Problem with working on 2.5V

[Artem Dmitriev]

Posted on March 01, 2018 at 11:10

Hello!

On the PCB from LDO(

MCP1700T-3302E

) 3.3V are supplied 2 chips:

GPS

:

SIM

68

M,

STM32L475RET6.

GPS power supply is controlled by STM.

When STM turn on GPS module voltage temporarily drops to 2,5V.

At this time MCU works incorrectly if frequency is 80MHz (hang or HardFault handler are called). If frequency 40Mhz all fine.

What could be the problem?

Note: this post was migrated and contained many threaded conversations, some content may be missing.

[Andrew Neil]

Posted on March 05, 2018 at 11:51

dvorak.peter wrote:

The better solution is the R C modification to the VT9 switch mentioned earlier.

I disagree: that's just masking the symptoms - not curing the fault.

The fault is that the power supply is inadequate - that needs to be fixed.

Giving a slow supply rise to the GPS may create more problems - as the GPS may not be designed to cope with that...

[Dvorak.Peter]

Posted on March 05, 2018 at 15:35

The power supply is just fine.

The problem is that the low resistance of the VT9 switch is capable of creating > 10 Ampere surge in charging

 C29 (

 4.7 uF) 

capacitor from 3.3V source.

I would design the switch with 200us rise time.

An alternate solution is to move C29 to the un-switched side. The VT9 resistance is will not degrade the filtering function. (However  the GPS module may still have significant internal decoupling capacitance to cause problems).

Peter

[AvaTar]

Posted on March 05, 2018 at 15:51

The problem is that the low resistance of the VT9 switch is capable of creating > 10 Ampere surge in charging

 C29 (

 4.7 uF) 

capacitor from 3.3V source.

Not sure if the series resistance of C29 is that low.

And if 4.7uF is necessary at all - perhaps a smaller value could be used.

An alternate solution is to move C29 to the un-switched side.

I would anyway suggest to buffer Vcc for the MCU with an appropriate cap.

Just measure the duration of the drop caused by the GPS module switch, and add a capacitor of size I * t * safety_factor.

'I' is max. expected current of the MCU, 't' the measured time, and 'safety_factor' - just that.

Or do a full-blown SPICE simulation ...

By the way, TTL electronics in the '70 and '80 used to contain ceramic capacitors between Vcc and GND, directly at the chip, to buffer Vcc during such surges. Electrolytic caps are too slow.

[Dvorak.Peter]

Posted on March 05, 2018 at 22:59

AvaTar wrote:

The problem is that the low resistance of the VT9 switch is capable of creating > 10 Ampere surge in charging

 C29 (

 4.7 uF) 

capacitor from 3.3V source.

Not sure if the series resistance of C29 is that low.

And if 4.7uF is necessary at all - perhaps a smaller value could be used.

 The ESR of a typical 4.7uF 6.3V  decoupling capacitor (MLC) is below 0.1 ohms.

I suspect that these $0.02 USD parts are generously used inside the GPS module.

Peter 

[Artem Dmitriev]

Posted on March 06, 2018 at 07:03

We tried to solder L475re on nucleo board. On  the board we see the same situation if we connect GPS module.

Also we tried to connect GPS module to

32

L

476

GDISCOVERY board. In this case all work fine.

We added additional capacitor near gate and this helped to improve step response, MCU works. But the questions is why MCU works incorrectly om 2.5V if it must work down to 1.8V?

[AvaTar]

Posted on March 06, 2018 at 07:19

But the questions is why MCU works incorrectly om 2.5V if it must work down to 1.8V?

It would have helped if you observed Vcc with a scope in the critical case.

I'm pretty sure it dropped far below 1.8V for at least 50 microseconds in this case - a so-called brown-out.

And second, the minimum operating voltage often implies other restrictions (like reduced core clock), you would need to check for the L475 MCU.

[Artem Dmitriev]

Posted on March 06, 2018 at 07:24

I tried to find info in datasheet, but I don'd understand where this info is located. Didn't find it.

[AvaTar]

Posted on March 06, 2018 at 08:09

I don't know the L475, so I can't speak of specifics. Just my experience from other MCUs.

If you want to operate the MCU at 1.8V, I would suggest to check the data sheet and reference manual for conditions and constraints.

[Artem Dmitriev]

Posted on March 06, 2018 at 09:18

We don't see any spikes on oscillograph. 

MCU works from 1.8V with connected GPS module. But GPS working voltages are 2.8 �?´�?¾ 4.3 V.  On voltages up to 2.8V GPS module doesn't work, voltage drop exists, but MCU works fine. If we connect GPS using voltages 2.8 to 3.6 MCU works incorrectly. 

[AvaTar]

Posted on March 06, 2018 at 10:21

Honestly, I'm no hardware guy, I can only guess.

There may still be some issues with the power supply or the buffering, either with the MCU or the GPS module.

MCU works from 1.8V with connected GPS module. But GPS working voltages are 2.8 �?´�?¾ 4.3 V.  On voltages up to 2.8V GPS module doesn't work, voltage drop exists, but MCU works fine. If we connect GPS using voltages 2.8 to 3.6 MCU works incorrectly. 

Sounds to me like you have two choices.

Either have separate supplies (1.8V for the MCU, 2.8V .. 4.3V for GPS), wich might require level shifters for interface lines.

Or you can pick a voltage both units can operate with, i.e. 2.8V .. 3.6V.

You would need to investigate the power consumption for both cases, guessing you want to run from battery, and optimize for low power.

At this time MCU works incorrectly if frequency is 80MHz (hang or HardFault handler are called). If frequency 40Mhz all fine.

BTW, have you really ruled out a software issue ?

Perhaps a bug, or your Flash interface setup (waitstates ?) for 80MHz is incorrect.