Flyback fails due to line noise: help!

[GonzaloEE]

Hi,
I'm charged to put a flyback SMPS to work with an alarm VHF transceiver for a surveillance system. The PS is an off-line, 40W, 230VAC/14VDC flyback with optocoupler feedback. Due to time constraints, I can't layout a new design but only change components in the current one.

The transceiver fails to send messages due to an annoying humming sound at the PS output. Moreover, there's no GND wire on field, so we can't ground the PS though we shipped them in metal boxes, so I'm thinking of a line noise and EMI issue.

Up to now, we have added a common mode choke + X2 shunt cap for line filtering, along with bigger output caps, even adding more gain and bandwidth to the feedback loop with no results.

I'm not a power supply guru (got digital h/w background), so any hints/ideas would be highly appreciated.

 

[Skogsgurra]

If you are in a time pinch, I think that you should get someone that is doing SMPs for a living involved. The cut and try process without adequate knowledge is risky and usually takes very long time.

Also, if you think that a new PCB would help, but you do not have the time to make one, then apply the changes you would include on the new PCB as kludges on the existing one and see if that helps.

Re the gnd wire. Does it help if you ground the box when testing on the bench? And, secondly, are you feeding that thing with 230 V AC without having a proper ground?

Gunnar Englund

http://www.gke.org

--------------------------------------
100 % recycled posting: Electrons, ideas, finger-tips have been used over and over again...

 

[treez]

Hello
You say there is an annoying humming sound....are you sure the clamp on the core is tight enough?

Anyway, are you sure that your primary current sense resistor is high enough value?.....maybe the primary current is rising well high before the controller is turning the gate drive off...and this is humming out the ferrite......

can it do more than 40watts?

if so you could use a low dropout regulator on the output and change the feedback resistors a bit to slow things down.

are your output caps of low enough esr..

and would you like to cheat and just put an inductor on the output to smooth things out there?

what about more RC snubbering across the FET and output diode.....

...is their a huge FET turn off spike messing things up

whats the leakage inductance like on the transformer.

 

[GonzaloEE]

Gunnar:
Yep, the field setup is 2 wire 230VAC, so we're connecting the case to -VDC out. creepage is OK.
The board has some parts soldered on-the-air to see what happens.
We have added a lot of components to the upset boards

Treez:
The 1ary. clamp is a UF2005 head2head with a P6KE200 200V Zener. Both fast switching. Zener with a RC snubber.
The current sense resistor looks as suggested in the controller's datasheet (TOP246Y).
Yes, it could deliver up to 70W (smoke), but the load works at 12V, 2A short bursts during tx.
I'm ok with using a LDO in our next design, though adding it to our current boards would be risky.
Output caps are a couple of low-ESR 2200uF, right after the sec. diode. Then a 3uH coil and another 1000uF cap.
I'm OK with a coil at the output.
The switcher IC has an embedded FET with RC snubbering. We have no external snubbers as.
Output diode has a RC snubber across it. The scope shows little overshoot at that point.
Xformer 1ary inductance is 876uH, 20uH leakage.

Thanks to all for your assistance!

 

[treez]

Hello,

I've had a quick look at the 52 page datasheet...

http://www.farnell.com/datasheets/67126.pdf

..for TOP246Y,

....I need to look longer but i don't see how this controller is limiting the primary current. I don't see any low value primary current sense resistors........nor did i see a current transformer to give away the primary current.

The flyback is usually done with current mode control, and especially if you are in discontinuous mode, its good to ensure that your primary current ramp is truncated before core saturation occurs.
I trust you have your Txformer core datasheet and have used the core area (Amin) to calculate the saturation current (which will give saturation flux density).

Flybacks usually have gapped cores......or otherwise the core would have to be unduly large...a too big gap can cause interference....a gap in the centre post of a potcore is nicer.

I'm now guessing from the datasheet though that with a near 876uH primary and 132KHz(?) switching frequency you are using continuous mode?

I hope you are managing it well if you are in continuous....you need to definetely ensure that your primary current doesn't take the core into saturation.

Humming is common in saturating cores.

I'd check the primary current is not taking the core into saturation.

It looks like the current limiting is provided by the controller noting the volts across the FET itself when it's on......but how does that help the user to see what the current is doing if you don't know the ON resistance of the FET?

....Also, have you enough capacitance after the line input diode bridge?.....because if your line is sinking down in volts then the flyback will be trying to increase the primary current to get the power throughput...and this could be saturating you....same goes if you have a lower line in than you thought.

you say "The transceiver fails to send messages due to an annoying humming sound at the PS output"

..can you expand on this?.....i don't see the problem of an audible noise in a circuit

Also, i'm sure other readers will be able to ponder on whether you have insufficient "slope compensation".....and whether you are having problems with "limit cycling" and "subharmonic oscillations"....three scourges which i believe are pestulent in flybacks.

How did you calculate the turns ratio......did you do it from datasheet formulae?.....i think you need to work it through and use Lenz Law and Faradays to see about saturation.

I hope if you are in continuous that your secondary diode can handle the recovery as it gets "smashed" off before its current has died to zero.

BTW, at 40W you can get a special controller which will do PFC and output volts regulation in one stage...you dont need PFC for 40w but it saves on the pulse currents from the Diode bridge.

Also, you did good PCB layout with power traces as short/wide as poss and a decent ground and power plane?.....earths connected to the right place where poss...eg terminals of filter caps. Minimissation of current loops by taking power/signal close to the return path.

