-
Single stage 55W offline Flyback (fully dimmable) with no electrolytic capacitor
Posted by ocset on 18 Feb, 2018 18:53 -
Hello,
We are doing an offline 55W PFC’d Flyback SMPS LED driver. (fully dimmable)
It is a single stage converter, with a constant off time flyback, and uses no electrolytic capacitors.
The schematic and LTspice simulation are as attached. One of the issues which is of concern is that the inverting opamp U11 in the pdf schematic (MCP6001U) , may for a short instant get a voltage on its inverting input that is more negative than minus 300mV. The datasheet doesn’t hint towards problems, but do you believe that this could cause reliability issues?
MCP6001U datasheet:-
https://www.mouser.com/ds/2/268/21733h-26779.pdf
-
That's a lot of shit for not a lot of glow.
Tim -
Thanks...the spec of
1...no electrolytics allowed.
2....No significant film capacitor banks allowed due to their expense.
2...PFC from 10W to 55W
3....full dimmability from 3W to 55W
4...Cant use eg TNYswitch for the bias supply because its internal FET is only 725V rated.
5...The capacitive transient protection sub-circuit is needed to give the high transient withstand spec.
...unfortunately means that this is the way to do it.
I admit its a lot of components.
-
And what, prey tell, is wrong with a nice and highly integrated PI chip?
-
Schematic says 240VAC. MOV says maximum clamping 710V. Looks fine for a TNY.
Tim -
Also: there are electrolytics available with a long life.
-
Quote
Schematic says 240VAC. MOV says maximum clamping 710V. Looks fine for a TNY.
Thanks, sorry "240VAC" is just the nominal.....i should have said 200-285VAC. (this is on Littelfuse advice).
We use a littelfuse TMOV module ahead of the above schem. Littelfuse assure us that with 6kV mains transients, there is 1100V of "let-through". There is no way of avoiding that. Using any MOV with UK mains means that the let-through with such a transient is always up to 1100V.
Littelfuse have assured us that every once in 20 years the UK mains goes up to 285VAC. And as Littelfuse told us......"thats not often, but 2018 could be THAT year"QuoteAlso: there are electrolytics available with a long life.
Yes but in our case, not long enough. They are still "wear out" components, unlike film caps. We need to see 10 years life out of this thing. -
So? 1100V clamping on the MOV does not mean 1100V on your FET. Look at some of the PI reference designs. They all comply to 285V(+) mains and are surge rated. Granted, perhaps only to 4kV but it shouldn't be all that hard to modify for 6kV (if that is your requirement).QuoteSchematic says 240VAC. MOV says maximum clamping 710V. Looks fine for a TNY.
Thanks, sorry "240VAC" is just the nominal.....i should have said 200-285VAC. (this is on Littelfuse advice).
We use a littelfuse TMOV module ahead of the above schem. Littelfuse assure us that with 6kV mains transients, there is 1100V of "let-through". There is no way of avoiding that. Using any MOV with UK mains means that the let-through with such a transient is always up to 1100V.
You are building a very complex circuit for something that the industry has figured out already. Get a PI chip that does it all for you. Unless you are an ace, your 'solution' will be more complex, more expensive, more difficult to build, less efficient and most likely, there's a 'whoopsie' somewhere in it that will kick you in the face like a mule somewhere down the line.QuoteQuoteAlso: there are electrolytics available with a long life.
Yes but in our case, not long enough. They are still "wear out" components, unlike film caps. We need to see 10 years life out of this thing.
If you wish you can build with ceramics off course (PI has reference designs for that) but there's no reason at all why electrolytics can't survive 10 years (unless your thermal design is beyond poor in which case your LEDs will go down the tube as fast or perhaps faster). -
That seems ridiculously complicated for a LED driver.
What's with the triple parallel transistors in the OVP circuit? A zener to the current feedback will usually do the trick.
And overcurrent protection on a constant-current supply...?QuoteYou are building a very complex circuit for something that the industry has figured out already. Get a PI chip that does it all for you. Unless you are an ace, your 'solution' will be more complex, more expensive, more difficult to build, less efficient and most likely, there's a 'whoopsie' somewhere in it that will kick you in the face like a mule somewhere down the line.
This.
If you can keep them cool, there is little wrong with electrolytics. The only time they present reliability issues is if they're run too hot.
-
4...Cant use eg TNYswitch for the bias supply because its internal FET is only 725V rated.
So add some protection instead of reinventing the wheel. -
Quote
And overcurrent protection on a constant-current supply...?
This is just in case the software engineer commands too much current during initial prototyping. The load might be a 55w, 100v banks of leds, but may also just be a 10w, 10v bank of leds in other cases.
As you know, the regulation is done ultimately by the half sinusoidal reference in the upstream bit....and this is adjusted by the micro....Current gets shovelled into the output and ends up being what it is in the LEDs. The micro measures the current in the leds (the average) and pegs back the sinusoidal demand accordingly. The micro in fact will start off with a low demand and edge the current up to the required value...bit by bit....thats the kind of regulation it is...slow...so a fast overcurrent trip was seen as a safety measure for prototyping where software may initially have bugs in it.QuoteGet a PI chip that does it all for you.
There are no PI chips that do a single stage, offline, PFC’d led driver which is dimmable from 55w down to 10w, and pfc’d to >0.9 at all those powers…And does this with LED loads that could be 115V down to 10V. (the maximum load is 55w, but some other lamps will just be 10V, 10w, or 30v, 30w, etc.)
As you know, Quasi resonant controllers are poor for this because their switching frequency gets too high at the lower dimming levels. (when you are also having PFC>0.9 at those low dimming levels).QuoteWhat's with the triple parallel transistors in the OVP circuit?
Thanks, that’s true, theyre only small SOT23 but I have been a little profligate with them.
Regarding electrolytics, we can use them and keep them cool, but this inevitably means more space needed.QuoteAnd what, prey tell, is wrong with a nice and highly integrated PI chip?
To be honest, i think we will indeed succumb and use a TNY284 for the 3W bias flyback.
-
Quote
I'm fairly certain they will get a lot closer than you. Why do you think you can do all that with this design? My guess is trying to shoehorn a 10V/10W in the same design as a 115V/55W will be more expensive, time consuming, fragile, regulatory unsavory and problems ridden than rolling two or three standard PI designs while at the same time being less efficient. But hey, that's just me.QuoteGet a PI chip that does it all for you.
There are no PI chips that do a single stage, offline, PFC’d led driver which is dimmable from 55w down to 10w, and pfc’d to >0.9 at all those powers…And does this with LED loads that could be 115V down to 10V. (the maximum load is 55w, but some other lamps will just be 10V, 10w, or 30v, 30w, etc.)QuoteAs you know, Quasi resonant controllers are poor for this because their switching frequency gets too high at the lower dimming levels. (when you are also having PFC>0.9 at those low dimming levels).
PI supports a lot of topologies. Not sure why you drag Quasi resonant into it. -
Quote
PI supports a lot of topologies. Not sure why you drag Quasi resonant into it.
Thanks, As you know, ..by 'Quasi Resonant', i am referring to what is the “Boundary conduction mode flyback”. As you know, this is commonly used offline, as a single stage converter, with very little DC bus capacitance, to give a single stage , high power factor conversion.
As you know, the “quasi resonant” name comes from the fact that it is often set up, such that the FET gets switched on as its drain is at the valley minimum of its off state ring.
Power.com's quasi resonant Lytswitch (6?) is the nearest thing for our job...but falls short...it wouldnt be able to handle the wide dimming range and stipulation of high power factor right from 55w down to 10w.QuoteWhy do you think you can do all that with this design?
The simulation isnt perfect, but in this case it does show that this is a go-er for our spec.
After all, the shown design is very simple. Its just a C.O.T. controller, which is commanded by the external error amplifier to make the current follow the half sinusoidal reference. As you know, the UCC28C43 has been hacked to make it operate in Constant off time.
I must admit we would prefer it if there were some off-the-shelf controllers which incorporated this...so we could get our component count down.
