SMPS PSU articles

[i4004]

most about repair, some about working principles, some ic datasheets...

(can't upload it here as forum is protesting:
There seems to have been a slight problem with the Badcaps Forums database.
Please try again by pressing the refresh button in your browser.
An E-Mail has been dispatched to our Technical Staff, who you can also contact if the problem persists.

We apologise for any inconvenience.)

here's the one file version
http://bbsrv.wks.gorlaeus.net/~i4004..._articles_.rar

and here 2 file, splitted under max size this type of attachment forum should accept (if mods wanna try to attach it to this message, they can feel free)
http://i4004.net/i4004/?d=psu&s=&r=

both single and 2-part rar archives have same content.

i cannot say how long will these links stay online. can be a month, 6 month, a year...

[davmax]

Re: SMPS PSU articles

I have observed all along that you are very keen to present good information to share in the forum. Your photos of failed SMPSs, your input on GP electros in consumer electronics and now the posting of information covering SMPSs. A good article included is the "Understanding switch mode power supplies" that supports the information I have given you but does not address some of the specifics you raised such as the transfer of energy from primary to secondary.

I am putting together an article that fills in the gaps to address your specifics. Hopefully this will help you and others better understand and assess failures.
I hope to compress the detail in the article mentioned above to some very simple reqirements "musts" for computer SMPSs.

[davmax]

Re: SMPS PSU articles

It will take some days to complete my paper/article in that there is family business to attend to and I want to simply explain the complexity of energy transfer in a SMPS transformer, it is very different to the familiar AC transformer in that energy is transferred with different processes in the MOSFET ON and OFF phases.

[starfury1]

Re: SMPS PSU articles

no worries Davmax

I am sure it will be an interesting and informative read take your time with it.

I know it can be hard to take something that is really more complex in nature
then try and present it in a simpler light so it quite understandable.
This subject is a complex one and has many aspects to it...
not like a DC light bulb circuit

Family does come first thought

Cheers

[gonzo0815]

Re: SMPS PSU articles

Many thx i4004 for posting the PowrSupplyRepair.pdf, i have lost it and never found on my hdd again. Nice to have it handy again.

[davmax]

Re: SMPS PSU articles

Finally I put together a paper dealing with a practical understanding of the SMPS in the ATX context. Hopefully it will fill the missing gaps of knowledge that I have often seen in different publications.

I have proof read it but mistakes may still have been missed. So please feedback any detected problems.

See attached:

[starfury1]

Re: SMPS PSU articles

Thanks Davmax for your hard work on this
I have it now and pdf-ed it
read some but not all as yet

Cheers

[i4004]

Re: SMPS PSU articles

starfury, attach that pdf here too, i don't have word on laptop, and laptop (lcd) is better for reading than crt on my main rig.

tia

davmax, first of all thanks for the effort, it's appreciated, i'm sure.

now for some corrections and additional thoughts:

The main switching power supply typically will switch a high voltage DC (approx 320 volts) with a least one power MOSFET driven by a controller chip at a fixed clock speed eg 50 Khz. The power output is controlled by adjusting the duty cycle of each clock cycle. The duty cycle is determined by comparing a DC sample of the output voltage with a reference voltage. The DC sample is in fact a fast negative feedback signal that passes through an optical isolator (to isolate the output from the high voltage input switcher circuit).

wouldn't really say it that way: output voltage is fed to pwm chip and there it steers its (pwm ics) reaction. pwm ic is usually on the output section so yes, it needs to be physically separated from hi-voltage primary.

this means controlling the duty cycle of switchers, yes.

and it is hooked to primary side via optocouplers OR transformers!



feedback signal (essentialy just the output voltage) doesn't pass thru the optocoupler or transformer(depending on the design...optocoupler design means you don't have 2 small trafos(one for stby power and another for this pwm ic->switchers link) and trafo design means you do).

what is passing thru the trafo or optocoupler are impulse shapes that determine duty cycle of switcher(s).

