3

I want to design a power supply based on a line frequency transformer (50 Hz in my case), and provide a stabilized voltage (15 V) after the rectifier (a classic bridge rectifier with tank capacitor).

To create this stabilized voltage, I would like to use a buck converter such as the AP63200 from Diodes Inc.

I don't want to put a giant tank capacitor, so the ripple will reach up to 3 V (over a dozen of volts) at the input of the converter.

My questions :

  • is this ripple problematic for a DC/DC converter?
  • why Power supply rejection ration (PSRR) is never specified for DC/DC? Is this value meaningless?
  • why people always use LDO after a line frequency bridge rectifier?

Do you think it doesn't make sense to design a power supply using 50 Hz transformer, rectifier and buck? Should I better use a high frequency transformer like in most AC/DC supplies?

I would like to stay on a 50 Hz transformer for robustness against line transients, endurance and EMI concerns.

Thank you for all you what you can tell me about power supply designs based on line filter transformer!

Altheo
  • 31
  • 1
  • 2
    *why Power supply rejection ration (PSRR) is never specified for DC/DC?* because PSRR is the term we use for amplifiers, opamps etc. For DCDC converters we use **line regulation** although that's not exactly the same but the response to a **slowly** changing input voltage. – Bimpelrekkie Apr 16 '21 at 07:45
  • *why people always use LDO after a line frequency bridge rectifier?* What makes you think this? It might more common to use a **linear regulator** (which doesn't have to be an LDO) after a line transformer + bridge rectifier but that doesn't mean it has to be like that. In most products such an "analog supply" isn't used anymore and replaced by a high voltage to low voltage (often isolated) power supply (an SMPS). – Bimpelrekkie Apr 16 '21 at 07:49
  • 1
    *Should I better use a high frequency transformer like in most AC/DC supplies?* The transformer isn't the only difference there. The whole working principle of such a **flyback converter** is quite different. **You would only design that if you have some experience**. The line transformer + bridge rectifier + DCDC (module) is much more beginner friendly. You could also consider using an existing SMPS (like a 19V laptop supply) to make the input voltage for your DCDC. – Bimpelrekkie Apr 16 '21 at 07:52
  • *I would like to stay on a 50 Hz transformer for robustness against line transients, endurance and EMI concerns.* These issues can be and are solved in existing SMPSs, that's why you should not design your own unless you have the experience to solve these issues. That doesn't mean you can't use an SMPS module that already exists instead of your line transformer + bridge rectifier + smoothing cap. – Bimpelrekkie Apr 16 '21 at 07:55
  • Hello Bimpelrekkie, thanks for all your comments! As I understand : - line regulation is used to specify dc/dc converter behaviour to "slow" line changes (how slow ? is 1kV/s considered to be slow ?) - nowaday, if you need to design an isolated power supply, you use a flyback topology or similar, not a line transformer+bridge followed by a buck. – Altheo Apr 21 '21 at 08:00
  • My problem is, I need to design a small (5..10W) 12 V power supply, with 4 kV isolation, 4 kV mains transient voltage, EN55022 class B compliance, with its secondary connected to eart. I've experienced many of the shelf supplies (meanwell, recom, xppower, and every time, we have had to put an expensive filter (common mode choke, X and Y capacitor) at the input to solve EMI problem (and even with the filter, the margin was not big, thats why I was looking for an easier to filter alternative. (And I don't want to use an LDO for regulation after the bridge because of it's really low efficiency. – Altheo Apr 21 '21 at 08:02
  • *I need to design a small (5..10W) 12 V power supply, with 4 kV isolation, 4 kV mains transient voltage, EN55022 class B compliance....* Sounds like a challenge! I would first have a look if supplies / supply modules like that **even exist** also consider that what you want might come at the cost of size. My point: if a supply has all the specs you want except it is **too large** then **maybe** it is not feasible to fit all that is a small size. I mean: if "the experts" cannot do it, what makes you think that you can? I mention this to prevent you from trying to design the impossible. – Bimpelrekkie Apr 21 '21 at 11:57
  • @Bimpelrekkie How is "line regulation" specified? I saw it given as "-0.05%/V to %0.05/V". What does this mean exactly? What is considered *slowly* (wrt to switching frequency)? What do I do with higher frequencies? I have an input waveform given (=broadband noise) and want to know why well my converter rejects this noise. How do I figure this out? – divB Dec 10 '21 at 09:33
  • @divB Read: https://en.wikipedia.org/wiki/Line_regulation and also the link at the bottom of that page. "Slowly" means, so slow that the feedback loop in the regulator can adjust for those changes. So line reg. then becomes a parameter for **DC**. For **AC** the frequency needs to be specified as the feedback loop (keeping the output voltage constant) can only work up to a certain speed. Above that (higher frequencies) we use **decoupling capacitors**. *how well my converter rejects* Read the datasheet and do simulations. – Bimpelrekkie Dec 10 '21 at 09:59
  • @Bimpelrekkie It seems PSSR for switching regulators can be defined, see: https://www.analog.com/en/technical-articles/comprehensively-understand-and-analyze-switching-regulator-noise.html . This is the only place I found it. Are you aware of any literaure/references that explains this in more detail? I am still struggling with the same problem: my buck input voltage has *arbitrary* Noise(t) and I want to know how it affects the output voltage (in frequency range below switching frequency). – divB Jan 08 '22 at 02:27

