LM5113: LM5113 destructive glitch question

Well, as I said, I can virtually guarantee that I will see a 0ns pulse-width.

Out of curiosity, why do your GaN drivers behave differently than your MOSFET drivers? I have never had problems at clipping with, say, your UCC27200.

John,

The LMG family and the rest of the MOSFET drivers use different processes and technologies.

Our GaN drivers are obviously suited for much higher switching frequencies and narrower pulse widths.

Regards,

-Mamadou

My deepest apologies, Mamadou, for not making my question clear.

In the thread linked in my first question, the original writer pointed out that the LM5113 would output a glitch when the input pulse width fell below a few ns. This glitch resulted in turning on the upper transistor while the lower one was still on, resulting in the destruction of the transistors. This is a normal occurrence in a Class-D audio amplifier. For example, in a typical application where the audio signal is compared with a tri-wave, the value of the audio signal may approach and then exceed the maximum peak of the tri-wave. This results in amplifier clipping. Therefore, the pulse-width will get narrower and narrower and eventually go away altogether. As long as the bootstrap capacitor is large enough, the high-side transistor will stay on and the low-side will stay off and the output voltage will limit at the high-side rail voltage.

In my experience, conventional half-bridge MOSFET drivers like your UCC27200 do not have any problem with this. However, also in my experience, the LM5113 GaN driver has a problem. When the amp hits clipping and the pulse width approaches zero, the half-bridge cross-conducts and destroys both devices. I had attributed this to the mechanism explained in the earlier-cited thread. So this is a matter of the much-slower driver behaving correctly, i.e. no glitch, while the faster device behaves incorrectly.

I was led to believe that this was anomalous to the LM5113 and wanted to verify that the LMG1205/10 did not behave like this. Can you verify?

I hope I am making my question clear. Please let me know if you would like further discussion.

Thanks John for the details,

Allow us to further look into this, we will get back to you soon.

Thanks for your patience.

Regards,

-Mamadou 

Thanks, Mamadou.

John, I expect the the LMG family to handle narrower pulse widths than the LM5113 which you saw the glitch issue at 5.3ns pulse width. As previously mentioned, LMG devices are our latest and most optimized drivers targeted for 1.8ns-4ns pulse widths. Now in your application there might still be a risk when the amplifier clips and the pulse width falls below 1.8ns in which case the glitch might occur.

As for the UCC27200, the IC structure is different than the LMG family and the LM5113 from a design standpoint with different input stage structures which might/ might not implement deglitch filters.

I need to put the LMG1210EVM on the bench to see what happens at when PWM <= 8ns. 

Regards,

-Mamadou

John, so far I have not been able to reproduce the previous issue described showing any glitches using the LMG1210EVM. 

The setup is with the LMG1210 running in PWM mode with as little as 10ns pulses showing no signs of glitches on HO and LO.

Waveform below shows the HO and LO with respect to GND. I will try to get to 1.8ns pulse width to verify the datasheet specification.

Regards,

-Mamadou

That is good news, Mamadou! Please see what happens as you go a bit lower. IIRC, the original poster found problems in the 3-8ns area. I hope all goes well!

John, Meanwhile, please refer to the Section of 6.6 where we specify pulse width operation at 1.8ns for LMG1210.

Regards,

-Mamadou

Hi, Mamadou. I do understand the minimum pulse width. When an audio amp just starts to clip, the pulse width will go to zero, but it doesn't matter that the output no longer responds. That's the nature of clipping. I just want to point out that this is a normal occurrence, so small pulse widths should be expected. They just shouldn't do anything destructive.

John, understood on the nature of the audio amp clipping.

Attached is waveform on the same EVM with 8ns pulse width in which case HO false low when the pulse width is trending toward 0.

I have not observed any destructive behavior in such conditions using the aforementioned LMG1210EVM in PWM mode with the GaN FETs populated. 

Regards,

-Mamadou

OK. Thanks, Mamadou. I will pick up an EVM to verify in my application.

Sounds good John.

Meanwhile, let us know if additional questions.

Regards,

-Mamadou