TNY274 Design - Improving Output Ripple Performance

Hello Chris,

The problem you described shows me TNY274 is not the source for this weird behavior. Worst case scenario is for output overload, but then you get a restart cycle, not high ripple.
This prompted me to check the optocoupler loop. HCPL-181 has photodiode forward voltage between 1.2V to 1.4V for 20mA forward current. Adding 2V from your zenner you need 3.2V to 3.4V to activate the optocoupler, not counting R5. But your whole output is 3.3V and the control loop is disabled. Pay attention to the control loop, you will get it right.

Cheers,
PI_Crusher

Hi PI_Crusher,

Thanks for your response and for the recommendations. I see how my control loop was not ideal and how there could be voltage regulation issues with that configuration based on the component tolerances. With that said, I'm not exactly sure what you mean by the control loop is disabled. Even with the worst case photodiode forward voltage and zener error, the larger control voltage would simply mean my output voltage would be greater than I initially expected. This would not cause extra ripple or am I missing something?

I revisited the control loop and made a couple changes, but still can't seem to solve this ripple issue. I've replaced the optocoupler with a low-current PCN367NJ0000F which has a much better current transfer ratio. This should improve the response time of the control loop and also minimize the voltage dropped in the resistor. I also tried changing to a 1.8V zener (MMSZ4678), but other than changing the output voltage I don't see how this would affect the ripple.

The changes did improve the ripple somewhat but I still see at least 200mV in just about any output configuration and load condition that I've tried. I've attached a couple waveforms showing the ripple that I see. These are both taken with a 6.8uH/47uF LC post filter (I've played with the inductor value between 2.2uH and 10uH, but this doesn't seem to have much affect). Without the LC filter there are significantly larger high-frequency spikes present on the rising edges of the sawtooth pattern (some at least 0.5V).

Just to investigate the potential differences, I also tried using only the 5V rail removing D4, connecting R5 from the control loop to the 5V output capacitor (prior to the LC filter), and using a 3.6V zener. The results were basically the same as all of my other tests and the ripple wasn't noticeably different... still at least 200mV across various ouptut and load conditions.

I'm just curious, but, what are typical ripple voltages for this type of supply based on the TinySwitchIII and what is the best ripple performance I could expect to achieve? If you have any further suggestions they would be greatly appreciated. I'm not an expert and I'm really starting to run out of ideas (and time).

Thanks again,
Chris

Hi PI_Crusher,

Thanks for your response and for the recommendations. I see how my control loop was not ideal and how there could be voltage regulation issues with that configuration based on the component tolerances. With that said, I'm not exactly sure what you mean by the control loop is disabled. Even with the worst case photodiode forward voltage and zener error, the larger control voltage would simply mean my output voltage would be greater than I initially expected. This would not cause extra ripple or am I missing something?

I revisited the control loop and made a couple changes, but still can't seem to solve this ripple issue. I've replaced the optocoupler with a low-current PCN367NJ0000F which has a much better current transfer ratio. This should improve the response time of the control loop and also minimize the voltage dropped in the resistor. I also tried changing to a 1.8V zener (MMSZ4678), but other than changing the output voltage I don't see how this would affect the ripple.

The changes did improve the ripple somewhat but I still see at least 200mV in just about any output configuration and load condition that I've tried. I've attached a couple waveforms showing the ripple that I see. These are both taken with a 6.8uH/47uF LC post filter (I've played with the inductor value between 2.2uH and 10uH, but this doesn't seem to have much affect). Without the LC filter there are significantly larger high-frequency spikes present on the rising edges of the sawtooth pattern (some at least 0.5V).

Just to investigate the potential differences, I also tried using only the 5V rail removing D4, connecting R5 from the control loop to the 5V output capacitor (prior to the LC filter), and using a 3.6V zener. The results were basically the same as all of my other tests and the ripple wasn't noticeably different... still at least 200mV across various ouptut and load conditions.

I'm just curious, but, what are typical ripple voltages for this type of supply based on the TinySwitchIII and what is the best ripple performance I could expect to achieve? If you have any further suggestions they would be greatly appreciated. I'm not an expert and I'm really starting to run out of ideas (and time).

Thanks again,
Chris

Hi PI_Crusher,

Thanks for your response and for the recommendations. I see how my control loop was not ideal and how there could be voltage regulation issues with that configuration based on the component tolerances. With that said, I'm not exactly sure what you mean by the control loop is disabled. Even with the worst case photodiode forward voltage and zener error, the larger control voltage would simply mean my output voltage would be greater than I initially expected. This would not cause extra ripple or am I missing something?

I revisited the control loop and made a couple changes, but still can't seem to solve this ripple issue. I've replaced the optocoupler with a low-current PCN367NJ0000F which has a much better current transfer ratio. This should improve the response time of the control loop and also minimize the voltage dropped in the resistor. I also tried changing to a 1.8V zener (MMSZ4678), but other than changing the output voltage I don't see how this would affect the ripple.

The changes did improve the ripple somewhat but I still see at least 200mV in just about any output configuration and load condition that I've tried. I've attached a couple waveforms showing the ripple that I see. These are both taken with a 6.8uH/47uF LC post filter (I've played with the inductor value between 2.2uH and 10uH, but this doesn't seem to have much affect). Without the LC filter there are significantly larger high-frequency spikes present on the rising edges of the sawtooth pattern (some at least 0.5V).

Just to investigate the potential differences, I also tried using only the 5V rail removing D4, connecting R5 from the control loop to the 5V output capacitor (prior to the LC filter), and using a 3.6V zener. The results were basically the same as all of my other tests and the ripple wasn't noticeably different... still at least 200mV across various ouptut and load conditions.

I'm just curious, but, what are typical ripple voltages for this type of supply based on the TinySwitchIII and what is the best ripple performance I could expect to achieve? If you have any further suggestions they would be greatly appreciated. I'm not an expert and I'm really starting to run out of ideas (and time).

Thanks again,
Chris

Hello Chris,

You’re right; the larger control voltage will give you higher output voltage. But there is another detail related to the zenner knee close to reverse conduction. If the current trough the optocoupler goes from high values to zero over a sharp knee, the optocoupler is not changing the feedback soft enough; rather you have the control loop going from conduction to disabled state. The waveforms you posted are a good hint to suspect this behavior. From time to time the output voltage exhibits a very sharp increase of more than 300mV.
You can try to soften the feedback knee. Use a higher value for R5; with incremental changes you can try go up to 400 ohm. Also, add a second resistor in parallel with the U3 diode (pin 1 to pin 2). You can start with around 5k and test multiple values incrementally lower, down to 1k.
You have to adjust the values of these two resistors in order to have smooth switching cycles, without big variations. I estimate the output ripple to go significantly lower than 100mV, and even less with output filter tweaking.

Cheers,
PI_Crusher

Hi PI_Crusher,

Thank you for the recommendations and guidelines as to what I can expect. I'll try to implement these changes and hopefully this will resolve my ripple issues.

Just as an aside, are there any documents that deals with control loop design? Perhaps outlining design procedures, common problems, or best practices for the TinySwitchIII or even in general?

Chris

Hi Chris,

We provide many documents you can consult online, with all these details you need to improve your application. Check the specifications for the part you use and reference designs.
An example: http://www.powerint.com/sites/default/files/PDFFiles/rdr91.pdf, page 9.

Cheers,
PI_Crusher