UCC28950: DC level in current transformer

For the moment I can not buy a specialized current transformer, for this reason I try to work with what I find in the local market.
I have built a new current sensor with an ETD29 and a 3c94 material, but the problem still remains. Also change the RC filter by a 100 ohm resistor and a capacitance of 870pF.
In the next image is the startup time. In blue it can be seen how the signal of the CS pin does not reset to 0 volts, although the useful cycle is very small. The circuit activates over current protection, and starts over, but never set.

The connected load is 20 ohms at the output, which corresponds to 90% of the load.
The circuit always works in VMC. If I disconnect the CS signal, the circuit works, but in DCM. I do not know why the circuit prefers to run in DCM, and not reduce the useful cycle (this should ask in another post?).
I attach the schematic of the circuit.

Hi Yelsing

I don't see any ground connection for the CS transformer - at the left hand side of R18 for example - am I missing something because without this connection the CS signal will be floating and the controller won't get a proper current sense signal.

A second point - the problem is that the CS signal isn't correct on a cycle-by-cycle basis. For this reason you need to look at it at a much faster timebase - showing two or three switching cycles.

Please double check the grounding and let me know how you get on.

Regards

Colin

Hi colin.

I had not updated the schematic. The previous images were taken with the ground connection at the node connecting to R18, R19 and C9.
I do not understand the second point. Is taking a picture like this?

This image is the start-up.
I have found something curious. If I disconnect the current sensor circuit from ucc28950, the DC level problem disappears. What could it be?

I have looked at the circuit of example PMP8606. The current sensor is implemented in the transformer. When I run the simulation with the sensor in the transformer, the simulation does not work. What should I keep in mind to use this setting?
Thanks

Hi Yelsing
this screenshot is much more informative - at least now you can see the current peak and the individual switching cycles. If you could zoom into two or three of the switching cycles where the DC offset is present that may be even more informative. For example, the slope of the current signal is predictable - based on the output inductor and the magnetizing inductance of the transformer. You should see something like that on the two wider pulses at the left hand side of the plot above for example. Now, in the much narrower pulses, you can see that the current gets to 2V very quickly indeed - maybe due to some capacitive effect - the capacitance from primary to secondary of your CT for example ro possibly due to simultaneous conduction of the top and bottom switches in the bridge (shoot through). This then terminates the cycle prematurely. I'm not sure where the DC offset is coming from but I'd check the phasing of your CT - I think it's ok but it would be good to double check it. I would also check the voltage at the secondary of your CT - the voltage across the 51k resistor. The 1N4148 is a fast diode, which is good but is rated for 75V only. It may be getting avalanched during the reset time. We used a 4.87k resistor on our EVM for the UCC28950 and while the resistor value needed will depend on the CT and the time available 51k seems a bit high. Check the peak reset voltage and reduce the 51k resistor if possible.

I'd also check that you have allowed enough dead time (DELAB and DELCD) to prevent shoot through - its better to have too much dead time than too little - it can always be optimised later.

Putting the CT in the transformer is also a possibility - but you should be able to make your existing arrangement work ok so I wouldn't change to that yet.
Regards
Colin

I have changed the resistance value from 51K to one of 3.9k. I put such a great value of resistance because I read in a book that the bigger the value, the reset of the transformer will be faster, but the voltage peak in the diode will be bigger.
I put a start up image with a smaller time scale.

I take a picture of the voltage at resistor R18 (now 3.9K). I think the voltage is lower than the breakdown voltage of the diode.

I connected the ADEL pin to ground. The results are the same.

Hi Yelsing

Can you confirm that the UCC28950 has a ground connection at the output - the schematic shows a different symbol for primary ground (0V) and secondary ground. The CS signal ground needs to be the same ground as the UCC28950 is using of course.
the output potential divider is set for about 50V rather than 250V. If this is true then the system will command Dmin when Vout is about 50V rather than 250V and will operate at a much lower duty cycle than you expect.

From the waveform above - the PWM cycle is being terminated when the CS signal is getting to 2V - 2V is the 'second over current limit' at which the Cycle-by-Cycle current limiting comparator starts to operate.

In your second plot - the negative reset voltage isn't clear because the 'scope' has run out of range (you can see the flat bottoms on the waveform) but you are correct, there is a correlation between the resistance, the reset voltage. For now, I'd leave it at 3.9k.

Please double check the CS grounding.

Then - for diagnostics - you can run the system with the voltage loop open circuited ( Remove R66). Turn Vbias on and slowly increase Vin. the controller should drive the power stage at Dmax. You need to monitor Vout - but once it gets to about 30% of your target (50V or 250V) you should stop increasing Vin.
Then look at the CS signal -

Please let me know how you get on -

Regards
Colin

I have verified the connection between the ground of the output voltage, and the ground of the CS. It's ok.

I have a daughter card, where the UCC28950 is located, for this reason they appear with different earth symbols. Anyway I have verified the connection between the ground of exit, and the ground of the UCC28950, and this is ok.

The hardware is designed for a 250V output and a nominal input of 150V, but to test for the moment I am applying a voltage of 30V to the input, a 65V output, with a maximum power of 90W.

I've done what you told me. I have removed the resistance R66, I applied to the input 12V, and to the output I get 40V. I have measured the DC current at the input with a multimeter, and this mark 2.36A. At the output the current marks 0.650A.

Apparently under these conditions the sensor is reset, but I'm not sure if the current reading is correct.

Hi Yelsing

So - in the picture above the controller seems to be running at Dmax - that's fine for now.

There is a lot of undamped ringing on the CS signal at about 1MHz, but it looks 'reasonable' if you ignore the ringing. The ringing will mask the correct current sense signal and prevent the controller from operating correctly so the task is to clean up this signal.
Noise like this is usually due to poor grounding arrangements somewhere - the ground connection from the current transformer back to the controller should be low impedance - most easily achieved by a short, wide PCB trace. It may also be due to stray inductances in the path from the current transformer to the controller. It may also be caused by common mode currents flowing from primary to secondary of the power transformer - these will be worse if the power transformer has a lot of primary to secondary capacitance -
Do you have access to a clamp on current probe so that you can look at the current in the primary of your current transformer ? and compare it to the current in the secondary (the voltage waveform above).

Peak current mode control is useful because it simplifies feedback loop design and enables you to get better transient response. You always have the option to change to voltage mode control - you will need to add a DC blocking capacitor to the power transformer primary of course - but even if you do change to VMC you will still need a clean current signal from your current transformer to provide accurate Over Current Protection.

Regards
Colin

Hello colin
I have simulated the circuit with the cores that I have used, and with all these the offset error is presented. I have researched and this is due to the remanence of the core. No matter how much the resistor R18 increases, the core can not be reset.
The cores with the lowest remanence, are the iron cores. I have used a 125u sendust core of permeability, with 150 turns.
With this core there are no DC offset problems, but there is still a ringing in the signal. I have wound the toroid by hand, so I think that the capacitances and inductances parasites are what affect the signal. I decided to buy the current sensor B82801C2245A200, but it takes two weeks to arrive so I have to wait :(
I do not have a current probe available at the operating frequency, so it's up to me to blindly trust the CS signal.
When the CS arrives, I comment the results.
Thank you.

Hi Yelsing
That makes sense - I hadn't thought of that because CTs don't normally exhibit significant remenance.
Let me know how you get on.
Regards
Colin

Hi collin
I bought the current sensor B82801C2245A200, and I no longer have the problem of offset in the signal.
Now I want to eliminate the oscillation in the current. As you can see in the image below (blue signal), the current oscillates a lot. I do not think it's ringing due to poor grounding, because I made a single board for the current sensor, and I'm not connected to the ucc28950 CS signal. I have done simulations, and it is apparently by the parasitic inductor of the transformer, but I am not sure. What else can cause this oscillation in the current, and how can I avoid it?
Thank you.

Hi Yelsing

Glad to see you are making progress. There are a couple of things that could be causing this ringing so I'd suggest you try the following

Add a capacitor from the secondary ground at the source of the SRs to the primary ground directly at the 0V end of the H bridge - this is effectively a Y cap and its function is to prevent any common mode currents generated in the transformer from flowing in the CT primary circuit.

Double check your layout - have you got any capacitance or stray inductance in the CT primary circuit - especially capacitance at the end closest to the H bridge.

Are you using the clamp diodes shown as D5 and D11 on page 63 of the DS ?

Regards
Colin

The capacitor you tell me to add is the capacitor C13 of the application note SLUA560B? What value should the capacitor be? What is the voltage of this capacitor?
The image that I publish before, of the oscillation of the current is no longer the same. The current is now as follows:

I re-soldered all the points and the oscillation changed. The current I have now, I think is a bit more consistent than the previous one. But still has a rather large oscillatory current.
When I turn on the circuit, an oscilloscope that is about 5 meters detects noise. I think my circuit makes a lot of noise, and maybe this affects the current measurement. For example the heatsink should be landed at some point? And also the transformer should be grounded? I think this may help to lower the EMI, but I do not know at what point to land, should it be the primary land or the secondary land?

I am not using the diode clamp because I do not have a shim inductor.

Hi Yelsing

Looks like you had an open circuit somewhere.

You should connect the heatsinks to an AC ground somewhere - the best place is normally  the source of the primary low side MOSFETS on the primary side of the transformer. What happens is that the high dv/dt rates at the MOSFET tabs couple current into the heatsink. These currents then return through stray external capacitances and by radiating onto nearby surfaces - all of which increases noise. If you ground the heatsink, these currents return to ground by a much more direct path.

You should also look at grounding the transformer core - again to the source of the primary low side MOSFET. A copper 'belly band' will also help to reduce noise - again, connect the belly band to the MOSFET sources.

The same goes for the tabs on the secondary side rectifiers - ground their heatsink to 0V out.

Regards

Colin