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.

Push-Pull start up issue?

Status
Not open for further replies.

bowman1710

Full Member level 3
Joined
Nov 8, 2014
Messages
183
Helped
6
Reputation
12
Reaction score
6
Trophy points
18
Activity points
1,604
Hi Guys,

Bit of a long shot with the amount of information I can give but just asking in case anyone else had had similar issues

I have a push-pull design that is taking 60V and stepping it up to 360V at 360KHz, Now my issue is when I power the board at start up the board makes a bit of a noise then runs fine. Ive tried playing around with the soft start to see if that helps but with no luck there. Looking at the switching waveforms on the primary of the transformer the waveforms go erratic then sort themselves out, there is no jitter what so ever. Any ideas on where is best to start looking for this start up issue?

The design is similar setup to (page 9) apart from I am not using the secondary rectification on the LT3723 just a full bridge rectifier with inductor/caps, feedback on the output.

http://cds.linear.com/docs/en/demo-board-manual/dc541A.pdf
 

I trust you understand the dynamic load impedance on startup and the effect of any Remanence from abrupt shutdown can have on saturation.
The load impedance = {ESR + j X(f)} / N² is a simplification for low reactance of L and Caps will be essential low Z with ESR only as the pulse current is rich in harmonics)

Determine it Remanence exists, then determine if you can soften the load on secondary somehow during startup.

Can you scope current with a 50mV shunt on primary and secondary using coax probe tip/ring and no long leads?

After thinking about it, I believe the loading problem is that a simple bridge is an "asynchronous" vs synch. converter. The RdsON is only low when the primary secondary switches are in synch, otherwise the bridge reacts to time delay in filter and secondary still rectifies the power signal when the switch has just opened up thus causing a more rapid load regulation error and circulating currents of back EMF.
If both primary and secondary are in synch then the impedances ought to switch in synch and me somewhat more matched for maximum power transfer. Although you generally want the source to be much lower than the load impedance for maximum efficiency and with an asynch switched load, during charging this does not happen.

It does however happen during normal discharge operation but charge current peaks are much higher than average so synch switch design feature is more important here to me. ( If I understand all the issues... and I dont ;)
 
Last edited:
Maybe you should post some of the waveforms for us to look at.
 

Do you also have the auxiliary bias ckt (D12, D14, L4 in page 9)?

How does the voltage at VCC pin 5 looks like during startup?
 
I trust you understand the dynamic load impedance on startup and the effect of any Remanence from abrupt shutdown can have on saturation.

My thoughts also...

One simple thing you can try is to introduce a very small air gap into the transformer, say .001 inch (.025mm).
 
you have no current limit or soft start at start up? charging big o/p caps at start up is a common problem if there is no current limiting on the driving switches or proper soft start, the currents can be big giving lots of turn off dv/dt which can interfere with the control, not to mention the di/dt at turn on...

Got the right snubbers on the fets and you o/p diodes...?
 
please put what you have done into ltspice free simulator and send it here, then youre problem will be solved in no time....here is some example smps with ltc3721, quite similar to ltc3723

- - - Updated - - -

Also, are you using the ltc3723-1 (current mode) or the ltc3723-2 (voltage mode)? If you mix these up it could mean problems

- - - Updated - - -

The push pull is good in that it allows all low side fet drive, but the fet off-state voltages are more than double what they would be with a full bridge or 2TF. The leakage spike has to be dissipated out, unless you have some very complicated active snubber. A full bridge or 2TF has no more than vin on the fets in the off time.

- - - Updated - - -

You also say that you have set it up like page 9…….so that has a split secondary…which means that there is a high off state voltage on your secondary diodes (because after all your vout is 360v)……are you getting voltage overshoot and terrible reverse recovery problems during startup.

What is your power level?…if its low then just use a flyback.
Did you need isolation?…..if not boost converter ,or interleaved boost is better.

Push-pull is essentially a throw back to when power fets had enormous Qg…and driving high side fets through the leakage L of a pulse transformer gave overly high switching losses…but we have more modern fets now…so push pull is a little “historic”

- - - Updated - - -

if your vin and vout are fixed then what about LLC

- - - Updated - - -

In your pushpull you obviously want low leakage L….You also would like the txfmr manufacture to be very repeatable as regards leakage L…but lets just think how we achieve low leakage in a pushpull converter. Usually we achieve low leakage by interleaving the secondary inbetween two halves of the primary…..but think how you do that with a push pull….there are effectively two primaries…….so you have to split both of those into two and interleave the respective secondary in between that lot…that’s a very complicated and messy transformer wind…..sure you don’t fancy a half bridge LLC or full bridge instead?…the leakage isn’t too critical in a full bridge…..it just slams itself out through the fet intrinsic diodes.
Yes you still have to snub the secondary diodes because of the leakage L…but as we’ve seen elsewhere in this forum, bigger leakage L in a full bridge smps doesn’t actually mean that much more snubbing needed across the secondary diodes.