2 Answers2

1

With this answer, I want to focus on the part that was not answered yet:

why Power supply rejection ration (PSRR) is never specified for DC/DC? Is this value meaningless?

It is neither meaningless nor unspecified but depends on external components such as feedback loop, L, C, load.

In a small-signal, ac model, the PSSR can be shown to be

$$ \operatorname{PSSR}(f) = \frac{v_o(f)}{v_i(f)} = \frac{G_{oi}(f)}{1+L(f)} $$

where \$G_{oi}(f)\$ is the small signal open loop transfer function of the regulator, and \$L(f)\$ is the loop gain (which is product of sensor gain, compensator transfer function and small signal DutyCycle-to-output transfer function).

In the bandwidth of the feedback loop, \$|L(f)| \gg 1\$ and hence the loop greatly rejects noise.

I recommend Fundamentals of Power Electronics by D. Erickson for details and great reference.

divB
  • 1,312
  • 14
  • 32
0

To answer your questions:

  1. "Is this ripple problematic for a DC/DC converter?"

The ripple at the output is directly related to your output capacitor and your output inductor. If your ripple is too high that means the current through the inductor is too high or your output capacitor is too low.

In the case that your output cap is too low there's not issue from a converter perspective given that you're not exceeding the maximum stress voltage on your transistors (highly unlikely). However, just because there's not problem from a converter perspective doesn't mean that you have a solution. You'll have to figure out if your load is sensitive to this ripple or not.

In the case that your inductor current is too high (the current grows linearly at a rate of (Vin-Vout)/L * t) then your inductor might saturate (if magnetic) and straight up burn whether magnetic or air core.

  1. "Why Power supply rejection ration (PSRR) is never specified for DC/DC? Is this value meaningless?"

The ripple at the output is a direct function of output inductor, output cap, and switching frequency. The control loop takes care of noise in your input supply (provided the noise is within the loop bandwidth). If it's higher you wouldn't see it.

  1. "Why people always use LDO after a line frequency bridge rectifier?:

It's not "always", it's application specific. If your load can handle the output ripple you can get away with using a dc/dc converter. But in general applications the manufacturer doesn't know how it'll be used to they air on the side of caution and give a steady output.

Also, it may be cost related. dc/dc converters are always more expensive due to it's control and inductor. Many times more expensive than a simple LDO.

ByTor_Hardware
  • 197
  • 7
  • Hello ByTor! Thank you for your reply :) My question was really not about the output ripple, which is a well documented topic, but about the effect of an high input ripple from a line frequency bridge rectifier (100 or 120 Hz ripple, 10-20 % of the avg voltage). But maybe, as you wrote it : if the control loop of the DC/DC converter is fast enough, this 100-120 Hz input ripple will not be a problem, as long as the dutycycle remains in an acceptable range. Maybe should I just pay attention to the "line regulation" characteristic ? Why do people never put a buck after a line freq. rectifier? – Altheo Apr 21 '21 at 07:48
  • The control loop bandwidth should take care of that, but it creates "noise" as well. Remember that the current through the inductor is given as (Vin-Vout)/L*t. Let's use silly numbers to illustrate a point. Assume Vin 2V, Vout = 1V and L = 1nH. That means for every nS you're building 1A of current. In your situation your L is not varying but your Vin is. You'll could have fluctuations in current through the inductor. Again the control loop should take care of that. The main reason is cost. – ByTor_Hardware Apr 21 '21 at 17:49