This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

WEBENCH® Tools/UCC2813-3-Q1: IC not starting up

Part Number: UCC2813-3-Q1
Other Parts Discussed in Thread: UCC2813-2, UCC2813-3,

Tool/software: WEBENCH® Design Tools

Hi,

I'm attempting to use a UCC2813 chip to make a AC->24Vdc supply.  I used WEBENCH to start the design and basically copied its output directly:

However I don't get any output from the supply.  What I'm finding is that Vcc of the UCC2813 is very low - it hovers around 3.8-4V, which is right around the UVLO threshold of the chip and also right at the hairy edge of Vgs on the switching FET.

The circuit straight from WEBENCH (as shown in the schematic above) doesn't start up at all, because Vcc is around 3.6V.  I removed D13 (Zener), increased C13 from 15uF to 120uF, and decreased R2+R3 from 500k to 300k.  Doing those will get the UCC2813 to switch, but Vcc is still hovering around 4V, and is apparently not enough to switch the MOSFET:

In both screenshots here, the green trace is the gate of the MOSFET and the yellow trace is the source pin of the MOSFET.  It looks like the UCC2813 tries to switch through a normal startup - when you zoom in you can see the output PWM start at a very low duty cycle and ramp up - and it switches until Vcc falls below the UVLO, then it gives up; then C13 recharges and it tries again.  But nothing ever gets through the transformer.

One caveat on the scope screenshots is that I took those with 50Vdc across the input capacitor, rather than rectified AC.  What seems interesting is that Vcc is the same if the input is 50Vdc or rectified 120Vac and the behavior of the circuit doesn't change.  I was actually able to get a stable output from the supply in this configuration - I reduced R2+R3 to 82k, and with 50Vdc input I got a solid 5ish volts output.  When I tried to scale this back up, something smoked and the MOSFET shorted internally.  Oops!  But I think what this tells me is that something is dragging down Vcc, but I haven't figured out what it could be.  Any help would be greatly appreciated! 

Thanks

  • Hi Elliot! 

    Thanks for reaching out and for using our Webench tool! Nothing jumps out at me here, but let me retrace your footsteps and see if I encounter the same issues in simulation. I will get back to you with an update on Thursday. 

    Thanks again for getting in touch, I look forward to talking more!

    Aidan Davidson

  • Excellent, thanks!

  • Hi Elliot! 

    I took a look at your schematic, thanks again for sharing! It definitely seems like the bootstrap circuit is having issues starting up and keeping the UCC2813 IC running. 

    Can you try lifting up the VCC pin from the board, isolating it, and providing VCC with a bench DC supply? Try starting the power supply with different load conditions and observe the time it takes for voltage to appear on the aux winding. I suspect you may need an even larger capacitor than the 120uF you changed over. 

    Also, if you're concerned about current leaking back through R2 & R3 to the HVDC rail, you can add a diode in series to prevent leakage. 

    Be careful about reducing R2 & R3 too far, you may provide too much current to the device. However, from the datasheet, the recommended max current at the VCC pin is 25mA and with 50VDC and R2+R3=83kOhm I don't think you should be coming close to that. 

    Thanks again for reaching out, give that test a try and think about possibly increasing C13's value further. 

    Aidan Davidson

  • Hi!  OK I tried a couple of things.

    First, I replaced C13 from 120uF to 200uF; it didn't seem to have any impact.

    Next, I removed D5 and R3, and powered the UCC2813 from a benchtop supply (10V, current limited to 50mA).  I applied AC and then powered up the UCC2813 and got 24V, briefly (maybe 4-5 seconds), then the AC input fuse (F3, it says 1A on the schematic but it's actually 2A) opened.  I had no load on the output (I meant to plug in a 60 ohm resistor but forgot).  Nothing appears to be damaged, but I'm not sure why the fuse opened so I stopped there.  Any thoughts?  The current draw on the benchtop supply never hit that 50mA limit but I didn't see what it was doing.  If nothing leaps to mind, on Monday I'll try using the 50V benchtop supply instead of the AC input and another 10V supply for the UCC2813, that way I can safely probe it with the oscilloscope while it's starting up.