 

[GonzaloEE]

Thanks for the lead on this. The problem was solved at a higher level in the last moment, and the customer agreed to use those PS for some other use. Anyway, I'd like to learn more on this.

The TOP has an internal opamp for IDS sense, based on the known RDS of integrated MOSFET, and some external current limiting resistor, placed between pin X and 1ary VDC rail (p.40, Fig 54b), along with internal safety limiter (2.7A typ, p.36). There was a 12.2kohm there when I first checked it. RdsON range goes from 2.6 to 5ohm (25ºC-100ºC) so I think the TOP current sense stage has some internal temp. compensation to show such current limit curves.

The transformer was a hard-to-find custom design, so I couldn't get any info other than L1, Np:Ns, etc. As said before, we had no chance to re-design the board, so we were about to cut the xformer open to check core diameter and gap, and use it in our design spreadsheet.

The big cap was chosen 68uF (Some AppNotes suggest 1uF/W out), then we increased it up to 220uF,all low-ESR type.

Reg the 'humming sound', a technician reported he made a quick test setting the PS, the transmitter (load) and a VHF receiver to compare signals as heard on the receiver. The signal sounded pretty clean when the transmitter was powered by batteries, though a 100Hz humming sound came up when switching from battery to our PS. I'm not sure, but I'd expect any saturated core humming to be around some subharmonic of switching freq, not mains (50Hz here), isn't it?

2ary diode is a 10A/100V shottky with proper RC snubbering across it. Found no weird scope curves other than core resonance. Ripple sink Caps are 2x2200uF low ESR, a 3.3uH choke and 1000uF low-esr cap after it.

Though I have some complaints about the layout, there are no dangerous loops, and all power traces look like wide enough (xformer, switch, 2ary diode and LC filters).

Thank you.

 

[Skogsgurra]

100 Hz is what you get when rectifying 50 Hz. Sounds quite obvious that the capacitor after rectifier is too small - or voltage too low.

Gunnar Englund

http://www.gke.org

--------------------------------------
100 % recycled posting: Electrons, ideas, finger-tips have been used over and over again...

 

[GonzaloEE]

We also started thinking of something like line noise getting through the xformer, so we changed the big cap from the 68uF (suggested in the datasheet for our line voltage and load power) to 100uF, even 220uF with no results. The ripple disappeared when the PS was connected to a 1A resistor.
To add more fun to this, we found that the radio set also produced current pulses of up to 1A, in bursts near 100Hz during transmissions. We were ordered to stop the tests before we could get any spectrum analyzer just to get over with this sort of enigma.

Thanks anyway.

 

[Skogsgurra]

Typical values are in the 1000 uF and up range...

Gunnar Englund

http://www.gke.org

--------------------------------------
100 % recycled posting: Electrons, ideas, finger-tips have been used over and over again...

 

[Skogsgurra]

Sorry, was thinking of low voltage section. Your caps seem OK.

Gunnar Englund

http://www.gke.org

--------------------------------------
100 % recycled posting: Electrons, ideas, finger-tips have been used over and over again...

 

[GonzaloEE]

The low side caps were 1000uF, as you said. We tried also 2x1000uF, then a 3.3uH ferrite choke, and another 1000uF ate the output.

The transmitter didn't work with anything other than a battery, even using an old linear PS (GND shielded), so I'm thinking of some very tight transmitter supply tolerance, our PS weren't able to stand.

Any ideas/vendors of SPS for VHF/wireless systems will be appreciated. Thanks to all for your help!

 

[itsmoked]

Can you tell us exactly why it didn't work?

Keith Cress
kcress -

http://www.flaminsystems.com

 

[GonzaloEE]

"Can you tell us exactly why it didn't work?"
Well, that why I started this thread. I'm only guessing.

As said before: when using battery, everything is OK. When using a linear PS, transmissions failed and a 100mV/100Hz ripple showed up at VCC bornes, with the typical cap charge/discharge shape.
When using a 40W off-line flyback, transmissions didn't work neither, while VCC dropped some 0.5V, showing a 100Hz period, 150mVpp ripple, looking much like this:

----/\/\/\/----/\/\/\/----...

Thanks again.

 

[treez]

i think i know what your problem is.

the key is as you said....it works with a battery...even though your flyback has been designed just fine...it doesn't make it work (transmitter).

Thinking of the DC bus on the primary side of your flyback.......the return rail of this is actually varying at 100Hz with respect to earth ground....-thinking of the switching of the bridge diodes from one line to the other.....i believe this "noise" is coupling in to your transmitter....i know that your secondary is isolated but it will still be capacitively coupled to the primary by winding inductance.....i think you need to put a cap across the transformer...from dot to dot.

if not that then some addition of capacitance to get round the winding capacitance.

I think your problem is to do with "earthing in power circuits" and winding capacitance and you may get some good mileage putting it to the power engineers forum.

 

[GonzaloEE]

We eventually contacted the radio device manufacturer -a Mid-East company. They said their device was meant 'to be powered from batteries during operation', and the PS behind should act as the charger during silences (most of the time), with 'high ripple rejection' our PS apparently couldn't stand.

I'm not allowed to post schematics, though the PS had a 2.2nF, Y1 cap between xformer pri no-dot and sec dot pins. This means from the 1ary bulk cap (+)pin to 2ary Schottky anode, as suggested in the switcher IC datasheet.

I'm not clear why to connect the cap dot-2-dot instead, is there something to do with current phases?

Thanks!