I have a DC/DC converter with +5V input and ±15V dual output to supply power to the analog portion of a board (just two operational amplifiers).

The +5V DC input comes from an external box that uses an AC/DC switching converter. I have an AWG-16 cable connecting the +5V from that external box to the power-in connector of my board.

Problem I'm having:

  • The DC/DC converter fails to provide power if I connect the already-powered cable directly to the board's power-in connector.
  • The DC/DC converter does work properly and does provide power if I plug the cable and then provide AC power to the external box.
  • When it fails, I measure approx. +1.5V at the positive output, and approx. -1.5V at the negative output.

Schematic (relevant portion) is included at the bottom; some details of the board are as follows:

  • The power input to the board is a "thick"(ish) 4-pin connector carrying two grounds, +3.3V, and +5V.
  • The +5V go to the input of the DC/DC. It has decoupling (0.1uF X7R) and a bulk capacitor (47uF Aluminum Polymer) right next to the power-in connector.
  • The power-in connector is approx. 3cm away (in trace length) from the input pin of the DC/DC converter.
  • The DC/DC converter does not have its own decoupling capacitor next to its input pins.

This is the schematic (all 0.1uF capacitors are ceramic, 50V, X7R):

enter image description here

This is a (synthetic) view of the board (3D view produced by KiCAD), to give you an idea of the physical layout of things (the aluminum capacitor at the top, next to CP4, that one is CB5):

enter image description here

These are the attempts to fix the issue:

  • Tried adding a ferrite bead at the input of the DC/DC converter — had some Murata Electronics BLM21PG331SN1D hanging around, so I cut the trace, pealed a bit of the solder mask and soldered the ferrite bead between the bulk capacitor and the DC/DC converter's input pin. No difference whatsoever.
  • Replaced the DC/DC converter with a pin-compatible model. Instead of the CUI PDS1-S5-D15-M-TR, I installed an XP Power ISA0515 (on a separate board; that is, the ferrite bead is in one board with its original CUI converter, and I put the XP Power converter on another board, without a ferrite bead). This actually made the problem far worse. Now, I only get power once every 5 or 10 times that I plug the AC power (seemingly randomly).

Any feedback/critique on the schematic? Any advice on typical strategy to make the circuit more reliable? Alternatively, any advice on what to try next to figure out the cause of the issue?

I added 3.6k resistors in parallel with CP1 and CM1 (directly loading +15V and ‒15V). I confirm with an ohmmeter; resistance between either +15V or -15V and GND is around 3k. I did this on the board with the replaced DC/DC (i.e., with the XP Power converter).

Did not make any difference.

Regarding the oscilloscope captures. Yesterday, I captured the 5V input in instances where the ±15V came up and also instances where it didn't. The traces for both cases are essentially identical. Something like this:

enter image description here

Now, this morning, with one of the original/unaltered boards (with the CUI converter, no added load, no added inductor), the +15V rampup shows this weird behavior: (horizontal scale is 10 ms/div)

enter image description here

However, this one does come up reliably (as long as I power-up the external power supply that delivers the +5V, as opposed to plugging the cable carrying the +5V to my board).

Any ideas?


[UPDATE 2 — presumably "problem solved"]
Strange that with the XP Power model, the added load did not make a difference (and the datasheet also specifies a minimum load of 10%, which would be 3.3mA).

With the original part (the CUI converter), without any alterations (i.e., no added inductor), adding the 3.6k resistors now makes it reliably come up, regardless of whether I power up the "upstream" +5V supply, or whether I unplug and plug the already-powered cable to my board's input connector.

