Minimize Ripple in Rectifier by Waveform, Polarity, Bridge-Type, Inductor, Trafo

[theboom]

In a rectifier and RC filter, which combination of waveform, bi-polarity, inductor/cap, and transformer-type will give the least ripple in the output?

My source is 100 kHz AC supplying a transformer. I have control over the supply voltage, wave shape, and polarity, over the transformer type, and over the rectifier / filter design. This question is about how to minimize ripple through selection of those parameters.

Assume there is no regulator involved. Just AC source, transformer, rectifier, filter.

Ie, how is ripple affected by:

• an AC source that goes negative on the troughs, vs an AC source that only goes positive, never below 0v?
• sine vs square source?
• pulse transformer vs non-pulse transformer?
• full-bridge vs half-bridge?
• Use an inductor in the filter? Or just a cap?

ChatGPT says:
To achieve the least ripple in the output of a bridge rectifier and RC filter:

• Use a bipolar AC source (e.g., a sine wave) rather than a unipolar one, as it results in a higher ripple frequency. [specifically, 2 x f-in]
• Use a sine wave input rather than a square wave for a smoother, more manageable ripple. [specifically, no harmonics]
• Use a standard non-pulse transformer designed for power applications rather than a pulse transformer. [i need to understand more]
• Use a full-bridge rectifier instead of a half-bridge rectifier to take advantage of the higher ripple frequency produced by using both halves of the AC waveform. [specifically, 2 x f-in]

This isn't for safety. My source is already isolated from mains. This is for operational needs.

My load wants 5V @ 3A (after the filter).

 

[BradtheRad]

My source is 100 kHz AC supplying a transformer. . This question is about how to minimize ripple through selection of those parameters.
Assume there is no regulator involved. Just AC source, transformer, rectifier, filter.

A power transformer gives most satisfactory performance from being fed bipolar sine waves.

On the other hand a full-bridge rectifier provides smoothest DC if you feed it square waves where the positive waveforms are equal amplitude as the negative.

A capacitive filter tends to make output rise near the peaks of pulsating DC. A choke filter tends to average the same pulsating DC.

 

[theboom]

Thanks for these tips.

A power transformer gives most satisfactory performance from being fed bipolar sine waves.

"performance"? You mean efficiency?

You mean compared to a pulse trafo?

a full-bridge rectifier provides smoothest DC if you feed it square waves where the positive waveforms are equal amplitude as the negative.

You mean, a bipolar square that goes negative on the troughs?

So which one will have more ripple?

• bipolar sine into power trafo?
• bipolar square into power trafo?

A capacitive filter tends to make output rise near the peaks of pulsating DC. A choke filter tends to average the same pulsating DC.

Which one will have less ripple?
Thx!

 

[KlausST]

Hi,

bipolar square wave --> full bridge rectifier --> C Filter, RC filter or LC filter --> load

For sure all components need to be selected carefully.

Klaus

 

[theboom]

Hi,

bipolar square wave --> full bridge rectifier --> C Filter, RC filter or LC filter --> load

For sure all components need to be selected carefully.

Klaus

Thx, but did you forget the transformer?
What about all those square harmonics?

 

[KlausST]

Thx, but did you forget the transformer?

Sorry, you are right.

Update:
bipolar square wave --> transformer --> full bridge rectifier --> C Filter, RC filter or LC filter --> load

What about all those square harmonics?

In ideal case the rectifier output is pure flat DC. No fundamental, no harmonics. In ideal case not even a filter is needed.
For sure real life is not ideal, thus I wrote that parts need to be selected carefully.

If you use sine instead ... the rectifier current will only be on the peak parts of the sine. In the rest of the sine there is no current at all. During this time the filter capacitor needs to supply the load. This causes ripple.

Please use a simulation tool to play around to see the resulting waveforms and ripple.
There are online tools as well as free downloadable software. Easy to use. No excuse not to use them.

Klaus

 

[theboom]

In ideal case the rectifier output is pure flat DC. No fundamental, no harmonics. In ideal case not even a filter is needed.

That's just a matter of centering the squarewave vertically before it enters the rectifier, correct?

For sure real life is not ideal, thus I wrote that parts need to be selected carefully.

Square reduces trafo efficiency, correct?

thx

 

[D.A.(Tony)Stewart]

LC stores much more energy than the load in order to reduce ripple.

I think the least energy + ripple solution involves multi-phase Trafos and drivers.

 

[theboom]

multi-phase Trafos and drivers

something new for me to learn about! Something like this?

 

[KlausST]

That's just a matter of centering the squarewave vertically before it enters the rectifier, correct?

When talking about a bipolar square wave ... we are not discussing unsymmetric bipolar sqaure wave.

Square reduces trafo efficiency, correct?

Why do you think so? Provide your idea. Also what magnitude of efficiency/loss are you talking about.
In post#1 you ask about low ripple. Thus I answered about low ripple, not about high efficiency.

You also talk about RC filter ... here "R" is an equivalent for "loss". Every current through it will cause loss according P = I * I * R.
If you use square wave .... you (ideally) have no R (at the RC filter) at all, so no loss at this filter.

