Continue to Site

Welcome to EDAboard.com

Welcome to our site! EDAboard.com is an international Electronics Discussion Forum focused on EDA software, circuits, schematics, books, theory, papers, asic, pld, 8051, DSP, Network, RF, Analog Design, PCB, Service Manuals... and a whole lot more! To participate you need to register. Registration is free. Click here to register now.

Slow diode used in primary clamp of Offline Flyback SMPS...Why so popular?

Status
Not open for further replies.
T

treez

Guest
Hello,
With offline flyback converters, why is there so much support given to the use of slow diodes in the primary RCD clamp network?

I do appreciate that when a slow diode is used, the RCD clamp capacitor no longer has to “swallow” all of the ½ L(lk)I^2 energy contained in the leakage inductor, -because due to the slow turn-off of the slow diode, the clamp capacitor can simply take part in a resonance with the leakage inductance, and indeed some of the energy can be transferred to the output during this resonance, and some can be dissipated in a series resistor. Also, the resonance, which takes place inside the “reverse recovery window” of the slower diode simply dissipates energy in a resistor in series with the clamp diode. As such, you subsequently end up with less voltage across the RCD clamp capacitor, and therefore less power dissipated in the Resistor of the RCD clamp. So I can see the attractions of using a slow diode in the RCD clamps of DCM flyback converters.

However, a slow diode in the RCD clamp is very bad during those moments when the flyback SMPS goes into continuous conduction mode (CCM) …..CCM can occur during output overload or short circuit, or during startup especially if a large output capacitor is used with the flyback SMPS. In CCM, the use of a slow recovery diode in the RCD clamp circuit can be disastrous, with overly large surge currents coming through the diode when the FET switches ON.
So why is everyone clammering to use slow diodes in the Flyback’s RCD clamp circuit? The web is full to the brim of it.

Eg, this app note…
**broken link removed**

- - - Updated - - -

Also, do you agee that if a slow diode is used, then ideally the reverse recovery time of the slow diode should be equal to the period of the resonance between The clamp capacitor and the leakage inductance?
 

Funnily enough I was repairing a small SMPS while your message was downloading and had just discovered a 1N4002 fitted as clamp diode. I too was puzzled by the slow diode and in particular why such a low voltage rated diode was used. It has 320V on the hot side and the RC snubber is 82K and 1nF/1KV so I'm guessing I've discovered why it has been intermittently blowing fuses every few weeks!

The document explains it well - but I think Mr Chinese who built this one has fitted the wrong part. Forget the notion of EMC reduction, this PSU wipes out half the frequency spectrum and it is only rated at 10W! The only supression it has is a ferrite bead threaded on each AC leg.

Brian.
 
  • Like
Reactions: treez

    T

    Points: 2
    Helpful Answer Positive Rating
Thanks, that was part of a big volume build?
I have known many chinese smps that have bad stuff in them and they need checking over.
As you know, its not safe to just put a chinese smps into the market without a massive health check,
yet i am amazed by the number of British co's that do just that, often paying the price in field failures.

As you correctly say, these Chinese supplies dont even meet the regs...but some British middle men are making small fortunes by
"Middle-manning" this trash into the British Market, and along with their influential friends, are making sure that the
indigenous British engineerig companies go out of business, so that everything is Chinese....and they are quids in...

The below is related to this, but dont bother reading it unless you want to...

..............................................................................................................................
The suicide of UK owned industry from 1952 to present day (2016)
In the early 1950s, the UK was an industrial giant. Today, it is an industrial pygmy. In 1952, UK owned companies made a quarter of the world’s manufacturing exports. Today (2016), the UK makes up less than 2 per cent of the world’s manufacturing exports, and even that percentage is mostly made up from foreign owned companies operating within the UK.
The UK now has a National Debt of 1.7 trillion pounds, and has been spending much more than it earns for every year over the 33 years up to 2016, and its trade deficit is dramatically worsening. The supply of North Sea Oil on which UK has become reliant is about to run out. >50% of UK industry is foreign owned; >40% of UK vital services (electricity, gas, infrastructure etc) are foreign owned; >66% of the fuel for the UK economy is foreign owned. UK is not capable of building its own nuclear power stations but relies on China to do it whilst China is already busy enough building islands just off the coasts of Vietnam, Brunei, Malaysia, Taiwan and Phillipines. >90% of UK schoolchildren stop progressing in maths & science at the age of 16. The OECD recently declared that a quarter of UK’s adults have the maths skills of a 10 year old. The brunt of the UK’s Naval fleet is powered by Electric Motors & Drives designed and built by the French. In the UK in 2016, >70% of London’s ‘zone 1’ office space is foreign owned.
The UK’s downward slide is moving at an alarming pace....In 1952, the UK was an industrial powerhouse. In 1952, UK brought out the world’s first commercial airliner, and was a close second to the USA in the development of the computer. In 1956, UK was amongst the first to bring out a Nuclear power station. In the 1950’s, Britain invented nuclear power, led the world in its application and developed enough stations to sustain a sizeable home-grown industry with a strong skills base. In the 1950s, UK owned companies were the second-largest manufacturers of cars in the world (after the United States) and the world’s largest exporter of cars.
The UK’s Sir Frank Whittle invented the turbojet engine in the 1930’s. In 1940, the UK had developed a Cavity Magnetron a thousand times more powerful than anything in the USA. The Mini, as developed by the British Motor Corporation in the 1960’s was voted the 2nd most influential car of the 20th century.
However, the UK flogged most of Rolls Royce to the Germans, Jaguar Land Rover to India, Bentley to the Germans, Rover to the Germans, Triumph to the Germans, and now has virtually no car industry of its own (Morgan cars have BMW engines). The UK has 25.8 million cars on its roads, and each one represents lots of money flowing overseas, worsening the already disastrous UK trade deficit. The UK cannot blame the arrival of the Far East into the world economy for UK’s poor performance, because after all Germany is managing just fine. In 2014, Germany was the world’s biggest exporter by Capital value.
The UK privatised its rail network so as to bring the benefits of private ownership...but then the German government bought a huge chunk of it (Arriva trains)...so its back into state-ownership...but owned by the state of a foreign country. In December 2015, the UK derived 17% of its energy from wind power. In Oct 2016, there were 6594 wind turbines in the UK. None of these wind turbines is designed or manufactured in the UK. Virtually all of them are designed and manufactured by Siemens of Germany or other foreign companies.

The UK payed £4.4bn net to the EU in 2009/10, but this rose to £8.8bn in 2014/15....the EU actually took 12.8 billion from UK, but gave back £4bn to be spent in ways decided entirely by the EU. The UK recently flogged off Admiralty Arch (the glorious gateway to Buckingham Palace) and The Old War Office (on Whitehall) to Spain & India respectively, for conversion into Hotels/Flats. In 2002, the UK came up with the "Enterprise Act", freeing its government from the duty of intervention when its own UK owned industries of great value were about to be sold into Foreign ownership...subsequently, from 1997 to 2007, foreign ownership of the UK’s firms rose from a third to a half, and foreign ownership of its vital services (electricity, gas infrastructure etc) rose to 40 per cent. Other countries (eg USA, Germany, France, Spain etc etc) have legislation which stops them from selling off their country’s prized assets in the way that the UK does.
Foreign companies acquired £30billion worth of the UK's enterprises in 2009. In 2010, that rose to a value of £54.5 billion. In 2016, the UK flogged off the magnificent ARM company to the Japanese for £24.3 billion, this had been the jewel in the UK's crown, one of the greatest electronics companies in the world. The UK flogged Boots the Chemists to the Italians, and Boots stores remain in the UK, and then the UK found that under Italian ownership, the UK received just £9 million in Tax from Boots, rather than the £90 million per year that it received before flogging it off (due to Boots getting "brass plated" by the Italians to Zug in Switzerland)......The same story of reduced Tax revenue is prevalent with most of the UK's other multitudinous industrial sales to foreign companies. In 2016, the Chief Advisor to the Turkish Prime Minister indicated that all the UK now does is to “Produce Cadbury’s chocolates and Maltesers”. (But in fact, Cadbury’s actually was sold to the Kraft foods company of the USA in 2010). In 2012, Nikolas Sarkozy, Prime Minister of France, declared that “The United Kingdom has no industry any more”. England is the 5th (fifth) most densely populated country in the world.
Between 2002 and 2008 the UK suffered a 50% drop in the number of its own domiciled school leavers opting for Electronics/Electrical Engineering degrees. Each year much less than 500 of UK school leavers "enrol" for an electronics degree, the majority of these choose all software modules or transfer to the School of Computer Science at the end of year two. In UK it is virtually impossible for almost anyone to assert the exact number of UK_citizen school leavers that end up actually graduating with an electronics degree each year....As far as Electronics Hardware centred graduates are concerned, the number is thought to be under 100 per year. A paltry amount.
When foreign owned companies in UK are a success, the profits flow overseas, when they do less well, there are more likely to be job losses. R&D spending gets notably less when UK industries get sold overseas. Tax revenue to UK is dramatically less. The UK is poorly placed to pay off its huge National Debt, since it has shed the once UK owned Engineering industry so desperately needed to pay off its debt.
The UK National Debt of UK is 1.7 trillion pounds in 2016. However, this figure is often “watered down” by expressing it as “National Debt as a percentage of GDP”. This comes out as 90% for UK…hardly a cause for comfort in itself, but in any case, GDP has dubious meaning for a country like UK, where >50% of industry & services are foreign owned. This is partly because for instance, GDP does not take into account profits earned in a nation by overseas companies that are remitted back to foreign investors. This can overstate a country's actual economic output.
Like in no other country, Britain has sold off more than half of its companies to Foreign owners (stated by Alex Brummer in book “Britain for Sale”). Nearly two thirds of UK manufacturing businesses that employ over 500 are now owned by foreign companies claimed business minister Baroness Neville-Rolfe in 2016.
----------------------------------------------------------------------------------------------------------
....so its not surprise that you have a trash smps in front of you that was shipped in.....thats all we in Britain do these days.
 
Last edited by a moderator:
The strange thing is that on the Ltspice simulator, the advantage of the slow diode vs fast diode is not really seen.
The LTspice simulation attached (pdf schematic also attached) shows the same flyback but one with a slow diode based clamp, and the other with a fast (30ns) diode based clamp.
The slow one dissipates 876mW in the clamp components and the fast one dissipates 920mW in the clamp components. –So not much difference.
(the fast one also has slightly more (300mW more) FET switching losses, due to its drain voltage going higher)…again though, its not a great difference.
 

Attachments

  • Flyback primary clamp_schematic_1.pdf
    31.1 KB · Views: 116
  • Flyback primary clamp_1.txt
    15.3 KB · Views: 64

The slow diode was an invention of Power integrations, but doesn't really work (hence they don't reccommend it now) if the input volts are low, e.g. 115Vac and the flyback volts are high e.g. 120V then at fet turn on some energy goes backwards thru the slow diode into the Vcc, from the snubber cap.

However the energy recovered this way is pretty small and the diode gets hot of course.

And if you ckt goes into CCM the diode will go west.

So always better to use the fastest diode your budget will allow.
 

And if you ckt goes into CCM the diode will go west.
Thanks, i think i said the same above, but i believe now that ccm doesnt matter so much, because after all, as long as the slow diode has completely reverse recovered when the fet switched on, then ccm shouldnt matter.
 

well just remember the above when you slow diode fails in the field...
 

The benefit of using a lossy diode clamp is only suitable for low power designs, under 20W. It helps to control radiated EMI problems.
 

in some article i have seen recommendation of using diodes with less than 500us trr in rcd clamp circuit, so 1N4002 is not suitable...
 

if the input volts are low, e.g. 115Vac and the flyback volts are high e.g. 120V then at fet turn on some energy goes backwards thru the slow diode into the Vcc, from the snubber cap.
Thanks, but i cant make out the actual loop of current thats involved here...surely if the DC bus capacitor was getting charged up then the current couldnt possibly be going backwards through the slow diode?...because that direction would be discharging the DC bus capacitor?

I can however see a certain situation where the DC bus capacitor would get charged up by the energy in the leakage inductance when a slow diode is used……as follows….
The FET turns off and then the leakage inductance current starts flowing through the slow diode in the normal way……it then resonantes with the primary clamp capacitor and starts coming backward through the slow diode…..but then, as the ringing current is coming backwards through the slow diode, the slow diode shuts off, and the current in the leakage inductor therefore forward bias’s the diode intrinsic to the power fet and starts charging into the DC bus capacitor.
I wonder though, for this to happen you would have to have a very tightly toleranced ‘ trr’ on the slow diode….plus tight tolerancing on the leakage inductance. Perhaps it may happen by accident, and just be a bonus.

well just remember the above when you slow diode fails in the field...
The Chinese have designed a slow diode (trr = 500ns) into our 65W flyback. It does indeed go into CCM for certain load conditions in the capacitor charging circuit which we have...this is because the FAN6300A has a maximum off time of 17us, and so when vout is low, we end up with CCM.....so you are telling us to beware with our slow diode?
Its an FR207.

FR207 datasheet
**broken link removed**

- - - Updated - - -

The benefit of using a lossy diode clamp is only suitable for low power designs, under 20W.
Why are these benefits only applicable for <20W designs?


in some article i have seen recommendation of using diodes with less than 500us trr in rcd clamp circuit, so 1N4002 is not suitable...
(I believe you actually mean 500 “nanoseconds”)
Thanks, but the following design by st.com uses an S1M diode in the RCD clamp for their 65W flyback, and the S1M diode has a trr of 1.8us……….

https://www.st.com/content/ccc/reso...df/jcr:content/translations/en.CD00252755.pdf

S1M diode datasheet
https://www.vishay.com/docs/88711/s1.pdf

...Also, this article states that the slow diode can be any value as long as its trr is greater than half the resonant period of the RCD clamp capacitor and the leakage inductance....
**broken link removed**

This article shows onsemi.com happily using a 1N4007 slow diode in their 65W Flyback SMPS RCD clamp circuit.....
https://www.arrow.com/en/reference-...for-notebook/7e6576217c08764155b69a13b3fee18b

- - - Updated - - -

Bottom right of page 7 of the attached pdf states how very slow diodes like 1N4007 can be destroyed if the flyback SMPS opeates in CCM with high duty cycle….It attributes this to “cross conduction”.
“Cross conduction” is not explained though.
 

Attachments

  • 1N4007 DESTROYED.pdf
    3.9 MB · Views: 147
Last edited by a moderator:

I read it as the diode's slow recovery is being used to dampen the ringing by presenting a low impedance. In other words using it's recovery period as a shunt across the signal. It increases dissipation in the diode but possibly still showing a benefit over using a fast diode and dumping more power into the RC network.

I think the 'cross conduction' is a term they use for the current flow through the diode during it's recovery period, when it behaves more like a resistor then a uni-directional conductor.

Brian.
 
  • Like
Reactions: treez

    T

    Points: 2
    Helpful Answer Positive Rating
And if you ckt goes into CCM the diode will go west.
Thanks, this is interesting.....the attached simulation in ltspice, (and the pdf schematic of it is here too) of a slow diode vs fast diode in Flyback in CCM, shows that their appears to be no problems with the slow diode and CCM. Also, the Slow diode has overall some 30% less dissipation in the clamp components and the FET. The slow diode version also has less dissipation in the diode itself.

I am yet to hear of a genuine disadvantage of the "slow diode method".

(obviously really high duty cycle and CCM would kill the "slow diode method" dead though...that goes without saying...presumably that is what Easy Peasy meant(?))

I am very suspicious of LTspice here though, as i doubt it really simulates reverse recovery and carrier recombination in PN junctions properly at all.
 

Attachments

  • Slow diode vs Fast Diode.pdf
    18.1 KB · Views: 115
  • Flyback primary clamp_ccm2.txt
    8.6 KB · Views: 50

Two issues here. One is the quality of any device model
you may get for free off the Internet, especially in the
"less important" aspects of operation. Parameter values
may be left to default, etc. The other is, even if you
used and properly set all of the available params, is the
LTSpice compact model able to fit the entire "surface"
of recovery behavior (across from:to current variations,
dI/dt, temperature, pulsed self-heating and so on)?
Not my experience with other, way more expensive
simulators so I'm disinclined to trust a free one either.
I think you might see a roughly-right charge and time
at some point(s) but a good fit across decades of
forward current, I doubt. And if you don't see a
temperature fitting param-set higher than first order
for TR, and a separate series resistance term for
reverse recovery (which may not be at all the same
path physically as forward conduction or reverse
leakage) you can figure it's only a crude approximation,
at best.

And looking at the D model docs from LT I don't see a
TR param at all.
 
  • Like
Reactions: treez

    T

    Points: 2
    Helpful Answer Positive Rating
And if you ckt goes into CCM the diode will go west.
And if you ckt goes into CCM the diode will go west.
Easy Peasy is correct…..The reason a slow diode is bad for the CCM flyback is that in CCM, the slow diode never really gets a decent ‘long’ interval of reverse voltage across it in order for it to properly reverse recover…..then in CCM, when the FET turns on, the reverse voltage cross the slow diode builds up rapidly and causes a significant, unwanted current spike, due to the “plasma” of un-re-combined minority carries in the PN junction.
Whereas in DCM, when the FET turns ON, the build up of reverse voltage across the slow diode has far less dv/dt, and so the reverse recovery is far milder. –Not only that, but in deep DCM, when the FET is OFF, the voltage across the slow diode resonates up and down at a low-ish frequency and this is long enough for the slow diode to more properly reverse recover before the FET finally turns on.
(I use the words ‘long’ and “low” above , but I am speaking of those intervals in electronics terms, I am sure you get the picture)
The LTspice simulation attached shows the situation…but the simulation doesn’t show the high spike of current in the slow diode, because the simulator doesn’t properly simulate reverse recovery.
 

Attachments

  • Slow diode vs Fast Diode.pdf
    18.1 KB · Views: 114
  • Flyback primary clamp_ccm2.txt
    8.6 KB · Views: 63

And if you ckt goes into CCM the diode will go west.
Thanks, I can see the logic in this. However, most flyback SMPS’s transition through an interval of CCM operation at startup, due to the intial lowness of the output voltage. Also, even Quasi -resonant flybacks sometimes work in CCM due to their maximum OFF time feature.
And as you know, all flybacks transition into CCM when their Vout drops due to being in overload or in output short circuit.
So if we are saying that the flyback should not use the “resonant snubber” technique if they go into CCM, then that means that virtually no flyback should ever utilise the “resonant snubber” technique.
[The “resonant snubber” technique is also known as the “slow diode” technique.]
 

A very simple test is to measure the temp with a fine current probe, compare a slow diode to a fast one, the hotter one is under greater Irev stress and turn off V stress.

Also if the snubber cap voltage is greater than the HVDC rail, then at fet turn on a slow diode allows the snubber cap to discharge into the HVDC rail via the slow diode, thus recovering some energy - at cost of heat in the slow diode, PI used to recommend this but nearly all traces of this recommendation have been removed due to the variability of slow diodes (crappiness) and the fact they will just randomly die over time with the high rev peak currents and assoc stress - even if they don't overheat under "normal" operation...