A power supply must be available for the controller chip to start and run the main power supply. Typically this will be supplied by the standby power supply and the power to the controller chip will be controlled in response to the voltage level of PS_ON (the green wire in the power supply output harness). When PS_ON goes to zero the power supply will switch ON, if the signal goes high the power will turn OFF.

don't think so. design with tl494 (db494) we were inspecting has stby power supply suplpying the ic all the time(ie any time the psu mains plug is plugged in to mains). it's probably the same with new designs.

this means psu startup is something pwm ic does while it's turned on(supplied by stby p.s.).


as for 1.73V spike(you mentioned in transformers section)...wouldn't smoothing cap soak most of that? let alone inductor.

i could post image of bottom side of that psu pcb(so you can see traces) and then also top side but with sides switched so it's easier to see what's what.

now about
7.0 Misunderstandings encountered, with responses

1-i think we have a slight misunderstanding here regarding what caps i was talking about and what caps you thought i was talking about.
offcourse there are no lytics "between primary and secondary". but you DO have lytics in BOTH pwm-ic circuit AND you have them on primary side in switching circuit!
also, i see you're again using "feedback path" in weird fashion. feedback is nothing but a piece of wire connecting output and particular pins of pwm-ic. this is irrelevant and it can't really fail. there is no failure mode there at all.
(what's more, i say failing output caps won't cause catastrophic failure as these will only induce voltages that will latch pwm-ic to standby operation...ie no pulses to switchers, as pwm-ic has over/under voltage protection...)

failure mode can occur if the lytics mentioned just above(those placed around ic) mangle with pulses from ic, OR influnce ic in such a way to produce weird pulses.
inspect this image to see lytics as passive components in pwm-ic circuit for my fortron psu :
https://www.badcaps.net/forum/attach...achmentid=7340
we don't even have to go into details, we know high esr won't affect a circuit in any good way, it can only bring bad things.

going further than that we can usually find 2 lytic caps on the secondary side(it is hi-voltage side of the psu) of the transformer that hooks primary and secondary side(when optocouplers are not used). while these probably can't influence core timing of pulses they can probably mis-shape them. the psu that blown the machine (case 5 from 'bad design' thread) has this caps with grossly increased esr.

2-i mentioned lack of emi filters as a hazard, but backed down when you stated how big primary caps will smooth those possible voltage variations anyway. emi filters in the end don't have much impact on reliability of psu.

3-

b. Energy only flows across the transformer in response to load.

this is true offcourse, but it doesn't really matter because if load (resistance) stays the same and you have increase of voltage you also have higher current and possibility to wreak destruction of load.

4-

If for any reason the switching transistor fails short there is plenty of power to burn out the small transformer, break down insulation and allow high voltage to get through to the switcher controller.

that's one possible scenario. another possibility:
-switching frequency of this circuit changes and causes voltage on the secondary to go too high...this affects the pwm ic (its suply went too high) and pwm ic is making pulses of wrong length...ultimately this overloads the trafo and it's windings melt (pulse current can also do damage, it just has to have big amount of dc component).
offcourse, this also affects main switchers which are producing too high voltage for prolonged periods of time and destroy the pc attached to it.

finally

c. An example of a 5Vstb supply:
http://www.onsemi.com/PowerSolutions...=NCP1027ATXEVB

we can only wish ALL psus have such 5vstb supply.
better ones do have it, offcourse.

it is a nice design with small amount of passive components, optocoupler to link feedback to primary side and ic that has plenty of protection mechanisms.
all things that my blown example didn't have...hehe...



about the figure 1: figure in my book (flyback topology chapter)

disagrees with your caption "current in t1 secondary".
that's diode current.
current in t1 secondary should be alternating current because secondary voltage is alternating voltage. that's why we have rectifying diodes. if output was DC pulse waveform as on primary side we would only need smoothing capacitors, but not the diodes. but we do have them.

[i4004]

Re: SMPS PSU articles

to make this clear(as we already had misunderstanding here):

-switching frequency of this circuit changes and causes voltage on the secondary to go too high...this affects the pwm ic (its suply went too high) and pwm ic is making pulses of wrong length...ultimately this overloads the trafo and it's windings melt (pulse current can also do damage, it just has to have big amount of dc component).
offcourse, this also affects main switchers which are producing too high voltage for prolonged periods of time and destroy the pc attached to it.


ultimately this overloads the 5v stby trafo and it's windings melt

is better, and

this also affects main switchers which are producing too long pulses for prolonged periods of time and destroy the pc attached to secondary.

is better than stuff in the post above.