- - - Updated - - -

One simple thing you can try is to introduce a very small air gap into the transformer, say .001 inch (.025mm).
Thankyou thankyou warpspeed for coming out with this, I have been saying words to this effect on this forum for ages and all have slammed me down.
 

Attachments

  • Full Bridge +-40V _LTC3721 _NON-ISOL.txt
    14.2 KB · Views: 90
  • 200W Two transistor forward converter.txt
    10.7 KB · Views: 78

Think how you do that with a push pull….there are effectively two primaries…….so you have to split both of those into two and interleave the respective secondary in between that lot…that’s a very complicated and messy transformer wind
Its actually not too bad.

The two foil primaries go on together at the same time.
You need two rolls of foil and two rolls of mylar, and wind all four on together.
Interleaving both halves of the primary that way can result in astonishingly low leakage inductance between primary halves. Both halves are pretty well matched for dc resistance as well.

Naturally half the secondary goes on first.
Then both primary halves are wound on together at the same time.
Then the other half of the secondary on top.
 
Right guys,

I have been on holiday so sorry for no replies,
I trust you understand the dynamic load impedance on startup and the effect of any Remanence from abrupt shutdown can have on saturation.
I have not considered this and will look at this to see what I can get from the calculations and measurements.

Do you also have the auxiliary bias ckt (D12, D14, L4 in page 9)?

I have a aux supply but its not from a transformer its coming externally from another device, I have tried changing it over for a bench a supply too to try and see if there was and issue for that.

One simple thing you can try is to introduce a very small air gap into the transformer, say .001 inch (.025mm).
I did think that myself but wasnt to sure if you were meant to put an air gap with a push-pull supply, or how it would affect the operation in any way?

Got the right snubbers on the fets and you o/p diodes...?

Yeah the dv/dt could be an issue there is a lot of capacitance on the output of our supply which could be causing some issues with the dv/dt or di/dt from the control point of view, thats why i tried playing around with the soft start of the device but it didnt really seem to change anything. Would the output snubbers have any effect on the start up because at the moment I have no snubbers on the output?


The push pull is good in that it allows all low side fet drive, but the fet off-state voltages are more than double what they would be with a full bridge or 2TF. The leakage spike has to be dissipated out, unless you have some very complicated active snubber. A full bridge or 2TF has no more than vin on the fets in the off time.

In reply to all of your questions no I haven't got a split secondary sorry should of said the output is as per picture

output.png

The power level is to be 200W+ and yes I do need the isolation

The leakage in the transformer is fairly low as we are using a large planar design so it is quite repeatable and easier to keep the leakage down for this design, as warspeed has said it can be done with foil if not which helps a lot with the winding, i will look at the the things that others have suggested then put it into LTspice and send it in, if all else fails to solve the issue!
 

I did think that myself but wasnt to sure if you were meant to put an air gap with a push-pull supply, or how it would affect the operation in any way?
Adding a very small air gap is not a normal thing to do with a push pull forward converter, but it is sometimes necessary.

What can happen s that when you switch off the power, the core can remain slightly magnetised (remnant flux). Next time the power is turned on, there is a 50/50 chance the core flux will be driven in the same direction as the stored flux that is already there.

This can very easily saturate the core and result in extremely high peak current.
Its called "flux doubling".
By introducing an extremely small air gap that greatly reduces the permanent magnetisation effect, and usually solves the problem.

The only down side with doing this, is that it reduces the primary inductance, and that will increase the magnetising current and lead to a small but noticeable increase of zero load running current.

That is why the gap must be kept very small.
If it fixes the turn on surge problem, the benefit far outweighs any disadvantage.
Anyhow its just something you can try.
It may or may not fix the problem, which may be caused by something completely different.
 
Last edited:

Its actually not too bad.

The two foil primaries go on together at the same time.
You need two rolls of foil and two rolls of mylar, and wind all four on together.
Interleaving both halves of the primary that way can result in astonishingly low leakage inductance between primary halves. Both halves are pretty well matched for dc resistance as well.

Naturally half the secondary goes on first.
Then both primary halves are wound on together at the same time.
Then the other half of the secondary on top.

So are you saying that each primary goes across the width of the bobbin minus any safety margin for sec, pri, pri, sec. Or are both primary's wound next to each other with a safety margin also in the middle sec, two primary's, sec?

