Taming an unstable power rail

[TechnoGilles]

Hi all,

I experience a strange behaviour with my project and I need some help to continue my investigation.

The project (see picture) :

- Was conceived and built entirely by me
- Has the following power rails : VD3, VC3, VSD, are 3.3V; VD5 is 5V. They are all regulated by Micrel LDO regulators powered by a main 6VDC, 3A supply;
- Has the following devices (and their power sources) : MCU (VD3), 4 CPLDs (VD3), SD Card (VSD), USB Module (VD5), 2 Audio DACs (VD3, VC3), LCD (VD3), Keypad (none), Real Time Clock (VD3), Beeper (VD5), and an interface for an extension device (octal transceivers on VD5);
- The PCB is a 4 layer PCB that was designed by me and manufactured professionally by Eurocircuits.com. The inner layers are one ground place (no sections) and one power plane (divided in sections for each power rail);
- Most chips are surface mount that I hand soldered.

The problem :
- First of all, let's say that, overall, the project is indeed working ! All the devices are performing properly and the project is actually usable to do what I want;
- But there is an instability that cause glitches once in a while; most of the time it's benign. Occasionally, it disrupt my USB communication.

The symptoms :
- I first noticed the problem because the LED backlight of my LCD was flickering. The backlight is 60mA and powered from the LCD power source which is VD5. It is controlled from one of the CPLD through a MOSFET transistor. The CPLD generates a PWM signal so I can control the intensity of the backlight (4 intensity from OFF to fully ON). I see the flickering all all intensities, the dimmest being the worst;
- When I connect my logic probe to the VD5 power rail, it picks up glitches and voltage variation from its power supply and beeps, even though the tip is not probing anything. The "rythm" of the glitchy beeps matches the LCD backlight flickers;
- Both the LCD backlight flickering and logic probe glitches are mild when my system is idle. They become severe when the system is at full capacity (reading from SD, transmitting/receiving USB, playing audio, controlling external device);
- When I monitor the voltage of the VD5 rail with my voltmeter (I unfortunately do not have a scope !), it fluctuates from 4.96V when my system is idle down to 4.80V when the system is running. The regulator never gets hot AT ALL. It is rated at 500mA but should not have to deliver more than around 200mA (USB rated at 75mA, LCD at 2.5mA + 60mA for bakclight, octal transceiver at 15mA x 2, beeper disabled in this case, logic probe at 20mA). All other 3.3V rails remains stable at approx 3.29V no matter what. Main power supply never goes below its rated 6V;
- Once in a while my USB module (on VD5) stop working and/or goes out of sequence and looses a byte causing either my project or the host PC to wait infinitely for the missing byte. This is the only major problem. The above issues, although annoying, are not causing malfunctions;
- The external device on the other side of the octal transceivers has its own 5V power and do not seem to be affected at all by the fluctuation on the transceivers VD5 power rail (although it might be the one drawing excess current from the transceiver... but then again, I'v seen the problem even when the external device is off !).

The question :
- So my guess is that something somehow is sometime causing a short or is drawing excess current on that VD5 power rail. But I failed to find what so far... And I do not know how to tackle the problem... So any suggestion at all would be more than welcome !!

Thanks for reading and helping !!!
TechnoGilles

 

[BobK]

Hi, and welcome to the forum!

Here are things that come to mind.

Do you have the recommended capacitor(s) at the input and output of the 5V regulator?

Do you have decoupling capacitors at all of the ICs?

You say the power plane is in sections. Are there multiple sections connect via traces for the VD5 rail? If so, is this a star configuration or daisy chained?
Bob

 

[TechnoGilles]

Hi ! Many thanks for the answer !

The regulator is a MIC5209-5V. I have a 10uF input cap, both 10uF and 0.1uF output caps, and a 470pF bypass cap. From the datasheet, I feel I am exceeding the recommended minimums.

Interestingly enough, the VD5 rail is the only one that does NOT have a power plane. It is indeed a 20 mil wide trace that provides power to the following daisy chain :

• Beeper (not drawing current when off)
• Logic Probe (20mA)
• Voltage Regulator (providing up to 500mA)
• USB Module (75mA)
• LCD Module and backlight (2.5mA + 60mA)
• 2 Octal Transceivers (approx 2 x 15mA max)

The 2 transceiver ICs have the standard 0.1uF decoupling caps; other modules do not.

In an attempt to solve the issue, here are two things that I did :

• I manually added 10uF caps near the following devices:
  • the logic probe connector
  • the LCD Module
  • the two transceivers
• I added straight wires at the back of the PCB directly from the regulator to the power pins of these devices:
  • the USB Module
  • the LCD Module
  • the first Transceiver (both are side by side)

I did not see any improvements.

 

[duke37]

Electrolytic capacitors can arc over internally particularly if they are overvoltaged. This can be quite exciting with large high voltage capacitors.The output voltage will therefore drop and the capacitor will recharge. Check for warm capacitors.
If this is the cause, then adding capacitors in parallel will not help but may slow the flicker.

You may be able to trace the fault by adding LEDs at all likely points and looking for flicker.

 

[BobK]

I would add 100nF caps close to power and ground of each of the ICs. 10uF caps will not serve the same function because of higher internal resistance and inductance.

It is too bad you did not start out with a star configuration for the power. Next time you probably will (I learned my lesson this way.)

I am surprised, however that all of the measures you took made no difference.

Bob

 

[TechnoGilles]

@duke37, I checked and replaced the caps on the VD5 rail. It did not help but thanks a lot for the advice.

@BobK, the transceivers IC already have100nF caps. The USB and LCD modules are autonomus modules plugged through a socket for the USB and a ribbon connector for the LCD. They both have their own onboard regulators and decoupling caps.

Here is a diagram of the power planes as well as the VD5 trace. Maybe this will give a better view of my design.


Thanks again to all. Any idea is welcome !!

 

[TechnoGilles]

I feel like one of the module/IC is drawing excess current. Does anybody has any idea as to how I can measure current consumption of a circuit that is soldered in place ?

 

[duke37]

An infra-red temperature detector could estimate excess current consumption.

 

[TechnoGilles]

ALL'S WELL THAT ENDS WELL !!

Although the bug was a stupid mistake I made from the beginning. All along I though I was getting power to my system from my 6V-3A DC source... WRONG !! ... I mistakenly used my 6V-1.2A DC source which is well below the 2A that my system needs at peak usage... How stupid !

So DC wise, I never saw the source go below 6V... But at high currents, the ripple was so high that the 5V regulator could not stand it. Although the 3.3V regulators were always ok.

Well that explains everything ! Switched the source and boom, no more bug !!

I'm so glad though cause now my system is complete and perfectly running !! Yeah !!

Thanks to all who took the time to read and comment !! Your insights pulled me in the right direction !!

Regards,
TechnoGilles