LDO RF behavior

[E_Blue]

Hello, I have a PIC18F67J50 at 3.3V through a TC2186 (150mA LDO) and the PIC is connected directly to the TX/RX port of a SIM928A(Quad band GSM module with GPS).
All the circuit is powered by a Li-Ion cell.

The battery is connected directly to the GSM module through a P Mosfet that the PIC uses to shutdown the GSM module and reduce power consumption.

I'm having some issues with the LDO, that only supply the PIC and some pullups, I'm getting low voltage pulses(2V) of about 500µS every 4mS when the GSM module is registering or sending a GPRS packet.

Thats low voltage pulses triggers the µC BrownOut circuit.

The PCB have four layers, all with GND pour.
The TC2186 have 2.2µF tantalum at input and output.

I changed the LDO by a TS5205 and has been running for 20 hours without a reset; but I need the ERROR flag of TC2186.

I'm attaching a capture at the output of TC2186; at the input the voltage variation on GSM registering is about 200mV.
The voltage at TC2186 input at that moment was 3.85V.

I put 2.2uF tantalum close to the input and output of TC2186 and 10uF tantalum at PIC internal LDO output.

The effect is worse if I connect the USB port to a PC.

The power consumption is about 20 to 25mA, so is very away from the 150mA limit of TC2186.

What can I do to improve the RF behavior of TC2186?

P1 is a 500mA picofusible



_________________
Electric Blue

[temtronic]

Sounds like you have insufficient power 'reserve'. Your battery can't supply the power( current), so voltage goes down. Classic.....
Options include bigger battery(cap unknown now) and adding bigger filter capacitor. I hate tantalum caps( they tend to self destruct and cause fires...) and 2.2mfd ain't enough anyway. I'd put in 100 mfd minimum, run some tests. Maybe a 'super cap' can be used as well.
Power from USB ports isn't exactly clean and steady, so additional filtering is needed.
It'd be nice to know what the GSM unit draws then build power supply for 3X that amount.

Jay

[E_Blue]

The battery is not the problem, I have the same circuit but with two more inputs, one more output and a TC1047 temperature sensor and works ok.

Also I put the oscilloscope at TC2186 input and get a maximum of 150mV to 200mV pulses when the GSM is transmitting.

3.85-0.2=3.65V and I think that's enough voltage to supply the LDO

Anyway, the voltaje at TC2186 never goes sow low to get 2V at the TC2186 output.

I have a DC-DC power supply with a LM2576(4.3V 3A) that supplies this board at VMIX an have 2x330uF electrolytic low ESR caps and make no difference if connected to this board or not.

I even put a 1000uFx10V electrolytic capacitor at GSM VDD module and make no difference either.

I run out of ideas, that's why I'm here now.
_________________
Electric Blue

[asmboy]

can you use a second regulator - JUST for the GSM module - not shared with the pic ??

[E_Blue]

The TC2186 is not shared, only supplies the µC and some pull-ups.

The TC2186 can't supply the GSM that module have current pulses of 2A for about 500µS every 4.5mS when transmitting and the Tc2186 have a limit of 150mA.
_________________
Electric Blue

[temtronic]

hmm reread your OP.
As the TS5205 worked fine, might be worth the effort to look at it with the scope and see what happens. Obviously if that's the only change there is 'something' different between the two LDO regs.
I saw in the uChp doc that they rate everything at Vout+1V, wonder if TI does the same?It was one of those 'notes' that caught my eye. Maybe nothing but.....

Perhaps the addition of a small choke and cap ( L-C ) filter to feed the LDO reg for the PIC will help ? In the past I've used C-R-C-R-C filters to 'clean up' feeds to analog sections.

You're now into the 'books don't help - test on the bench' phase. Write options down,test, eliminate them one by one.

It does look like the GSM is 'starving' the PIC for power though.

Jay

[asmboy]

Ok then this circuit change:

Between testpoint Vmix and C8 , cut trace and install a BAT54 or other schottky diode cathode on the right side.

Change c8 to a 100uf / 6 or 10v D- Tantalum
and make the output cap on the regulator be 100uf too.

When the supply dips below the new diode Vf
the bigger capacitor will better carry the droop.

[Ttelmah]

As a completely 'off the wall' suggestion, try reducing the pull-up on the shutdown input. As it stands, depending on the wire involved, and length of connection to this pin, if the RF induces some voltage on this, the regulator could simply by shutting down, it then recovers quickly when this goes high again. Testing what is actually on this pin at the moment when the problem occurs, might be worth doing. However, since there is no disadvantage to using a lower resistor, this is a simple thing to try.