9
\$\begingroup\$

I had problems with DS1307 based real time clocks, the culprit seems to be the solder we used. Let me explain.

We were using an Adafruit DS1307 based RTC I2C module on a stock Raspberry Pi. The clocks were built - and found to be communicating ok (eg. I2C was good), but not "ticking", eg. the second was not advancing. Building them on a breadboard (not soldering) led to working clocks, so the module itself was out of the question.
Also the electrical connection between crystal and IC, as tested by a simple multimeter, passed on all units.

Further investigation and narrowing things down with my trusted bus pirate (a god sent device for us without heavy testing equipment!) lead me to re-solder the connections with my own (ie. what I use at home) solder - and voilà, they worked. Further narrowing down showed that I actually only had to clean and re-solder the connections of the crystal quartz, not even the IC side had to be re soldered.

Does anyone know or have an idea or explanation why
S-Pb60Sn40 does not,
S-Sn60Pb36Cu1 does work in connecting the crystal quartz to the PCB well, eg. in a way that the DS1307 starts ticking?
The large difference between the solder used seems to be in approximately swapped proportions of lead and tin used, but maybe the 1% copper does it's job somehow?
My further guess would be this has something to do with the choice of solder somehow damping the oscillator, but the "how" of it still leaves me baffled.

For the statistics: We have soldered 87 units with the "bad" solder, none of them working (eg. "ticking"). I have manually re-soldered 53 of them now, all working. For 48 of those I only de-, the re-soldered the crystal oscillator connections.

Edit 1 - Cleaning the solder pads
As proposed by some of you, I manually cleaned a number of PCBs with 70% isopropyl alcohol and/or pure water - no avail, the clocks still did not "tick" away. Only de-, then re-soldering the two crystal connectors makes them start ticking.

Edit 2 - The type of solder I used
The solder that did not work is this product from Stannol, the one that does work is this product from Felder Löttechnik GmbH (pdf in german, sorry, they don't seem to have an english version online).

Edit 3 - The flux that is in the solder (aka. getting closer...)
The non-working Stannol solder uses flux of type 2.1.2.A - that is, according to wikipedia an organic, water-soluble liquid (?? this seems to be wrong!) flux with halides. The working solder from Felder uses flux of type 1.1.2.B, that is resin, rosin based solid flux with halides.
\$\Longrightarrow\$ The culprit could either be the unknown "organic, water soluble" flux from Stannol or the different solder formula with added copper from Felder.

\$\endgroup\$
15
  • 4
    \$\begingroup\$ Did your solder by any chance have water-soluble flux that was not cleaned from the board? \$\endgroup\$
    – Tut
    Apr 29, 2014 at 15:15
  • \$\begingroup\$ @tut - That would be my vote, based on similar experience of unclean crystals failing to start. \$\endgroup\$
    – John U
    Apr 29, 2014 at 15:44
  • \$\begingroup\$ @JohnU Yes, we ran into this a long time ago at a place I worked where the manager rushed out an order of motherboards without cleaning the water-soluble flux. The RTC oscillators began failing on the first rainy day. \$\endgroup\$
    – Tut
    Apr 29, 2014 at 15:48
  • 1
    \$\begingroup\$ You might try manually resoldering a few using the "bad" solder. You may find that it's about the connection being made, rather than the material. Mostly for curiosity, since the answers regarding flux and so forth still apply, and you'll want to use the proper solder in the automated process. \$\endgroup\$
    – gbarry
    Apr 29, 2014 at 17:17
  • 1
    \$\begingroup\$ @Christian RTC oscillators should not be this finicky. It may be time to re-evaluate the oscillator design including: exact components used (including part#s), board layout (and material?), solder and anything else that may be pertinent. This would probably better be served in another question. Also how accurate are the working oscillators? Fast or slow and by how much? This could provide a clue regarding component selection such as capacitors. \$\endgroup\$
    – Tut
    Apr 30, 2014 at 11:03

2 Answers 2

Reset to default
14
\$\begingroup\$

Your problem is almost certainly not a direct issue with the solder itself but with the flux in the solder. Many fluxes leave a conductive film on the board which must be removed for circuits such as yours.

Note that RTC chips try to be extremely low power so as to preserve the battery as much as possible. As such, the biasing of the 32.768KHz crystal for the oscillator is designed to be rather low current (i.e. high impedance circuit). The flux conductivity can sometimes be low enough to completely swamp out and literally short out the oscillator preventing it from working.

\$\endgroup\$
4
  • \$\begingroup\$ Just out of curiosity, would you mind listing a couple flux types that are conductive? I was only aware of muriatic acid fluxes being conductive, but they aren't typically used for electronics. \$\endgroup\$
    – TimH - Codidact
    Apr 29, 2014 at 19:49
  • \$\begingroup\$ I tried cleaning 12 random PCBs with 70% isopropyl alcohol and 14 random PCBs with simple water; none of the > 40 PCBs I visually inspected had visible deposits of flux between the solder pads of the crystal. Cleaning the PCB did not change anything, ie. the clocks were still not ticking. Though your assumption makes a lot of sense, it seems to be something in the composition of the solder itself. \$\endgroup\$
    – Christian
    Apr 30, 2014 at 7:50
  • \$\begingroup\$ @TimH The "conductive" fluxes I am familiar with are those that are water-soluble which become partially conductive as they absorb moisture from the air. An example of a water-soluble flux is Kester 331 Organic Flux. \$\endgroup\$
    – Tut
    Apr 30, 2014 at 11:14
  • \$\begingroup\$ I see what you mean. Organic acid fluxes also will really damage the board (bad corrosion) if they are left on the PCB, so proper a assembly process should always include cleaning off the OA flux. \$\endgroup\$
    – TimH - Codidact
    Apr 30, 2014 at 21:38
2
\$\begingroup\$

You might want to use hot glue to fix the crystal case to the PCB in order to inhibit parasitic vibration. Soldering the case to the PCB is more thorough if somebody was smart enough to give you a pad for that, but it doesn't look like it.

If the can is just touching the board such that it can possibly buzz, that's about the worst case for parasitic loss of energy.

\$\endgroup\$
1
  • \$\begingroup\$ Thanks, but as strange as it sounds, the issue was indeed wrong solder. Not the flux, but the solder... \$\endgroup\$
    – Christian
    Mar 11, 2015 at 14:02

Your Answer

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

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