[i4004]

Re: SMPS PSU articles

dave, something to consider: feedback is not the same for +5vstby supply(WHEN +5vstby has it!!! my case no.5 didn't have it! no optocoupler in +5vstby circuit at all!!!) and main supply. +5vstby supply indeed has this direct feedback, but you mentioned direct feedback it in the context of main suppy, i'm afraid.

[davmax]

Re: SMPS PSU articles

wouldn't really say it that way: output voltage is fed to pwm chip and there it steers its (pwm ics) reaction. pwm ic is usually on the output section so yes, it needs to be physically separated from hi-voltage primary.


this means controlling the duty cycle of switchers, yes.

and it is hooked to primary side via optocouplers OR transformers!

Steering needs to be got right here. Steering is used within the pwm ic to direct the drive pulses alternately to two driven MMOSFETs. The DC feedback voltage sample from the SMPS output controls the clocked pulse width of the drive pulses. The feedback is not EVER via any transformer because it is DC plus ripple, only opto isolators are used for isolation

feedback signal (essentialy just the output voltage) doesn't pass thru the optocoupler or transformer(depending on the design...optocoupler design means you don't have 2 small trafos(one for stby power and another for this pwm ic->switchers link) and trafo design means you do).

what is passing thru the trafo or optocoupler are impulse shapes that determine duty cycle of switcher(s).

This is totally WRONG.

as for 1.73V spike(you mentioned in transformers section)...wouldn't smoothing cap soak most of that? let alone inductor.

I covered this quite clearly in terms of total energy percentage, time duration, even mention the inductor. I went to lot of effort to make to demonstrated that the spike is very small compared with the total energy drawn by the load.

(what's more, i say failing output caps won't cause catastrophic failure as these will only induce voltages that will latch pwm-ic to standby operation...ie no pulses to switchers, as pwm-ic has over/under voltage protection...)

You have a point. However nowhere did I state that failing output caps would cause catastrophic failure.

Many have seen failing caps reduce output voltage. True most SMPS have over voltage protection that would protect the output. If the scenario shown in my Fig 3 happened ie increased ripple due to failed caps the controller will try to restore the output to what the sample indicates. In the example it is sampling low output but producing long pulses and missing a cycle. So indeed the output ripple will remain high and the output low if the condition is maintained.

this is true offcourse, but it doesn't really matter because if load (resistance) stays the same and you have increase of voltage you also have higher current and possibility to wreak destruction of load.

Where is the increased voltage coming from? If from the primary side this is compensated by the controller.

that's one possible scenario. another possibility:
-switching frequency of this circuit changes and causes voltage on the secondary to go too high...this affects the pwm ic (its suply went too high) and pwm ic is making pulses of wrong length...ultimately this overloads the trafo and it's windings melt (pulse current can also do damage, it just has to have big amount of dc component).

Switching frequency in itself will not increase any voltage. In fact if it is too high the transformer will causea drop in output.

about the figure 1: figure in my book (flyback topology chapter)

disagrees with your caption "current in t1 secondary".
that's diode current.
current in t1 secondary should be alternating current because secondary voltage is alternating voltage. that's why we have rectifying diodes. if output was DC pulse waveform as on primary side we would only need smoothing capacitors, but not the diodes. but we do have them.

Sorry but all very WRONG.
In a flyback circuit the secondary current only flows for half the cycle, controlled by the diode. This was explained. This curent is drawing energy from the transformer that resets the core magnetisation. There is no AC in the seconday only DC pulses allowed by the single diode, for AC to flow there would need to be a fullwave rectifier.

[davmax]

Re: SMPS PSU articles

to make this clear(as we already had misunderstanding here):

-switching frequency of this circuit changes and causes voltage on the secondary to go too high...this affects the pwm ic (its suply went too high) and pwm ic is making pulses of wrong length...ultimately this overloads the trafo and it's windings melt (pulse current can also do damage, it just has to have big amount of dc component).
offcourse, this also affects main switchers which are producing too high voltage for prolonged periods of time and destroy the pc attached to it.


ultimately this overloads the 5v stby trafo and it's windings melt

is better, and

this also affects main switchers which are producing too long pulses for prolonged periods of time and destroy the pc attached to secondary.

is better than stuff in the post above.

I am not very clear what you are saying here.

1. The pwm controller accepts a wide range of supply voltage without any effect.
2. Main switcher producing long pulses will raise output voltage and this will be protected by the over voltage circuit.
3. With standby transformer burnout and high voltage going to the pwm, this is most likely a rapid event that I have already covered. It will either remove switcher drive or cause switcher to remain ON destroying the primary circuit.
4. Some overvoltage protection circuits use the pwm chip. Clearly this is not good if this chip is damaged. This is why the ATX spec states the overvoltage circuit must be seperate from the pwm chip.
5. DC component? A DC component cannot pass the switcher transformer.

[starfury1]

Re: SMPS PSU articles

ok haven't read any of the above
but since you asked I'll attach it PDF in ZIP file (as site don't allow PDF's)

hope you don't mind Davmax.

Computer ATX Switch Mode Power Supplies.PDF zipped

Original doc as posted By Davmax above

let me know if everything is all right with it

Cheers

[davmax]

Re: SMPS PSU articles

I4004
Just a quick re-visit of the flyback SMPS. I think you may have been referring to voltages rather than currents.

In flyback config the transformer is essentially a pulse transformer, transferring energy in pulses. The well defined electronics theory explains in terms of energy transfer and no energy flows without current, therefore the operation is described in terms of current rather than voltage. The actaul voltages are od less importance. If we look at the voltages the primary voltage across the switch it is a unipolar pulse whose amplitude during the OFF cycle is greater than the the supply, the magnitude and duration of this voltage will depend on the secondary loading through the diode conduction. Yes the secondary voltage is AC and the diode only conducts during the switch OFF cycle. As said before, it is the power transfer that is relevant, not the AC voltage, but the current.

dave, something to consider: feedback is not the same for +5vstby supply(WHEN +5vstby has it!!! my case no.5 didn't have it! no optocoupler in +5vstby circuit at all!!!) and main supply. +5vstby supply indeed has this direct feedback, but you mentioned direct feedback it in the context of main suppy, i'm afraid.

I don't get this. The main switcher MUST have direct feedback or else there will be no stabilised output. It must also be DC so that the DC output can be compared and adjusted.

I can understand that some 5Vstb supplies will not have feedback for regulation. Where the pwm has wide tolerance and also the 7805 it is not necessary. HOWEVER where 5Vstb is designed for high output current eg 2.5A it is very necessary to regulate the standby supply so that the heat loss in the linear regulator is controlled to a sustainable level. ie the input to the linear reg is held constant at the minimum permissible level consistent with meeting the 2.5A load.

[i4004]

Re: SMPS PSU articles

dave, i now feel just like in time i told you you can't read electric diagram.
transformer stuff aside(which seems to be your forte and you're right there) you're not really understanding smps working principles.

first about trafos; yes, you're right about frequency, and you're right about current in the secondary. i didn't really put my thoughts to words in succesfull manner.
what i was aiming at was that if you would put just the resistive load to secondary you would have ac voltage and ac current. ie that you have possibility of ac current there, when you don't put a diode, but again, rather bad wording. awfull really.
frequency? just wrong.
when i speaked about dc component making damage i was thinking of primary current. ie if you have pulses of such shape that they have much dc component that's heating the primary windings.

now about (what i percept as) your misconception on operating principles. with this pic i'll explain terminology as i use it.
or..hold on, i'll just use pic i posted:

you can see there what is feedback to me, and what is control to me.
control box here is pwm-ic on the psu pcb.
(don't be mistaken; here control is placed near primary, but pwm-ic actually resides on secondary side!)

now we just need to establish what's where in detail:
looking at the pix(of different supplies) you could establish this
-pwm ic is on the secondary, low voltage side
-switchers are on the primary(hi-voltage) side
this means some link has to be established between the control circuit(that controls switchers) that is on the low voltage side(because it's part of the pwm-ic(!)) and switchers that are on primary side.
control signal is pulse signal, not dc current. (feedback signal is dc current(!))
in pix you see 3 transformers, and what would you guess 3rd trafo does?
we have main trafo(big, pieces=1), small one(+5vstby and pwm ic supply, pieces=1) so what are we left with for 3rd trafo?
that's right, it passed this control pulses to primary.
again, it OR optocouplers, as you can have design with optocoupler instead of that trafo.
also, inspect this
http://www.hardwaresecrets.com/article/327/8

something more: let's go back to this pic

you can find "FEEDBACK" caption to the right of the pwm-ic.
that's exaclty how i used it too; just the link from output to pwm-ic.

