10

In the germanium transistor era, the RF transistors were PNP small signal devices. Some of these small metal cased BJTs had 4 leads instead of the usual 3. The metal can was isolated and the 4th lead was called the screen and was connected to an internal screen. Common european example part numbers of such devices are:

  • AF116
  • AF117
  • AF126

I have even seen such transistors in early solid state MW AM car radios, and the popular Phillips 1960s electronics set had an AF116.

Sure there were some reliability issues with the screens shorting due to tin whiskers but they were not that bad. The long leads of the day would undermine the effectiveness of the screen at HF but the lead could be cut shorter.

  • Why don't they make SMD screened Si or GaAs or GaN devices today?
  • Is a screen expensive to manufacture?
  • Would there still be reliability issues?
JRE
  • 67,678
  • 8
  • 104
  • 179
Autistic
  • 14,235
  • 2
  • 27
  • 65

3 Answers3

11

The normal approach for small SMT components is to put a single big shield or "can" over a number of components:enter image description here

(image from this link)

pjc50
  • 46,540
  • 4
  • 64
  • 126
  • 2
    I use beautiful cans +1 .What if you had high gain a multistage RF amplifier inside the can ? – Autistic Feb 06 '20 at 10:06
  • The cans in that pic don't look particularly RF proof, they are wide open. – Lundin Feb 06 '20 at 13:37
  • 2
    @Lundin Wide open how? I think the metal covers the components completely - there are some odd reflections, though. – IronEagle Feb 06 '20 at 23:53
  • 3
    @IronEagle Calculate the gap between the clips on the short side and long side and there you have the wave length of RF emissions that the cap lets in through the gap in the bottom. It's going to be several GHz up, but good thing there's absolutely no 2.4GHz radio with harmonics anywhere these days... right? These have to be attached to the PCB by soldering. Leave as wide gaps between solder joints as your upper limit for RF filtering. >18GHz is a good rule of thumb corresponding to radiated susceptibility standards. – Lundin Feb 07 '20 at 07:28
  • 3
    The PCB edge on the lower left suggests a rendering, as does the lack of markings on the chip on the right side. – MSalters Feb 07 '20 at 12:25
  • 2
    It's obviously a rendering - the clips on the left edge are unnaturally thin and not held onto the board by anything, there's no solder paste visible. In practice either the cans would be soldered around the edge or there would be a metal "wall" soldered to the board for them to clip onto. – pjc50 Feb 07 '20 at 16:21
7

They did make it into the silicon era ... just. Look for any transistor in the TO-72 package for example : it's a 4-lead version of the TO-18, with the extra lead connecting to the case.

A few examples : BF115 (perhaps following the AF116?) but also BF173, and 2N4416 JFET.

And the 2N5179 - still (just?) available from Farnell.

I believe what killed them was the move away from metal cans (TO-18, TO-72, TO-5 etc) to epoxy (TO-92 etc) packaging, where screening would cost extra.

Inside those shiny cans in pjc50's answer you will often find further internal partitions, between stages, so the idea hasn't gone away altogether.

  • 3
    Not only would screening cost extra on SMD, it's also a lot more expensive to do correctly for minimum benefit. Either you put it in a can or you don't, but partial screening (and you can't fully screen without using a can or your screen will effectively *be* a can) simply isn't worth it. – Mast Feb 06 '20 at 19:53
6

RF screening makes sense if any of the elements come even close to a length where RF of relevant frequencies can couple in.

For SMD components, that would only happen in the millimeter wave regions.

For millimeter wave circuits, you basically need to consider your whole circuit including the circuit board as RF element that you'd want to simulate to know what's happening: you'd not expect any "stray" radiation coming from the outside, since you'd typically RF-encase the whole circuit, and no stray radiation from the circuitry itself, since everything needs to be designed as transmission lines rather than simple current-carrying traces, anyway.

Marcus Müller
  • 88,280
  • 5
  • 131
  • 237
  • What about MF electrostatic coupling?They did make an AF117 which is rare and I only have one .This was mainly used in high end low noise audio. – Autistic Feb 06 '20 at 09:46
  • I don't know what MF is? – Marcus Müller Feb 06 '20 at 14:57
  • 1
    @MarcusMüller It's right between LF and HF. –  Feb 06 '20 at 15:53
  • @BrianDrummond but how's something RF "electrostatic"? – Marcus Müller Feb 06 '20 at 16:10
  • @Autistic not quite sure what *electrostatic coupling* is (that seems self-contradictory to me), but as said: MF wavelengths really don't matter to any transistor I've ever seen; screening won't help if your interference is already in your signal path. – Marcus Müller Feb 07 '20 at 09:23
  • MF electric field coupling, probably. (As opposed to magnetic coupling) – user253751 Feb 07 '20 at 12:10
  • yes, but as said, where's the relevance? that coupling happens either on the whole circuit, in which case it has no influence on the transistor, or it happens into the signal lines, in which case screening doesn't help. Why would it couple into the transistor, if the transistor is about 1/100.000 of the length of the wave? – Marcus Müller Feb 07 '20 at 15:10