Author Topic: Is anything inherently wrong with thi supply circuit for stm32f103rct6?  (Read 1082 times)

0 Members and 2 Guests are viewing this topic.

Offline teslinius

  • Regular Contributor
  • *
  • Posts: 57
  • Country: nl
Does anything jump out immediatelly?, or with the supply circuit, because i designed a pcb with stm32f103rct6 and the uc has fried 3 times already and i have no idea why, i powered it with lab bench power supply limited to 50mA and it short circuited, i have no clue what's the problem, all the supply circuit looks fine to me, also there aren't any visible problems on the pcb, this is driving me crazy, can't figure it out, all the reference schematics i watched had same supply circuit. Only thing i notice is the reference schematic is using 100k pullup/pulldown and i'm using 10k, but i don't think that's the problem. |O
 

Offline thm_w

  • Super Contributor
  • ***
  • Posts: 3925
  • Country: ca
  • Non-expert
What voltage did you supply with the lab power supply, to where?
Did you verify it with a multimeter?
After the uC blew, did you remove it from the PCB, and then power on the power supply and measure what voltage it receives?
Anything externally connected?
 

Offline bugnate

  • Contributor
  • Posts: 48
  • Country: us
Agree with previous poster than you need to provide a lot more debugging info if you are hoping for any help. That said, as far as "jumps out at me", and without hearing your side of things... You might reflect on Q11 being an NMOS part on the high-side. Likewise L3 and L4 may not be a good idea. I'd short all of them out and see what happens.
 

Offline DavidAlfa

  • Super Contributor
  • ***
  • Posts: 2823
  • Country: es
This looks very wrong, I've never seen anything like this before!


Where did you saw that or got the idea from?
Normally you add capacitors to the LDO output, put the series inductor and then more capacitors.
The LDO output must always have direct connection to a capacitor!
But like that, the output has a high impedance path when dealing with transients, there's no capacitor to properly filter the changes, but a coil that will block these fast current spikes, causing the ldo to do very weird things, or even charge up and oscillate, creating a LC tank circuit.

As they said, short them out and try again.
This is the usual way of using a filtering inductor:
« Last Edit: May 11, 2022, 11:05:31 pm by DavidAlfa »
Hantek DSO2x1x            Drive        FAQ          DON'T BUY HANTEK! (Aka HALF-MADE)
Stm32 Soldering FW      Forum      Github      Donate
 

Offline T3sl4co1l

  • Super Contributor
  • ***
  • Posts: 19607
  • Country: us
  • Expert, Analog Electronics, PCB Layout, EMC
    • Seven Transistor Labs
Only one side of the transformer is fused.

It's a 18VCT transformer, right?

Why ferrite bead supply the MCU?  Do you know you'll need extra filtering between, whatever else is on +3.3V?  Or filtering to, as the case may be?  It can be better without, as the supply is bypassed by everything connected to it.  This is more helpful on 4+ layer inner-plane PCB designs, but doesn't make much difference on 2-layer designs with only stitched ground fill (where +3.3V has to be routed as just any other trace, and so there isn't direct communication between bypass caps on the supply).

As above, large inductors on linear regulators, is just wrong.  Easy fix.

If you're going to the trouble of burning three separate MCUs, why not consider probing the supply voltage with an oscilloscope while you blow one of them?  Set it to rising trigger, 2.5V threshold, single shot acquisition, scales around 1ms/div and 1V/div, and see what's happening.  As thm said, this is probably better done without an MCU (and other chips for that matter) installed, so you aren't burning up your supply of parts...

Tim
Seven Transistor Labs, LLC
Electronic design, from concept to prototype.
Bringing a project to life?  Send me a message!
 

Online pcprogrammer

  • Super Contributor
  • ***
  • Posts: 1351
  • Country: nl
Another observation, why are you using AMS1117 regulators when you have a high enough voltage at the input for using no low drop voltage regulators?

Depending on the transformer you are using the 12V can be higher then the 15V maximum input voltage of the AMS1117. A 12V AC rated transformer will give ~17V DC when rectified.

With such a big voltage drop across the regulator it might dissipate to much heat for it to handle.

Using switching converters would be a better solution. For digital circuitry these are no problem and have less energy loss. Look into the LM2596 or other step down converters in the market.