-----------------------------
i must admit , we tooK apart a 75w offline led driver from a big well known company.....we havent got all the thick black potting off it yet, but there appears to be a PFC inductor, and two major switching (ferrite) transformers and a bias transformer too. Its a big bit of kit for 75W.
-
Quote
You are building a very complex circuit for something that the industry has figured out already.
Thanks, i assume you mean eg the NCL30001 controller?
..however, the NCL30001, and its brethren, have "issues" as follows, would you agree?...................................
NCL30001 controller
www.onsemi.com/pub/Collateral/NCL30001-D.PDF
(NCP1652 is of a similar ilk)
https://www.google.co.uk/url?sa=t&rct=j&q=&esrc=s&source=web&cd=1&ved=0ahUKEwiDnoWL9oLaAhWFF8AKHdrqBS4QFggnMAA&url=http%3A%2F%2Fwww.onsemi.com%2Fpub%2FCollateral%2FNCP1652-D.pdf&usg=AOvVaw0TAkjnyuXiO3Q0ozSmGSeH
The NCL30001 controller implements an offline PFC’d single stage flyback converter. This operates in fixed frequency, and is intended for CCM.
The schematic for it involves only a FET source sense resistor. It does not include a current sensing resistor in the return path. As you know, the source sense resistor signal represents the Flyback FET current , and this is not the actual total current.
As you know, virtually all PFC Boost converter controllers involve a sense resistor which is actually in the return path. This is , obviously because in order to make the input current follow the input voltage (sinusoidal), its obviously necessary to sense the actual input current. If its not sensed, then we cannot know what it is.
As such, the PFC operation of the NCL30001 will have significant limitations over varying line and load. Would you agree? I am wondering if they have implemented some fantastic algorithm in the NCL30001 which somehow overcomes this? I very much doubt it, because if they have, then its “goodnight Vienna” to all the PFC controllers which comprise a current sense resistor in the return path. –Obviously, such a sense resistor breaks up the ground…it “splits” the ground, which is something which is of some level of inconvenience. If controllers existed which obviated the need for such a “ground-splitting” sense resistor, then all such controllers would die off overnight….since they have obviously not all dies off, I put forward that the NCL30001 has a level of inflexibility in its PFC operation…would you agree?
Certainly, all the app notes suggest correctly PFC’d operation only over a 2:1 output voltage range.
NCL30001 datasheet:
-
Hello,
The following (attached) is a 150W offline Flyback ( abbreviated schem attached) with an output of 300V. [its the circuit of the top post but now with threee secondaries, which get their outputs stacked].
The three secondaries would each be spread across their own layer layer in the transformer. The three secondaries would be interleaved between the two primary halves to reduce leakage inductance.
-So, Given that the transformer has five layers of coils, to what degree do you believe that “Proximity Loss” would reduce the efficiency?
(it’s a constant off time flyback and switching frequency is around, but not exactly 90kHz.)
-
Back on the subject of single stage pfc'd flybacks if i may...
The attached Single stage offline PFC’d BCM flyback simulations in LTspice show just why Boundary conduction mode is so terrible for single stage PFC solutions if the switching frequency is not limited.
At 60W, the switching frequency inside a single 10ms half sine ranges from 65khz to 190khz.
This is really wider than liked, but is something you have to accept with BCM operation.
At 25W, the same stage has its switching frequency ranging from 145khz to 350khz….this is far too high frequency for an offline flyback.
I am sure you agree that unless a BCM flyback PFC stage is limited in its upper frequency level, then it is not a viable solution. The problem is, none of the offtheshelf single stage flyback PFC chips show any evidence of switching frequency limitation in their datasheets. Do you know why this is?
The limitation of switching frequency can be done by having a minimum off time. However, eg the L6562 datasheet shows no such limitations.
Do you know why?
L6562 datasheet:
https://www.google.co.uk/url?sa=t&rct=j&q=&esrc=s&source=web&cd=1&ved=0ahUKEwj7hKDfoqjaAhXJAsAKHSTNAmQQFggpMAA&url=http%3A%2F%2Fwww.st.com%2Fresource%2Fen%2Fdatasheet%2Fl6562.pdf&usg=AOvVaw3Wr6KcQFud0TVMTNZLDEA6