Author Topic: Help with choosing diode for boost converter  (Read 237 times)

0 Members and 1 Guest are viewing this topic.

Offline fabiodl

  • Regular Contributor
  • *
  • Posts: 94
Help with choosing diode for boost converter
« on: May 23, 2020, 11:22:57 am »
I am building a boost converter 5V to 15.5V using an mt3608.
As a reference, I use this commercial module.
Figures refOpen and refLoad show its output (the oscilloscope is ac coupled) when it is open and when there's a 103mA load (a 150 ohm resistor used as a load).

To be sure it is not due to layout differences, I stripped the parts out of a module and soldered my own parts on it.
When only the inductor and the capacitors are replaced (with   this inductor SRR1050A-220Y
and  these caps C3216X5R1V226M160AC) I get the outputs in figures clOpen and clLoad for open output and 150 ohm load, respectively. The ripple when the output is open is slighly smaller, the one over load is slightly worse.

The problem rises when I replace the original diode (with marking SS34) with  this diode SSA34HE3_A/H. As shown in figures dOpen and dLoad, the output ripple reaches 4Vpp when under load.

I also tried  this diode SS34FA and an 1N5817, which I found in the datasheet for another boost converer (the mc34063ad) but they are even worse.

With what criteria should I choose the diode? I thought I just needed something with low Vf, enough high reverse V and enough current capability, but I am clearly missing something
 

Offline trobbins

  • Regular Contributor
  • *
  • Posts: 162
  • Country: au
Re: Help with choosing diode for boost converter
« Reply #1 on: May 23, 2020, 11:34:44 am »
You don't provide info on what the C and L were, and why you chose the alternatives.   Also no info on how and where you probed for the waveforms.
 

Offline fabiodl

  • Regular Contributor
  • *
  • Posts: 94
Re: Help with choosing diode for boost converter
« Reply #2 on: May 23, 2020, 11:36:39 am »
The original C and L were 10uf and 22uH, respectively.
I probed the waveforms at the module's output.
 

Offline nctnico

  • Super Contributor
  • ***
  • Posts: 19404
  • Country: nl
    • NCT Developments
Re: Help with choosing diode for boost converter
« Reply #3 on: May 23, 2020, 11:46:59 am »
How are you probing? Likely you are picking up the switching noise which does depend on the speed and recovery parameters of the diode. A soft-recovery diode combined with an RC snubber can reduce the noise a lot. However you have to realise that the switching noise is picked up by your probe through radiated emissions; the load doesn't see it.
There are small lies, big lies and then there is what is on the screen of your oscilloscope.
 

Offline Siwastaja

  • Super Contributor
  • ***
  • Posts: 2514
  • Country: fi
Re: Help with choosing diode for boost converter
« Reply #4 on: May 23, 2020, 12:21:31 pm »
Reverse recovery time (and possibly sharpness, if it is of a type which has reverse recovery time) is absolutely critical. Think about it: if the diode still conducts for some time when the MOSFET has turned on, the diode and MOSFET form a total short circuit over the low-impedance capacitance you have at the output!

Hence, a schottky type tends to be the only sensible choice; traditional Schottky for low voltages, SiC for high voltages. Schottky diodes have no reverse recovery time (although they do have junction capacitance which causes a similar switching loss element, but this usually isn't a huge problem; by all means calculate it to be pedantic).

Before SiC diodes, we were forced to balance  between minimal reverse recovery time and softness factor, and possibly add RC snubbers to dampen the oscillation caused by the short circuit current charging layout parasitic L; or just slow down the MOSFET turn-on time. All such measures caused quite some heavy switching loss. Now with SiC diodes available, with zero reverse recovery time, such practices are usually not needed at all.

Of course, with your 15.5V output voltage, SiC is not needed, a standard schottky (for switching applications!) does better, and such parts have been available for several decades now.

But keep your eye on the reverse leakage spec of the schottky diodes. Lowest Vf types, looking appealing at first sight, may have very high off-time leakage, possibly causing thermal runaway. Likely not a huge problem at Vr=15.5V, unless you run very close to maximum ratings or fail at cooling the devices.

Don't fall into the trap of wondering where the reverse recovery times are listed. Ignore the parameter in parametric searches and don't look for it in the datasheets: it's not listed because schottky diodes do not have this thing at all. (Took me a year to figure this out...)
« Last Edit: May 23, 2020, 12:26:31 pm by Siwastaja »
 
The following users thanked this post: fabiodl

Offline fabiodl

  • Regular Contributor
  • *
  • Posts: 94
Re: Help with choosing diode for boost converter
« Reply #5 on: May 23, 2020, 03:18:52 pm »
How are you probing? Likely you are picking up the switching noise which does depend on the speed and recovery parameters of the diode. A soft-recovery diode combined with an RC snubber can reduce the noise a lot. However you have to realise that the switching noise is picked up by your probe through radiated emissions; the load doesn't see it.

The reason I started investigating this is that this voltage spikes seem sufficient to kill other ICs on the board.
Specifically, I use this boost converter to generate voltages for programming a PEEL18CV8.
While using the module worked fine, using the alternative diode already killed two of them.

I probe it with normal x10 oscilloscope probes
 

Offline fabiodl

  • Regular Contributor
  • *
  • Posts: 94
Re: Help with choosing diode for boost converter
« Reply #6 on: May 23, 2020, 03:35:37 pm »
Reverse recovery time (and possibly sharpness, if it is of a type which has reverse recovery time) is absolutely critical. Think about it: if the diode still conducts for some time when the MOSFET has turned on, the diode and MOSFET form a total short circuit over the low-impedance capacitance you have at the output!

