12
\$\begingroup\$

I goofed and didn't notice this until the board was already fab'd and assembled. The board is an RF amplifier; the portion I have pictured is a part of the DC control pat (so no RF is nearby but we're talking 100MHz-1GHz so it's surely floating everywhere). See the possibly disastrous screenshot marked by 'missing a via here'. (The fab removed by hand the huge trace to nowhere, before anyone asks). I really need to be more careful with altium's polygon pours...

I'm really kicking myself right now about this stupid error, it's a 20 board run and money is really tight; I'm in academia, so these boards are not getting remade. The problem is C18 is a 100nF bypass cap for a high speed op-amp. It seems to me that without a via to the ground plane, theres only that ultra tiny slice of the pour connecting it to a via 'very far' away. I might be wrong, but from everything I've read, the cap might as well not even be there because the inductance will be so large. I don't have the boards back yet, so the fab might even have eliminated that small trace entirely! It's only a few mils thick.

enter image description here

Maybe I'm being overly worried, as I don't yet know if this will cause issues. But is there anything I can do 'by hand' to improve the decoupling? Would soldering a small wire to ground be effective? I guess my main concern is oscillation with RF signal floating around everywhere; the op amp I'm decoupling is the LME49990, and I've seen this thing oscillate when bypass caps arent placed right.

\$\endgroup\$
  • \$\begingroup\$ Are there vias in the ground pour north of the thick V-15 track? \$\endgroup\$ – The Photon Jun 2 '15 at 4:34
  • \$\begingroup\$ And when you say "high speed", what do you mean? What frequency signals will be passing through this op-amp? \$\endgroup\$ – The Photon Jun 2 '15 at 4:35
  • 5
    \$\begingroup\$ On the picture I can't see which pin of the amp is its ground. Assuming it has one, I would try to mount a through-hole capacitor (with leads cut to the minimum) on top of the chip. \$\endgroup\$ – Wouter van Ooijen Jun 2 '15 at 7:46
  • 1
    \$\begingroup\$ As it's 20 boards, hand modifications won't take significant time. 20,000 boards would be a different matter. \$\endgroup\$ – Brian Drummond Jun 2 '15 at 10:25
  • \$\begingroup\$ Photon: highest speed signal through this chip is ~1uS. \$\endgroup\$ – Paul L Jun 2 '15 at 17:40
18
\$\begingroup\$

Drill a small hole that will allow you to simply run a free thin piece of wire from the ground plane on the opposite side of the board to the capacitor.

Stripped wire-wrap wire works well. It's 30 AWG, so the hole that you drill can be very small.

Use a Dremel tool in a drill press if you have such available - I routinely drill #78 holes reliably without breaking the bits.

Your can modify all of your boards in fairly short order.

\$\endgroup\$
  • \$\begingroup\$ If there's no rotary tool with a press then this handheld thing will do diy.stackexchange.com/a/41908/807 \$\endgroup\$ – sharptooth Jun 2 '15 at 13:53
  • \$\begingroup\$ I think this is probably the best option. Adding a bodge wire like Spehro suggested would add extra inductance and resistance, and it could also act like a loop antenna that could pick up noise. Keeping the wire as short as possible is ideal. \$\endgroup\$ – DerStrom8 Jun 2 '15 at 16:07
  • \$\begingroup\$ Back in the day when plated thru holes cost extra, people did this all the time. We called them "c wires" because the symbol for them on the assembly drawing looked like a C (which I assume was meant to represent a loop of wire, though of course the actual wire is straight). \$\endgroup\$ – Jeanne Pindar Jun 3 '15 at 4:08
16
\$\begingroup\$

I think you can most likely add a little fly wire from the existing GND via (or a closer one above) to the capacitor.

enter image description here

If you're having problems, do what Dwayne suggests and drill a hole and scrape the resist off, but you've got little to lose in trying the easy way first, especially if this op-amp is not directly handling the high frequency signals.

\$\endgroup\$
  • \$\begingroup\$ The best solution! Via to ground is not recommended when bypassing. You need to keep HF current off the ground plane. Just \$\endgroup\$ – Bip Jun 2 '15 at 23:01
  • \$\begingroup\$ @PredragPejic -- HF current is fine on a ground plane provided it doesn't mix with low level signal references, which is handled by proper board layout. \$\endgroup\$ – ThreePhaseEel Jun 3 '15 at 0:19
  • \$\begingroup\$ @ThreePhaseEel HF current isn't fine on the ground plane because when you have HF bypass current on the big ground plance you don't have ground plane but center fed patch antenna. A good explanation is given here electronics.stackexchange.com/questions/15135/… \$\endgroup\$ – Bip Jun 3 '15 at 4:57
15
\$\begingroup\$

Given how close the nearest ground region is, you probably don't need to use any wire or drill holes.

In the picture below, the points in the green circle are both Ground, in which case all you need do is scratch off the solder mask nearby and add a solder bridge. This will create a nice large (low inductance!) connection between the C18 ground pad and the rest of the plane.

Ground Points

\$\endgroup\$
4
\$\begingroup\$

The inductance between a bypass capacitor and the power supply is nearly irrelevant.

The entire point of a bypass capacitor is to supply brief bursts of energy to some IC through a low-inductance connection. It's the inductance between the IC and it's associated capacitor that is important.

The suggestion to "mount a through-hole capacitor (with leads cut to the minimum) on top of the chip." (-- Wouter van Ooijen) gives close to the minimum loop area, and will be the best you can do -- it appears that will give even less inductance than adding a via to your design and refabricating your board.

I see there is already a good connection between the high side of the capacitor and one power pin of your IC. I agree with Spehro Pefhany that a short bit of wire on the low side of the capacitor will almost certainly fix your problem, but I would connect the other end of that wire to the GND pin of the IC to minimize the inductance between the capacitor and the IC. Such jumper wires are extremely common in commercial PCBs and in space-qualified hardware; one hopes that if it's good enough for NASA, it's good enough for your application -- see "Is Rework Unacceptable?" and p. 1 of "NASA Workmanship: jumper wires". It is unclear if piggybacked parts are good enough for NASA -- p. 16 of "NASA Workmanship Standards: through-hole soldering" seems to say no, while p. 3 of "NASA Workmanship Standards: Deadbugs" seems to say yes.

If you are lucky, that GND pin on that IC already has a connection to the power supply that is more than adequate for slowly recharging the bypass capacitor between bursts; if not, a second wire attached according to (a) Spehro Pefhany's suggestion or (b) directly between that GND pin on that IC to some nearby GND point may be needed -- the difference between (a) and (b) is nearly irrelevant.

EDIT:

Many chips don't have a GND pin, or even when some pin happens to be connected to GND, they get their power from some other pin. For such chips -- for example, an opamp that gets its power from +15VDC and -15VDC -- the bypass capacitor should go directly across the power pins -- in this example, the capacitor should go directly from the +15VDC pin to the -15VDC pin.

\$\endgroup\$
  • 1
    \$\begingroup\$ Unforunately there is no ground pin on this chip; it's an op amp. \$\endgroup\$ – Paul L Jun 2 '15 at 17:41
3
\$\begingroup\$

First, try as it is. If you have problems, content a wire from other pad. The dremmel advice is cool, but before you ruin your board, try patching conventional patches. I think all you may need is a wire from other gnd pad and maybe additional capacitor. But probably it will just work. Many designers put there a serial ferrite or inductor, so just close your eyes and imagine this is what you did.

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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