Practical Understanding of Reverse Recovery

I don't have a totally coherent understanding of reverse recovery and I need to.

Here's what I understand:

• RR happens when a diode is conducting forward and then gets reverse biased
• RR losses are roughly proportional to current for a given device
• SiC, GAN, and Schottky devices don't suffer from rr

So this should mean:

• RR should matter in any hard switched application
• In a buck or boost converter the diode will have rr losses (same for boost)
• In a synchronous buck or boost the synchronous fet will have rr losses (since right after its turn off the diode has to take up the current)
• Class D amplifiers are variations of synchronous buck and will therefore have rr losses
• Soft switched/resonant topologies still need to supply the rr charge but it shouldn't show up as a loss

So what I don't understand (with possible answers):

1. Why isn't rr more often mentioned in the context of standard low voltage buck converters? (because low voltage fets don't have large rr losses?).
2. Why in the context of totem pole PFCs is it declared that silicon mosfets can't be used when there should also be rr losses in the standard alternative boost converter? (because discrete diodes in the boost perform better than fet intrinsic diodes?).
3. Why are mosfets with fast recovery diodes such as Infineon CFD2 marketed for 'soft switching' or resonant applications when its hard switching applications that should suffer most from rr losses?

So again I'm just looking for rough rules of thumb here such as "below X-Y voltage rr losses probably don't matter" or "discrete diode are usually X-Y times better than most fet body diodes". Etc.

Minority carriers pushed into the device have a lifetime
and this is the time signature of reverse recovery. In a
"40V" device this can be microseconds, higher voltage
means lower doping means higher lifetime. The reverse
current*voltage*time is your energy slug per cycle that
goes to switching losses.

High voltage simple bucks used to use a plain diode or
perhaps a Schottky for the low side circulating path.
Sync bucks replace this with a commutated FET, but
this may or may not be a win for switching losses
depending on the FET and the incumbent diode.

GaN diodes and silicon Schottky diodes are majority
carrier devices. I've seen SiC BJT-like devices referred
to as "quasi-majority-carrier" but couldn't say whether
this is physics or marketing or a stew. Silicon Schottky
devices beyond maybe 60V will have a P-doped guard
ring to keep the Shottky junction's tiny radius of curvature
from causing premature breakdown. Downside is, if you
try to run too much current and get to silicon Vf, the
Schottky benefit of zero charge storage goes away and
the recovery may not be fast at all. Conversely you can
make silicon diodes faster by device design (carrier
recombination "plugs", material abuse to increase
recombination, etc.).

A diode without much minority carrier injection will
have little conductivity modulation and so will show
more conduction losses instead.

Soft switching is for sync buck or other commutated
low side topologies. A buck with a low side diode -is-
soft switched (inductor does the work). But a sync
buck's goal is to catch that fall right about zero, where
the diode will go negative before it conducts. And that
negative pedestal is conduction losses.

FET intrinsic diodes may or may not be engineered to
suit, depending on expected application and diligence.
Making the body "fast recovery" requires adding features
where you would rather not; adding to the drift region
again would steal face area from normal conduction.
In a cut-throat market you would not do this every
time necause it would kill pricing position; instead you
make it a "feature" for folks who will pay more for it.

What people declare regarding PFC and stuff, and why,
I have no ideas.

Before SiC diodes/mosfets and now GaN reverse recovery was (and still is) a real problem with any hard switched converter and fets/igbts used to be switched on more slowly to reduce the peak reverse current and assoc noise and interference to control...

Low voltage fets ( <40) have very fast intrinsic diodes and so can be used in HF buck converters...

the high dv/dt of modern switching devices is also a problem ...

asdf44 said:

Why are mosfets with fast recovery diodes such as Infineon CFD2 marketed for 'soft switching' or resonant applications when its hard switching applications that should suffer most from rr losses?

Click to expand...

Reverse Recovery time increases as reverse voltage applied to the diode decreases.
In some soft switching apps, e.g. ZVS apps (such as PSFB):
1) the body diode conducts before FET is ON, enabling low voltage across the FET prior to its turn ON. Current flows through body diode.
2) FET turned ON, diverting current from the body diode to the channel of the FET
3) External circuitry reverses the current flow to Drain-to-Source => applies small reverse voltage (Rds(on)*Ids) to the body diode
4) Body diode must be fully recovered before turning OFF the FET (and hence the FET will see across it BUS voltage i.e. higher voltage than the conduction voltage drop) because otherwise high reverse current might flow through the diode turning on the parasitic BJT and hence destruction of the FET.

In summary, low reverse recovery time diodes are needed in ZVS apps. The reason is not RR power loss, but FET destruction due to parasitic BJT turn ON.