  • 1
    \$\begingroup\$ Sounds like you are hitting an inrush limiter upstream. I would put a PTC or positive coefficient resistor (inrush limiter) . Ferrite beads suppress transients but do appreciably limit DC inrush current . Do you have scope traces of what the 5V and +/-15V rails are doing? That can help pinpoint the issue. See also murata knowledge base article on inrush and ferrites murata.com/en-us/products/emiconfun/emc/2012/10/29/… \$\endgroup\$ – crasic Oct 22 '19 at 19:03
  • \$\begingroup\$ @crasic ‒ upstream?? do you mean a limiter in the +5V supply feeding the input of my DC/DC converter? The +5V at the input are present in all instances. I'll try to get some traces of the input's waveform at power-up. \$\endgroup\$ – Cal-linux Oct 22 '19 at 19:14
  • 1
    \$\begingroup\$ Guessing: 2 op amps may be insufficient load causing spiky output. Try increasing CP1/CM1 and/or adding a 2K-5K load on both the plus 15 and minus 15. \$\endgroup\$ – John Birckhead Oct 22 '19 at 21:39
  • \$\begingroup\$ @JohnBirckhead ‒ interesting suggestion/guess.... I will definitely try that tomorrow when I'm back at the office. One of the puzzling aspects is that almost everything in this design is essentially identical to previous designs where I've used similar DC/DC converters..... But now that you pointed this out, it's a good candidate for a discriminating factor: I don't think any of the previous boards had such low load on any of the DC/DC converters! \$\endgroup\$ – Cal-linux Oct 23 '19 at 0:10
  • 2
    \$\begingroup\$ This is a common SMPS stability issue . Typically when you roll your own supplies from PWM controllers they have extra pins to strap some additional feedback compensation, required for stability of the feedback loop. Part of the problem with these all-in-one dc bricks is that they require you to stay within their operating impedance range , which is required for current feedback stability. You may want to add some additional impedance besides just the resistor and retest. Good that you confirmed this isn't 5 V ratcheting due to inrush \$\endgroup\$ – crasic Oct 23 '19 at 15:37

Your load current and decoupling is inadequate to maintain the stability of the internal switching regulator feedback, and is causing unreliable startup.

Things to try

  1. Add more load
  2. Switch input cap to 4.7uF
  3. Add 1nF cap between V- and Input GND
  4. Increase output caps to 1uF
  5. Alternatively remove output caps.
  6. Add 6uH input choke
  7. See Below

You must ensure at least 3mA of current on each rail, according to the datasheet, I would go a little more above this to 10mA load and go down from there. The datasheet says that the minimal load should be no less than 10% of the maximum power

enter image description here

The datasheet recommends a slightly different input and output capacitor arrangement plus a 6uH choke. With additional capacitance of 1nF between V- and input Ground. And a recommended 4.7uF of input capacitance.

enter image description here

enter image description here

This device does not require any output filters, it already has them, but the datasheet is unclear, The way I read Table 2 and Table 3 is that it is recommended to have 1uF of output capacitance, but no more than 100uF per channel.

enter image description here

I would remove them first. Although 0.1uf Should not cause any problems, then add them back in as 1uF parts.

Additionally, ensure that your op-amp decoupling does not overload the device capacitatively, excess output capacitance will cause startup issues

All or some of these may be contributing to your startup issues

These DC/DC converter bricks are Integrated Switching Mode Power Supply (SMPS) Converters. That is how they achieve such small size, high efficiency, and stable output.

SMPS Converters generally use current and voltage feedback to establish the duty cycle of the switching device (PWM).

These feedback loops are sensitive to load impedance changes and require you to maintain a minimum operating current.

When you roll your own SMPS, whether from discrete components or using an SMPS PWM Controller (Bring your own inductor and feedback loop) you tune this feedback for the size of the switching inductor, expected load, minimum load, and desired regulation/rippler

On other DC/DC bricks they may provide a feedback compensation pin to strap on more impedance to tune this behavior slightly, these all-in-ones do not, but will still be susceptible to the same kind of issues common to all SMPS designs.


2 op amps may be insufficient load causing spiky output. Try increasing CP1/CM1 and/or adding a 2K-5K load on both the plus 15 and minus 15.


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.