Offline T3sl4co1l

  • Super Contributor
  • ***
  • Posts: 19607
  • Country: us
  • Expert, Analog Electronics, PCB Layout, EMC
    • Seven Transistor Labs
LDO vs. HDO isn't really a problem, though there are better performing / behaved options than ye olde '1117.  The unloaded or high-mains voltage is an excellent point though!

Tim
Seven Transistor Labs, LLC
Electronic design, from concept to prototype.
Bringing a project to life?  Send me a message!
 

Offline teslinius

  • Regular Contributor
  • *
  • Posts: 57
  • Country: nl
Oh so that's probably the problem, yeah i saw that circuit and thought that cap before the inductor is just for extra filtering, was kinda worried that it may create voltage spikes, but didn't find anything about it on the internet, so i let it be. Gonna remove the inductors and see if this fixes the problem. Also gonna measure for any voltage spikes on the oscilloscope.
 

Offline teslinius

  • Regular Contributor
  • *
  • Posts: 57
  • Country: nl
I used linear regulators because they are less noisy then switching, i have 16 bit adc and some opamps so i want the supply voltage to be clean.
 

Online pcprogrammer

  • Super Contributor
  • ***
  • Posts: 1351
  • Country: nl
I used linear regulators because they are less noisy then switching, i have 16 bit adc and some opamps so i want the supply voltage to be clean.

Sure then it is a good idea to use linear regulators, but also best to use separate ones for the analog part and use a separation between the digital and analog ground. Use separate areas on the PCB that have their own ground planes and use a pi filter with a ferrite bead and 1nf capacitors to connect the two. Also use different value capacitors like 1nF, 10nF and 100nF to decouple the power rails to make a clean as possible supply.

Offline mon2

  • Frequent Contributor
  • **
  • Posts: 341
  • Country: ca
With the industry wide shortage of semiconductors, are you using a true and 100% original stm32f103rct6 from ST?

Or is it one of the many 'clones'. Many are not PIN for PIN compatible.

Was it sourced from an authorized distributor or from offshore?

What is the full part number of the referenced LDO? These sot-223 packaged LDO offer 2 different pinouts. A normal and reversed pinout. See Diodes Inc. datasheets of similar devices. Do confirm which one you have onboard.

Who did the PCBA? Yourself or was this outsourced?
 

Online wraper

  • Supporter
  • ****
  • Posts: 14526
  • Country: lv
I used linear regulators because they are less noisy then switching, i have 16 bit adc and some opamps so i want the supply voltage to be clean.

Sure then it is a good idea to use linear regulators, but also best to use separate ones for the analog part and use a separation between the digital and analog ground. Use separate areas on the PCB that have their own ground planes and use a pi filter with a ferrite bead and 1nf capacitors to connect the two. Also use different value capacitors like 1nF, 10nF and 100nF to decouple the power rails to make a clean as possible supply.
Separating the ground plane generally is not a good idea. Especially if you don't know what you are doing and traces are crossing the ground split.
 
The following users thanked this post: Bassman59

Offline teslinius

  • Regular Contributor
  • *
  • Posts: 57
  • Country: nl
Hmm, not sure if it's 100% original, because i ordered it from aliexpress, but it's from a seller i trust, i ordered before from him and i allways got working chips. Also before i powered the circuit with the lab bench power supply i connected it to stlink and ran st-info --probe command and i got correct flash and ram values also uploaded test code to blink led and it worked. Yeah i designed the pcb as a hobby project.
 

Offline teslinius

  • Regular Contributor
  • *
  • Posts: 57
  • Country: nl
Thanks, gonna consider that on the next pcb design.
 

Offline Doctorandus_P

  • Super Contributor
  • ***
  • Posts: 2302
  • Country: nl
I'm also always very suspicious about LC combinations and tend to add a bit of series resistance to lower their Q and suppress oscillation tendencies.
 

Offline teslinius

  • Regular Contributor
  • *
  • Posts: 57
  • Country: nl
Removed the inductors and now everything works great, well i'll know not to repeat this mistake on the next design, this also explains abnormal behaviour of the mcu and why linear regulators went short circuit. Thanks a lot for the help everyone  :-+
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf