Author Topic: MT3608 going wild  (Read 1308 times)

0 Members and 1 Guest are viewing this topic.

Offline fabiodlTopic starter

  • Frequent Contributor
  • **
  • Posts: 282
MT3608 going wild
« on: May 21, 2020, 12:33:11 am »
On a board I need 15.5V, that I generate from the USB bus.
In a previous revision I used a MT3608 module.
In this revision I integrated the chip on my board, and at a 40mA draw the noise is way worse than with the module (Vpp in the order over 1 V).
I used
this inductor    SRR1050A-220Y

this diode SS34FA

these caps C3216X5R1V226M160AC

and the layout is the one in the image.R23=82K, R24=3K3, RV1 is not installed.
Is there anything wrong in the components choice or the board layout?
 

Offline fabiodlTopic starter

  • Frequent Contributor
  • **
  • Posts: 282
Re: MT3608 going wild
« Reply #1 on: May 21, 2020, 05:29:16 am »
Here is a pic of the oscilloscope trace (infinite persistence enabled)
I also tried moving R23 next to the chip (I soldered it between R24 and C13) and it made no difference whatsoever
« Last Edit: May 21, 2020, 07:29:49 am by fabiodl »
 

Offline awallin

  • Frequent Contributor
  • **
  • Posts: 694
Re: MT3608 going wild
« Reply #2 on: May 21, 2020, 08:11:14 am »
without persistence-mode, is that a switching-glitch at ~1.1 MHz?
I guess that is expected and only goes away with more LC filtering on the output?
(you could try to probe the +5VDC input and see how that looks. more input capacitance might help?)
 
The following users thanked this post: fabiodl

Offline fabiodlTopic starter

  • Frequent Contributor
  • **
  • Posts: 282
Re: MT3608 going wild
« Reply #3 on: May 21, 2020, 09:00:26 am »
awallin, thank you, I suspected that too ( the padauk free programmer seems to have a 47 uF output cap) so I checked the module I have.
It is even less. Both input and output are 7uF (maybe that's a low 10uF?)
I started replacing my parts with the module ones, in the following order

output cap-> no change
diode -> no change
inductor-> bingo!
now I will try to restore my output cap and diode and more importantly I need to understand why mine was unsuited and
 

Offline fabiodlTopic starter

  • Frequent Contributor
  • **
  • Posts: 282
Re: MT3608 going wild
« Reply #4 on: May 21, 2020, 09:02:07 am »
Check your inductor or IC soldering. Some boost converters are so amazing that they might be able to work from just stray inductance when the proper inductor is accidentally shorted, just with much higher noise and quiescent current. I had that mistake done on TPS61170.

I will test the inductor and try to resolder it, thanks for the tip!
edit: the inductor measures 20uH, 0.4ohm
« Last Edit: May 21, 2020, 09:09:15 am by fabiodl »
 

Offline fabiodlTopic starter

  • Frequent Contributor
  • **
  • Posts: 282
Re: MT3608 going wild
« Reply #5 on: May 21, 2020, 10:22:44 am »
it seems it was a combination of poorly soldered inductor and diode choice.
I have no idea on what diode would be ok then.

Here is the complete set of test I took
my cap, my diode, my inductor (initial soldering)= bad
their cap, my diode, my inductor (initial soldering) = bad
their cap, their diode, my inductor (initial soldering) = bad
their cap, their diode, their inductor  = good
my cap, their diode, their inductor  = good
my cap, my diode, their inductor  = bad
my cap, new diode SSA34HE3_A/H   , their inductor = bad
my cap, their diode, their inducutor = good
my cap, their diode, my inductor (resoldered) = good,but theirs may be slighrly better

Why is it not working with the SS34FA or the SSA34HE3_A/H but their diode (with silkscreen SS34) does?
updates:
with a 1N1819 it really goes mad, Vpp is 2V
adding a 100n cap to the output makes no difference
adding a 100u cap on the top of c13 makes no difference, but adding it closer to the load (where the two pads are) makes even the SS34FA work reasonably  well



« Last Edit: May 21, 2020, 01:57:17 pm by fabiodl »
 

Offline Doctorandus_P

  • Super Contributor
  • ***
  • Posts: 4004
  • Country: nl
Re: MT3608 going wild
« Reply #6 on: May 23, 2020, 11:51:01 am »
I assume this is a 4 layer PCB with a full GND layer which is turned of for viewing.
Correct?

Analog has a nice writeup about the "hot loop" of a SMPS circuit, and the loop area of this loop has to be minimized. The "hot loop" is the loop in the circuit that has rapid current changes and generally is the (mosfet) switch, diode, and filter capacitor. The inductor does not have high current transients and placement is much less critical.

https://www.analog.com/en/technical-articles/layout-for-power-designs-1-hot-loops.html#

Is C13 in the right top corner?
Then put it closer to the IC.

Oops...
C13 seems to be between U4 and D8 which looks quite optimal.
What I do not understand is that you seem to have a lot of unconnected vias (layers turned off), but SMT pads for the GND net also do not seem to go anywhere.
« Last Edit: May 23, 2020, 11:59:50 am by Doctorandus_P »
 

Offline fabiodlTopic starter

  • Frequent Contributor
  • **
  • Posts: 282
Re: MT3608 going wild
« Reply #7 on: May 23, 2020, 03:12:38 pm »
This is a 2 layer board, with gnd copper fill. I hid it to make the bottom traces visible.
Thank you for the link, indeed C13 is the closest part to the chip
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf