TPS5410 Startup problem

Hi John,

Below is my schematic.  Is this not the same as Datasheet circuit page 10,  with the exception of the feedback voltage setting resistors?  Is there a problem with trimming it for 17V instead of 12V?

L908 is an EMI filter,  essentially 0 ohm below 1MHZ.

Successful Startup

Failed Startup

  Blue = current through inductor at startup (varies,  each time lightly different.  I don't have a current probe so this is with a 0.68ohm shunt which results in always successful startup.  So positive peak represents approx 150mA

Thanks

Hmm,  no sure how this image paste works here , sorry,  here is another attempt to past the images

current

good startup

Failed Startup

You may have to use the insert images or insert file tool rather than cut'n'paste...

I think I solved the issue.  It seems that the 1.2A current limit is triggered during startup when the regulator tries to charge the 100uF ouput EL cap.  Maybe because I am using 17V output the current spikes are higher than in typical applications?  I'm using a Panasonic EEE series cap which doesn't specify what the ESR is so I can't be sure.  I'm measuring 1A current spikes across a 0.1 ohm series R.  Maybe the inductor is saturating?

The solution seems to be to insert 0.5 ohm series resistance before the 100uF output cap.  Current spikes stay below 0.25A and regulator starts fine.

That sounds very likely.  There is some inrush current required at the start of the voltage ramp up.  After that the current required is constant I = C * Vout / Tss.

If you post waveforms, I can verify your assumption.

Hi Steve

I've had the exact same problem. I have 12-24V in and 5V out. In a new design I needed some more current so I added some capacitors. If I have one 330uF Aluminium Electrolytic it works fine. If I use 2x330uF it sometimes doesn’t start. I also thought I saturated my inductor (68uF) due to start up current, so I increased the size of the inductor but that did not help.

I don’t like the idea of a resistor in series. My design draws current very uneven, so every time the current consumption goes up a resistor will give a voltage drop.

I wonder if a stronger version s.a. TPS5420 or TPS5430 can handle higher inrush current. That could be a solution.

It's probably not the inductor.  Must inductors have a relatively soft saturation characteristic, so the inductance does decrease to rapidly above the specified saturation current.  More likely you are triggering the internal current limit protection of the IC.  The IC must provide the charge current for the capacitors (I chg = Cout * Vout / Tss) plus the load current during start up.  There may also be an initial spike in teh current when voltage is initially applied to the output capacitors.

Hi John

I'm seeing almost the same thing. During start-up the regulator goes into hiccup short circuit protection mode and never comes up. (Judged from the interval of around 15ms between start-up attempts). Also when I reduce the output capacitance I am able to add additional load (resistor) on the output before hiccup occurs.

However the output current stays under 1A, and the minimum current limit according to the datasheet is 1.2A, so how is it possible for the regulator to enter hiccup mode? The datasheet is very vague on this.

Also I'm considering changing to the TPS5420. But that is only a solution, if entering the hiccup mode happens after pulse-by-pulse current limiting has started. Is this the case?

Can you give a reference describing current limiting and hiccup mode functionality in more detail?

Thx.

The internal comparator measures PEAK current, so you would nee to monitor the switch (or inductor current) waveform rather than the load current.  The TPS5410 series uses pulse by pulse peak current detect.  There is about 100 nsec of blanking time during which the current limit is not monitored.  After that time if the peak current exceeds the current limit, the high side on time is terminated.  If this occurs for 3 consecutive switching cycles, the device enters hiccup mode.  For soft over current conditions, on the first cycle that reaches peak current limit detect, the duty cycle is shortened.  Ther is a longer "off" time durinfg which the swich current is ramping down.  So now the valley of the switch current waveform is lower and on the next cycle the peak current may not reach the limit during the on time so the hiccup mode timer is reset.  You may go thru many of these periods where one or two cycles reach peak current limit, then one or do do not, and hiccup mode is never entered.  I may have additional PPT somewhere, but it has been several years since I worked on those devices.

John,

I rented a 4054, current probe and diff probe and measured the switch voltage, inductor current and output voltage during startup. the 4054 has 10M memory depth so you can measure 20ms startup at high sample rate to zoom in and see details. My waveforms show something wrong with soft starty behavior of TOS5410. By your formula average current during startup for 17V over 8ms should be .2A. What I see with TPS5410 is a lag where switching does not start as Vref begins ramping so there is a spike when switching commences to charge cap up in initial step. Then charge current oscilates up and down during soft start period with peaks hitting current limit threshold and sometimes triggering hiccup. I lost the wavefomrs for 17V version but have them for 12V version. I do not see this behavior with TPS 5420 or TPS5450

here is zoom around peak at .8ms from start

Please post your schematic and some waveforms that show your issue.

hi, John,

already create another topic: LM2679 start up issue relate to current limit

https://e2e.ti.com/support/power_management/non-isolated_dcdc/f/196/t/588972

In the early discussion, you mention the load is just some op amps. If these op amps also have a negative power supply rail connected to them then probably the negative rail is starting slightly sooner than the positive. This can pull the positive supply output slightly negative and it will often fail to start. Could this be the situation? Adding a reverse biased Schottky diode across Vout will resolve this issue. This will clamp the negative excursion and allow reliable startup.

hi, Alan,

please looking at the new creat topic about LM2679 start up.

e2e.ti.com/.../588972