Flyback controller output voltage and amps under load

[thomasx]

Hi

I have designed an AC/DC flyback controller using the TI UCC28C54 controller, very much like the typical application example in the datasheet linked below.

230 VAC input
12 VDC, 7 A output.

The UCC28C54 features cycle-by-cycle over-current limiting, designed in this case to kick in at 7 A.


So, I have a couple of things I wonder about
 
It seems to be working ok, but I am not quite satisfied with the voltage drop as the load increases.

At 100mA load I have 12,2 volts, at 6A load I have 8,8volts, that's a 3,4 volt, 28% drop.

So I wonder, is what I am getting what could be expected, or should I look for something in the design that is not quite right? Please see the attached screen-shot/diagram

[ Attachment Invalid Or Does Not Exist ]

One other thing that puzzles me a bit is that when the load increases to draw more than 7A, I can see that the current sense voltage doesn't increase above 1 volt, so that part is working, but the output amps still goes higher, even if the voltage drop increases. I use a B&K 8601 electronic load, so if I set it to draw 9 amps, the AC/DC flyback converter actually supplies 9 amps, even though it should be current limited to 7 amps, the output voltage though drops to 5,5 volts as can be seen in the diagram.


Datasheet
https://www.ti.com/lit/ds/symlink/ucc28c54-q1.pdf?ts=1714685064113&ref_url=https%253A%252F%252Fwww.ti.com%252Fproduct%252FUCC28C54-Q1

[thomasx]

Ok, so basically in both DCM and CCM the current can go higher. In this case the converter transits from DCM to CCM at about 40W load  (if I remember correctly).

My biggest concern however, is the drop in voltage. This is supposed to be an 84W power supply, 12 V x 7 A, but the highest output I can reach is 57W at about 8 amps 7 volts, which is totally off the chart in all ways.

[thomasx]

Here's the diagram that didn't show up in the original post and for some reason can't be added to it. Voltage drops all the way.



I'll get back with the schematics.

[Faringdon]

And you are measuring vout right at the output capacitors of the SMPS?

Also, it sounds to me like you have a too high resistance sense resistor....at full load, you shouldnt be peaking at 1v with that..it should be less than one volt.
You can sort this stuff out by making a ltspice sim of it and playing about with it.
Is your COMP pin max'd out.?..if so, get the sense res down.

Also, attached is a 12v 7a sim of flyback, if you want you can play with this or change it to your set up....only note is that the ltspice opto's dont have similar responses to real optos.
IMHO, always get a  representative sim going before you go to the bench. If  you cant get a rep sim going first, then the bench one will never work.

And how big is your input cap?...are you getting too low input volts at the higher loading?

[thomasx]

And you are measuring vout right at the output capacitors of the SMPS?

Yes.

Also, it sounds to me like you have a too high resistance sense resistor....at full load, you shouldnt be peaking at 1v with that..it should be less than one volt.
You can sort this stuff out by making a ltspice sim of it and playing about with it.
Is your COMP pin max'd out.?..if so, get the sense res down.

You mean if CS is max'd out? It's not.
Sense resistor is calculated to 604mOhm and measurements have shown it to be the correct value. CS is < 1 volt all the way up to 7 amp output load, just like intended.

Also, attached is a 12v 7a sim of flyback, if you want you can play with this or change it to your set up....only note is that the ltspice opto's dont have similar responses to real optos.
IMHO, always get a  representative sim going before you go to the bench. If  you cant get a rep sim going first, then the bench one will never work.

Thanks. I hade a look at your sim and it works fine for me.

I have attached a sim of mine, but I cannot get it to work at all. It's like the controller never starts up. Feels like I have missed something "obvious".

And how big is your input cap?...are you getting too low input volts at the higher loading?

Input cap is calculated to about 3uF. I am running 10uF. I know it sounds small but I've done the calculations over more than twice. Please note the min input voltage is 220VAC RMS, and my tests are run at 230VAC RMS. But I will double check what the Vbulk looks like. Though the voltage drop starts already from "start" with low loads.

 UCC28C54-12V7A.asc (4.63 kB - downloaded 12 times.)

[thomasx]

The voltage begins to drop at low loads. Possibly an unstable loop or limited DC gain.
The schematic might indicate something.