Let's just say for this discussion that the foil is 1.5 inches wide (3.8 cm), how do you make external connections to these foil winding's?
 

Different situations require different solutions.

The last time I did this (funnily enough only a few days ago) was on a pair of E65/32/27 EE cores for an novel experimental dc/dc converter. It required 15 + 15 turns of .005" foil interleaved, with about 220 working volts between the two foils.

The .005" mylar was cut to 40mm width to be a very snug fit in the bobbin.
The .005" foil was cut to 34mm width, to carry about 16A rms.

The foil being much narrower than the mylar insulation. If carefully wound, provides excellent insulation at the edges.

To carry 16 amps I used some coarse seven stranded pvc insulated electricians wire.
The seven strands being fanned out flat (like the fingers on a hand) the bare strands being about 30mm long. This was soldered to the foil spreading the current across the whole width of the foil and avoiding a hot spot.

This does produce a very slight bump in the winding, but it is definitely not a problem.

I have not carried out a high voltage insulation test, as this is just a quickie just for fun prototype, to investigate an experimental dc/dc topology I wished to try.
An off the wall idea of mine I wanted to try out, and take some measurements from.
 

An uncontrolled (flux) push pull may well benefit from an air gap, a current mode controlled push pull does not require any gap.
 
  • Like
Reactions: treez

    T

    Points: 2
    Helpful Answer Positive Rating

    azadfalah

    Points: 2
    Helpful Answer Positive Rating
What can happen s that when you switch off the power, the core can remain slightly magnetised (remnant flux). Next time the power is turned on, there is a 50/50 chance the core flux will be driven in the same direction as the stored flux that is already there.

This can very easily saturate the core and result in extremely high peak current.
Its called "flux doubling".
By introducing an extremely small air gap that greatly reduces the permanent magnetisation effect, and usually solves the problem.
..so the above (from post#10) is wrong? , at least with regard to current mode SMPS's?

By uncontrolled flux type , you mean "voltage mode"?
 

Including limiting the peak currents at start up..! and yes uncontrolled = voltage mode (controls Vo only, not Iout or Iin)
 
  • Like
Reactions: treez

    T

    Points: 2
    Helpful Answer Positive Rating
Regards gapping full bridge /pushpull / forward transformers.

Supposing you have designed for a duty cycle of 0.2 at max load/min vin, then you pick your core so that the magnetising current peaks up to “x” amps, and you assure that ‘x’ amps is far enough away from saturation level of the transformer.

..All well and good, no gap needed, everything’s fine…..however…

What about those occasions when the error amplifier gets railed… (eg return from overload, sudden restart without soft start, etc etc)….then your duty cycle goes to max, which could be 0.8, for example………that’s a lot more than 0.2……and at those albeit short intervals where the duty cycle is 0.8, your magnetising current could then be going into saturation level….not good, better to be gapped and have the magnetising current going that high because if you’re not gapped, then you have far less resilience against runaway saturation , and BOOM!

So surely, put a gap in there, and be safe…you agree?

Also, when you’ve got a gap, you have better definement of your exact primary inductance, which means you have a more exact knowledge of your magnetising current ramp, which is like slope compensation……and that value is needed to be known for your feedback loop equation, so that you can do the stability calculation (calculate gain and phase margin). So use a gap, and have a more accurate feedback loop calculation, which is better than a less accurate feedback loop calculation.
Surely you agree?

I also think the remenance situation mentioned by Warpspeed is relevant, because the core flux is higher and the delay of the current mode filter means that you cannot always catch such flyaway current peaks……..and you are better “insured” against flyaway current peaks if you have a little gap.

Don’t get me wrong, I know we can get away without a gap, but the smps will be less reliable, but that doesn’t matter…because then it fails and we can “sell ‘em another one” ..ummmm

Reminds us of the Phoebus cartel, that made light bulbs that were designed to fail so that they could profit from sale of replacements...
https://en.wikipedia.org/wiki/Phoebus_cartel

QUOTE
"Members' bulbs were regularly tested and fines were levied for bulbs that lasted more than 1000 hours"
UNQUOTE
 
Last edited by a moderator:

in a peak current mode controller the flux can never exceed the 0.2 you mention above.
 

I think in certain special circumstances it can....due to staircasing of inductor current which can creep up the flux due to the leading edge blanking or current sense filtering that one inevitably has to put up with with real current mode converters.
 

Status
Not open for further replies.

Similar threads

Part and Inventory Search

Welcome to EDABoard.com

Sponsor

Back
Top