TOPSwitch-JX application phase margin too low

Would you mind pasting the design file so I can review.

Thanks PI-Chekov

Hello, please find attached the design file as requested...
Regards, David.

There are a couple of things about your design that I can point out
1) The output voltage is 48 V which will prohibit using TL431 (which has a max voltage of 36 V I think. So you'll have to use an op-amp instead. The model of the loop will be the same its just the limitation of the TL431 that's all
2) You can try to limit your crossover frequency down to a little below 1 kHz entering something like 700 Hz as your Crossover target frequency
3) Try increasing the value of CF1 to 68 or 100 nF.

Hope that helps.

Hi PI_Spock

You kindly responded to me earlier in the year and we are now ready to move forward with the changes you suggested.
We have managed to get the phase margin to 44.88% using your suggestion, which is good, however we had not reviewed the design at that time and noticed that the TL431 was only a 36V part. Is there a relevant application note or schematic that details the implementation of the opamp shunt regulator as we are now ready to move forward with this design.

Best Regards

David

Hi PI_Spock

Would you kindly respond to the last post left on 21st May, this is now urgent to allow us to move forward.

Best Regards

David

There are a couple of ways to address this
1) You can add a series zener diode with the opto-coupler to shit the DC voltage as seen by the TL431. This will not require a tweak to the loop design other than adjusting the value of the resistor in series with optocoupler to macth the DC gain that you are seeing right now.
Does that make sense?.

2) Or - You can use an traditional opto-coupler as your error amplifier - It will need its own bias voltage and may require recompensating the loop.

Hi PI-Spock

Option 1 is by far the simplest so we will go with this one. So what you are saying is that we can insert a 15V zener [to reduce the 48V currently present on the TL431 to 33V] into the feedback chain between the TL431 and the opto-coupler and then adjust RF6 [shunt regulator bias resistor] to compensate for this addition? What would be the calculation to determine the new shunt resistor and keep the current gain ?

Regards

David

I wish I could give you a straight answer to this one... I'm afraid the easiest way is trial and error. Try to maintain the same DC bias point in both cases. What I mean is that if currently your RF6 is 12 k (say) then this 12 K will correspond to some DC bias current (48 - 2.5 - 1.1) / 12 k = 3.7 mA
Now if you use a 15 V zener you should try to size the resistor such that you have the same DC bias current. Thus you may have something like this
(48 - 15 - 2.5 - 1.1) / 3.7 mA = 7.9 k

Does that make sense?

Hi PI-Spock

We have integrated the zener diode, see attached schematic picture, and just wanted to double check that this is correct. The current value of R26 was calculated by PI Expert so I guess from your calculation that the output voltage with the 15V zener in series will be 48-15 = 33V. We will then need to adjust R26 to increase the loop gain and raise the output voltage back up to 48V, is this about right ?

Regards

David

Yes it looks right. You wont have to adjust the output voltage - only the loop gain. This is done by changing values of R26 (and subsiquently R27 and C31)

Let me know how it goes...