Buck Converter PWM problems

[turtlepokerman]

Good Evening Forum,

I am currently working on designing a buck converter with the ir2117 bootstrap converter that feeds into a buck converter, a MIC4420 amplifier that feeds into the ir2117, and a pwm signal that comes from an arduino uno. I am monitoring the the voltage cycles at the output of the buck converter and they take 10 of the pwm cycles to produce one of the output cycles. Does anyone know why this is happening? I wondered if it had to do with the feedback network on the ir2117 only being able to trigger when the Vs is high, but I am not sure. Any help is greatly appreciated. As for measurements I have a PWM signal at 50kHz and an output of 1.1kHz.

Thanks,

 

[D.A.(Tony)Stewart]

have you read this app note?

Layout considerations and probe locations may not be overlooked.
Yes Vs must be high enough to permit low Ron.

 

[turtlepokerman]

Yes I have several times. One thing that I noticed that my circuit operated differently than the application note is the capacitor from Vcc to COM completely destroys my output signal. So I removed that capacitor. Could this be the cause for my troubles?

 

[D.A.(Tony)Stewart]

partly.

You will need to include schematic and layout for specific help, I think with some waveforms would help you explain what is wrong.

 

[turtlepokerman]

Another problem that my buck converter output voltage is highly dependent upon the frequency and capacitance at the output. Whenever I increase the capacitance in the buck converter the output voltage decreases with the gate voltage. I tried stepping down the frequency, but that only makes the ripple worse. How do I remove the ripple voltage without increasing the capacitance or the pwm frequency?

 

[BradtheRad]

I am monitoring the the voltage cycles at the output of the buck converter and they take 10 of the pwm cycles to produce one of the output cycles.

There is such a thing as 'hiccup mode' in a switched-coil converter. A single cycle makes the output level rise high enough, to be above the threshold for the switching device to turn on. As a result the switching device stays off for a few cycles. Hence the hiccup effect.

A delay somewhere in your feedback loop could cause this. Something that extends switch-On time, and thus increases ripple.

See if you can tighten up the feedback hysteresis envelope. That will reduce ripple.

Or see what happens if you reduce the smoothing (output) capacitor value. Or if you remove it. The operating frequency should get much faster.

 

[turtlepokerman]

BradtheRad, Do you have any documentation or articles discussing hiccup effect? I have looked on the web and cannot find any.

You are correct in your assumption that removing the output capacitance increases my switching frequency, but how am I supposed to reduce ripple from a higher frequency signal if I cannot use capacitance?

You also mentioned tightening up my feedback hysteresis envelope. How do I do this or do you have any documentation on how to do this?

Thanks for all of your help

 

[BradtheRad]

buck converter with the ir2117 bootstrap converter that feeds into a buck converter, a MIC4420 amplifier that feeds into the ir2117, and a pwm signal that comes from an arduino uno.

You have several devices in your control block. I have no experience with Arduino devices but if software is controlling the length of switch-On, then a time lag could come from cpu cycles occupied by unknown processes.

I wonder if your output waveform rises quickly, and falls slowly? That would allow your feedback devices to switch-On at a leisurely pace, but switch-Off will be late and will cause overvoltage at the load and load capacitor.

I believe you want to switch-Off earlier. This will reduce the hysteresis envelope, and thus reduce ripple.

 

[D.A.(Tony)Stewart]

have you read this app note?

Layout considerations and probe locations may not be overlooked.
Yes Vs must be high enough to permit low Ron.

The better your question, the better the answers.
Include full schematic, values, layout photos and measurements.

 

[turtlepokerman]

I got my circuit working, but I had to remove the large electrolytic capacitance at the output of the buck converter. I also found that using smaller non-electrolytic capacitors helped with output smoothing. This with choosing different capacitor values for the IR2117 between Vb and Vs helped increase the range of the voltages. The final question that I have for you is if you can put an electrolytic capacitor in series with a non electrolytic capacitor and still have the polarizing effects?

Thanks for all of your help so far.

 

[turtlepokerman]

Okay I had the circuit working, but as soon as I turned off all of the power supplies and restarted my circuit. The Buck Converter stopped working properly. When testing the gate and source voltage of the Nmos device for the buck converter, I notice that they are at the same voltage.

I also noticed that the output of the bootstrap amplifier is supplying the output of the buck converter. Should it be doing this or should their be isolation between the two sources?

Changing the capacitance out of either the bootstrap amplifier or the buck converter seems does not seem to have an effect anymore. Any help is greatly appreciated.

 

[FvM]

The better your question, the better the answers.
Include full schematic, values, layout photos and measurements.

Very true. For the time being, I'll assume serious mistreatment of poor IR2117 drivers as the most likely problem.

 

[BradtheRad]

I got my circuit working, but I had to remove the large electrolytic capacitance at the output of the buck converter. I also found that using smaller non-electrolytic capacitors helped with output smoothing. This with choosing different capacitor values for the IR2117 between Vb and Vs helped increase the range of the voltages. The final question that I have for you is if you can put an electrolytic capacitor in series with a non electrolytic capacitor and still have the polarizing effects?

Thanks for all of your help so far.

This is progress. Good going. Try different things and you're likely to hit on an improvement somehow.

If your operating frequency is high, and load is light, then you can get by with a smaller value smoothing capacitor. You may not need a big electrolytic.

It is typical for capacitors in series to acquire unpredictable charge levels when a DC waveform is applied. Hence you're better off using one capacitor, or capacitors in parallel.

 

[D.A.(Tony)Stewart]

THe ESR of the cap is also very important as well as ratios of boost and shunt caps.

 

[turtlepokerman]

Thanks everyone for the help so far,

I believe that I burnt out my ir2117, so it will be a couple of days until I can get a new one in the mail. It was strange though how the device burnt out. I had the circuit operating and then when I turned off all of the power supplies and turned them all back on the device did not want to work anymore. So far I do not have any good reason for why this happened. If anyone would have any idea please let me know so that I can avoid this from happening in the future. I will keep everyone updated once I can begin testing again.

Thanks again