Obviously 130kHz operation is far more detrimental than say 20kHz operation to a slow diode...
 

Also if the snubber cap voltage is greater than the HVDC rail, then at FET turn on a slow diode allows the snubber cap to discharge into the HVDC rail via the slow diode, thus recovering some energy

Thanks, now I see what you mean.

The thing is, surely the diode would have to be extremely slow to allow that to happen? I mean, surely by the time the FET is due to turn ON, even a slow diode should have fully reverse recovered by then? (unless of course the duty cycle was very high)…..But with a good few microseconds between FET turn off and turn back on, then surely say a 500ns diode would have totally reverse recovered by the time the FET turns back ON?

My company has just gone into 7k units per week production with a 67W offline Quasi-Resonant flyback (FAN6300A controller) and a 500ns diode in the primary clamp. The diode is an FR207 diode….

FR207 datasheet
**broken link removed**

The primary clamp is a 2n2 capacitor in series with a 51R axial power resistor. And in parallel with this RC circuit is a 100kOhm resistance made up of four 1206 100K resistors in series_parallel. At 100VAC input the primary peak current is 2.5A peak. The maximum voltage peak seen across the 51R_2n2 series combo was 248V. Voltage reflected to primary was 170V max.

Since the clamp diode has a trr of 500ns, this means that the “ideal value” for the leakage inductance would be any value less than 11.6 microHenries. If it was any more than 11.6 microHenries, then the leakage inductance’s resonating current would be running back in to the clamp capacitor during the latter stages of the diode reverse recovery. This would mean the clamp capacitor taking on a higher voltage and thus giving rise to higher FET switch OFF switching loss.
As you know, the best case scenario with a slow diode is that the leakage inductance current goes in to the clamp capacitor in its first half period of resonance, and then resonates back out again , going backwards through the slow clamp diode….and then the slow diode shuts off just as the leakage inductance current is going back through the zero crossing at the end of its second half period of resonance. –Because in this case, there is then minimal charge-up of the clamp capacitor.

The Chinese designers tell us that the transformer primary leakage inductance is 25uH at 1khz, (this isnt very useful since the flyback operates around 70khz or so when vin is 100VAC)

Anyway, this flyback operates in continuous mode during the start of the recharge period (it’s a xenon beacon flash lamp) and this doesnt seem to give rise to problems even though the clamp diode is relatively slow. I never got time to do thermal tests on this PSU though. I did run it at max power on the bench at room temperature with the PCB “Open-frame” and the 51R clamp resistor was mad hot to the touch.

I got moved to another job so this PSU , which is a bespoke PSU designed in China for us, has gone into mass production without us even testing it.
 
Last edited by a moderator:

lets hope this converter does not run at 100% in the field... for a resistor to dissipate 1W say, at 51 ohms the rms current is 140mA if this is in a spike of 2% say of the period of the converter then this would be roughly 7A peak, you can see why a faster diode is better, also a faster diode gives rise to less stress and heating in the fet - simple temp tests reveal all this ...

- - - Updated - - -

you should look at the current in the diode (small current clamp or toroid) alongside the Vds waveform - and zoom in - you will find it very interesting - then put in a fast diode...!

- - - Updated - - -

p.s. if you look at the direction of current flow when the fet turns on, if HVDC is < Vsnub cap, then current flows in the forward direction of the diode as Csnub discharges into HVDC

- - - Updated - - -

In fact this happens naturally for a converter in DCM as the fet has a diode in the same direction...
 
  • Like
Reactions: treez

    T

    Points: 2
    Helpful Answer Positive Rating
Status
Not open for further replies.

Similar threads

Part and Inventory Search

Welcome to EDABoard.com

Sponsor

Back
Top