Again: simulation tools also tell you efficiency / loss.

Klaus

 

[theboom]

When talking about a bipolar square wave ... we are not discussing unsymmetric bipolar sqaure wave.

"In ideal case the rectifier output is pure flat DC."

You used the word "ideal" a couple times. What causes deviation from the pure flat?

Why do you think so? Provide your idea. Also what magnitude of efficiency/loss are you talking about.

"Square waves induce more Eddy Currents in the transformer core - generates heat and reduces efficiency. That's why we have Thinner laminations in Audio Transformers (how many know that there was such a thing ! ) and Powdered iron/ Ferrite in RF / Switching transformers."

https://www.reddit.com/r/AskElectronics/comments/ifw12d/comment/g2seuez

So i guess i should use Ferrite Core?

In post#1 you ask about low ripple. Thus I answered about low ripple, not about high efficiency.

True, my first goal is low ripple. But then, it has to be practical. If there's more than one way to minimize ripple, then my choice will be based on other factors, like efficiency and noise.

You also talk about RC filter ... here "R" is an equivalent for "loss". Every current through it will cause loss according P = I * I * R.
If you use square wave .... you (ideally) have no R (at the RC filter) at all, so no loss at this filter.

 

[BradtheRad]

"performance"? You mean efficiency?

Performance in efficiency and more criteria. Abruptly stopping current in a transformer generates a spike, possibly arcing. (It's characteristic of inductors.) Sinewaves are kinder in comparison.

There are different types of power supply which do apply square waves to transformers and inductors. A snubbing network is a common solution. It's a tradeoff which loses efficiency in one way yet improves efficiency in other ways.

You mean compared to a pulse trafo?

I'm convinced transformer manufacturers keep secrets regarding both types. Power transformers have a lot of ingenious technology built into them to make them useful.

You mean, a bipolar square that goes negative on the troughs?

Yes. You're asking two questions, one about transformer waveforms, the other about rectifying AC. In fact there are ways to get symmetrical AC square waves out of a transformer and turn them into smooth DC. On the other hand transfomers thrive on sine waves yet sine are full of ripple.

So which one will have more ripple?

• bipolar sine into power trafo?
• bipolar square into power trafo?

In earlier decades hobbyists and professionals built power supplies based on a transformer stepping down mains AC (sinusoidal). We barely thought twice about it. Generally the solution to ripple was to smooth it with caps with large Farad value. That's the usual solution in modern times too. However faster frequencies can make do with smaller cap values.

 

[theboom]

Abruptly stopping current in a transformer generates a spike

Same with reversing the current (as in a bipolar squarewave)?

A snubbing network is a common solution. It's a tradeoff which loses efficiency in one way yet improves efficiency in other ways.

The improved efficiency is trafo performance, but the snubber steals some current?

faster frequencies can make do with smaller cap values.

@KlausST says with a square, i may be able to eliminate the need for caps.

 

[KlausST]

So i guess i should use Ferrite Core?

Again: what magnitude of loss are you talking about?
0.1%? 1%? 10%? 50%?

I mean: we don´t need to waste time by talking about 1% of transformer loss ... while the full bridge rectifier will have about 30% loss.



Klaus

--- Updated Sep 16, 2024 ---

Same with reversing the current (as in a bipolar squarewave)?

Not quite correct.
An ideal tranformer makes the output waveform to be exactly like the input waveform.

You may reverse the input voltage ... this does not mean the input CURRENT reverses at exactly the same time. Current in an inductance is lagging.

You get a spike when the input and output goes high impedances ... while there is energy stored in the transformer core.
An inductance wants the current to maintain at the same level. But this are inductor basics ... and if you drive the transformer input LOW IMPEDANCE all the time it never will happen.

Klaus

 

[theboom]

Again: what magnitude of loss are you talking about?

I'll have to learn and research a bit more before I can answer that. Also important is: how much matters? I assume that has something to do with available heat sinking?

My use case is a sealed box without fans, so I can only depend on metal-core PCB, if that works, or some other passive heat removal.

we don´t need to waste time by talking about 1% of transformer loss ... while the full bridge rectifier will have about 30% loss

Feeding a square wave into the rectifier, will that be so inefficient? I'm guessing that since the current is spread out across the waveform, efficiency in the diodes will be less than with a sine wave, correct?

 

[BradtheRad]

Same with reversing the current (as in a bipolar squarewave)?

Yes, current wants to continue in either case, whether impedance is abruptly high or low. Each case has its hurdles.

The improved efficiency is trafo performance, but the snubber steals some current?

Another method that was frequently used: resistive drop (or linear drop). It's inefficient and generates heat. It's fallen out of favor as switching (PWM) topologies became popular. These families have their different waveforms and different hurdles and different methods to deal with inefficiency.

@KlausST says with a square, i may be able to eliminate the need for caps.

True, by starting with symmetrical AC square waves, then rectifying them.

 

[KlausST]

Feeding a square wave into the rectifier, will that be so inefficient? I'm guessing that since the current is spread out across the waveform, efficiency in the diodes will be less than with a sine wave, correct?