hope that we now can differentiate control and feedback portions.
----------------
good thing you agreed on output caps not being able to wreak havoc, which brings us to this: we have 2 options for catastrophic disasters
1-pwm-ic supply failure(going too high)
2-passive elements in the control circuit(not feedback circuit!) of pwm-ic failing, causing pulses with wrong timing/wrong amplitude.

so upon inspecting such failed psus(that burned mobos etc.) we should be able to find either bulging caps around the pwm-ic or failure of supply, or both.

[starfury1]

Re: SMPS PSU articles

Block diagram of TL 494
for reference



TL494 PDF link

other pdf's in other thread Here

you have to remember there's whole lot going on it that little black box

From the schematic of the TL494 PSU I4004 posted
All I can assume is that
feedback is to the IC which then controllers the mains side via T2

The Mains side (from what I understand)
should always be "Electrically Isolated" from the secondary side
this is achieved by the use of Transformers and Opto couplers

Agree, there has to be some form of accurate feedback from secondary to mains side for control of main side input
(normally a small portion of sampled output)

If this feedback information is in accurate then trouble must follow

what defines "Feedback information" I suppose is depended on the circuit design
if it be Via the medium of current, voltage or a series of varying time pulses


these links below more for my reference really but might be useful for others
(at lest I'll know were to find the )
they are related to switching technology

http://en.wikipedia.org/wiki/Flyback_converter

http://www.dos4ever.com/flyback/flyback.html

http://www.powerdesigners.com/InfoWe...converter.shtm







http://www.hills2.u-net.com/electron/smps.htm



cheers

[i4004]

Re: SMPS PSU articles

yeah, in above diagram u can see feedback and output ctrl are 2 different stages.
sure one influences the other, but feedback is never fed straight to primary side.
it makes no sense.

so again, feedback doesn't go to primary side...control (influenced by feedback) does.

[davmax]

Re: SMPS PSU articles

See next. miss posted

[davmax]

Re: SMPS PSU articles

I can see that I have misread what you have been saying. I have been at a disadvantage for some time because the SMPS diagram you posted is a mess. Just take a look at your link for yourself. It is impossible to read the diagram effectively. You are right I could not read the diagram because of the mess. Previously I wondered how you could see the the direct standby supply connection to the TL494.

Starfury has posted me a clear version. Now I can see what you are talking about. Yes T2 is the transistor pulse driving transformer that isolates the primary voltage from the secondaries and therefore no need of an opto anywhere. Yes the DC feedback voltage is there connected to pin 1 of TL494. I might add here that I do not favour the pulse transformer drive system because the emitter of Q1 swings from the mid voltage point to the high voltage point at the switching rate. With such a large voltage swing and at those frequencies the stray capacitance in the transformer has to be very low. Not easy. I prefer other high side drivers.

Yes I can now see in this example that the TL494 is connected to the standby power. I can also see that the Overvoltage circuit and the Power On circuits can bypass the TL494 power to shut it down.

I note the unlabelled inductor in the output that has windings for all five voltage rails and it is helpful to see the phasing dot on each winding. I am not really sure what this achieves in that L1 to L4 are certainly placed as output filter inductors.

In the standby power supply I can now see that the TL494 and 78L05 have seperate supply voltages, this would help with the fact that there is no regulation of the standby switcher oscillator. Gee only a 78L05 that is low current (rated at 100mA)!!!!!!!!!!!!!

Fancy saying I do not understand SMPS principles, your mistake here. Yes I misunderstood you and the "no read diagram" that is very different.

Hopefully we understand one another now. Please check your diagram posts and I will try to make sure that I receive clear info.

[davmax]

Re: SMPS PSU articles

I4004. I thank you for your inputs and I have made some alterations to my doc.

However Fig. 1 starts as correct and also section 3.6.