    Thanks!

  • Hi Elliot! 

    Just to confirm, your test with the benchtop supply was with the VCC pin isolated from the rest of the board, correct? Were you able to observe the timing of the aux winding kicking in? 

    Aidan Davidson

  • Hi!  I did not lift the VCC pin from the board, the idea was that by removing R3 and D5 from the board, and applying voltage across C12 (which was convenient), it would power the UCC2813 without interacting with the rest of the circuit.  Was this the right thing to do?

    Unfortunately I didn't catch any of the other timing but will try it tomorrow.

    Thanks!

  • Hi Elliot! 

    It is true that removing R3 and D5 and applying voltage across C12 accomplishes isolating the VCC pin, but it also disconnects the bootstrap capacitor from the AUX winding which maybe might have something to do with your AC fuse blowing. Can you try replacing R3 and D5 and carefully lifting and isolating the VCC pin? This way the bootstrap capacitor can stay connected to the AUX winding and you can get a good impression of the timings involved here. 

    Let me know how your testing goes with the DC supply also! 

    Aidan Davidson


  • Hello!

    OK, I replaced D5 and R3 and lifted the VCC pin to power it separately, and replaced the fuse.  And ... it works!  There's an audible squeal and the output (into a 20 ohm load) is a bit messy, but it was there and pretty solid (scope_6.png).

    I put a 56 ohm resistor across C13 to give it some load (otherwise, with no load, the voltage crept up >20V) and then checked the time it took to reach about 5V; that looked like 35-40ms (scope_7.png):

    ... though that was with the 50Vdc supply and R2+R3=82k.  Gate-Source captures, if they're helpful, are also attached, scope_4 and scope_5.

    I wonder about the audible noise and the less-than-perfectly stable output.  Is it possible that the transformer I have is not made correctly?

    Thanks!

  • Hi Elliot! 

    I'm very excited that we're getting output!

    First, after reading into this a bit more it might actually be better for you to use the UCC2813-2 instead of the UCC2813-3. This device has a wider gap between the UVLO on/off threshold and that may be just what you need to overcome your issue with the PWM IC startup. 

    Generally the 2 main sources of audible noise in designs like this one are the transformer and ceramic capacitors. Noise in the transformer can generally be resolved with varnish which prevents the core from vibrating. Audible noise from ceramic capacitors is most commonly found in caps that experience high dV/dT swings like output, snubber, and clamp capacitors. This type of noise can be resolved by replacing the ceramic cap with a metal film cap or maybe a ceramic cap with a different dielectric. However, this noise may be caused by the instability on our output. 

    As you vary the load (both decreasing and increasing), does the stability / audible noise get worse or better? Can you take a couple more scope shots with varying loads? 

    Have you tried with the AC supply? When you do, does the fuse still blow? 

    Aidan Davidson

  • Hello!  I was very happy to see output too, I feel like a solution is near!

    I am a bit concerned about Vth of the MOSFET too - even with a UCC2813-2 device that can switch with a lower voltage, if I cannot drive the MOSFET efficiently enough then I don't know if startup will work anyway.  Maybe, I can certainly get the chip and try it out.

    As for the noise, I am using mostly ceramic capacitors so I will try switching to a metal film cap for the snubber to see if it helps.

    I switched back to the AC input and tried three different loads.  The good news is that the fuse did not blow!  The scope captures below are from the 24V output.

    The first load is the same 20 ohm resistor, it looked about the same, though the audible noise was quite loud and constant:

    The ripple seemed basically the same as before.  Then I tried a 250 ohm load, which is closer to steady state current consumption of the device.  In this case, the audible noise was much lower but the output ripple was huge:

    Last, I tried a brushed DC motor on a linear actuator, drawing about 1A:

    In this case I could not hear the audible noise over the motor, so if it was loud it didn't matter!

    What could cause such large spikes on the output?  I can add additional capacitors but this seems like a lot.

    Thanks!

  • Hi Elliot! 

