I'm pretty inexperienced at electronics (but learning!) and recently had some PCBs manufactured for a project I'm working on. I had 40 PCBs made (PCB Assembly done by the manufacturer) and received them earlier today.

While my PCB does several things, I'm only concerned about one specific part here; the power system. I power this system with a 3.0-3.6V battery (Temperature/Current-dependent), and it is supposed to boost this to a steady 3.6V to power a Particle Electron with (using two wires with a JST-PH2 plug, connected at W1/W2).

PCB Schematic PCB design Parts

  • U1: TPS610995DRVR

  • L1: 74404024022

  • C1/C2/C3: GRM188R60J106ME84D

  • W1/W2: JST-PH2 w/ wires (to be inserted directly into a Particle Electron)

I based my design off the TPS61099x datasheet and quite religiously followed everything in it (e.g. schematic, recommended parts). I thought it would work.

My issue:

  • For 10 of my PCBs, everything works fine. The Particle Electron is nicely powered up and everything runs as it should.

  • For the remaining 30 PCBs, as soon as I connect my Particle Electron (i.e. current is drawn) the output voltage immediately dips to ~0.9V.

I'm at a complete loss as how this happened. These PCBs were assembled by the manufacturer so I assume they should be identical/free of assembly errors. I also can't observe any visual differences that would hint at a wrong part, or part being placed incorrectly. The batteries can handle up to 2A, which is way more than a Particle Electron would possibly draw (max. 800mA). What could be a possible reason only some of my PCBs work? What should I look into first? I simply have no idea how to approach this issue and where to start with fixing it.

I can share my information if needed.

  • \$\begingroup\$ Test for shorts across the power rails. If you did not lay out your PCB correctly or the assembler was not prepared for the way you laid it out, you might have solder bridges. \$\endgroup\$
    – DKNguyen
    Apr 8 '19 at 23:13
  • \$\begingroup\$ Why do you have gnd at the top, convention is bottom. Why is EN tied to find, this disables, why is FB to gnd \$\endgroup\$
    – JonRB
    Apr 8 '19 at 23:14
  • 1
    \$\begingroup\$ Can you measure the VI curve with a linear supply using a slow ramp with current limit? On the good ones can you measure the startup load Current and also Tx pulse. GSM module on the Electron alone can consume a peak of 800mA to 1800mA of current during transmission \$\endgroup\$ Apr 8 '19 at 23:15
  • 1
    \$\begingroup\$ "What could be the reason" questions are almost always deeply flawed. Rather than asking, you should be exploring the circuit with a multimeter and scope, and identifying the most upstream deviation from design expectation in the problem units. What is the voltage on the battery under load? What happens if you substitute a supply and measure or limit the current? You need to work through your circuit and identify what is different between the good and bad examples. \$\endgroup\$ Apr 9 '19 at 4:48
  • 1
    \$\begingroup\$ Turned out the JST-PH2 plug at the end of the system had reversed polarity (wires swapped around). After fixing it, the PCB was working as expected. It's pretty easy to miss... \$\endgroup\$
    – Vitesze
    Apr 9 '19 at 16:20

I based my design off the TPS61099x datasheet and quite religiously followed everything in it (e.g. schematic, recommended parts). I thought it would work.

Your religious beliefs don't seem to be very inclusive. You completely ignored the suggested layout. Look at your design, and the recommended layout side-by-side:

enter image description here

Do you see how the recommended pattern on the switcher Vin pin is designed? A solid block of copper. Yours is a ratnest of skinny wires. Parasitic inductance will be everywhere. Look at the shape of ground: same thing, a solid extremely wide block of copper versus your skinny wires. I am sure there are a few other problem with grounds as well.

Switchers are sensitive to details of layout, all MHz-range switchers are fundamentally sensitive to implementation of "high-current" loops. You can't get decent (advertized) efficiency nor sometimes even basic functionality without reasonable implementation of these loops.

