• State Resolved
  • Locked Locked
  • Replies 7 replies
  • Answers 1 answer
  • Subscribers 62 subscribers
  • Views 918 views
  • Users 0 members are here
Support feedback
Options
Options
Related

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.

LMG5200: Bootstrap supply

beat Ronner
Intellectual 405 points
Part Number: LMG5200


Dear TI

During last week, I was detecting a small oscillation in my converter output voltage. The frequency is depending on the operating point and was in my case around 20 kHz. It took some time to follow it to what seems to me to be the origin, which is the bootstrap-voltage (between Pin 2 and 3 of the chip). When changing the capacitor connected to these Pins from 220n --> 100n, the frequency changed to approx. 30 kHz.

While the bootstrap voltage clearly shows this frequency component, the VCC voltage on the chip (Pins 6 to 7) does not show this frequency and is pretty clean other than some needles in the switching instants.

When looking at the waveform, I was surprised to find something almost triangular. It looks like the behavior of the bootstrap-circuit is changing at a distinct upper and lower level. See the two images above, once in the full recorded time scale, once zoomed in.

Is this something known to you? Any tips how I can prevent this?

As indicated above, I find the same frequency in my output voltage with about 20 mV pk-pk. This may not seem a lot, but for our high precision power supply for a particle accelarator, this is already a substantial distortion.

Regards

Beat Ronner

  • 0
    TI__Prodigy 645 points

    Hi Beat,

      I think there are two different things happening here that are superimposed on each other. One is the apparent step-wise variation of the bootstrap voltage as the switch node changes from low to high. The other is the slower triangular wave superimposed at a slower frequency which varies from 4.85 to 5.1V. 

    The high-frequency step-wise variation is most likely a measurement artifact. The bootstrap cap should be large enough to prevent such step-wise movements (i.e. producing these step-wise voltages on a the cap would require tens of amps of current or more during the transitions). Therefore, I suspect measurement errors. If you are measuring this via two single-ended probes and subtracting them, I'd recommend you check the compensation. If using a differential probe, you may be exceeding the common-mode range or other similar problem.

    The lower-frequency triangular wave superimposed on top is the normal behavior high-side bootstrap "clamp" circuit in the driver. The clamp is not truly a clamp in that it does not shunt current because that would result in high losses. Instead what it does it it detects when the bootstrap voltage is getting too high and shuts off a switch in series with the bootstrap diode, and then when the voltage drops back down, it turns it back on. The lower and upper limits of the triangular wave are the rising and falling thresholds of the clamp and the voltage goes up and down between them as the bootstrap diode turns on and off. Increasing the bootstrap cap will slow down the triangular wave frequency but won't change the magnitude. Changing the dead time or other switching parameters which affect how much charging is done per cycle will also affect the triangular wave.

    If the triangular wave is a problem in your application, you have a few choices:

    1.) Disable the "clamp" and provide some other means of limiting the bootstrap voltage externally. You can put an external bootstrap diode (which doesn't have the switch in series with it) to disable the clamp, but now you can over-charge the high side. You may be able to add an external shut regulator to limit the voltage with the penalty of the excess loss in the regulator.

    2.) Supply the high-side from an LDO or similar to provide a known fixed regulated high-side voltage. For example, maybe you have a 12V rail in your system, you could use an external bootstrap diode to make a HS+12V rail, then use a LDO to produce a HS+5V rail for the high-side.

    Regards,

    Nathan

  • 0 in reply to Nathan Schemm
    Intellectual 405 points

    Hello Nathan

    Thanks for your fast and competent answer.

    Yes, I was targeting the lower frequency triangular wave, and what you describe is about what I thought when I saw the waveform.

    What happens is that the on-switching of the FET driven by the bootstrap voltage either lasts a little longer when the bootstrap voltage is lower, or is delayed a little more. What I see is that the corresponding edge of the FET halfbridge output voltage shifts back and forth in the same rhythm as the bootstrap voltage rises or falls. This results finally in my output voltage varying by about 20 mV. This may not seem a lot, but for our application it is already well noticable.

    So yes, we will decide whether we can live with the variation, or whether we need to take one of the measures you proposed.

    Regards

    Beat

  • 0 in reply to beat Ronner
    Intellectual 405 points

    Hello Nathan

    Sorry of still getting back after having marked the issue as "resolved".

    I noticed that I only get into this mode where the bootstrap-voltage varies between the 2 levels above a certain current of my converter. For low modulation index and/or current, the bootstrap voltage is nicely constant. What's funny is that the level changes with duty-cycle (or current?). At duty cycle 50:50 (modulation index m1=0), the bootstrap voltage is at roughly 4.2.. 4.3 V, and rises towards 5V while the modulation index goes against 0.24. Above that, changing between the two levels starts.

    Do you have an idea why this could be, i.e. why is the bootstrap-voltage operation-point dependent?

    Does this have to do with how nice my VCC voltage is? It's already buffered with a C according to your layout guidelines, but maybe I can do even better and prevent uBootstrap from going into the 2-level mode?

    Regards

    Beat

  • 0 in reply to beat Ronner
    TI__Prodigy 645 points

    Hi Beat,

     I'm assuming this is acting as a buck converter? If so, the most common way that the bootstrap can get charged up to 5.1V for the bootstrap clamp to kick in is charging during the dead times. During the dead times, the SW drops below ground by a few volts and the current through the bootstrap diode will shoot up to an amp or more. Therefore whether the clamp gets engaged depends on how long the dead times are and how low the SW goes during that time. As the current goes up, the SW gets lower during the dead time increasing the bootstrap charging. Also, the soft-switching edge time gets shorter which makes the dead time during that edge longer. If you examine the time the SW spends below ground during the dead-times as you change your current and modulation index, I suspect you will see that you only see this with longer dead-times and when SW is more negative.

    Regards,

    Nathan

  • 0 in reply to Nathan Schemm
    Intellectual 405 points

    Hello Nathan

    Wow..... yes, I think you hit the bull's eye! Thanks for your competent answer. Makes sense.

    I have dead-time already as short as I dare (12 ns) considering the tolerances of the LMG5200 and my signal isolators, so unfortunately not a lot to get out there. But at least I understand now what's happening.

    Thanks for your competent help.

    Regards, Beat

  • 0 in reply to beat Ronner
    Intellectual 405 points

    Hello Nathan

    Can we just connect a voltage-limiting element (Z-diode, TVS, ...) to the bootstrap pins (HS, HB), or will this overload the internal bootstrap diode? What is the limit (RMS, ...) of the internal bootstrap diode?

    In case we need to connect an external diode: Can you give us a recommendation for a suitable element?
    What parameters are relevant? Sure low Qrr, but otherwise?

    Regards
    Beat

  • 0 in reply to beat Ronner
    TI__Prodigy 645 points

    Hi Beat,

      You probably can do that, as mentioned in option (1) above, you will want to use an external bootstrap diode and an external clamp. The external clamp will be fighting against the external bootstrap diode but probably not the internal bootstrap diode as that will disable itself. Typically a schottky diode is good for the bootstrap for Qrr reasons, but the low-voltage drop of the schottky is going to hurt you in this application, so maybe a small silicon diode would be better. You want low Qrr and low capacitance for the diode.

    For the clamping element, you need to stop HB-HS from going above 5.25 V for long-term reliability concerns of the Vgs of the GaN. I don't have any experience running the part in this config so I'm not quite sure what would work well, but as long as you can keep the HB-HS to 5.25V or lower, you should be good.

    Regards,

    Nathan