Could be something like that, I guess. But I followed the design process set out in the datasheet and use the values calculated for the compensation loop, so it's a bit hard to know where to go from there. The values also correspond pretty well to the values in the typical example in the datasheet, which isn't too far off from my design.  And I can't seem to see anything looking unstable during measurements on the signals in the control loop.

I posted an LTSpice sim/schematic in the previous post. A bit simplified as I cannot get that one to work at all.

[magic]

The way "overcurrent protection" works in simple offline PSUs is that primary current sensing limits primary power, secondary overcurrent drags down secondary voltage because P=U·I, secondary voltage droop is reflected in auxilliary voltage by the transformer, then the controller UVLOs and restarts after its supply is recharged by the startup resistor.

Regarding the almost linear drop in voltage with load current, maybe your feedback circuitry has extremely low loop gain or you have excessive resistance somewhere on the secondary, not compensated by feedback.

[Faringdon]

You say you have 3uF on the primary DC bus?
As Xavier already states, 3uF is nowhere near enough.

Your sim is now attached and working....you have chose very deep CCM....so maybe look into SiC sec diode. Or just make it more near to DCM
You chose bias winding reg...so will not be that accurate over the load range.
In the sim, you need to add the resistances of caps and L's

As you know, you chose K1=1, so no leakage, that will need adding in but i sure you know this

You also need to pick  component type and P/N...not just have the generic diode. I had to change your diodes as i did not have them in mine.
Be sure to choose ultra fsat diodes where needed.

[thomasx]

The input cap seems to be way too small. I'd expect 100uF.
Is there 100HZ ripple at the output at higher loads?

I agree it's too small. Despite it being 3 times the value calculated based on the TI datasheet design process

I added a 820 uF just to try, as I happened to have one on the desk. Less ripple, of course. But there is no change at all in the voltage drop, that issue is the same.

[thomasx]

The way "overcurrent protection" works in simple offline PSUs is that primary current sensing limits primary power, secondary overcurrent drags down secondary voltage because P=U·I, secondary voltage droop is reflected in auxilliary voltage by the transformer, then the controller UVLOs and restarts after its supply is recharged by the startup resistor.

Regarding the almost linear drop in voltage with load current, maybe your feedback circuitry has extremely low loop gain or you have excessive resistance somewhere on the secondary, not compensated by feedback.

There are no restart issues. aux-voltage is fine and suppling the controller as intended.

How can I have excessive resistance on the secondary that isn't compensated by the feedback? I need to understand that so I can secure I don't have an issue like that

[thomasx]

Thanks for the help, everyone. It is really appreciated!

You say you have 3uF on the primary DC bus?
As Xavier already states, 3uF is nowhere near enough.

It's 10uF. Yes, it's too small. I added 820 uF just to try, as I happened to have it at hand. Less ripple but no change at all in the output voltage drop issue.

Will likely use somewhere around 100uF in final solution.

Your sim is now attached and working....you have chose very deep CCM....so maybe look into SiC sec diode. Or just make it more near to DCM
You chose bias winding reg...so will not be that accurate over the load range.
In the sim, you need to add the resistances of caps and L's

As you know, you chose K1=1, so no leakage, that will need adding in but i sure you know this

You also need to pick  component type and P/N...not just have the generic diode. I had to change your diodes as i did not have them in mine.
Be sure to choose ultra fsat diodes where needed.

Thanks! Now I have my version working as well, but with one exception, the TI UCC28C54 model is NOT working. It works when I use the LT1243 however.
I noticed you made quite some changes to component values in your version and it seems to work fine at max load, but becomes unstable at lower loads.

It transits from DCM to CCM at about 55% load.

I also added the snubber and adjusted the coupling factor to 0.995.

Now to the interesting part. The simulation shows an output voltage of 10 volts at max load, while in IRL I see 8 volts
 
Question is how big difference it makes to use the LT1243 in the sim instead of the 28C54. The 28C54 has max Duty cycle = 50% and the switching frequency is half the RT/CT frequency. Though that can be compensated for in the sim to some extent, but might perhaps have effect on slope compensation (which isn't in included here though)

Here's my updated version that now works fine.

However, yours is about 10 times fasteri in simulating. Do you know why that is?

 UCC28C54-12V7A.asc (6.19 kB - downloaded 6 times.)