It´s the other way round.
flat DC would be the most efficient, due to ohmic losses.
A simulation would show you ... but it seems you don´t take me serious on this.

You come with unvalidated ideas .. and we have to rectify them. You can not excuse this with "low experience" since there are ways to find out on your own within seconds.
--> A forum can´t replace school, can´t replace doing own research, can´t replace doing your job...

Also important is: how much matters?

An electronics design should START with design goals.
In this case: NOT "as cool as possible" (this is no design goal) but with something like: "a max temperature rise of 40°C, nice to have: only 15°C ... for example.
Then you need to decide the enclusure size and material, vented or not, external heat sink or not, fan or not ... and the ambient temperature range. This enables you to get a clue about the max power dissipation = loss of your application.

One needs numbers. Especially when you want people work together - like here in the forum.
And one needs to focus on the part with the highest loss (= disspation) first.
***
And back to your application: A bridge rectifier for 3A average with standard diodes may dissipate about 0.93V x 3A x 2 = 5.5W. That´s a lot!
(I just chose an example diode BY 550-600)
***

@KlausST says with a square, i may be able to eliminate the need for caps.

Read carefully what I worte:

In ideal case the rectifier output is pure flat DC. No fundamental, no harmonics. In ideal case not even a filter is needed.
For sure real life is not ideal, thus I wrote that parts need to be selected carefully.

Since I guess you want to design for the real life .... it means you will need some kind of energy storage. It may be small ... maybe it´s already installed in your load....

We still don´t have the full idea of your design. We don´t know where the energy comes from, nor do we know where it goes.
You say "low ripple" but we have no numbers.

****

Don´t get me wrong: I like to help, but I don´t want to waste my time by playing guessing games. Thus I will be back when I see you do your job.

Klaus

 

[betwixt]

For clarification, you mentioned 100KHz but much of this thread seems to relate to lower frequencies, especially when it comes to losses. Can you confirm the frequency and maybe what the expected load would be.

Brian.

 

[D.A.(Tony)Stewart]

something new for me to learn about! Something like this?
View attachment 193890

This is 3 phase generally low f and sine.
Switched mode will be pulse mode voltages from switches. Diodes are also switches but loss is not just I*I*Rs but also I*Vf which is bigger and only exists in Darlington BJTs.

Massive power transformers are more efficient since the spectrum is very small +/- 10% for example compared to a SMPS. Due to harmonic content of pulses dictates more careful selection of magnetics since useful frequency range for low impedance or I*t energy storage are tradeoffs with saturation and Eddy loss. Both L, DCR are critical vales as well as Tau= L/DCR. But the math may have other complexities for closed loop stability.

When you restrict the mechanical design you must impose these thermal limits into concrete power limits at max internal ambient then choose a design that meets all variables in your design spec before you deign anything, not be surprised after.

When you need low EMI, low power, high effciency , no shielding , low mass magnetics, low cost AND high voltage , then you avoid SMPS and choose high frequency sine waves and linear control to regulate the voltage.

NOTE:
Because a sine FW bridge is only active during voltage peak detector the current pulse width tends to-be <10% duty cycle or as small as the % ripple needs to be. Now you have 9 harmonics ! So filtering is a nonlinear tradeoff between RLC 1st order lossy and 2nd order non-lossy LC series shunt attenuation and also more math or simulation required. (Which is far easier?)

 

[theboom]

something like: "a max temperature rise of 40°C, nice to have: only 15°C ... for example.
Then you need to decide the enclusure size and material, vented or not, external heat sink or not, fan or not

Unfortunately that's not an option. I have to design for a sealed enclosure without a fan, or give up. The interior of the enclosure will be at room temp, about 26c/80f. The chamber interior temp mustn't exceed 70c/158f deg. Let's assume the chamber will reach the same temp as the components. So that gives about a 55c/80f deg allowed rise above ambient. I don't know if that's the right way to figure this.

you mentioned 100KHz but much of this thread seems to relate to lower frequencies

I want to operate at 100 kHz min, preferably 150 kHz or higher.

choose a design that meets all variables in your design spec before you deign anything, not be surprised after.

You've brought up important issues regarding setting overall goals and constraints. Why do you assume i haven't? This post is asking about ripple.

When you need low EMI, low power, high effciency , no shielding , low mass magnetics, low cost AND high voltage , then you avoid SMPS and choose high frequency sine waves and linear control to regulate the voltage.

EMI, low-mass magnetics: I'm targeting 100 kHz - 400 kHz to avoid audio and commercial radio freq's, and to keep components small.
Low Voltage, Power: approx 5V at 5A. Does this qualify for high-f sine?
Efficiency: See temperatures above.
Cost: Will leave that aside for the moment. First i need to understand the available options.

"choose high frequency sine waves" - How high? Does any of this say "sine waves" to you?

Because a sine FW bridge is only active during voltage peak detector the current pulse width tends to-be <10% duty cycle or as small as the % ripple needs to be.

"as small as the % ripple needs to be"? A shorter-duration current pulse-width leads to less ripple?