STM32f7 - ripples on 5.0V supply

[reza]

Posted on September 18, 2017 at 18:20

I have noticed that STM32F7 microcontroller generates a digital noise (see the figures. It is not from transmission lines and bursts are repeating every 16ms) on the 5.0V supply voltage. 

I have observed this using both Discovery and Nucleo boards.

I have seen this using external power supply and the power over USB.

I have also tried different sample codes. It actually does not matter what code you run, because even before I start the code in my debugger, the ripples can be observed. The only time, I do not see the ripples, is when the microcontroller is kept in reset, or when I remove a jumper to power it down.

I have also noticed that, after couple of resets, the noise sometimes disappear. It comes back if I reset again.

Anyone else who has observed this issue?  Any thoughts?

#ripples #noise #stm32f7 #power-supply

[LMI2]

Posted on September 18, 2017 at 19:21

There is something else running in your board. Some HW counters for instance. Have you measured the 3.3V power supply of the CPU. That is critical. If you mean by 5V the supply from USB, it could be anything. I think minimum is 4.4V or so and cheap low quality cables can make it even worse.

[reza]

Posted on September 18, 2017 at 20:26

First, thank you for your response

Michael.Lei

‌.

The 3.3V is generated from the 5.0V on these boards, by means of a voltage regulator. So the 3.3V also suffers from the ripples.

As I mentioned in my first post, I have observed this, when using USB supply, external independent supply, and even a battery. I have also tried normal and good quality cables. Nothing helps.

It is moreplausible that a component on theboard or the microcontoller itselfgenerating those digital pulses. I have not been able to figure out the source yet.

[Manojkumar Subramaniam]

Posted on November 26, 2017 at 20:17

is the same with turning off most peripherals ?

[LMI2]

Posted on November 26, 2017 at 20:56

Digital circuits make noise in their power supplies, because their current consumption is not smooth but varies in time.

[T J]

Posted on November 28, 2017 at 03:37

I checked my homemade board, I have a 7805 (K7805-1000R3) style switcher on board showing 400mV spikes at 100KHz.

I have two 2A ferrites in series with those 2 power lines. (5V and Gnd)

I put caps everywhere I can, usually a 1uF and a 0.1uF where requested by the datasheets. the bigger the better. 0805's are easy to work with.

Under reset, the noise is from the power supply, 400mV spikes before and a neat 100mV after the ferrites.

release the reset and run my canbus/ethernet thing, 400mV noise before and a messy 100mV after the ferrites.

I drop the rail from 5V down through two diodes, approx 1.1V drop just to remove the heat from the linear reg.

I run the linear 3V reg from those diodes.  AP2114H-3.3TRG1

Noise on processor pins:

actually, if I put the scope leads to either side on any cap on the processor, it is approx 10mV noise in reset or not.

There are no repetitive spikes on the processor pins..above 20mV at any frequency running or not.

EDIT:  I don't have DRAM running on my board. what is the refresh frequency ?

the DRAM specs say there is a noisy path to pay special attention to.

[AVI-crak]

Posted on November 28, 2017 at 23:08

Modern digital oscilloscopes can detect a signal with a frequency of 100 MHz or higher. But this does not mean the use of old measuring methods with grounding by a long wire. On this loop, you can catch an external radio signal, you can enhance the interference from the source, you can introduce significant distortion in the measured circuit. A long wire is always inductance. At frequencies above 100 MHz, this becomes noticeable.

Solution.

In the kit with the measured oscilloscope probe should be small springs, which are worn on the metal part of the probe. It looks like a spring, it is dressed like a spring, but it works like a point of contact of the earth. Having very small dimensions, and very close ground contact - the minimum contact inductance is obtained. External interference disappears. You measure a useful signal in differential mode. A very long wire of the probe itself will play the role of a large inductor (for such a frequency). As a result, the useful signal and the ground point will have synchronous oscillations. And on the oscilloscope connector will come a signal shot at the point of contact.

[T J]

Posted on November 29, 2017 at 00:00

SDRAM specs

Alliance :- AS4C8M16SA

128M - 8M x 16 bit Synchronous DRAM (SDRAM)

ï‚· Auto Refresh and Self Refresh

ï‚· 4096 refresh cycles/64ms

Micron:

• PC100- and PC133-compliant

• Auto precharge, includes concurrent auto precharge

and auto refresh modes

• Self refresh mode (not available on AT devices)

• Auto refresh

– 64ms, 8192-cycle (commercial and industrial)

– 16ms, 8192-cycle (automotive)

• LVTTL-compatible inputs and outputs

• Single +3.3V ±0.3V power supply

there is your 16mS noise...

[Eugene Solo]

Posted on November 29, 2017 at 02:03

looks like bad decoupling on memory chips