Defluxing such that single-digit fA electrometers work
Posted: Fri Sep 19, 2025 10:32 am
I mentioned to another member about defluxing and it was suggested that I start a thread on the subject...
We start with soldering. "Proper" solder is 60/40, hopefully modified by 2-3% silver (makes much better joints). Years ago, there used to be adverts for Multicore solder with a picture of the five flux cores. The flux is there to chemically clean the surface of the parts to be soldered such that you get a good joint. You heat the joint, apply the solder and it melts, flows, and produces a perfect joint with zero angle of contact. But what happens as the solder melts? Before the metal melts, heat has flowed into the solder and the flux heats and would expand, but is constrained by the solder. Steam engine buffs refer to this as being superheated. When the metal melts, the pressure is released and there is an explosion of flux that throws solder droplets around. That pool of flux after soldering is flux containing solder droplets. That's the exact recipe for a thick film resistor; conductive droplets in an insulating binder. That's why you need to deflux - those pools of flux around your joints are surface leakage paths.
Defluxing doesn't mean vaguely spraying defluxer at a joint. The defluxer can soften/dissolve the flux, but it needs a good scrub to shift it. That's why some of the defluxer cans have a stiff brush on the end; you're meant to scrub vigorously. Your flux (and solder droplets) are now mobile and everywhere, making the situation even worse than before you started. You now need to hold the board vertical and dilute and wash away the contaminated board with copious isopropyl alcohol. A few layers of elephant's loo roll underneath is a good idea. Again, scrub as you wash, starting at the top and quickly working your way down. Speed is of the essence; don't dither.
Now wave your (pre-heated) hair dryer at the board to evaporate the alcohol whilst shaking the board. If you don't do this immediately, the isopropyl alcohol will absorb water from the air and leave the board dirty. With all this liquid swilling around, some will have got on the other side of the board. So deflux that too.
But in defluxing the reverse side of the board, some of its contamination will have made it onto the first side...
Using a fresh (clean) toothbrush, deflux the first side of the board all over again. Defluxer, then rinse, wash away with isopropyl alcohol, and dry with pre-heated hair dryer.
You've guessed it. The reverse side of the board now needs a second deflux. That's four entire deflux processes. Don't skimp.
Things to consider before even pressing the defluxer can:
Modern components survive defluxer; axial leaded components are from an earlier design era and probably don't. Suflex polystyrene capacitors don't even survive isopropyl alcohol, let alone flux cleaner
Flux cleaner often strips paint
That isopropyl alcohol had better be clean. Not stuff poured back in a bottle after it has washed something
Hold the board with the most sensitive parts at the top, not the bottom
If you're laying out a board, electrometer op-amps benefit from a hole routed undeneath them to prevent dirty flux being trapped undeneath
Wear protective gloves
The elephant's loo roll evaporates the solvents rather effectively, so deflux with plenty of fresh air and take the loo roll outside immediately afterwards
Once you're done, the board should look perfectly clean. If it doesn't, do both sides again. Subsequently hold the board only by the edges - finger grease causes leakage paths. Do not breathe on the board (humidity and contaminants in exhaled breath). Cleanliness is next to godliness.
Tip for BNCs at the input of an electrometer: Push a clean bit of paper over the pin you're going to solder, solder (exploding flux all over the paper), let the flux harden, tear the paper away, leaving the PTFE underneath uncontaminated. No need for defluxing.
Electrometers are really fussy about leakage paths. The slightest hint of dirt/fibres is enough to increase their noise. There's no such thing as "good enough"; it has to be perfect.
Using the described techniques, I soldered a batch of twenty electrometer boards that had to measure <2fA and they all worked first time...
We start with soldering. "Proper" solder is 60/40, hopefully modified by 2-3% silver (makes much better joints). Years ago, there used to be adverts for Multicore solder with a picture of the five flux cores. The flux is there to chemically clean the surface of the parts to be soldered such that you get a good joint. You heat the joint, apply the solder and it melts, flows, and produces a perfect joint with zero angle of contact. But what happens as the solder melts? Before the metal melts, heat has flowed into the solder and the flux heats and would expand, but is constrained by the solder. Steam engine buffs refer to this as being superheated. When the metal melts, the pressure is released and there is an explosion of flux that throws solder droplets around. That pool of flux after soldering is flux containing solder droplets. That's the exact recipe for a thick film resistor; conductive droplets in an insulating binder. That's why you need to deflux - those pools of flux around your joints are surface leakage paths.
Defluxing doesn't mean vaguely spraying defluxer at a joint. The defluxer can soften/dissolve the flux, but it needs a good scrub to shift it. That's why some of the defluxer cans have a stiff brush on the end; you're meant to scrub vigorously. Your flux (and solder droplets) are now mobile and everywhere, making the situation even worse than before you started. You now need to hold the board vertical and dilute and wash away the contaminated board with copious isopropyl alcohol. A few layers of elephant's loo roll underneath is a good idea. Again, scrub as you wash, starting at the top and quickly working your way down. Speed is of the essence; don't dither.
Now wave your (pre-heated) hair dryer at the board to evaporate the alcohol whilst shaking the board. If you don't do this immediately, the isopropyl alcohol will absorb water from the air and leave the board dirty. With all this liquid swilling around, some will have got on the other side of the board. So deflux that too.
But in defluxing the reverse side of the board, some of its contamination will have made it onto the first side...
Using a fresh (clean) toothbrush, deflux the first side of the board all over again. Defluxer, then rinse, wash away with isopropyl alcohol, and dry with pre-heated hair dryer.
You've guessed it. The reverse side of the board now needs a second deflux. That's four entire deflux processes. Don't skimp.
Things to consider before even pressing the defluxer can:
Modern components survive defluxer; axial leaded components are from an earlier design era and probably don't. Suflex polystyrene capacitors don't even survive isopropyl alcohol, let alone flux cleaner
Flux cleaner often strips paint
That isopropyl alcohol had better be clean. Not stuff poured back in a bottle after it has washed something
Hold the board with the most sensitive parts at the top, not the bottom
If you're laying out a board, electrometer op-amps benefit from a hole routed undeneath them to prevent dirty flux being trapped undeneath
Wear protective gloves
The elephant's loo roll evaporates the solvents rather effectively, so deflux with plenty of fresh air and take the loo roll outside immediately afterwards
Once you're done, the board should look perfectly clean. If it doesn't, do both sides again. Subsequently hold the board only by the edges - finger grease causes leakage paths. Do not breathe on the board (humidity and contaminants in exhaled breath). Cleanliness is next to godliness.
Tip for BNCs at the input of an electrometer: Push a clean bit of paper over the pin you're going to solder, solder (exploding flux all over the paper), let the flux harden, tear the paper away, leaving the PTFE underneath uncontaminated. No need for defluxing.
Electrometers are really fussy about leakage paths. The slightest hint of dirt/fibres is enough to increase their noise. There's no such thing as "good enough"; it has to be perfect.
Using the described techniques, I soldered a batch of twenty electrometer boards that had to measure <2fA and they all worked first time...