1
\$\begingroup\$

I have a design which uses a secret technique, and I would like it to remain secret, at least for a period of time. My plan was to pot the board in epoxy. Specifically, MG Chemicals 832TC is advertised as providing "superior protection from ..., and analysis".

But then I read on PottingSolutions.com that epoxys can shear off SMD parts due to thermal expansion or contraction, especially below the glass transition temperature. It seems that the very properties that make epoxy hard to remove are those which are incompatible with fragile SMD parts.

What is the recommended way to protect trade secrets on boards with SMD parts? I was considering applying a thin coat of silicone and then filling the rest with epoxy. If it matters, this will be inside of an aluminum box.

\$\endgroup\$
7
  • \$\begingroup\$ IME, pulling this kind of stuff just increases your costs, reduces your reliability and draws undue attention from the wrong people to your profit margin. \$\endgroup\$ Commented Sep 30, 2014 at 9:14
  • \$\begingroup\$ The first thing I wouldn't do is tell the world I had a valuable secret circuit worth protecting... \$\endgroup\$
    – user16324
    Commented Sep 30, 2014 at 16:59
  • 1
    \$\begingroup\$ Related: electronics.stackexchange.com/q/100078/2028 \$\endgroup\$
    – JYelton
    Commented Sep 30, 2014 at 17:26
  • \$\begingroup\$ There can be rational reasons to do this. For example the "Pacman Plus" was potted and I assume this help protect its secrets for some time. Did the epoxy make people curious? Sure, but this probably just added to the mystique. In my case, it is something that was obvious to me but apparently not to an entire (small) industry, even after many years. I'd like to buy myself a few months. I'd prefer to avoid a patent because that can take a while and the number of units sold probably doesn't justify that measure. \$\endgroup\$ Commented Oct 1, 2014 at 1:11
  • 1
    \$\begingroup\$ How big is your device? I don't expect shear failure in normal (small) potted devices. Also 'below the glass transition temperature' means 'at normal temperatures'. \$\endgroup\$
    – david
    Commented Oct 2, 2014 at 0:16

4 Answers 4

6
\$\begingroup\$

I find it extremely unlikely that your product could contain 'secret' methods that can be implemented with recognizable discrete components. If anything, actual innovative design that is worth keeping a secret is at least done on the chip level, but more commonly on a process level. Things that can be implemented with discrete or at least non-custom components are most likely just obvious designs.

That being said: you're not going to have any legal force by keeping stuff secret. This is exactly what patents were made for. You don't have to increase the cost of your product and screw over the customer with unnecessary protection measures and on the other hand you get the complete right to deny any of your competitors your technology. Otherwise they'll know there's something juicy in there, reverse engineer it within .4 of a femtosecond and hide it in their products. Even if you find out, you have no way whatsoever to stop them from using the same not-so-secret.

\$\endgroup\$
1
  • 2
    \$\begingroup\$ ^^^ This. You will never stop people reverse-engineering what you've done if there is any value whatsoever to it (coughCHINAcough), and no-one has yet invented a truly impenetrable protection system. $20 at a backstreet dentist gets you an X-ray of a board. If it's truly worth protecting, patent it and get a good lawyer. \$\endgroup\$
    – John U
    Commented Sep 30, 2014 at 8:33
2
\$\begingroup\$

I sneak my personal stock through the back door of a German producer, so I don't know any brands or numbers in an official manner, but there are low shrink and low warp epoxies as well. Although low is not zero, so if your parts are very, very fragile it may still not be safe enough.

You can add a thin layer of flexible material like silicone, but remember, this makes your epoxy much more easy to tear off of the PCB as well.

A good way to limit shrink and warp is preventing large thermal differentials over the entire area to be filled. This can be done by potting in layers, where you add a new layer when the last one is tough, but still lightly tacky.

\$\endgroup\$
2
  • \$\begingroup\$ How does layering prevent thermal differentials? Are you speaking of heat from the exothermic reaction of the epoxy setting? \$\endgroup\$ Commented Sep 30, 2014 at 4:57
  • 1
    \$\begingroup\$ It is to do with the peak exothermic point in setting. The epoxy is not extremely thermally conductive, so a 5mm layer will "vent" the core heat better than a 25mm layer. Or of course allow curing heat to creep in faster. \$\endgroup\$
    – Asmyldof
    Commented Sep 30, 2014 at 21:50
2
\$\begingroup\$

Instead of building an ultimately penetrable bunker around your widget, institute a poison pill or a scorched earth policy, make it smart enough to determine whether there's any hanky-panky going on and, if so, have it destroy itself and, possibly, its environs.

\$\endgroup\$
2
\$\begingroup\$

There's a couple of methods you can use to make reverse engineering more difficult that don't require expensive methods like potting. Here's a couple:

  1. Scratch off IC markings with an exacto blade. Even a sharpie is sometimes enough so someone can't read the IC marking (oh it's an LM314), see the company logo (oh it's a linear technology power supply), or read the processor family name (great they use the NXP1700 series).

  2. Design the board with all traces in middle layers. Now they can't reverse engineer connections without expensive X-ray equipment. Route everything through the middle and use vias near the pins to make the connections.

Again, this will only slow someone down. If they really wanted to and had the time and money, almost anything can be reverse-engineered.

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.