    I am glad to hear that the fuse is holding! This likely means that the fuse blowing had something to do with the removal of R3 and D5. You can increase the output capacitance, but with a 100uF cap already there it may be better to consider adding an RC snubber across the output diode. Please let me know if swapping in metal film caps for some of your high dv/dt caps makes a change in the noise! 

    Thank you! 

    Aidan

  • I'll get some caps to try out for the noise and let you know. 

    BUT - I have output here but only when powering the UCC2813 from a different power supply, so the main problem is still not resolved.  You mentioned ordering a different version of the chip, but as far as I can tell, the chip part number I have (UCC2813QDR-3Q1) has the lowest UVLO threshold (3.6V) and I am still not meeting it.  Is that correct or do I have the part number scheme wrong?

    Thanks!

  • Hi Elliot! 

    You're correct that the UCC2813-3-Q1 has the lowest UVLO threshold, the distinction (and why I recommended the other device) is in the difference in voltage between the UVLO_ON and UVLO_OFF thresholds. This gap is where your IC is making the transition between drawing VCC from the HV rail and drawing VCC from your bootstrap circuit powered by the AUX winding. A wider voltage gap there gives you more time for the bootstrap circuit to get up to the needed voltage level. 

    Aidan

  • Aah, OK, I get it.  I'm ordering a few UCC2813QDR-2Q1 and will how they look.  Thank you!

  • Hello again!

    Parts arrived, so I made a few small changes:

    1. Replaced the UCC2813 part with UCC2813QDR-2Q1

    2. Replaced C11, the 6.8nF snubber cap, with a film cap

    3. Added a 1.0uF capacitor in parallel with the output capacitors C19 and C24

    ... R2+R3 remains 331k and C13 remains 100uF (D13 is still removed).

    The good news: it starts up on its own!  With the chip fully in place, about 4 seconds after AC power is applied the DC output starts and is stable.  By reducing C13 to 47uF, that 4 second delay is reduced to about 2 seconds, but the startup is not reliable; sometimes it takes 2-3 attempts before the output is stable.  Increasing to 200uF didn't change anything noticeable, so I left it at 100uF.

    The bad news: it is still quite noisy.  The audible noise did not go away (I also tried wrapping the transformer in rubber bands, which did not dampen the audible noise at all).  The ripple on the 24V output is a bit less than before, but still not very good:

    ... this is into the 60 ohm resistive load.  I switched to the 250 ohm load, and the output won't stay on, it goes in hiccup mode:

    In the 250-300ms that it is running, the output looks about the same:

    I definitely feel like this is one step closer, I'm very happy that it starts up on its own (even though 4 seconds seems like quite a lot).  As far as I can tell from experimenting, changing C11 from a ceramic part to a film part had the most impact on the output ripple (the 1uF capacitor may have helped a small amount but not conclusively) even though it didn't really improve the audible noise at all.  Is there a chance that something is incorrect with the snubber?

    Thanks!

  • Hi Elliot! 

    I'm very glad to hear that it's starting up on its own with the replacement IC! As for the ringing, let me talk with a few teammates and get back to you Tuesday, I want to make sure I'm pointing you in the right direction here! 

    Best, 

    Aidan

  • Great, OK, I'll wait for your response!  Thank you!

  • Hi,

    Just wanted to see if you got any input about this?

    Thanks!

  • Hi Elliot! 

    Apologies for the delay here!

    1. If you replace D13, is there any change in the output stability? 

    2. Purely out of curiosity, is there any difference in output stability between AC in and DC in? 

    3. Can we try adding an RC snubber across the secondary side diodes (D6 & D7)? 

    Best, 

    Aidan

  • Hi, no worries! 

    OK when I turned it on today, and I honestly don't remember if I left it this way but with a 60 ohm load the thing starts up after ~4 seconds and is stable, but with less load (e.g. 250 ohms, or none) it "hiccups" and won't stay on.  I'll look more at what I might have changed last time.  I have done a lot of things to this PCB so I might need to just get a fresh one!