P.S. This is the second question for today where OP completely ignored suggested layout. I am wondering why people ignore good , manufacturer's validated layout advices? There is absolutely no reason to do anthing differently. Now you will pay with re-design, re-spin, and your time.


The batteries can handle up to 2A, which is way more than a Particle Electron would possibly draw (max. 800mA)

This might be the final clue to the problem: batteries can do 2000 mA, the load can take up to 800 mA, true, but the whole TPS610095 regulator can deliver 800 mA only under most favorable conditions: Vin=3.3V and PERFECT LAYOUT per Mfg's specifications. If the layout is suboptimal in few places, the switcher won't be able to deliver this kind of peak currents, and half of boards may shut off.

In summary, one margin (800 mA switcher) gets multiplied by marginal PCB layout, and the result is 75% failure rate. You need to increase your margins in this design.

  • \$\begingroup\$ While you are absolutely right, I agree with you, I'm curious whether it explains that 10 of boards work and 30 don't. \$\endgroup\$ Apr 9 '19 at 19:18
  • \$\begingroup\$ @Chupacabras, 50%+ failures indicate that the design is marginal, small differences in components swing the functionality one way or another. Which one is critical is difficult to say, the internals of IC contain dozens of high-speed comparators operating in 10+MHz range, the device operates at the boundary of Discontinuous Conduction Mode and Continuous Conduction Mode at start, so any ringing due to parasitics in layout may trigger some unwanted conditions preventing it to start properly. Who knows. That's why it isn't worth to risk development schedule by re-inventing own layout. \$\endgroup\$ Apr 9 '19 at 19:39

Murphy's Law

If anything can be reversed, it will be. enter image description here

  • your reversed design is swapped above.


  • FB only matters on the adjustable version TPS61099
  • FB is open on all the others TPS61099x
  • The IC FET and diode switch have a low resistance 80 mΩ

    • link http://gebc-energy.com/en-US/ProductClass-13-1.html
    • Choices: enter image description here
    • the green bobin type has low self discharge for 10yr low current
    • The battery has a high resistance (red flag) 1.5 Ω rising to > 3 Ω
    • the blue spiral type has lower ESR but shorter shelf life from leakage.
    • given the GSM module on the Electron alone can consume a peak of 800mA to 1800mA of current during transmission
      • this battery and booster IC is unsuitable for current >=800mA
  • Have you life tested this design over the life of the battery?

  • I expect it to fail earlier than the predicted mAh lifespan due to battery ESR to load Z ratios.
  • 1uH is 5 Ohms at 1MHz but this regulator has cycle-based current cut-off.
  • low ESR polymer ultracap ought to be attached to battery to improve ESR or use the Blue SPiral version so that dV/dt during transmission < 0.2V on battery.
  • If Battery ESR rises to 3 Ohms 0.8A = 2.4V drop in voltage. ( Regulator fails )
  • note the battery graph drops 300mV with 200mA or 1.5 Ohms when fresh.

enter image description here

  • Over what DoD ( depth of discharge) were you able to get 2A ?

This is inconsistent with battery VI drop curves for incremental resistance for the Green Bobbin battery.


Consider the 34615M ( blue one) which has lower ESR. (TBD)

  • \$\begingroup\$ The Bobbin cell has a max. current of 200mA, which is too little for the Particle Electron (drawing up to 800mA). The one I'm using right now is the 34615M, which is what you recommend. \$\endgroup\$
    – Vitesze
    Apr 9 '19 at 16:16
  • \$\begingroup\$ OK then you have wrong P/N on drawing and wrong orientation and who was the -1 low-life? \$\endgroup\$ Apr 9 '19 at 16:17
  • \$\begingroup\$ Yes apologies, the schematic is just off, I did not switch around the battery. The text is just misplaced.Anyway, turned out the JST-PH2 plug has reversed polarity, so it seems an easy fix. \$\endgroup\$
    – Vitesze
    Apr 9 '19 at 16:17
  • \$\begingroup\$ welcome to Murphy's Law \$\endgroup\$ Apr 9 '19 at 16:23

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.