Thanks!

[thomasx]

Please post your schematic.

Yes, sorry.
It's now in the LTSpice file attached to my previous post.

[thomasx]

Here's a sim that seems to be working really well.

I'll give that a try tomorrow. Way past bedtime here now.

 UCC28C54-12V7A-V2.asc (6.3 kB - downloaded 6 times.)

[thomasx]

Some improvements, I am now "only" down 2,5V at max load, 7A, from 12.2 to 9.7.

Looking at figure 9.21 in page 40 of the datasheet, it shows about 1 volt drop at a step to full load, (16V, 2.5A, 40W in that design).



So question is, should I also expect to get to a 1V drop at full load, or is 2.5V drop what I can expect as best in this case, with a 12V, 7A, 84 W converter?

Getting 9.7 volt from a 12 V supply just doesn't feel ok. Sure, I can up the low-load voltage to 13V and end up at 10.5V at full load, but it still doesn't feel good enough.

Losing 1V from 16V going from no to full load is a 6,25% loss. If I can get the same that would mean loosing about 0.8V in my case. I'd be happy with that, and 1V too. But more than that seems a bit much. Or?

https://www.ti.com/lit/ds/symlink/ucc28c54-q1.pdf?ts=1714685064113&ref_url=https%253A%252F%252Fwww.ti.com%252Fproduct%252FUCC28C54-Q1

[Faringdon]

you are using bias coil regulation....which is indurect....so will suffer esp from winding resistance etc etc..coupling, what have you.

[mtwieg]

My first guess would be that the controller chip is not the issue, but rather your method of voltage feedback. Keep in mind the chip is trying to regulate its FB pin voltage. When your Vout drops, is the FB pin voltage remaining constant? If so, then the chip is doing its job, and the problem is that your feedback circuitry is not properly conveying Vout back to FB. Getting decent regulation with back-emf feedback can be quite difficult as it's sensitive to transformer construction (leakage inductance, parasitic capacitances, etc). Is there a reason you're trying to use back-emf feedback as opposed to a more conventional optocoupler-based approach?

Also the first simulation you shared didn't include any symbol or model files for the UCC28C54, so it's impossible to try and troubleshoot it. In the past, when simulating pwm controller chips like this I created my own SPICE model from basic elements. Generally for these "UCC" type parts, the internals are fairly simple and the datasheets give sufficient detail to replicate their behavior quite closely (unlike many other pwm chips which have lots of ill-definted circuitry).

[thomasx]

Hi

Yes, it may very well be something other than the controller. I am open to all suggestions

Today I got some TVS-diodes I've been waiting for, to replace the RCD-snubber in order to get a bit better efficiency at lower loads.

When using the TVS's I instantly noticed a major improvement in voltage drop. I have trimmed the circuit with RCD to 14 volt out at no load, giving about 11.4 volt at full load, 7A. With the TVS's the output voltage is now 12.5 - 12.6 volt at full load 7A. So that's a great improvement. I am will stay satisfied with that for now. Need to do some more measurements though.

What I did notice however that surprised me a bit is that the TVS's gets a lot hotter than the RCD, att full load. But I assume for now this is related to the higher output wattage. More tests and measurements needed there.

I any case, the big issue right now is to dissipate the heat generated in the snubber, which is way more than what is dissipated by the swtiching MOSFET.

Any ideas why changing to TVS-diodes had such a big impact on the voltage regulation? Is it perhaps related to the capacitor in the RCD-snubber?

[mtwieg]

What I did notice however that surprised me a bit is that the TVS's gets a lot hotter than the RCD, att full load. But I assume for now this is related to the higher output wattage. More tests and measurements needed there.

In theory, the power dissipated in your snubber should not exceed the peak energy stored in the transformer's leakage inductance multiplied by the switching frequency. If it's dissipating more than that, than I would assume the breakdown voltage of the TVS is too low, and it's actually clamping the primary-referred output voltage. I suppose that might improve your output regulation, but that's not a good method of doing that!

[Faringdon]

..Yes, what Mtwieg said, the TVS voltage should be above the reflected output voltage...ie vout x np/ns
You can also do an RCDZ snubber where the resistor doesnt dicsharge  the RCD cap till voutxnp/ns has been exceeded.
This gives more eficiency at light load too..