CataM said:

Reverse Recovery time increases as reverse voltage applied to the diode decreases.
In some soft switching apps, e.g. ZVS apps (such as PSFB):
1) the body diode conducts before FET is ON, enabling low voltage across the FET prior to its turn ON. Current flows through body diode.
2) FET turned ON, diverting current from the body diode to the channel of the FET
3) External circuitry reverses the current flow to Drain-to-Source => applies small reverse voltage (Rds(on)*Ids) to the body diode
4) Body diode must be fully recovered before turning OFF the FET (and hence the FET will see across it BUS voltage i.e. higher voltage than the conduction voltage drop) because otherwise high reverse current might flow through the diode turning on the parasitic BJT and hence destruction of the FET.

In summary, low reverse recovery time diodes are needed in ZVS apps. The reason is not RR power loss, but FET destruction due to parasitic BJT turn ON.

Click to expand...

Ok, that's all well and good and I've seen this in literature but let me make sure I'm clear on the context.
(https://www.st.com/content/ccc/reso...df/jcr:content/translations/en.CD00171347.pdf)

The reason people bother to print "Hey watch out for RR in PSFB" is because you might not otherwise notice or anticipate this problem in a ZVS PSFB. On the other hand you wouldn't bother to print this warning in the context of a 100V Class D amplifier because that hard switches 100% of the time and you have zero chance of not noticing it?

Is that correct or am I missing a reason why this problem is actually worse in the soft switching example you gave versus a hard switched scenario.

But finally while it makes sense low RR helps in that ZVS scenario that doesn't answer for me why low RR parts are marketed towards soft switching when it seems they should also be better in any hard switched half bridge.

Actually low Trr are not needed in ZVS converters (low forward recovery time is useful) as the diodes in the fets are on as the fet turns on, as long as the converter STAYS in ZVS. Low Trr diodes ( and fets with low Trr diodes) are used in ZVS converters because oftentimes they operate out of ZVS (transient load steps) and the use of faster parts helps prevent a blow up as this is when the internal BJT can be triggered ( e.g. upper fet diode conducting and then you turn the lower fet on very fast).

If you can guarantee ZVS then you can use slower parts, one of the few topologies where ZVS is always there is the parallel loaded series resonant converter, or a well designed LLC with lots of resonant current ...

Yes so in CatM's example above even if it gets to 4 (shut off) prior to recovery the resonant component of the circuit will finish the recovery with no losses (with the recovery 'looking like' Coss) if it's operating in ZVS. Right?

But again, moving to a specific example if I was specifying parts for a 250VAC inverter that's basically a high voltage class D what type of parts should I look for? It's hard switching so it seems I want low rr such as Infineon CFD2 (or CFD7 now). Does that sound right? This is the issue that doesn't seem to match the marketing.

This is why IGBT's are used in inverters, the co-packaged diodes are fast and soft recovery (RFI / EMC) - igbt's can turn on very fast, so turn on gate resistors are often higher than you expect to limit dv/dt, also some/most IGBT's are more rugged than large mosfets ...

And you don't just mean IGBT's with silicon carbide diodes right? This goes back to an earlier point, is it correct that discrete diodes can be substantially better mosfet body diodes? Hence, for example PFC boost converters never have a synchronous fet because their inferior body diode would erase the conducted loss gains?

asdf44 said:

The reason people bother to print "Hey watch out for RR in PSFB" is because you might not otherwise notice or anticipate this problem in a ZVS PSFB. On the other hand you wouldn't bother to print this warning in the context of a 100V Class D amplifier because that hard switches 100% of the time and you have zero chance of not noticing it?

Click to expand...

Yes, that is what I think.

Hence, for example PFC boost converters never have a synchronous fet because their inferior body diode would erase the conducted loss gains?

Click to expand...

Because it needs high side driving and because Si FETs have high reverse recovery compared to SiC. At HV such as PFC apps, you want virtually zero reverse recovery, otherwise the reverse recovery power gets too high.
Moreover, at high dv/dt, as we mentioned before, parasitic BJT can be turned ON if using diode from a FET. There is no BJT failure possibility in a discrete diode, isn't it?

People are starting to use GaN & SiC in full bridge boosters, to get the advantage on low Ron, switching speeds are moderate ( 66 - 120kHz) - the fast intrinsic diode is the key ...

Why are mosfets with fast recovery diodes such as Infineon CFD2 marketed for 'soft switching' or resonant applications when its hard switching applications that should suffer most from rr losses?

Click to expand...

Yes, i often think this too...but then i found some app notes on failure modes of PSFBs and LLCs.........they can fail from RR....i will go and get the docs for you...

- - - Updated - - -

...it would only let me upload two documents.

- - - Updated - - -

some more on psfb

- - - Updated - - -

..ok i tried to upload several but again it only allowed two.....give me a shout if you want more.

- - - Updated - - -

Why in the context of totem pole PFCs is it declared that silicon mosfets can't be used when there should also be rr losses in the standard alternative boost converter? (because discrete diodes in the boost perform better than fet intrinsic diodes?).

Click to expand...

yes i believe you answered it well yourself.

I think they like to switch BTP-PFC fets hellishly fast in apps where passing emc isnt so important but efficiency is...eg some military uses