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

theboom

Member level 2
Joined
Jun 4, 2022
Messages
47
Helped
1
Reputation
2
Reaction score
0
Trophy points
6
Activity points
281
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).
 

Sort by date Sort by votes
theboom said:
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.
Click to expand...
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.
 
Upvote 1 Downvote
Thanks for these tips.
BradtheRad said:
A power transformer gives most satisfactory performance from being fed bipolar sine waves.
Click to expand...
"performance"? You mean efficiency?

You mean compared to a pulse trafo?

BradtheRad said:
a full-bridge rectifier provides smoothest DC if you feed it square waves where the positive waveforms are equal amplitude as the negative.
Click to expand...
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?

BradtheRad said:
A capacitive filter tends to make output rise near the peaks of pulsating DC. A choke filter tends to average the same pulsating DC.
Click to expand...
Which one will have less ripple?
Thx!
 

Upvote 0 Downvote
KlausST said:
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
Click to expand...
Thx, but did you forget the transformer?
What about all those square harmonics?
 

Upvote 0 Downvote
theboom said:
Thx, but did you forget the transformer?
Click to expand...
Sorry, you are right.

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

theboom said:
What about all those square harmonics?
Click to expand...
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
 
Upvote 1 Downvote
KlausST said:
In ideal case the rectifier output is pure flat DC. No fundamental, no harmonics. In ideal case not even a filter is needed.
Click to expand...
That's just a matter of centering the squarewave vertically before it enters the rectifier, correct?

KlausST said:
For sure real life is not ideal, thus I wrote that parts need to be selected carefully.
Click to expand...
Square reduces trafo efficiency, correct?

thx
 

Upvote 0 Downvote
theboom said:
That's just a matter of centering the squarewave vertically before it enters the rectifier, correct?
Click to expand...
When talking about a bipolar square wave ... we are not discussing unsymmetric bipolar sqaure wave.

theboom said:
Square reduces trafo efficiency, correct?
Click to expand...
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
 
Upvote 1 Downvote
KlausST said:
When talking about a bipolar square wave ... we are not discussing unsymmetric bipolar sqaure wave.
Click to expand...
"In ideal case the rectifier output is pure flat DC."
Click to expand...
You used the word "ideal" a couple times. What causes deviation from the pure flat?


KlausST said:
Why do you think so? Provide your idea. Also what magnitude of efficiency/loss are you talking about.
Click to expand...
So i guess i should use Ferrite Core?


KlausST said:
In post#1 you ask about low ripple. Thus I answered about low ripple, not about high efficiency.
Click to expand...

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.


KlausST said:
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.
Click to expand...
 

Upvote 0 Downvote
theboom said:
"performance"? You mean efficiency?
Click to expand...
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.

theboom said:
You mean compared to a pulse trafo?
Click to expand...
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.

theboom said:
You mean, a bipolar square that goes negative on the troughs?
Click to expand...
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.

theboom said:
So which one will have more ripple?
  • bipolar sine into power trafo?
  • bipolar square into power trafo?
Click to expand...
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.
 
Upvote 1 Downvote
BradtheRad said:
Abruptly stopping current in a transformer generates a spike
Click to expand...
Same with reversing the current (as in a bipolar squarewave)?

BradtheRad said:
A snubbing network is a common solution. It's a tradeoff which loses efficiency in one way yet improves efficiency in other ways.
Click to expand...
The improved efficiency is trafo performance, but the snubber steals some current?

BradtheRad said:
faster frequencies can make do with smaller cap values.
Click to expand...
@KlausST says with a square, i may be able to eliminate the need for caps.
 

Upvote 0 Downvote
theboom said:
So i guess i should use Ferrite Core?
Click to expand...
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 ---

theboom said:
Same with reversing the current (as in a bipolar squarewave)?
Click to expand...
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
 
Last edited:

Upvote 0 Downvote
KlausST said:
Again: what magnitude of loss are you talking about?
Click to expand...
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.

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

Upvote 0 Downvote
theboom said:
Same with reversing the current (as in a bipolar squarewave)?
Click to expand...
Yes, current wants to continue in either case, whether impedance is abruptly high or low. Each case has its hurdles.

theboom said:
The improved efficiency is trafo performance, but the snubber steals some current?
Click to expand...

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.

theboom said:
@KlausST says with a square, i may be able to eliminate the need for caps.
Click to expand...
True, by starting with symmetrical AC square waves, then rectifying them.
 
Upvote 0 Downvote
theboom said:
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?
Click to expand...
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...


theboom said:
Also important is: how much matters?
Click to expand...
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)
***

theboom said:
@KlausST says with a square, i may be able to eliminate the need for caps.
Click to expand...
Read carefully what I worte:
KlausST said:
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.
Click to expand...
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
 
Upvote 0 Downvote
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.
 
Upvote 0 Downvote
theboom said:
something new for me to learn about! Something like this?
View attachment 193890
Click to expand...
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?)
 
Last edited:
Upvote 1 Downvote
KlausST said:
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
Click to expand...
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.

betwixt said:
you mentioned 100KHz but much of this thread seems to relate to lower frequencies
Click to expand...
I want to operate at 100 kHz min, preferably 150 kHz or higher.

D.A.(Tony)Stewart said:
choose a design that meets all variables in your design spec before you deign anything, not be surprised after.
Click to expand...
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.

D.A.(Tony)Stewart said:
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.
Click to expand...
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?

D.A.(Tony)Stewart said:
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.
Click to expand...
"as small as the % ripple needs to be"? A shorter-duration current pulse-width leads to less ripple?
 

Upvote 0 Downvote
You must log in or register to reply here.

Similar threads

Menu
Log in
Register
Cookies are required to use this site. You must accept them to continue using the site. Learn more…