Hence, a schottky type tends to be the only sensible choice; traditional Schottky for low voltages, SiC for high voltages. Schottky diodes have no reverse recovery time (although they do have junction capacitance which causes a similar switching loss element, but this usually isn't a huge problem; by all means calculate it to be pedantic).

Before SiC diodes, we were forced to balance  between minimal reverse recovery time and softness factor, and possibly add RC snubbers to dampen the oscillation caused by the short circuit current charging layout parasitic L; or just slow down the MOSFET turn-on time. All such measures caused quite some heavy switching loss. Now with SiC diodes available, with zero reverse recovery time, such practices are usually not needed at all.

Of course, with your 15.5V output voltage, SiC is not needed, a standard schottky (for switching applications!) does better, and such parts have been available for several decades now.

But keep your eye on the reverse leakage spec of the schottky diodes. Lowest Vf types, looking appealing at first sight, may have very high off-time leakage, possibly causing thermal runaway. Likely not a huge problem at Vr=15.5V, unless you run very close to maximum ratings or fail at cooling the devices.

Don't fall into the trap of wondering where the reverse recovery times are listed. Ignore the parameter in parametric searches and don't look for it in the datasheets: it's not listed because schottky diodes do not have this thing at all. (Took me a year to figure this out...)

Thank you for this throughful answer, I even didn't know of the existance if SiC diodes, and the concept that a diode still conducting when the mosfet is on will discharge the cap makes a lot of sense!

I checked the datasheets of the diodes I used, and the SSA34 reports "TYPICAL APPLICATIONS For    use    in    low    voltage    high    frequency    inverters,    freewheeling,  DC/DC  converters,  and  polarity  protection  applications.", but still in practice it does not work  well with the mt3608. Do you have any suggestion of a particular diode I could try?
As for the snubber. do you mean a RC series from the inductor to ground as shown here http://www.ti.com/lit/an/slva255/slva255.pdf ?

I was thinking about adding a RC low pass filter after the ouput cap. I know it would be power wasted on the R all the time, but, as I wrote in the previous post, since this is killing other ICs, a decrease in efficiency is still the least painful choice.
 

Online uer166

  • Regular Contributor
  • *
  • Posts: 140
  • Country: us
Re: Help with choosing diode for boost converter
« Reply #7 on: May 23, 2020, 10:19:21 pm »
I probe it with normal x10 oscilloscope probes

There's the problem most likely, try using one of these:
 

Offline nctnico

  • Super Contributor
  • ***
  • Posts: 19404
  • Country: nl
    • NCT Developments
Re: Help with choosing diode for boost converter
« Reply #8 on: May 23, 2020, 10:39:15 pm »
Don't fall into the trap of wondering where the reverse recovery times are listed. Ignore the parameter in parametric searches and don't look for it in the datasheets: it's not listed because schottky diodes do not have this thing at all. (Took me a year to figure this out...)
Still schottky diodes do have some capacitance which needs to be discharged. Some of the bigger ones may have parasitic PN diodes which could become forward biased and exhibit reverse recovery. All in all the schottky diodes are less ideal then you might expect.
There are small lies, big lies and then there is what is on the screen of your oscilloscope.
 

Offline trobbins

  • Regular Contributor
  • *
  • Posts: 162
  • Country: au
Re: Help with choosing diode for boost converter
« Reply #9 on: May 24, 2020, 12:00:00 am »
The link to the new part is to a 22uF cap, and you indicate the original part used was a 10uF cap.  Were there no 10uF caps available on digikey, or did you you think that 22uF would be better and have no consequences?

You need to set a specific smt cap part to probe across when taking comparison measurements, and use the probe technique shown in post #7, and appreciate how your probe and scope's frequency response will influence what you are observing.  If the main 10uF cap is not very close to the output terminals, or you want to try and confirm what is presented at the output terminals, then you may need to add in an smt cap directly between the output terminals, as a form of LISN point to do comparative noise measurements.  Some smps manufacturers present such details as to how their spec noise levels are (and should be) measured.
 

Offline jonroger

  • Regular Contributor
  • *
  • Posts: 66
  • Country: us
Re: Help with choosing diode for boost converter
« Reply #10 on: May 24, 2020, 12:26:34 am »
To select parts for a DC-DC converter, I highly recommend that you use TI's WEBENCH - even if you aren't using one of their controllers.
I am available for custom hardware/firmware development.
 

Offline fabiodl

  • Regular Contributor
  • *
  • Posts: 94
Re: Help with choosing diode for boost converter
« Reply #11 on: May 24, 2020, 03:20:02 am »
I probe it with normal x10 oscilloscope probes

There's the problem most likely, try using one of these:
Thank you, I will try to improve the measurement setup. Is there a commercial name or did you make it yourself?
However, it does not seem to be only a measurement error. As I reported in #5, the spikes are actually destroying ICs on the board (when not measuring).
« Last Edit: May 24, 2020, 03:32:30 am by fabiodl »
 

Offline fabiodl

  • Regular Contributor
  • *
  • Posts: 94
Re: Help with choosing diode for boost converter
« Reply #12 on: May 24, 2020, 03:21:54 am »
The link to the new part is to a 22uF cap, and you indicate the original part used was a 10uF cap.  Were there no 10uF caps available on digikey, or did you you think that 22uF would be better and have no consequences?

The datasheet for the mt3608 reports 22uF as a suggested value, so my guess was the module used 10uF just because they were cheaper. Indeed, 22uF with the original diode works reasonably well, and with open load, 22uF caps provide a reduction in the ripple over the 10uF caps.
« Last Edit: May 24, 2020, 03:23:42 am by fabiodl »
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf