• State
  • Locked Locked
  • Replies 5 replies
  • Subscribers 63 subscribers
  • Views 917 views
  • Users 0 members are here
Support feedback
Options
Options
Similar topics

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.

UCC28950: problems with interleaving two UCC28950

Dmitry Noskov
Prodigy 40 points
Part Number: UCC28950

Hello,

We're currently designing a Full Bridge ZVS 4800 W converter with two UCC28950 (coupled as master/slave interleaving).
Power stage of the converter presents two PHFB DC/DC with center-tapped rectifier scheme tied together by the two output inductors.

We have an external voltage amplifier for feedback (based on MC33272ADG) and both UCC28950 are connected as voltage amplifier in a voltage follower configuration, as simplified shown below:


Both master and slave (when used as master) parts of converter work proper separately and each of them converts 800 V DC input into 24 V 100 A DC output. Both UCC28950 have exactly the same resistors (DELAB, DELCD, Tmin, Rsum and others) and parameters of master and slave power stages are very similar (Lmag, Lout, Rcs and others) – in fact we have two identical converters placed on joint PCB. The only different thing is that one part of converter starts drop output voltage at Iout = 105 А and another starts drop output voltage at Iout = 115 А (when they work separately). However, when we launch both parts in master/slave interleaving configuration, we have two main problems:

1) master and slave converter work without 90 degree phase shift, they have zero shift;

2) current sharing between master and slave converters is very bad.

Here are CS signals from master and slave parts, total ouput current is 55 A.


About the first problem: we have 90 degree phase shift between PWMAmaster and PWMAslave only if we extract control board (on control board we have 2 UCC28950, STM8 and some operational amplifiers) from power board and power this board with 12 V. But when we use control board connected to the power board the phases of PWMAmaster and PWMAslave are exactly the same. Is it possible to EMI to interfere synchronization process this way?


PWMAmaster and PWMAslave when control board is extracted


PWMAmaster and PWMAslave when control board is connected to the power board

About the second problem: we decided to break synchronization line between two UCC28950 and configure both of them as masters to check if they share current equally, but they don’t! It is confusing because EA+ level is the same for two UCC28950, additional ramp is the same (Rsum configured for 0.147 V/us slope compensation), but CS trigger level differs.


Master converter works on 81 A load separately


Slave converter (configured as master) works on 81 A load separately


Both converters work on 81 A load as masters (sync wire is disconnected). Phase shift between converters are constantly changes because each conveter has its own Rt and they are not synchronized.

Best regards, Dmitry.


  • 0
    Prodigy 40 points

    Got some issues with attaching pictures to original post, so I made separate PDF for them.oscillograms.pdf

  • 0 in reply to Dmitry Noskov
    TI__Mastermind 25570 points

    Hi Dmitry

    I'll have to think some more about the current sharing issue but here are some comments.

    The Slave unit should have a 825k resistor added between SS/EN and GND

    The UCC28950 SYNC signal has to be at least 1.8 times the controller's free running frequency. It might be that your slave device is losing synchronisation so I would suggest that you reduce the free running frequency of the slave by about 10% or so and see if its behaviour improves. I would also double check the amplitude and waveshape of the SYNC signal.

    It might be that there are differences in the GND potential of the two controllers so that the effective EA+ level is slightly different.

    Please let me know if you have any further information and in the meanwhile I'll think about this issue some more.

    Regards
    Colin

  • 0 in reply to Colin Gillmor
    Prodigy 40 points

    Hi Colin

    Thanks for your concernment

    Slave do has resistor 825k in parallel with SS capacitor.

    About timings: unfortunately I don’t have opportunity to place my oscillator probe to the SYNC signal when control board is inserted into power board due to power board’s construction complexity.

    But I get picture of SYNC signal when control board is extracted and supplied by external 12 V source and it’s ok.

    GND potential of UCC’s should be equal, because they placed close to each other and there is GND plane in control board in the whole inner layer. CS1 and CS2 signals traced as differential pairs with ground from RT1 and RT2 resistors on power plate to UCC’s inputs. Connections of CS1ground with ground plane and CS2ground with ground plane takes place near the UCCmaster and UCCslave respectively.

    I even wrapped the control board with copper foil and ground this screen before inserting into power plate – nothing changes.

    I found some new issues:

    1)      So when control card is extracted and supplied from 12 V lab source – phase shift is ok. Always. I turned on and turn off supply many times and there always was phase shift.

    2)      2) When control card is inserted into power board and 12 V lab source connected to the supply capacitor of power plate – phase shift may take the following values: zero, 90 degrees and 180 degrees. So when I turn on and turn off lab source many times I have one of this shiftes. And 90 degrees shift occurs very seldom – about one time per 5-8 supply resets. No 800 V input dc is applied in this experiment, so my own flyback converter (which gives 12 V supply in normal mode) is off and no EMI interference. The lab source is the same as in case 1.

    3)      When control card is inserted into power board and 800 V input dc is applied – corresponding transistors of master and slave converter work with 0 phase shift. I reset 800 V input about 10 times – nothing changes. They always work with 0 shift.

    Look at the attached file for oscillograms and photo of control card.

    Tomorrow I will decrease slave free running frequency and check whether anything is changing either.Figures.pdf

    Best regard, Dmitry.

  • 0 in reply to Dmitry Noskov
    TI__Mastermind 25570 points

    Hello Dmitry

    Thanks for the added information.

    It will be interesting to see what happens when you reduce the free running frequency.

    From the pictures you sent it looks like the two controllers do actually synchronise each time but not at the required 90deg phase shift.

    From the image of the board you sent it looks like the OUTx signals are routed under the IC, this isn't ideal. It is possible that these lines are coupling noise into the controller. I would check this by removing the controllers from the layout and placing a piece of (insulated) grounded copper underneath their locations and then putting the controllers back.

    Do try to probe the SYNC line if you can - and yes I know it can be difficult.

    Please let me know how you get on.

    Regards
    Colin

  • 0 in reply to Colin Gillmor
    Prodigy 40 points

    Hello Colin
    I reduced slave controller’s frequency down to 90 kHz (master runs at 97,6 kHz) and nothing changed.
    But now situation when PWMAmaster and PWAslave have 90deg shift become rarer (about one time per 10-20 supply resets).
    Have some oscillograms with SYNC signal.

    Experimental conditions are the same: control board is inserted into power board, no high voltage is applied, only 12 V lab supply for IC’s.

    Figures SYNC.pdf

    Regards
    Dmitry