high voltage drop under low load

[V_King]

Hello,

I am starting with atmels xmega mcus. I did a small breakout pcb for the atxmega64a4u. I also included a voltage regulator for the 3V3, required by the mcu, with enough juice (1A) to power other peripherals while prototyping. 

Unfortunately, I get an excessive voltage drop when connect a 3.2' tft.

The power supply configuration is:
USB 5V > shotky diode (1A rated, 0.4V drop) > NCP1117 linear regulator (1A rated, 1.2V drop max) > 10uH wire wound  power inductor (1.4A rated) > 3V3
I followed atmel's AVR1012: XMEGA A Schematic Checklist for the breakout, hence the inductor.

when just blinking a led etc, I get a 3.3V on the rails, but when I connected the lcd pannel, the voltage drops to 2.8V
the backlight uses 80mA and the lcd panel uses 20mA, xmega has 200mA max current rating. all in all, the current consumption is less than 300mA. So I should be ok, running the board from USB 5V, 500mA supply. I even bypassed the diode, but it did not make any difference.

where is my mistake ? any ideas 

thanks

[ConKbot]

Whats the voltage coming in at? With a weedy little 28 AWG cable, you could be running into voltage loss on the input under the load.  IIRC usb specs allow a 10% drop over cables and connectors at the 500mA, so even if the supply is accurate, your input could be coming in at 4.5V.  With the Schottky bypassed, 4.5V would be borderline for the 1.2v drop of the linear, and any resistance in the inductor could be making it sag under 3.3v. Why is that there, btw? the linear should be quiet enough that you shouldn't need an inductor on the output, plus any step-changes in current draw could also make the voltage dip, possibly enough to trigger a brown out reset if your voltage is marginal already. 

[lgbeno]

Use a dmm to measure the voltage at each node in the circuit and determine where the in expected drop is coming from.  Also verify your current to confirm that it is 80ma.  Inductor seems suspect and it really is not needed

[V_King]

with, or without load, the voltage from usb is 5.06-5.08V. I also tested with three different usb cables (one used for high power external hdd).

the voltage drop happens across the voltage regulator. when I disconnect tft pannel, I see 3.3V and when connected, it drops to 2.85V.

I measured the current consumption of the tft panel with my dmm. not the most accurate way to measure, but seem to match with the datasheet info.

I've put the indurctor, due to the atmels application note, which could be found here - www.atmel.com/Images/doc8278.pdf page 2
and I've followed the app note to the letter, as I want as low noise as possible to the analog part of the xmega circuitry. I also made the power line traces as wide and short as possible, without vias, so there should be minimum resistance there as well (the board is only 50x50mm).

the mcu is working ok, with or without voltage drop, as I have the brown out detector disabled by default, with a long startup time due to the inductor.

I don't have experience in doing the power supply side for mcus, so it would be great if you guys could share the best practises for the future (I've read about MOVs etc, but they seem to puzzle me).

[Precipice]

Can you dig out the datasheet for your Schottky diode? 0.4V sounds unlikely for a 1A current through a 1A (max) device. Look for the Vf vs If graph...
(For instance, Farnell sell the Avago HSMS-280x   as a 1A device - but that's the 1uS pulse abs max number. At 1A, Vf is over 0.9V, and not even on the graph!)

Also, can you confirm the LDO isn't getting hot enough to trigger its protection stuff? Do you get a few seconds of running before it fails?

[ConKbot]

with, or without load, the voltage from usb is 5.06-5.08V. I also tested with three different usb cables (one used for high power external hdd).

the voltage drop happens across the voltage regulator. when I disconnect tft pannel, I see 3.3V and when connected, it drops to 2.85V.

I measured the current consumption of the tft panel with my dmm. not the most accurate way to measure, but seem to match with the datasheet info.

I've put the indurctor, due to the atmels application note, which could be found here - www.atmel.com/Images/doc8278.pdf page 2
and I've followed the app note to the letter, as I want as low noise as possible to the analog part of the xmega circuitry. I also made the power line traces as wide and short as possible, without vias, so there should be minimum resistance there as well (the board is only 50x50mm).

the mcu is working ok, with or without voltage drop, as I have the brown out detector disabled by default, with a long startup time due to the inductor.

I don't have experience in doing the power supply side for mcus, so it would be great if you guys could share the best practises for the future (I've read about MOVs etc, but they seem to puzzle me).

Do you have a scope to probe your power rail and see if its clean or relatively noisy and the average is coming out to 2.85v?

Are you powering just the MCU from the 10uH inductor? or the LCD too? If its just the MCU it should be fine following the app-note. The display could have a boost converter for the LEDs (if they are all put in series in a string or a few series/parallel strings, as 3.3v is a tad low for white LEDs) Which obviously draws bursts of current, and with the series inductor,  your voltage could be swinging back and forth from 3.3v to something lower, with the average being 2.85v.

[V_King]

with, or without load, the voltage from usb is 5.06-5.08V. I also tested with three different usb cables (one used for high power external hdd).

the voltage drop happens across the voltage regulator. when I disconnect tft pannel, I see 3.3V and when connected, it drops to 2.85V.

I measured the current consumption of the tft panel with my dmm. not the most accurate way to measure, but seem to match with the datasheet info.

I've put the indurctor, due to the atmels application note, which could be found here - www.atmel.com/Images/doc8278.pdf page 2
and I've followed the app note to the letter, as I want as low noise as possible to the analog part of the xmega circuitry. I also made the power line traces as wide and short as possible, without vias, so there should be minimum resistance there as well (the board is only 50x50mm).

the mcu is working ok, with or without voltage drop, as I have the brown out detector disabled by default, with a long startup time due to the inductor.

I don't have experience in doing the power supply side for mcus, so it would be great if you guys could share the best practises for the future (I've read about MOVs etc, but they seem to puzzle me).

Do you have a scope to probe your power rail and see if its clean or relatively noisy and the average is coming out to 2.85v?

Are you powering just the MCU from the 10uH inductor? or the LCD too? If its just the MCU it should be fine following the app-note. The display could have a boost converter for the LEDs (if they are all put in series in a string or a few series/parallel strings, as 3.3v is a tad low for white LEDs) Which obviously draws bursts of current, and with the series inductor,  your voltage could be swinging back and forth from 3.3v to something lower, with the average being 2.85v.

well, you are probably right as got the problem solved. don't know why, but after bypassing the inductor, the voltage went back to 3.3V and even the back light got brighter (the screen was a bit dim). I was powering everything through the inductor. Putting inductor to filter out digital noise sort of make sense, but why I had such result I have no idea 

Anyways, will have access to the oscilloscope next week, so will hook the power supply to see what is happening.

Can you dig out the datasheet for your Schottky diode? 0.4V sounds unlikely for a 1A current through a 1A (max) device. Look for the Vf vs If graph...
(For instance, Farnell sell the Avago HSMS-280x   as a 1A device - but that's the 1uS pulse abs max number. At 1A, Vf is over 0.9V, and not even on the graph!)

I bypassed the diode this time, while looking for the problem. the datasheet is here http://www.farnell.com/datasheets/461158.pdf
It seem that it might be my mistake in selection as the Vf / If graph is only in mV / mA 

Also, can you confirm the LDO isn't getting hot enough to trigger its protection stuff? Do you get a few seconds of running before it fails?

the board was running ok, even with the low supply voltage and neither LDO, nor inductor was getting hot. LDO was getting hot when I supplied external 10V supply, but I did not hold it for long.

thanks everybody for replies