UCC28780: ucc28780

Hello Robin,

Thank you for choosing the UCC28780 ACF controller for your design.

I'm sorry to hear about the CPC3909 failure during open-loop testing.  Normally the UCC28780 maintains control of the switched-node sense Fet (let's call it Qs) by manipulating its gate voltage. First it allows Vg to follow Vs during the VDD charge-up phase to ~17.5V.  Then it pulls it down to 11V to perform switched-node voltage sensing for ZVS condition.

Without the controller in place, Vgs of Qs = 0V and there is nothing to control it when bulk voltage is applied.  Qs naturally turns on as intended and current charges up SWS cap C38 rapidly.  Per the TIDA-01622 design, there is a 20-V Zener clamp on SWS and a 24-V TVS on its gate to PGND. Since there is nothing to pull Vgs of Qs low to turn it off, Qs conducts to its normal 0-bias on-current of at least 300mA, limited by external conditions.

The CPC3909 is a 400-V device, so I assume that the peak switched-node voltage is lower than that. Assuming OV did not damage Qs, one can conclude that the high current, limited mainly by the 121R SWS resistor R35, and the bulk voltage during test overheated QS and blew it up. Probably as a short-circuit and that overheated R35 and burned it up too. For example, if Vbulk = 100V, then 300mA into 121R = 36V = 11W in R35 and 13.2W in Qs.

The power rating of Rsws in your design will determine the maximum steady-state voltage it can handle across it, and hence the maximum current through it. Since CPC3909 is only 6R when on, it does not limit current.

For open-loop test purposes in this particular case, I suggest to apply only a voltage to the bulk rail which is only a few volts above the SWS zener voltage to match the power rating of your Rsws. Then verify the correct current is drawn from your bulk source. For example, (~23V-20V)/121R = ~25mA which would confirm conduction of Qs and of Rsws and of the Zener. This is 0.5W in the Zener, so keep the test time down, too. That is the simplest method I can think of without resorting to in-circuit probing.

I forgot to mention that unfortunately, there is no easy way to get PWM from UCC28780 in an open-loop mode. It requires switching signals with specific timing edges at the VS and CS inputs. You pretty much have to emulate the usual signals. 

Regards,
Ulrich

Hello Robin,

I am not certain, but I think that it can "work" as you describe.  By work, I mean that you should be able to generate PWML and PWMH outputs if you can bring VDD up to 17.5V.  But all other components must already be in place, so how much can be gained by keeping out only the CPC3909?

Without the SWS signal, PWML will be fired by an internal time-out signal, and the PWMH tuner will extend the PWMH on-time in an attempt to get a signal at SWS, so the actual switching currents and voltages will be distorted from the expected ACF signals. To get PWMH, you have to be at a power level that reaches into ABM at least, so I'm concerned that the poor waveforms may cause some overstress somewhere. 

So although it may appear to "work" from a test point of view, it may have some unpredictable consequence.  To avoid this situation, you would have to emulate the waveforms at VS, CS and SWS inputs to the PWM signals without actually driving live MOSFETs. It's a lot of trouble to avoid tempting Murphy, but in the end, you still have to take the leap of faith and turn it on with the controller in place and he can still cause trouble even with all the pretesting you may have done. . 

I can say that our charger/adapter customers make this leap everyday thousands of times with each unit test, because they can't afford the time to prove out each subsection separately.  They just go for it.  If there were test failures we would hear about it, but (after design bugs are driven out) there are none.

So I suggest that, if you've already proven most of the circuit in an open-loop manner, then install both controller and CPC3909 and "go for it".
What I do is start at 3Vrms.  No smoke?... go to 10Vrms, then 20V then 50V while triggering on Vsw.  By this point, you should see some test pulses in the Vsw waveform. It shouldn't start until you reach the programmed "brown-in" level, but the 3 test pulses should give you confidence to keep increasing until it runs.  If  you increase Vac more than 10~20% above the expected brown-in level and it hasn't attempted more than 3 pulses yet, then investigate the VS brown-in settings. 

Good luck,

Ulrich

Hi Robin,

This is the same as the other post which I just answered:

So I will close this one.

Regards,
Ulrich