? Suggestions? > >Thanks
What are you actually testing, and how are you doing it? A reversable effect is a useful indicator, but if it involves Joe Blow breaking and making wire connections with his fingers on an opened box, it's not going to tell you much. 10MHz is approaching conducted limits and engenders radiated complications, as you seem to be noting. Stainless steel is a poor material to employ in any electro-mechanical iteration, due to its hostile relationship to machining and rework. Get it working in some other material, then and only then, if you must have stainless, see what happens when the substitution is made. RL
On 03/10/2015 15:09, legg wrote:
> On Fri, 2 Oct 2015 10:06:32 +0100, Syd Rumpo <usenet@nononono.co.uk> > wrote: > >> I'm working with some equipment which contains a commercial SMPS which >> has been enclosed in a stainless steel box to screen it. >> >> However, we still get some noise from it, experiments suggest that this >> is radiated. The noise interferes with some sensitive signal >> conditioning and manifests as noise on a channel at half the switching >> frequency which is about 130kHz, although messing about with lengths of >> wire suggests the real problem is in the tens of MHz region. >> >> [Connect screening box top to a nearby ground (although it's already >> grounded) with a 1m piece of wire and the noise reduces. Great! We're >> getting somewhere! Make that a few cm so it can be permanent. Oh, the >> noise has become worse. That's repeatable, long wire good, short wire >> bad. That's why I think it's tens of MHz.] >> >> I don't have the expertise or equipment for this, and we'll probably get >> someone in who specialises in the black arts, but in the interim, I >> suspect that we could improve matters by using something other than >> stainless steel for the box - this was only used as there was some in >> the workshop. >> >> The SMPS is about 120mm x 150mm x 50mm. Would copper be much better? >> How thick? Tinplate? Suggestions? >> >> Thanks > > What are you actually testing, and how are you doing it?
It's a sensitive modem which uses multiple frequencies, and the problem manifests as noise on only one frequency channel, related to the SMPS switching frequency.
> A reversable effect is a useful indicator, but if it involves Joe Blow > breaking and making wire connections with his fingers on an opened > box, it's not going to tell you much. > > 10MHz is approaching conducted limits and engenders radiated > complications, as you seem to be noting.
Yes, it indicates that we need someone with equipment and experience.
> Stainless steel is a poor material to employ in any electro-mechanical > iteration, due to its hostile relationship to machining and rework. > Get it working in some other material, then and only then, if you must > have stainless, see what happens when the substitution is made. > > RL >
Actually, stainless is unnecessary - it's all inside a rack mounted enclosure and could be made of anything really - but S/S was used because it was available, and the machine shop just bought a laser cutter which makes it quick, easy and fun to use. Apparently, copper is more difficult. Cheers -- Syd
On 10/3/2015 9:33 AM, Syd Rumpo wrote:
> On 03/10/2015 15:09, legg wrote: >> On Fri, 2 Oct 2015 10:06:32 +0100, Syd Rumpo <usenet@nononono.co.uk> >> wrote: >> >>> I'm working with some equipment which contains a commercial SMPS which >>> has been enclosed in a stainless steel box to screen it. >>> >>> However, we still get some noise from it, experiments suggest that this >>> is radiated. The noise interferes with some sensitive signal >>> conditioning and manifests as noise on a channel at half the switching >>> frequency which is about 130kHz, although messing about with lengths of >>> wire suggests the real problem is in the tens of MHz region. >>> >>> [Connect screening box top to a nearby ground (although it's already >>> grounded) with a 1m piece of wire and the noise reduces. Great! We're >>> getting somewhere! Make that a few cm so it can be permanent. Oh, the >>> noise has become worse. That's repeatable, long wire good, short wire >>> bad. That's why I think it's tens of MHz.] >>> >>> I don't have the expertise or equipment for this, and we'll probably get >>> someone in who specialises in the black arts, but in the interim, I >>> suspect that we could improve matters by using something other than >>> stainless steel for the box - this was only used as there was some in >>> the workshop. >>> >>> The SMPS is about 120mm x 150mm x 50mm. Would copper be much better? >>> How thick? Tinplate? Suggestions? >>> >>> Thanks >> >> What are you actually testing, and how are you doing it? > > It's a sensitive modem which uses multiple frequencies, and the problem > manifests as noise on only one frequency channel, related to the SMPS > switching frequency. > >> A reversable effect is a useful indicator, but if it involves Joe Blow >> breaking and making wire connections with his fingers on an opened >> box, it's not going to tell you much. >> >> 10MHz is approaching conducted limits and engenders radiated >> complications, as you seem to be noting. > > Yes, it indicates that we need someone with equipment and experience. > >> Stainless steel is a poor material to employ in any electro-mechanical >> iteration, due to its hostile relationship to machining and rework. >> Get it working in some other material, then and only then, if you must >> have stainless, see what happens when the substitution is made. >> >> RL >> > Actually, stainless is unnecessary - it's all inside a rack mounted > enclosure and could be made of anything really - but S/S was used > because it was available, and the machine shop just bought a laser > cutter which makes it quick, easy and fun to use. Apparently, copper is > more difficult. > > Cheers
Hey, Sid - Take my comments as inexperienced conjecture. 1. A really nice copper enclosure may increase your problem. I have read that the internal signal just reflects off the walls many more times than something made of, say, steel because of the permeability. 2. Ideally, you would want some RF absorbent stuff inside the enclosure. That may reduce any RF circulating currents on the skins of the metal. 3. The other suggestions here are golden. It will actually be somewhat difficult to decide whether it is leaking from the box or radiated from the exiting leads. If you have time to read, I recommend: <http://emcesd.com/> Best of luck to you.