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.

UCC3895 Problem with PWM Duty Cyle linearity

Other Parts Discussed in Thread: UCC3895, UCC21520, UCC2895, UCC28951-Q1, UCC28950

Hello All,

I have a question abouzt the UCC3595. We using it at our 5kW Power supply and it work in Voltage Mode (nor current feedback for the Ramp, but at moment I also try to operate in current mode). For Compensation we using external Operationampilifier circuits. The supply can change his output voltage and output current from 0% to 100%. Theoretical all work fine. But I found one problem, At some load points I get hard problems with switching noise, first I was thinking is is a problem with may layout and the compensator, any way (I can not exclude the Layout at this moment, but I will make a redesign after reading SULA501).

No I have make it very simple for testing and evaluating. Instead of my compensation and control circuit, i using a simple resistiver divider to set the duty cycle to check the behavior at different load points. Now I found one point which I not can easy explain.

If I change the control voltage from 1.33V up to 1.35V I get an jump at the duty cycle on time from 1.3µs to 3.9µs (periode time is about 12.5µs). At the voltage range lower 1.33V and higher then 1.35V there is no jump I can easy adjust the duty cycle. This is a very smal range of set voltage, but it has a high influence of the output.

If you want I can also send you some more information about my circuit. I can also send you some scope picture of the above descripted behavior.

I hope you can help me with this issue, or you can give me a advice or idea which is the possible cause of me problems.

Best regards, and a nice weekend.

Sascha

  • Hello Sascha,

    I will alert the appropriate support team to address your question. I do believe that screenshots of the behavior and circuit information will be helpful, if you can provide that.  Click on the "Rich text format" roll-over at the bottom-right of the post window to enable the paperclip button for attaching files.
     
    Regards,
    Ulrich

  • Dear Ulrich,

    thank you for fast reply.

    I found the solution for the above descripted Problem, I can solve it with the document SULA275, if I add the offset to the ramp this behavior at my simpel circuit was missed, this is good.

    But now back to my main Problem I still have this behavior at higher voltage devices (Vout 1500V). I have now also read the SULA501 and I get now a new Idea what could be the problem. At my design I have only one Goundplane, no seperation between control and power ground, but I have a low impedance between VCC and GND (distance between plane 100µm). (I think I get this kind of problem because I have very high voltage and current transients at my application).

    Here are some points about my desing:

    - I using the UCC3895 for generate the pulses

    - My control circuit consist of Operation ampilfieres and I have over all 3 Controlloops (Output current, Output voltage, power balancing for synchron devices)

    - I have MOSFET driver direct on the control card (but they dont have to deliver high power)

    - I have the high power MOSFET driver on the mother board how drive the Transformer for the bridge MOSFET's (distance to UCC3895 may be 3cm)

    - My hole control circuit gound is referenc to the secundary side

    What I found now was, if I remove the driver on the control card and put some resistors with 10kOhm in series to the driver of the mother board I manage to reduce the fault behavior, but it is still there.

    One thing I have already done, I make a new power module (mother board) with isolatet gate driver (UCC21520).  with this I think it is possible to reduce the noise at the Control card side because the Driver current will now referenc to Primary side (also it is possible to reduce the distance between driver and bridge MOSFET will hopefully improve the switching behavior :-) ). I will test this next week.

    Now back to the control card, I will make the redesign like I was intendet above, but may be it could be also work if I put some high impedance (buffer circuit) between driver and UCC3895. In this case I have no high current at the control card side.

    What do you think is this a possible way or do you recomende to do the seperation between the Gound planes? May be you can also recommend me any application note with more EMC issue how are related to my problem.


    Here are the picture which I was make during my search: yellow VCC; magenta Driver current; green bridge voltage; blue primary current

    Same behavior, but measured of ramp(yellow) and driver signal(magenta) (for your information I will also try to start working in current mode; see at ramp)

    Thank you for help and have a nice weekend.

    Regards

    Sascha

  • Hi Sascha,
    If I understand you have a 1500V output and you have no separation between the power ground and the signal ground.
    You have a control card with the UCC2895 that drives a power stage through a connector.
    When you open the control loop and generate a variable voltage on EA+ you have a point of operation where you get the assymetric signal on C4 above. Is that correct?

    Can you please disconnect the synchronization feature. Run the test at a fixed load and vary the EA+ voltage.
    You should try to determine which control feature is causing the problem.
    Is the issue worse at high or lower loads ?
    Does it just happen at a particular input voltage?
    Is it related to synchronizing the devices?
    I have not used this particular part before but I think this issue is most likely related to the high voltage switching noise on the output coupling into the low voltage control lines. I would look at decoupling of the control signals and of course keeping high dV/dt signals away from low voltage bias supply and control lines.
    Regards,
    John
  • Hi John,

    thank you for reply my question, I was a littel busy with the new board, but now I'm at the same point. First I answer your question, I copy your question.

    If I understand you have a 1500V output and you have no separation between the power ground and the signal ground.  -> YES

    You have a control card with the UCC2895 that drives a power stage through a connector.  ->YES

    When you open the control loop and generate a variable voltage on EA+ you have a point of operation where you get the assymetric signal on C4 above. Is that correct?  ->YES

    Can you please disconnect the synchronization feature. Run the test at a fixed load and vary the EA+ voltage.  -> Done

    I have disconnect my control loop and had only add a potentiometer for control my duty cycle, I get by different load, different Input voltage the similar problem. -> I get the same Problem how is descripted at SLUA275, but I can not fix it with the recomendet solution.

    Is the issue worse at high or lower loads ? -> same Problem

    Does it just happen at a particular input voltage? -> Input was change from 500V up to 622V -> same problem

    Is it related to synchronizing the devices? -> I have disconnect the syncronisation -> no change

    Conclusion:

    I have 2 Problem

    1. I get some EMC Problem if I close my loop, but this I think like you it is because my control signal get some spikes from the high voltage switching, I have also a long line (about 40cm) how connect my control card with the controller (output voltage and output current controller) (I need it because we parralle more then 10 units...). But for this issue I will try to make a differential signal transfer for improve the noise immunity... may be this solve this problem.

    2. I get this duty cyle jump and between this jump, I get the asymetrical Signals OUT C and OUT D. This is my main problem at the moment because it looks like is depending from the IC. Here are some of my trys with circuit diagramm an Waveforms.

    I working only at voltage mode. This transistor circuit should add the ramp with an DC offset to the ramp pin

    Here I have increase the DC-Offset like it was suggested.... By the way may be you can expalin why I get this negative spike at the ramp, If I reduce the 680Ohm resistor I can reduce this spike, but I don't know why it is there...

    Here you see my change of the EAP-Voltage (1)

    Change of the EAP-Voltage. You see I have write the output voltage change of about 60V by change the EAP-Voltage only about 60mV.

    Of cource I have a big transformer ratio this will result in a big output voltage change by a littel EAP-Voltage change. But the change are not linear it will change with a jump from one point to the other. If I want to control the voltage betwen this point, I will get some serious problems.

    Here I have change my circuit, that it work like recomended at the SLUA275, but I get the same behavior.

    I also try rigth now to remove my Transistor circuit and add the offset direct to the RAMP by a singel resistor (24kOhm resistor at the picture above) and connect the CT Pin direct to the RAMP pin but in this case the operation frequency will change also.... I will look tomorrow closer to this issue...

    But may be you have any idea what could be the reason for this streange behavior of the PWM IC. May be you can also give me some more detail information about the RAMP input circuit, may be this will help to find a solution.

    I hope you can help me with this issue.

    Best regards

    Sascha

  • Hi John,

    I have made now some more tests and it looks like this is the problem of the UCC3895. If I want to change the Duty cycle between 0% up to 100%, at some regions I get this no linearity this will make the output instable. Of course the controller (my compensator) try to set the correct voltage but the PWM jump with his duty cycle and a stable operation at this point is not possible.

    Special if you have a big output voltage range and so a high “resolution” for different duty cycle adjustment then this problem will influence the output stability…

    Now my question is, is it possible to solve this issue (it is offerd at the data sheet)? If yes how can I make it?

    If not:

    TI also offers a newer chip UCC28950 and UCC28951-Q1, they have 2 additional output for the secondary side and the have some light load management function like burst mode. Is it possible to reduce the duty cycle to 0% at this device or do the burst mode not allowed this? Next point is here also any regions how the duty cycle jumps?

    I hope you can give me any advice with this issue.

     

    Regards

    Sascha

  • Hi every one,

    can you give me any advice about my problem. I need following information from your side:
    1. Do the UCC28950 have a similar problem with the duty cycle at low ratios?
    2. Have you any idea what I can do to solve the Problem by the UCC3895?
    3. If you don't know a solution for the UCC3895, may be you can give me a more detailed
    circuit diagram of the input circuit of the RAMP Pin. This will help to understand the behavior more
    and maybe I can find a solution by myself.

    I hope you can help me.

    Regards
    Sascha
  • Hi Sascha,

    Sorry to hear that you are still experience problems with the UCC3895.

    One think I notice is that you are generating the ramp signal from an AC coupled oscillator timing capacitor.
    Any you apply this signal to the base of a bipolar transistor that has a DC bias also on the base.

    This is significantly different from the EVM circuit where the oscillator signal is applied directly to the base of the bipolar and a fixed DC bias of about .05V is applied to the ramp pin. It may be that the ramp you are generating is not quite so linear.

    Have you tried the circuit used in the EVM?
    The comparator circuit has an offset of 0.8V nominal so once the EA output reaches 0.8V you should get a linear 0 to 100% duty cycles as EA out increases.

    If this is not happening it could be:

    1. The PGND noise is entering the low level GND signal
    2. The ramp signal is not linear.
    3. The EA out signal is not stable.

    These are just the ideas that come to mind and they should be investigated.

    If you want to email me the schematic and any other test waveforms you can contact me at johngriffin@ti.com