    Anyway:

    1. I can't find D13 and the only Zener diodes I have on hand are 11V or less, so I can't test this today.  Another reason to get a fresh board!  I can probably get one tomorrow.

    2. Answer as part of #3.

    3. I tried two snubber combinations: 1nF + 15ohms, and 2.2nf + 22 ohms.  The 1nF/15 ohm combination seemed best:

    1nF + 15ohm with 60 ohm load:

    1nF + 15 ohms into 250 ohm load (captured during the second or so of stable output between hiccups):

    ... these results are from AC power.  With DC 50V input:

    Overall I think the snubber helped the output, though it needs some more careful tuning (I did not make any calculations here, just picked some parts).  I'll try to find a Zener diode to test #1.

    Thanks!

  • New captures today!  Because the board I was working on was starting to be a bit abused, I got the other prototype, which is pristine, and made the following changes:

    • U1 changed to UCC2813QDR-2Q1
    • Input fuse changed from 1A to 2A
    • C13 is 15uF + 100uF in parallel
    • R2+R3 = 305k Ohms
    • RC snubber (15 Ohms + 6.8nF) across D6//D7

    Otherwise no changes; D13 is still in place.  The results:

    Into 250 ohm resistive load with 1uF parallel cap:

    Into 60 Ohm resistive load with 1uF parallel film cap:

    With no load, but across the 1uF film cap - the supply would not stay on, but gave this hiccup:

    ... The audible noise with the 250 ohm load was much better than before.

  • Hi, just wanted to see if you have any more thoughts on this, with the new scope captures?

    Thanks!

  • Hi Elliot! 

    Apologies for the delay here. Since the suggestions I've given so far have not resolved the output stability issue, I'm inclined to point fingers at the compensation network. Do you have the equipment needed to obtain a bode plot from your board in its current state? 

    Check out these closed-loop design resources: Flyback Closed Loop Design, Flyback Compensation Techniques

    Thanks, 

    Aidan

  • Hi,

    Sadly I do not have test equipment fancy enough for this, but I will spend a bit of time studying the compensation/feedback parts of the circuit and see if I can find anything obvious.  Hints would be welcome, of course; there's a reason I used WEBENCH for this :)

    Thanks

  • Hi Elliot! 

    Here's something I should've asked forever ago: can you take a scope shot with VCC, COMP, OUT (PWM from IC), and VOUT (your output voltage)? Particularly I would want to see how the PWM duty cycle is changing as the output voltage comes up. 

    Thanks, 

    Aidan

  • Hi!  I'm playing with this but have two issues: first, for me to use the oscilloscope I have to power the system from the 50Vdc power supply, and when I do that the output is clean and stable and quite nice.  Second, I'm having a bit of trouble getting a useful amount of time on the scope reading.  Is startup (right when OUT starts switching) most useful, or steady state?

    I'm looking for a good way to at least power the board with a higher DC voltage to give more life-like operating conditions.

    Thanks,

  • Hi Elliot! 

    To confirm, you're saying that the DC output is clean now? In previous scope shots that has not been the case, what did you change to resolve that? 

    For the scope shots, if you could get startup & steady state separately that would be wonderful. If only one is possible, then steady state is a good place to start but I would want to revisit the startup. 

    Thanks! 

    Aidan

  • Hi,

    Sorry for the late reply on this, I've been busy with another deadline.  For the DC output - I didn't change anything, other than powering the supply from a 50Vdc input instead of 120Vac.  So I guess the change is that the duty cycle to maintain the 24Vdc output will be much higher, and also there is no bridge rectifier or input filter (AC side) in the circuit.

    I need to go to the office for some other testing and there is a higher voltage DC supply there so I will try to make the same measurements with a higher input voltage, to see what differences there are.

  • Hi Elliot! 

    It may be worthwhile to explore ramping your DC input voltage and seeing whether or not the output starts ringing above a certain voltage. Keep me posted with what you find! 

    Thanks, 

    Aidan

  • Hi, sorry it's been a couple of weeks but I still haven't gotten around to this! It's high on the list though and I hope to get to it between holidays.